Home

Measuring the Electron Beam Energy in a Magnetic

1. i i i i i i l i Mg i l i i al 0 0 5 1 1 5 2 2 5 3 3 5 4 4 5 5 0 0 5 1 1 5 2 2 5 3 3 5 4 4 5 5 ns ns 1 m mn A eg i man T FE i ei Alf Try fi a S i 1d S e 14 N II I I i N II i Ly RAAT A All CK AWA AAA N N I I hI AAU WE I Ii T i HH f f HVA Ih NAAT af Mi WAM II WTI I I i WTI H
2. 1 5 i i i i l i i i i 1 5 i i i i i i i i l 0 0 5 1 1 5 2 2 5 3 3 5 4 4 5 5 0 0 5 1 1 5 2 2 5 3 3 5 4 4 5 5 ns ns Figure 7 1 4 RF signal composed by sum of two different frequencies mixed is with the LO red the output of this mixer is shown in green As the phase of the LO is changed by a few degrees about the zero crossing of the RF signal shape of the output of the mixer changes in a manner that 1s similar to that which is observed in the pulsed case from Fig 7 1 3 The blue line is the result of low pass filtering the green signal If the amplitude of one of the signals shown in either Fi
3. 260 1 degree room ee eer temperature change cable drift ih 25 um 15 um u HH 2 u Ih i CANT IIA 5 y A a VAa Ao a o INN Il yobon I JUN 111 i 190 180 170 160 0 1 2 3 4 5 6 7 hours Figure 7 2 6 Resolution of the 10 4 GHz front end with long cables This was evaluated by splitting a signal from the beam pickup directly after the pickup sending the signal over 30 meter long cables to the front end chassis and measuring the difference between the arrival times of the pulses The influence of the temperature dependent drift of the long cables is frequently smaller than the influence of the temperature stability of the front end chassis The temperature control of the chassis was not operating during this measurement From Fig 7 2 6 t can be seen that the influence of the temperature dependent drift of the 30 meter long cables can be smaller than the influence of the temperature stability of the front end chassis To be fair the temperature control of the chassis was not working properly during this measurement and a large drift of two cables that are of the same type and run parallel to one another would not be expected since the majority of the drifts w
4. 2 systematic error 15 statistical error linear fit fo simulation D 1 i O N O V o5 O w 0 0 5 Q O Q E 1 gt Mm 1 5 2 4 3 2 1 0 1 2 3 4 Bump in 9ACC1 mm Figure 4 3 4 Beam tilts in the middle of the first bunch compressor measured and simulated Each tilt corresponds to a defined orbit bump through the first accelerator section The individual measurement points for two different measurements are represented as blue and black solid dots The beam is not only tilted in the x z plane but also in the y z plane After the second bunch compressor the beam is streaked vertically in the transverse deflecting cavity On the streaked beam image one can see that beam 1s tilted resulting n a slice emittance that is a factor of two smaller than the projected emittance Measurements conducted with the transverse deflecting cavity indicate that the head of the bunch is separated from the tail by 200 um In 32 sources for this tilt originating downstream of the first bunch compressor were simulated and only a third of the tilt measured with the transverse deflecting cavity could be accounted for when a non tilted distribution in the first bunch compressor was used This simulation propagated a particle distribution from the exit of the first bunch compressor through the transverse deflecting cavity This 48 leads one to sus
5. electron 1 beam copper cavities coaxial coupler waveguide Figure 2 1 1 Simplified cross section of the RF Gun not to scale A laser pulse is impinged upon a cesium telluride cathode and the electrons that are ejected are accelerated in an RF cavity a solenoid provides additional focusing In order to produce the high charge density necessary for the FEL lasing process the gun is designed to produce as low an emittance as possible Emittance represents the transverse extent of the beam and when it increases the charge density decreases In terms of the horizontal particle distribution of the beam x the emittance is defined as E x x xx 2 1 1 and the width of the beam is given in terms of the emittance and the beta function a function which describes how much the beam is focused at a given phase in the beam transport lattice o Be 2 1 2 The same definitions also apply to the vertical particle distribution A long list of effects must be correctly balanced in order for the non relativistic beam to emerge from the photoinjector with a low emittance non linear space charge effects the gradient at the cathode RF field distortions residual magnetic fields and wakefield kicks name the largest effects 9 Each of these effects can distort the shape of the bunch creating and exacerbating asymmetries The solenoid field counteracts some of the space charge effects by focusing the beam radially wh
6. 3 2 1 0 1 2 y position mm es Simulated Vertical Position Sensitivit 90 l 80 70 60 j so Bottom pickup Top pickup 40 l Fi b Pd 30 20 o o 10 ee eee al en a 3 2 1 0 1 2 3 4 y position mm Measurement top and simulation bottom of chicane BPM pickup signal amplitude response to changes in y position If the top and bottom signals of the monitor are combined with equal length cables a cancellation of this dependence will occur While it is possible to reduce the dependence experience with signal combination suggests that it is not possible to entirely remove it T1 While the impact of the beam charge and y position is fairly easy to predict the impact of the horizontal position spread of the beam is not as obvious A measurement of the influence of the phase of the upstream accelerator section on the slope of the pickup signal is shown in Fig 5 5 10 The amplitude changes of the pickup signal due to RF phase changes are weak and will not impact the resolution of the monitor by more than a few percent BPM slope 1 5 h iN j 2k l slope at zero crossing V ps se 4 ae 20 15 10 5 0 phase deg Figure 5 5 10 Dependence of the slope of pickup signal on the phase of the upstream accelerator section The upstream phase determines the energy spread of the beam and therefore the beam width in the chicane Right blue and left red pickup output slopes plott
7. x 0 tail 5 5 4 3 2 1 0 1 2 3 4 35 z mm Figure 6 5 1 Particle tracking simulation of a nicely matched beam at the location of the BC2 BPM A beam tilted in the x z plane is depicted alongside a bar representing the pickup in Fig 6 5 2 One might naively think that since it appears that the head of the bunch would arrive 14 ps earlier than the tail there would then be a 4 mm error in the measurement of the center of the beam but this would be mistaken error 14 ps Incorrect interpretation gt 4mm Figure 6 5 2 The beam tilted in x z plane relative to the pickup A naive estimate of the effect of this tilt is sketched alongside the bunch While it is true that the bunch is tilted in this plane and that the separation between the head and tail in the x z plane is about 4 mm the interpretation shown in Fig 6 5 2 ignores the Doppler like effect of the tilted beam coupling to the pickup 91 The Doppler effect refers to the change in the frequency of a signal that is measured by an observer which is moving relative to a source While the frequency of the signal changes the phase does not If the beam is tilted it appears that the beam is traveling more quickly in the reference frame of one end of the pickup while in the reference frame of the other end of the pickup it appears that the beam is traveling more slowly If this still doesn t seem plausible imagine the beam divided into many small
8. AQ Co 1 E E j i u pp a er k 1 rf X Api where C C C gt Cok the result Eq 3 2 2 for the net timing jitter after two bunch compressor stages 1s l l E Oy R E E 2 r ea 2 Ru ERs 142 1 To 0 E v Co E V C Cok rf 171 Appendix B A description of how the one can measure and remove the effect of reflections in the pickup from the RF beam position measurement follows There is a constant for each pickup output that is a measure of the effect of the reflections within the pickup on the arrival times measured at the outputs of each pickup This is given for the right R and left L pickup outputs in terms of the beam arrival time t beam position x and as ar Ay and At IN owa OR a 1 C C At At arrival In terms of the beam energy and constants of the chicane the beam arrival time and position are given by At Af and Ax Ri E 2 arrival Ru AE AE 2 E E where fg is the arrival time before the chicane For higher energies the arrival time is earlier and the beam position moved to the right in the positive direction Writing Eqs 1 in matrix format At l A A 3 At 1 a Axle one can calculate the inverse of the 2x2 matrix in order to find the beam arrival time and position in terms of the measured arrival times and reflection coefficients Al crival l AR l At 4 Ale a a a 1 At 172 References
9. 4 lt BC2 L a Fil a 8 89ps 3 gt BC2 R zZ Fit a 15 83ps 29 am N a I gt O O0 a ep amp T Z 2 lt 3 g 4 0 2 0 15 0 1 0 05 0 0 05 0 1 0 15 0 2 ACC1 voltage change Figure 10 4 1 Verification of the cal bration of optical front end beam position measurement The rat o of the slopes of the energy dependent arr val time change of the signal from the left side of the pickup to that coming from the right side of the pickup should be equal to the ratio of R56 2 R 6 Rs6 2 R16 Both calibrations could be still be wrong in the same proportion say by a factor of two error or so but checking the measurements relative to an external reference can rule that 158 out This was done in two ways by changing the setpoint of the accelerating gradient and checking that the beam position changed by an expected amount Fig 10 4 2 10 4 3 and by deriving a measurement of the beam arrival time upstream of the chicane from the beam position and arrival time and comparing that to a measurement of the beam arrival time measured upstream of the chicane with a button type BAM monitor Fig 10 4 4 In Fig s 10 4 2 and 10 4 3 the accuracy of the chicane BPM optical front end was checked by changing the energy of the beam with the accelerating gradient setpoint and measuring how much the beam position changed In Fig 10 4 2 the beam energy was changed b
10. 8 10 12 14 16 18 20 22 hours Figure 10 2 2 10 4 GHz chicane BPM front end measurement and photomultiplier tube PMT BPM measurement in good agreement 155 Benchmarking RF chicane BPM against PMT BPM 0 4 0 2 o N ww BO2 PMT BC2 BPM energy change FY Y VIEW a g W YA oth E Veo VAY yor ee a l 0 8 g 5 10 15 20 25 hours Figure 10 2 3 10 4 GHz chicane BPM front end measurement and photomultiplier tube PMT BPM measurement in poor agreement One or both calibrations are wrong In Fig 10 2 3 either the calibration constant of the PMT BPM is too small or the calibration constant of the RF chicane BPM is too large In general the PMT BPM is calibrated by scanning the ACC1 gradient setpoint and the RF chicane BPM is calibrated by scanning an RF reference with a vector modulator As was described in Section 7 2 this sort of calibration is subject to errors due to reflections in the pickup unless the effect of the reflections is measured and removed The PMT BPM has also been calibrated by scanning a the position of a stage but when comparisons of calibrations done with ACC1 gradient setpoints dipole current and stage position differences of 10 20 were observed due to a high level of energy drift that impacts the repeatability of the calibrations 56 The result is that neither monitor provides a high level of confidence
11. Fine HF front ends position measurement and photomultiplier tube position measurement in good agreement Fine HF front ends position measurement and photomultiplier tube position measurement in poor agreement Fine HF front ends position measurement and photomultiplier tube position measurement Optical EOM front end position measurement and photomultiplier tube position measurement along with a time of flight measurement involving 2 BAMs and a line showing how the setpoint of the gradient changed The beam energy was changed by 0 3 with the accelerator gradient setpoint and the beam energy measured by the chicane BPM changed by a comparable amount Optical EOM front end chicane BPM measurement and photomultiplier tube BPM measurement along with a time of flight measurement involving 2 BAMs and a line showing how the setpoint of the gradient predicted an energy change of 0 1 The beam arrival time upstream of the chicane measured with both the transversely mounted stripline BPM installed in the chicane and with a button type pickup BAM installed upstream of the chicane Optical EOM front end position measurement 10 4 GHz front end measurement photomultiplier tube position measurement time of flight measurement involving 2 BAMs and the setpoint of the gradient are plotted together over several hours 1 Introduction Knowing the exact position of an electron beam under the influence of a magnetic field has been at the heart of
12. Il Mh BC2 x position mm 0 5 1 5 0 0 5 1 1 5 2 2 5 change ACC1 gradient Figure 7 2 8 Scanning the gradient of the first accelerating module and measuring the change in the position of the beam with the chicane BPM The position measurement is only in range for the first few data points Zooming in on the first 0 5 of the scan shows the mm linear range of the monitor Fig 7 2 9 One percent energy change times the R s of the chicane gives 3 5 mm of expected position change The position change measured by the monitor in the linear range was 3 5 0 1 mm Die 0 8 BC2 x position mm o o gt O I 2 N MIL IM 0 1 0 2 0 3 0 4 0 5 0 6 0 7 change ACC1 gradient oO oO Figure 7 2 9 Beam position change corresponding to a small energy change 113 When good agreement with expectations is observed at one measurement location it is tempting to assume that such agreement will aslso be observed at other locations but this is not the case for this RF front end Due to reflections within the pickup which are one third to one half the amplitude of the initial beam transient the calibrations produced by scanning the vector modulator are wrong by a factor of two For some beam positions the errors in the calibrations exactly cancel out giving a correct measurement of the beam position but for other beam positions the measurements done with these
13. ER pickup u C Cy G d 9 TC 5 1 25 pickup pickup where d is the diameter of the button and r is the distance from the beam This is of course not the case in an FEL where each bunch is less than a millimeter in length in 58 some locations Nevertheless it is convenient to ignore the time dependence at this stage as the equation is true at least for wavelengths longer than the button size and it gives us a useful rule if the vacuum chamber diameter quadruples the button size only needs to double in order to conserve the amplitude of the signal Now we consider the load R determined by the cable impedance and the termination of the cable In the frequency domain the voltage at the load is given by 45 R e R 1 ioC pickup V o V 5 1 26 where Vpickup 18 given by Eq 5 1 19 Then according to Ohm s law the transfer impedance is given by 2 A E l IL 5 1 27 Src C eg R 1 i C ae Taking the absolute value of Z we see the behavior of a high pass filter 2 IZ ee 5 1 28 M ORC u We also see that the transfer impedance is more sensitive to the button diameter than to the capacitance of the pickup which is generally less than 10 pF and can be calculated from Eq 5 5 23 a multiplied by the length of the button This is the reason that button pickup BPM front ends that use wavelengths which are longer than the size of the button use buttons with a diameter that is maximiz
14. based system because beam based measurements can often provide a more accurate measure of how much energy the beam gained in the cavity than a measurement of how much energy the cavity lost when the beam traversed it The reason for this is that the beam based measurement can often be accomplished in one location with one pickup and one set of electronics while the cavity measurement requires a sum of signals from many different pickups in an RF vector sum The beam based measurements can also be drift free relative to an optical reference which is used to synchronize various laser systems through optical cross correlation It is not possible to use an RF reference to synchronize the beam to lasers with femtosecond precision The beam position monitor developed in this thesis is ideally suited for use in a beam based feedback system because it provides beam energy measurements with lt le 5 resolution for every bunch in the train as well as a measurement of the beam arrival time in the chicane A pair of beam arrival time monitors installed up and downstream of the chicane provides lower resolution information about the energy changes of the beam and higher resolution information about the arrival time of the beam at the chicane Because of the systematic errors that both the chicane BPM and the arrival time monitor suffer from when the beam shape changes an ideal solution is to use both measurements to cross check one another No other existing monito
15. reference These signals are shown below n Fig 8 3 2 139 BP filtered and amplified 10GHz PD signal 20 e 0 1215 Amp 1150 Amp 20 M 2 w 40 D 100 0 2 4 6 8 Frequency GHz Figure 8 3 2 One frequency is filtered out of the frequency comb of pulsed laser signal on photodetector The noise in the low frequency part of the red trace is an artifact of the spectrum analyzer Before worrying about the drift and noise of the photodetector a picture of the noise density and drift of the phase measurement circuit will be made Using a signal generator as shown in Fig 8 3 3 the noise contribution of the amplifiers voltage regulators and power supplies could be studied By mixing the signal generator s output with itself we can see the performance of the circuit without the influence of the noise from the photodetector or from the MO signal The formula for measuring the Kg of the circuit is also shown in the figure The Kg increases when the signal is amplified For a low gains in the last amplifier stage LNA the signal to noise ratio of the circuit s output is poor but above a certain gain increasing the gain further will not produce any improvement in the signal to noise ratio The gain was selected so that the signal would not saturate the ADC signal generator Ky dV dd LEMG datalogger LNA Figure 8 3 3 Setup for measurement of
16. trajectories are bent less than those of low momentum particles the high momentum particles will travel a shorter path through the chicane and will arrive at the end of the chicane earlier than the low momentum particles We would like to know how a momentum change of a group of particles will affect the path length through the chicane and the horizontal position in the chicane To find this we will Taylor expand Eq s 1 5 in terms of a small change in momentum 6 Ap p This is given by d sc 64 eg Op 2 Op tReet Rs t Ipc PU 0 lee p 1 7 O pc 54 PO Asc O Op 2 p x t Ret Reg 0 Xpc PUL 0 X_0 P where Rss Rs66 Ris and Rie6 are functions of the magnetic field and the effective length of each magnet They are named after their locations n a transfer matrix used to calculate beam transport and they are used to predict the arr val time and x position changes of particles traveling through dipole fields for given momentum changes For short bunches like those in FELs the R566 and R ss terms are typically small compared to the R56 and R s terms So for the majority of the calculations in this thesis only the first order terms will be used The first order terms are frequently referred to as momentum compaction Rs6 0 and linear dispersion R75 D Doing the first order derivatives of Eqs 1 6 one finds 4 2d eBL p eBl en _ 2d leBley p 1 p Paes Oe P R N eB
17. 10 cross talk of a pair of photomultiplier tubes this gives an estimate for the position uncertainty ranging from 7 um for a 4 mm wide on crest beam to 34 um for a 2 cm wide off crest beam Since the resolution of the monitor as measured by comparing the relative jitter of neighboring bunches and by comparing the measurement to that of the synchrotron light camera is 4e 5 15 um for an on crest beam and 9e 5 30 um for an off crest beam 57 only slightly worse than the best case resolution predicted by photon statistics alone it is fair to judge that this resolution is limited primarily by shot noise The monitor has been calibrated with dipole scans accelerator gradient scans and motor scans Ideally all three calibrations would match with the dipole scan being the gold standard While the three calibrations were sometimes within the error bars of one another differences of 10 20 between the three different calibration routines have frequently been observed It is suspected that this is due to a high level of energy jitter and drift that impact the repeatability of the calibrations 57 151 10 Energy Measurement Benchmarking The chicane BPM Ch 7 the arrival time measurements around the chicane Ch 8 the photomultiplier monitor Ch 9 and the out of loop vector sum Ch 3 all provide measurements of the beam energy A table comparing these measurements is shown below The monitors constructed during this thesis are highli
18. 5 1 10 5 1 11 ILIZ 5 1 13 5 1 14 5 1 15 3 241 Ss 5 4 1 54 2 5 5 1 3 9 2 5 5 3 5 5 4 5 5 5 5 5 6 3 37 5 5 8 5 5 9 Beam tilts in the middle of the first bunch compressor measured and simulated A cross section of button pickups in a round vacuum chamber left and button pickups in a flat vacuum chamber right The electric field lines of charged particle beam moving at much less than the speed of light left and close to the speed of light right Coordinate system for a circular vacuum chamber Dimensions of pickups in three different configurations Sensitivities of button pickups in flat chamber and round chamber configurations Electrical circuit representation of a small section of transmission line Button geometry that keeps the impedance constant keeps the ratio between the inner and outer conductor constant Button pickup equivalent circuit Approximation that bunch is longer than button radius allows for integration over the beam current in steps of At Definition of two port S parameters in terms of incoming wave A and outgoing wave B Comparison of frequency and time domain simulations of BAM and BPM pickups Cross sections of the old a and new b beam arrival time pickups Ring pickup output with red and without blue limiter Position dependence of pickup output slope with and without combiner BAM pickup designs Side view cavity BPM Longitudinally oriented stripline BPM princi
19. Fig 7 2 11 Fitting a line to the middle portion of the left hand plot and to this first half of the right hand plot gives a net arrival time change of 2 3 0 2 ps degree phase change For one degree of phase shift the sum of the laser and RF phase shifts should produce 2 125 picoseconds of arrival time change This is within the error bars of the measurement 114 3 0 0 2 ps deg 0 7 0 2 ps deg Chicane BPM 7 linear range BC2 arrival time ps fe O1 BC2 arrival time ps i i L I l 4 10 12 12 5 13 13 5 14 14 5 15 15 5 16 160 160 5 161 161 5 162 162 5 163 163 5 164 GUN phase deg LASER phase deg Figure 7 2 11 Measurements of the beam arrival time changes resulting from scans of the RF GUN and laser phases in the photo injector For one degree of phase shift the sum of the laser and RF phase shifts should produce 2 125 picoseconds of arrival time change Fitting a line to the middle portion of the left hand plot and to this first half of the right hand plot
20. al 42 40 l l l 0 0 2 0 4 0 6 0 8 1 2 1 4 hours Figure 7 2 5 Resolution of the 10 4 GHz BPM front end chassis This was evaluated by splitting a signal from the beam pickup directly in front of the chassis and measuring the difference between the arrival times of the pulses If the front end had perfect resolution the standard deviation of the difference between the arrival times of the two identical signals would be equal to zero The rms jitter of the difference of the split signals over the course of an hour is 3 3 um Based on this one could say that the RF front end can achieve 3 3 um resolution if the cable lengths from the pickup to the chassis are minimized At the beginning of the measurement there was a change in the split signal difference measurement that was correlated with a change n the vertical position of the beam This was marked with a red line in Fig 7 2 5 It occurred because the monitor was not periodically re calibrated during this measurement Given periodic active re calibrations using a quick scan of the vector modulator this sort of correlation is not observed Alternatively a calibration constant based on a measurement of the vertical position of the beam using BPMs upstream and downstream of the chicane can be determined and this can be multiplied by the measurements of the beam arrival time made with the chicane BPM in order to without active re calibration passively remove the influence
21. calibrations can be wrong by a factor of two If the monitor 1s calibrated with a beam based reference instead of the LO reference these errors do not occur but it 1s unfortunate because it is more convenient to calibrate a monitor with a parasitic reference that does not disturb the operation of the machine Scanning the position of the beam in the chicane in order to calibrate the monitor disturbs the machine operation Beam based calibrations aside the beam arrival time change corresponding to a beam energy change can also be measured with the monitor by adding the change of the vector modulator to the average of the pulse arrival times measured Fig 7 2 10 The arrival time change that would be expected for a 1 energy change s equal to 3 0 ps The arrival time change measured for this energy change was 2 8 0 4 ps The error bars of this measurement are large because the beam arrival time jitter is large 10 mixer phase VM phase 8 mixer VM linear _ 6 VM phase Era zz P E E un S sits Zu ae T i Er OL 25 N a z O aln 2 a y 2 8 x 2 6 T 6 T l l l 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 change ACC1 gradient Figure 7 2 10 Beam arrival time change corresponding to a small energy change Using the chicane BPM RF front end the arrival time changes from the injector can also be measured
22. is used the dynamic range of the measurement increases while the resolution decreases 120 If a limiter or nothing at all is used the highest resolution 1s achieved The maximum resolution that has been achieved to date was lt 2 um with 6dB attenuation and a 3 nC beam With 6dB attenuation and a 1 nC beam the best resolution was lt 4 um These measurements were conducted with an in tunnel installation of the optical front end With an out of tunnel installation of the front end the best resolution achieved was 4 um with a 1 nC beam and a non split signal The reduction in resolution for the out of tunnel measurement can be attributed to signal attenuation over the 30 meter long cable This resolution was achieved through the use of a signal phase measurement technique that takes short pulses from an optical reference and uses them to sample the zero crossing of the beam transient signal from the pickup This sampling is enabled through a compact 5x10x50 mm device called a Mach Zehnder Electro Optical Modulator MZ EOM In Fig 7 3 1 the EOM is depicted as a rectangular box with red lines representing a laser beam going through it and an electrical signal coming in from the top The polarization of the laser light must be adjusted so that it is in alignment with the polarization axis of the EOM The incoming laser beam is split and it travels through two Lithium Niobate crystals which are under the influence of the electric field from the e
23. most noticeable difference between the two curves shown in Fig 5 5 5 1s the additional ringing that one sees on the oscilloscope signal This is an artifact of the oscilloscope caused by the interaction of frequencies higher than 8 GHz with capacitive elements in the oscilloscope simulation and oscilliscope readout simulation and oscilliscope readout N 30 scope 200 gt N N simu ai N 100 A f LRA TE 4 E j p 10 u A I fh Lp wi nl 2 A TINA AAN IAW N 2 Ors NF Kar sO ha N MLA T oY a IN N if hy mn AN fe va VU Af t gt w l j I a gt hi l N We iy Y 10 005 V IE 00 w 20 i yi 200 30 0 0 5 1 1 5 2 0 0 2 0 4 0 6 0 8 1 time ns time ns Figure 5 5 5 The simulated blue and measured red performance of the pickup below 8 GHz left and below 50GHz right Measurements were done with an 8 GHz oscilloscope There is a concern about the length of time that the signal from the pickup rings The XFEL bunch spacing is only 200 ns and the ringing from the first bunch must be lt 0 01 of the peak voltage by the time the second bunch comes When the signal was measured in FLASH with a higher bandwidth setup the optical front end to be described in chapter 7 the ringing is gone before the next bunch comes 1 us but it looks like there is still a significant amount of ringing at 200 ns after the bunch transien
24. s a cal bration constant It was a relat vely easy design to quickly construct and install with components that were already on hand and since there was a hurry to get even a low resolution measurement up and running the button pickups were installed with an eye to arranging them as closely together as possible but without detailed simulations of their interaction with the beam As it turned out due to the size of the 50 vacuum feedthroughs the buttons were not close enough together in order to measure a significant voltage difference signal between two buttons Due to the poor performance of the system and limited potential for improvement the concept was abandoned The reasons for this failure and the limitations of the method along with the principles of button pickup operation are described in the following paragraphs These principles are also useful in describing how other types of pickups work In a round chamber button pickups work well because the electric field lines from the electron beam spread out radially in the transverse plane and terminate over the entire surface of the chamber Figure 5 1 1 This means that in a round chamber the button pickup will always experience the image current of the beam as it passes and will always generate a signal In the case of the wide flat chamber in the middle of the bunch compressor the electric field lines tend to concentrate in regions directly above and below the location of the beam
25. 1 2 10 11 12 13 P Castro Beam trajectory calculations in bunch compressors of TTF2 DESY TECHNICAL NOTE 2003 01 R Brinkmann et al TESLA XFEL First Stage of the X Ray Laser Laboratory Technical Design Report Supplement DESY TESLA FEL 2002 09 W Ackermann et al Operation of a free electron laser from the extreme ultraviolet to the water window Nature Photon 1 336 342 2007 A Zholents M Zolotorev W Wan Generation of Attosecond Bunches PAC 2001 N Stojanovic Status and plans of THz IR beamline FLASH Seminar Feb 2010 Massimo Altarell et al XFEL The European X Ray Free Electron Laser Technical design report DES Y 06 097 A Azima et al Status of sSFLASH the Seeding Experiment at FLASH Proceedings of the IPAC 2010 Kyoto Japan M Krasilnikov et al Experimental characterization of the electron source at the photo injector test facility at DESY Zeuthen PAC 2003 Portland May 2003 p 2114 S Schreiber et al Measurement of space charge effects and laser bunch length in the TTF RF gun using the phase scan technique FEL 1999 p 69 S Schreiber et al Experience w th the Photoinjector Laser at FLASH FEL 2006 BESSY Berlin Germany C Schmidt RF System Modeling and Controller Design for the European XFEL Ph D Thesis University of Hamburg 2010 K L Brown A First and Second order Matrix Th
26. 1 7 0 0 Substituting in Eq 6 1 5 this becomes 8 1 U t G t 1 x Ae td de 6 1 8 Q o If the bandwidth of the beam is much larger than that of the pickup or likewise o lt lt 4t we can write the result for the pickup output for a wide beam 2X0 U G t x c A x dx 6 1 9 For a beam that is tilted in the x y plane an additional function y x must be incorporated into the equation U t G t x c A x t y x dx 6 1 10 X y x weights the individual slices of the beam in terms of how their y position influences the amplitude that the pencil like beam would produce on the pickup For a beam that is tilted in the x z plane the pulses traveling to the left on the pickup and the pulses traveling to the right on the pickup need to be given separate treatments If the side of the beam arrives earlier than the side of the beam the arrival times of the pulses traveling to the side will be condensed with respect to one another and the arrival times of the pulses traveling to the side of the pickup will be spread out with respect to one another This is described by A x t dx 6 1 11 tilt U t G t x c t and U_ t G_ t x c ty A x dx 6 1 12 Further complications from an asymmetric charge distribution can also easily be accommodated with this method The goal of this is to determine how the measurement is affected by wide tilted beams and
27. 9 ad nom 2 1 T E Rs nom 2 and a particle with a position z after the first bunch compressor will be located at zz after the second bunch compressor Z 2Z T E Z Ty Bion SAFE ea 2 elk Ts ZZ WFT ER 2 55 where the reference particle for each chicane has been set equal to the nominal position This eliminates any constant offset due to injecting the bunch at an energy which is not the nominal energy of the chicane This can be used to derive the sensitivity of the arrival time to the voltage phase and incoming timing jitter in the same fashion that was used in the single bunch compressor case Starting with 170 ap av eee z cos k Z AV 0 Rs E amp z amp 2 AV Co E nom 2 so E E AV Co E V where the nominal energy was assumed to be equal to the reference energy Ot U Ag T E z sin k Z o AQ 09 Co Nr E z E z T amp 2 AQ Co ky z0 where the energy chirps prior to both bunch compressors are assumed to be approximately equal Ot l Oz AV T E z E z AV av a 2 E 21 E Day l Re Ei Z R E Eo 2 AV CoE iis Las V l E AV gt Lf kos E ru Co v where the nominal energy was assumed to be equal to the reference energy and the initial energy Eo is small compared to the energy after the first accelerating section pol OZ T E z E z op
28. Finally space considerations can be prohibitive for a large cavity installation Likewise 20 cm longitudinal space considerations limit the possibility of taking a smaller existing cavity monitor and putting it on movers so that it slides from side to side along the flat chamber Long stretches of bellows would be required n order to limit the discontinuities in this wakefield sensitive area 5 3 Stripline Pickups Longitudinally oriented stripline pickups are used frequently throughout the FLASH linac because they deliver a better resolution than button pickups When optimized they have achieved 5 um resolution at FLASH This is due to the larger voltages that they produce at the frequencies for which the front end electronics are designed Both button and stripline front ends rely on measuring the differences in amplitudes between pickups on opposite sides of the vacuum chamber Striplines however cannot be used everywhere because they take up a lot of longitudinal space they can require 100 1000 mm compared to the button s 20 mm Stripline pickups consist of a channel carved out of the vacuum pipe in which a metal rod or strip is suspended and terminated on one end with the characteristic impedance of the stripline Fig 5 3 1 Because the width of the stripline is only a fraction of the beam pipe circumference the stripline will not carry all of the image current The fraction of the image current not carried by the stripline will t
29. N fof l NN di a y N S S A AA A f St Ss 1 A J 7 ice A f Fi l Tilted beam signal right Tilted beam signal left 45 J A A Lf i 5 2 i q Un tilted beam signal right Un tilted beam signal left 20 80 60 40 20 0 20 40 60 80 time ps Figure 6 3 4 Pickup outputs for tilted un tilted beams with flat charge distribution Difference between arrival times of zero crossings of tilted and un tilted beams gives measurement error 88 Such an x y tilt was measured with the beam image on the synchrotron light monitor screen and the resulting changes in the beam position measured with the beam pickup and an oscilloscope were recorded Fig 6 3 5 The position of the beam was held constant and the tilt of the beam was changed and measured on a screen The sum of the pickup signals arrival times is constant while the difference of the pickup signals arrival times has a dependence on the tilt of the beam The measurement provides an opportunity to cross check the simulation presented in Fig 6 2 4 In the simulation 5 degrees of tilt of a 4 mm FWHM wide Gaussian beam produces 200 um of position measurement error and 5 degrees of tilt in the measurement would produce 250 um of position measurement error Since the beam shape is more complicated than a simple Gaussian Ch 4 this can be considered good agreement Position in Chicane a 18 0 deg 0
30. Th s means that 0 07 s a rather pessimistic value for the phase jitter of the first accelerator section since it represents more the jitter between the phase of the gun laser and gun RF The true value of the first accelerator section jitter could be anywhere between 0 01 and 0 07 but based on principles of the regulation algorithm the amplitude jitter should always be 1 8 times the phase jitter 19 so for future calculations we will assume a phase jitter of le 4 or 0 02 degrees These numbers are also only applicable for short bunch trains and 1 nC bunch charge When long bunch trains with 3 nC were used in the 9 mA experiment for ILC research due to the large beam loading effect the energy jitter of the bunches at the end of the bunch train was ten times worse than the energy jitter of the bunches at the beginning of the bunch train 24 While the combination of a new down conversion front end 22 reference injection 21 and system identification 11 could conceivably stabilize the cavity amplitude jitter and drift to within le 5 for short bunch trains in a well tuned machine it is unclear what the system performance would be under more typical circumstances Using the present machine as an example the best results are different from the typical results for short bunch trains by about a factor of 2 For long bunch trains the difference can be a factor of 10 30 A beam based feedback system is being developed to complement this cavity
31. The microbunching instability is more likely in the second chicane due to the added bending of the double chicane design and the higher energy of the beam 13 In general the gain of the instability increases with the inverse of the characteristic wavelength of the modulation squared It is cut off for wavelengths that are shorter than Rs6 C 6 For the first bunch compressor R56 C and 6 are larger than for the second bunch compressor and so the cut off occurs at longer wavelengths making the microbunching less effective With careful balancing of CSR and other effects this can be avoided and the second chicane can be used to compensate for CSR based emittance growth generated in the upstream bunch compressor 17 16 a b Figure 2 3 4 Single chicane a and symmetric double chicane b The higher order dispersion terms R566 and Rj66 are opposite for bunch compressors of the single and double symmetric types adding another opportunity for canceling out destructive effects These cancellation effects only become possible with the third harmonic module n operation to linearize the compression process 2 4 Undulator Section The electron bunch with a high current density is sent through a series of undulator magnets n order to create a short and high energy pulse of light Undulator magnets with a period of A and a magnetic field of B bend the path of the electrons back and forth many times and cause them to radiate synch
32. The low pass filtered signal is amplified in order to make it match the 1 V range of the Struck ADC which is clocked by 108 MHz delivered from the Master Oscillator So far this description has avoided mention of the various phase shifters shown in the diagram There is a motorized trombone phase shifter on the lower phase measurement arm that must move to account for changes n the difference of the arrival times of the pulses at the mixers There is an electrical phase shifter called a vector modulator labled VM that shifts the phase of the 1 3 GHz signal from the MO in order to account for any changes in the sum of the arrival times of the pulses from the pickup Lastly there is a vector modulator phase shifter that can shift the 108 MHz in order to adjust the sampling time of the ADC The weakness of the scheme shown in Fig 7 1 1 for measuring the sum of the arr val times of the pulses is that all of the noise on the LO will be part of the measurement The phase of the MO signal drifts on the 30 meter cable with a temperature coefficient of 3 ps deg C Table 8 1 1 While the drifts of the cable could be compensated with a reflectrometry scheme to within 100 fs 46 noise picked up on the long cable would limit the measurement resolution to more than 20 fs rms This was measured by comparing the resolution of the phase measurement for a short cable to that of a long cable The strength of the scheme shown in Fig 7 1 1 for measuring the
33. The measurement was done over the coarse of 17 hours with both trombone and vector modulator feedbacks on meaning that both feedbacks attempted to keep the system measuring at the zero crossings of the two signals The position changes measured by the trombone are subject to 100 um errors from mechanical hysteresis and backlash Fig 7 2 7 nevertheless for this 10 4 GHz measurement the trombone changes appear to have been appropriate since the changes measured by 10 4 GHz measurement match those of the gradient setpoint and those of the 1 3 GHz measurement During the eight hour period in the right hand figure over which no gradient setpoint changes were made it appears that the gradient regulation was drifting by 0 1 In the left hand figure it drifts by as much as 0 2 before operators react Not all machine energy changes depend on the gradient setting of the first accelerator section however A change in the phase of the gun RF impacted the left hand measurement around hour 32 long term drift long term drift 1 3 GHz 10 4 GHz ACC setpoint 0 5 gt 0 2L 0 4 0 1 0 3 0 2 energy change oO g energy change Figure 10 1 1 Measurements of energy stability in the chicane taken by the coarse and fine HF front ends of the chicane BPM plotted with energy setpoint values from the ups
34. also be possible with this design Typical beam position measurements achieve few micron resolution over a few millimeter range The monitor system developed within this thesis work represents a new and unique tool in the spectrum of accelerator diagnostics Several types of pickups were considered for this task a pin that detects the ringing in a cavity an array of closely spaced striplines oriented in the direction of the beam and a rod 1n a coaxially shaped channel oriented perpendicularly to the direction of the beam In the end the perpendicularly oriented pickup was constructed and utilized with two different types of pulse measurement systems one of which made use of high frequency electronic techniques and the other which made use of an optical technique involving Mach Zehnder electro optical modulators and the pulses from a master laser oscillator Both techniques delivered the required resolution but each had unique advantages and disadvantages Particular emphasis 1s placed on the cost and robustness of the high frequency electronic system compared to the lower potential for systematic error of the optical system Understanding the influence of the shape of the beam on chicane beam position measurements 1s also critical For this reason simulations of the beam transport from the start of the accelerator up to the middle of the chicane were undertaken in order to predict the likely beam properties at the Free electron LASer in Hamburg
35. alternative to this scheme using two arrival time monitors up and downstream of a chicane to calculate corrections to the beam energy were also described with a view to a future in which cross checks of all of these independent high resolution measurements will anchor the machine to a stable reference 164 Appendix A The derivation of Eqs 3 2 1 and 3 2 1 will be presented here The equations describe the arrival time jitter of the beam after a bunch compressor A first order derivation of the arrival time jitter 1s carried out first and an investigation of the complications posed by higher order terms follows 56 Starting with the equations for the energy of a particle with position z within the bunch F z and the first order energy chirp after the first acceleration section 7 z we have E z E z V cos k z 9 1 and E z E z Vik sin k Z Q 2 where the energy and energy chirp from before the first acceleration section are given by Eo z and Ey z These terms take into account the initial correlated energy distribution entering the first acceleration section as well as any energy variations generated by collective effects experienced by the bunch before it reaches the entrance to the chicane The energy dependent path length through a chicane was derived n chapter 1 and can be expressed in terms of a nominal path length L a nominal energy Enom and the R56 T E L Ry un 3 The first
36. amplitude These parameters vary depending on the beam charge and the cables used to deliver the signals The signals come from each side of the pickup and the difference between their arrival times is 97 proportional to the beam position They are each sent through a 4 pole band pass filter with a center frequency of 10 4 GHz and a 200 MHz FWHM bandwidth The filter has four poles so that the group delay of the signal in the filter is flat in the pass band The bandwidth is large so that the group delay doesn t respond too dramatically to temperature changes the smaller the filter bandwidth is the more sensitive it is to temperature changes Even though the pulse from the pickup is more than a hundred volts and less than 100 ps long after the 30 meter cable the bandpass filter and the mixer only a few mV remain To compensate for this loss the filtered signal 1s amplified and then mixed with what is known n mixer terminology as the Local Oscillator LO The LO is generated from the Master Oscillator MO reference frequency of 1 3 GHz by multiplying it by 8 in a Hittite frequency multiplier to make 10 4 GHz The frequency multiplier also provides some amplification to the LO signal through an active component The output from the Minicircuits mixers 1s then low pass filtered with a cut off frequency of 30 MHz This serves 2 purposes it removes some high frequency noise and it broadens the signal so that it is easier to sample with an ADC
37. and charge The influence of these amplitude changes on the resolution of the measurement cannot be removed but the impact on the resolution is typically small and is only of significance when the charge of the beam is dramatically changed The dependence of the resolution on the beam charge is linear and if for example a lower charge of 0 2 nC were to be used the resolution would be less than a quarter of what it would be for a 1 nC beam The effect of changing the beam charge by large amounts can be compensated by adding or removing attenuators on the measurement front end When the beam is vertically centered in the vacuum chamber the BPM resolution will not be dramatically affected if the vertical position of the beam jitters by a few hundred microns Fig 5 5 9 but for any static position changes the measurement will need to be re calibrated The y position sensitivity appears to provide the option that for low charge levels the beam could be steered close to the pickup n order to improve the resolution of the measurement If the beam gets within a couple of millimeters of the pickup however the calibration will become unstable due to the high sensitivity to vertical position changes 76 signal amplitude V Signal amplitude V Figure 5 5 9 Measured Vertical Position Sensitivity 35 n 30 Bottom pickup Top pickup 25 _ ge ae k 20 p 2 15 P A 10 P Fr
38. and it is possible that if the button pickup is located at a distance from the beam it will produce no signal at all Figure 5 1 1 Figure 5 1 1 A cross section of button pickups in a round vacuum chamber left and button pickups in a flat vacuum chamber right If the electric field lines from the beam do not terminate on the pickup no signal will be produced As a beam travels through a vacuum chamber an image current mirroring the beam travels along the walls of the vacuum chamber In the absence of wakefields when the beam velocity s approaching the speed of light the electric field lines are longitudinally concentrated above and below the beam Fig 5 1 2 1 By X V lt lt c Figure 5 1 2 The electric field lines of charged particle beam moving at much less than the speed of light left and close to the speed of light right 51 One can numerically calculate the distribution of the image currents in the plane perpendicular to the beam s direction of motion with a 2 d magneto or electrostatic simulation of the vacuum chamber cross section or for a simple geometry like a circle the distribution of the image current density can be found analytically from the static 2 D version of Ampere s law ee VxH 5 1 4 where the curl of the magnetic field in units of Amperes per meter is equal to jpcam the free current density of the beam The displacement current term is zero because the electric fiel
39. arrival time jitter with an accelerating section followed by a bunch compressor Layout of synchronization sensitive components at FLASH along with desired feedback loops Measurement of the arrival time of the injector laser pulse relative to a timing laser reference MLO using a two color single crystal balanced optical cross correlator The shape of the beam in the bunch compressor for on crest operation as viewed on the synchrotron light monitor in the first bunch compressor Coupler kick concept Coupler geometry with pickups top and voltage of kick bottom Voltages acting on a beam as it travels in z through a cavity tilted in the y z plane The effect of a dipole force on a beam with an energy chirp A beam offset in the injector is magnified as it travels through the first accelerating section A lattice optimized for a space charge dominated beam magnifies beam tilts generated in the injector The tilt of the beam for various closed orbit bumps through the first accelerating section Images on the screen contain projections of the beam streaked out in a longitudinal direction Simulation results for energy spread of the beam coming out of the injector top left the horizontal beam path through the accelerating module top right and the resulting tilt in the first bunch compressor with bottom left and without bottom right the chromatic effect of the quadrupoles 5 1 3 5 1 4 5 1 5 5 1 6 SA7 5 1 8 5 1 9
40. average of the arrival times of the signals and hence noise or drifts of the incoming reference signal would affect the stability of the measurement Because the phase of the optical reference delivered on a length stabilized fiber can be made more stable than that of the 1 3 GHz RF reference delivered on an un stabilized RF cable the use of an optical reference could remove the impact of drifts from the delivery of the LO signal and the use of an optical delay line n lieu of the noisy vector modulator One method is to use a 10 GHz photo detector operated in saturation and filter out the desired frequency component from the frequency comb 10 4 GHz and 1 3 GHz would be filtered out for example and then a non saturated photo detector would be used n order to actively stabilize the amplitude of the laser signal with DSP feedback on the laser diode driver responsible for the amplification of the laser signal Fig 7 2 2 This last step is necessary because the phase of the signal produced by a photodetector will change by 20 fs whenever the laser amplitude changes by 0 1 47 Thermal stability of the photodetectors is also important due to the 340fs degC thermal drift coefficient 48 From length Stabilized BP filter fiber link 1 3GHz To phase detection BP filter circuits 10 4 GHz For optical input of 108 MHz ADC clock To ADC DSP DAC feedback for fiber link amplification 216 MHz BP AD9510 Figure 7 2 2
41. bump through the first accelerator section When the beam has an energy chirp the effect that usually influences the beam tilt more than any other is that of dispersion When a beam with an energy chirp travels through a dipole field like that of a corrector an offset quadrupole magnet or an offset focusing field of an accelerating structure the lower energy particles in the head of the bunch are deflected more than the higher energy particles in the tail of the bunch and therefore acquire an offset relative to the lower energy particles The head and the tail of the bunch get further apart in that deflecting plane Fig 4 2 5 lt _ gt Fp high low energy energy Figure 4 2 5 The effect of a dipole magnet on a beam with an energy chirp 42 Upstream of the first accelerator section where there is a 1 energy chirp and there are no quadrupole magnets only a couple of correctors the contribution to the beam tilt from dispersion for a 2 mm offset out of the gun would only be a 20 um vertical separation between the head and tail of the beam For a beam of 2 mm length this corresponds to a tilt of 170 urad This is about as strong as the wakefield effects but even in combination with the wakefields and other effects it is still too small to create the gt 400 um of head tail separation seen in BC2 at the phase advance for which the tilt is maximized After some investigation simulating and measuring the effects listed abov
42. calculate this signal to noise threshold we need the beam current the transfer impedance of the pickup and the noise floor of the measurement The time dependence of the current of the short Gaussian electron bunches of an FEL can be approximated as 1 2 o Im gx 5 1 15 V27 O and the frequency response as l 2 o Isean 0 Qe 5 1 16 55 where o is the rms duration of the beam A Gaussian is not an accurate representation of a bunch for high frequencies because the tails of a Gaussian go out to infinity while a real bunch has a finite extent The result of this discrepancy is that any edges in the distribution will cause the beam current spectral density to fall off with 1 m at high frequencies To describe the transfer impedance it is helpful to first write an equivalent circuit for the pickup Each segment of a button pickup or any pickup can be described with some equations from transmission line theory 44 A small section of transmission line can be modeled as a series inductance with a shunt capacitance as shown in figure 5 1 6 I I Al V AV T GND Figure 5 1 6 Electrical circuit representation of a small section of transmission line The current going through the inductor is proportional to the voltage drop across the inductor BLZ i 5 1 17 OZ Ot and the current exiting the transmission line segment is the difference between the input current and the capacitance cu
43. calibrate the monitor prior to the arrival of each bunch It is designed to operate with 10 cm long stripline pickups with diameters of 25 mm in a 35 mm diameter beam pipe 42 It gets 2 3 um resolution for an 68 0 2 nC beam charge and can handle beam currents of up to 8 nC with the application of built n attenuators An array of 10 cm long stripline pickups shown in Fig 5 4 1 might have been able to achieve the required sub 5 um resolution in the chicane with an adaptation of the LCLS front end design even though the pickups in a flat chamber configuration have less sensitivity than the pickups had in their LCLS installation Fig 5 4 1 amp Fig 5 4 2 mm 35 mm 100 mm o 35 mm 070070 O Figure 5 4 1 Cross section of striplines in a round chamber and in a flat chamber not lt gt 25 mm gt to scale sensitivities of 25 mm stripline pickups 0 6 flat chamber round chamber 3 Ss gt gt D Cc eb N 10 5 0 5 10 mm Figure 5 4 2 Simulation of sensitivity of 25 mm diameter striplines in a round chamber configuration and in a flat chamber configuration The flat chamber has reduced sensitivity At the time of the design decision in 2005 this multi channel reference injection solution that was commissioned in 2008 by LCLS seemed more complicated to implement than 69 the two channel solution afforded by a transversely mounted stripline design In house supp
44. differences in signals amplitudes The concept requires that the speed of the signals on the pickup are equal to c and s based on the basic transmission line equations derived in section 5 70 a T1 T2 T3 Tl T2 X T3 c Figure 5 5 2 Cross section of a transversely mounted stripline pickup with tapering to vacuum feedthroughs The beam passes beneath or above the pickup and causes current pulses to travel to either end of the pickup The arrival times Tl and T2 of the pulses are measured and the position of the beam X can be determined The pickup shown in Fig 5 5 2 was chosen because of its ability to deliver the high bandwidth pulses that are desirable for optical and high frequency phase measurement techniques Using the design principles established in the previous sections the questions that arose during the design process were primarily practical in nature Existing striplines at DESY consisted of hollow rounded rods suspended at a diagonal from SMA feedthroughs Fig 5 5 3 In principle a tapered design would offer an improvement in impedance matching over the existing design The extent of the improvement needed to be simulated and verified The tapering required more welding stages and had more potential for failure due to the newness of the design If the older design had been sufficient then it could have been constructed with existing parts and know how Figure 5 5 3 Stripline feedthrough cross sections not t
45. enough to ignore an arrival time monitor upstream of the chicane must be used in any energy arrival time feedback scheme 3 8 Outlook While two beam arrival time monitors BAMs placed one after the other in a drift section have been shown to produce 9 fs jitter relative to one another over few minute time scales if the longitudinal profile of the beam changes slightly the arrival time measurement will change by up to 100 femtoseconds 20 While the bunch head is only 25 fs long at the end of the machine 32 the picosecond long beam tails that result from non linear compression 33 i e without the third harmonic module cause the arrival time measurement to be more sensitive to changes in the beam profile that it would be without a long tail which changes in length when the RF phase changes Changes in the beam profile limit the stability of the measurement under stable beam conditions to 30 fs rms over 7 hours 20 But the beam profile is not the only source of measurement error Although the BAM equipment is in a temperature stabilized enclosure the arrival time measurement will drift by 2 femtoseconds when the temperature of the fibers changes by 0 1 degree A beam energy measurement done by using BAMs to measure the time of flight through the chicane will be especially sensitive to these temperature changes because it is carried out with two measurements separated by gt 10 meters The chicane BPM is less sensitive to temperatu
46. field measurements The reference injection and tracking method reduced the impacts of temperature changes in the electronics racks from several degrees in phase and le 2 in amplitude to below 2e 4 n amplitude and 0 008 degrees in phase The amplitude drift was verified with a beam based measurement but the phase measurement s only relative to the 1 3 GHz reference delivered on an RF cable not relative to the beam While reference injection can remove the le 3 degC drift of the field detectors the phase of the reference signal on every meter of RF cable will still drift with temperature with a coefficient of le 4 degC After 10 meters of cable the cable drift is the same as the field detector drift With more expensive cables this phase drift can be reduced by a factor of 5 With active drift compensation using RF reflectrometry the phase drift on the long RF cables could be made as small as 0 005 degrees Cable drift compensation will be discussed in Chapters 7 and 8 in the context of beam arrival time measurements 22 The problem of cavity field measurement resolution resulting in bunch to bunch energy fluctuations has recently been approached with a new version of the feedback control algorithm that incorporates a system identification scheme n a so called Multiple Inputs Multiple Outputs MIMO framework 11 The advantage of the MIMO feedback is that it increases the gain for low frequency fluctuations without increasing sensitivity
47. gives a net arrival time change of 2 3 0 2 ps degree The main concern about the performance of this RF scheme as it is presently built is that the mechanical phase shifter from ATM does not have the position repeatability that one would desire in a measurement that could serve as a reference for measurements of the energy and arrival time of the beam The problems associated with the potentiometer based position readback of the ATM phase shifter are illustrated below in Fig 7 2 12 The position of the mechanical phase shifter is read by measuring the voltage of a potentiometer mounted to a gear in the motor assembly The correct scaling of the voltage measured across the potentiometer can be determined by comparing the change in the mixer output produced by changes in the mechanical phase shifter position to changes in the mixer output produced by changes in the phase setting of the vector modulator If the calibration factor for the potentiometer voltage is correct then the slope of the signal measured by scanning the phase of the vector modulator will be equal to the slope of the signal measured by scanning the phase with the mechanical phase shifter Backlash of the motor is seen as a discontinuity between data points measured after a forward movement of the stage red compared with the data taken after a backward movement of the stage blue The fact that the period of the signal measured with the scan of the mechanical phase shifter is not const
48. hundred femtoseconds with disturbances figure not shown 142 MLO RF Lock Drift gt O NO NO Mixer Output fs 11 12 13 14 15 16 17 18 Time hours 0 02 0 01 fh id IN Inn If Mu IN TE yt ul iy al il eh N I al IM AN N I wit Il Al II Wik WEIL Itt MIN MY ud tt i Fatir Vi Hm N M MN N y ily jy M g ALY IM yw Hy ld Mil y A id m rf NL Ih WY ANT yi l 0 01 Mh Ai Malt adil 0 02 hl 0 03 11 12 13 14 15 16 17 18 Time hours Temperature variation box deg Figure 8 3 6 RF phase measurement drift without temperature control without disturbances people in room Averaging over the jitter the signal drifts by 20 fs Due to cables that were 2 meters longer than they needed to be the jitter of the measurement 1s also larger than in the previous cases more than doubling the jitter through cable vibrations alone With the RF lock loop closed the drift of the lock as measured with an out of loop measurement is 77 fs pkpk without disturbances Fig 8 3 7 and several picoseconds with disturbances figure not shown The several picoseconds of drift that was seen when people were working in the room was due to the fact that the active temperature control system was not functioning at that time and the temperature of the room changed by a few degrees In
49. in the calibration The author s preference would be for using the calibration of the PMT BPM done with a motor scan as a benchmark This was not available during the studies done here Progress was made in terms of writing software to keep the monitors calibrated all of the time But because of the complexity of the algorithm required the 10 4 GHz measurement will not be immediately incorporated into a beam energy server The 1 3 GHz measurement however does not require constant adjustment and recalibration Because of its relative reliability it has been incorporated into an energy server that will be easy for the operators to use 156 10 3 Out of loop Vector Sum The out of loop vector sum should provide a measurement of the drifts of the in loop vector sum the gradient setpoint which result from drifts of the downconverters 21 It does not provide a measurement of drifts which occur on the long cables Cable drifts are a concern for phase stability but less of a concern for amplitude stability the measurement in question here The out of loop vector sum should also provide energy change resolution that 1s comparable to that of both the PMT monitor and the RF fine front end of the chicane BPM It appears however that it is subject to errors to which the PMT monitor and chicane BPM are not subject In Fig 10 3 1 there is a jump in the out of loop vector sum that is not seen on either the PMT monitor or the chicane BPM There is al
50. loop using multiple cavity pickups as the diagnostic references All of this would be accomplished with a newly developed uTCA crate system this is the crate system that will replace the existing VME infrastructure in the coming years The advantages of the system from Fig 3 3 1 a include e the ability to use a high level system identification algorithm to tune and stabilize the entire machine e built in cross checks and redundant measurements that use different techniques to measure the same quantities e robustness afforded by distributed cavity controllers that do not require the central controller to operate The disadvantage is that with a centralized decision making process the latency of the signal transport increases due to the multiple digital processors and long cables to and from the central decision making crate The latency problems can be offset with the use of a normal conducting cavity with a low quality factor in which the accelerating field can be changed quickly The alternative more expedient architecture using existing hardware and distributed control loops shown in Fig 3 3 1 b requires fewer digital processors and cable lengths that are shorter however the high quality factor of the super conducting cavities used as actuators limits the speed with which the energy of each bunch can be adjusted This means that the first 10 or more bunches of the train are un stabilized In Fig 3 3 1 b the photo injector laser phase i
51. many important experiments in the history of physics From the first cathode ray tube that was placed next to a piece of magnetized metal to the high energy beams of modern accelerators traveling through lattices of powerful electromagnets the position of the beam under the influence of a magnetic field gives information about the momentum of the beam High precision knowledge of the beam momentum can enable higher precision control of the beam Control of the beam momentum is critical for the stability of both the wavelength and arrival time of the light pulses generated by free electron lasers The measurement of the beam momentum in a free electron laser with magnetic bunch compressor chicanes is the topic of this thesis The bending radius r of an electron with charge e traveling through a magnetic dipole field perpendicular to the beam direction B depends on the momentum p of the particle 1 1 l e r S is an equation derived from the Lorentz force law F q x B and the relation between force and momentum F dp dt pv r For a rectangular dipole magnet in which the electron beam enters perpendicularly to one of the magnet s faces the path length of the electron s trajectory is given by Se 1 2 arc where a is the bending angle introduced by the magnet This allows us to write the effective length of the dipole g in terms of the bending angle lL r sina 1 3 eff As well as the x offset of the pa
52. measurement than the day night temperature changes that were experienced by the 30 meter long cables During this measurement the large temperature changes in the room on the first day were due to the opening and closing of doors The smaller temperature change observed on the second day is more typical of the changes that the measurement must withstand While day night temperature changes are not easy to see what is more apparent is a slow trend upward as the tunnel temperature warms by one degree over the course of three days This 1s a drift that could not be eliminated unless the chassis were moved closer to the pickup an option that while not too challenging to implement was not implemented The measurements of the resolution can have no bearing on reality unless they are accompanied by a measurement of the monitor s response to changes of the beam position In Fig 7 2 8 the position of the beam was changed by altering the energy of the beam with the first accelerator section amplitude The mechanical phase shifter position was held constant and the vector modulator kept the measurement centered about the zero crossing of one signal As the sampling position of the other signal moves further and further away from the zero crossing the measurement of the beam position becomes non linear and moves out of the range for which the calibration was valid The beam was off crest during the scan 112 Ill I UI A
53. measurement with two beam arrival time monitors and a synchrotron light monitor with two photomultiplier tubes The electron beam position measurement is required as part of a measurement of the electron beam energy and could be used in an intra bunch train beam based feedback system that would stabilize the amplitude of the accelerating field By stabilizing the accelerating field amplitude the arrival time of the electron beam can be made more stable By stabilizing the electron beam arrival time relative to a stable reference diagnostic seeding and beam manipulation lasers can be synchronized to the beam Zusammenfassung Im Rahmen dieser Doktorarbeit wurden an dem Freien Elektronen Laser FLASH in Hamburg zwei unterschiedliche Techniken zur Vermessung der transversalen Elektronenstrahlposition in magnetischen Schikanen mit einer Aufl sung von 5 um ber einen 10 cm breiten Me bereich entwickelt Eine diese Technik basiert auf der Bestimmung der Ankunftszeiten zweier kurzer elektrischer S gnale welche beim pass eren des Elektronenstrahls an einer Hochfrequenzantenne erzeugt werden mittels Hochfrequenzelektronik Die zweite Technik verwendet kurze Laserpulse d e zur Hochfrequenz des Beschleunigers synchronisiert sind um die Ankunftszeiten der elektrischen Antennensignale mit hoher Pr zision zu ermittelt Die Vor und Nachteile dieser beiden Methoden werden in dieser Arbeit theoretisch und experimentell untersucht und verglichen mit an
54. mounted Stripline 6 Impacts of Beam Shape and Orientation 6 1 Pickup Signals from a Wide Beam 6 2 Beam Width Changes 6 3 Tilted in x z plane 6 4 Tilted in y z plane 6 5 Tilted in x z plane 6 6 Asymmetric Charge Distribution Tilted 6 7 Fields from Previous Positions 6 8 Summary 7 Chicane BPM front end 7 1 RF Front end Concept 7 2 RF Front end Execution 7 3 Optical Front end Concept 7 4 Optical Front end Execution 7 5 Front end costs 8 Beam Arrival time Monitors 8 1 RF Front end 8 2 Optical Front end 8 3 MLO RF lock 9 Synchrotron Light Monitors 9 1 Profile Monitors 9 2 Photomultiplier Tube Monitors 10 Energy Measurement Benchmarking 10 1 RF BPM Measurements 10 2 Photomultipler Tube Monitor 10 3 Out of loop Vector Sum 10 4 Optical BPM Measurements 11 Conclusion and Outlook List of Tables 2 3 1 3 1 1 4 3 1 6 8 1 75 1 8 1 1 8 1 2 8 2 1 8 2 2 8 3 1 10 0 1 Rs6 and Rio values for FLASH and XFEL the corresponding dynamic apertures of the chicanes X range and the position spreads of the beam within the chicanes DX Module controller performance benchmarks 4 Contributions to first bunch compressor beam tilt for an off crest bunch Effects of beam tilts on beam position measured with transversely mounted stripline BPM Rough cost estimate of RF front end and optical front end for the chicane BPM Costs and performance of RF cables and optical fibers Rough cost estimate for a 20 fs resolution RF arrival
55. of the machine would then be 30 fs rms This calculation has however ignored the second accelerator section Because of the assumptions of Eq 3 2 1 one cannot simply use the equation recursively for the second accelerator section For the second accelerator section the incoming energy chirp is not small compared to the outgoing chirp and the incoming energy 1s not small compared to the outgoing energy Eq 3 2 2 describes the arrival time jitter after the second bunch compressor for a beam that is on crest in the second 26 accelerator section and it was first published in 26 the derivation of Eq 3 2 2 is written in the Appendix A 2 2 2 E oO E E 1 0 x 2 2 zu R 56 F R 56 a R 56 Z l z C a ay ned Bix CoA C ck uf Where the R 561s from the first bunch compressor and R255 1s from the second bunch compressor The RF amplitude A and beam energy E are given with indices corresponding to whether they refer to the first or second accelerator section C C C gt is the compression factor for the combination of both bunch compressors and op is the phase stability of the first accelerator section When the beam is on crest in the second accelerator section judging from Eq 3 2 2 an arrival time monitor after the second bunch compressor will measure primarily arrival time jitter caused by the amplitude fluctuations of the first accelerator section for the following reasons e Injector jitter will be compress
56. of the reflections phases together with the phase of the initial beam transient pulse If the amplitudes of the reflections are very small compared to the amplitude of the initial beam transient pulse then this will have no effect at all but since the amplitude of the first reflection in the pickup is about a third of the initial beam transient pulse it cannot be ignored When the beam position changes the beam transient pulse on one side of the pickup will arrive earlier but the reflection of that pulse will arrive later The change of the sum of the phases of the initial and reflected pulses will therefore be smaller than the change of the phase of the initial pulse alone This means that when the monitor is calibrated by scanning the LO with the vector modulator the calibration factor will not be accurate for measurements of the beam position or arrival time For the measurement of the beam position to be accurate the monitor must be calibrated by scanning the position of the beam either by changing the accelerating gradient setpoint or the chicane dipole current In the measurements presented in the following sections the calibration of the monitor was done by scanning the LO with the vector modulator phase and not by scanning the beam position with the accelerating gradient For most measurements with this calibration method the errors due to the reflected pulses were not apparent because the beam was close to the horizontal center of th
57. of vertical position changes on the measurements of the horizontal position of the beam performed with the chicane BPM 110 For the measurement shown in Fig 7 2 5 the pickup signal was split at the end of the 30 meter cable so one can only claim that the resolution of the beam position measurements produced by this RF front end would be 3 um 1f it had been installed in the tunnel with 1 2 meter long cables connecting the pickup to the chassis If the front end is installed out of the tunnel with 30 meter long cables connecting the pickup outputs to the chassis only 6 um resolution can be claimed based on the fast jitter shown in Fig 7 2 6 difference of split signals arrival times resolution
58. prevent thermal gradients over the RF circuit each Peltier element was in direct thermal contact with a 10x300x50 mm metal bar which was mounted to the aluminum plate and not in direct thermal contact with the filter itself or with the 4 mm thick aluminum plate to which the circuit elements were fixed This intermediate medium distributed the heating and cooling action of the Peltier over a larger surface reducing the possibility of noise in the temperature controller making its way into the RF circuit stability 107 Peltier Cooling block controller Voltage regulator Hittite x8 Joy ds 5 Hittite LPF HF 5 front end 2 Upper Level e LNA a z 1 3 Cooling block HF front end Lower Level Peltier ele metal bloc underneat Figure 7 2 4 The upper level and lower level of the HF front end chassis 108 A Peltier element acts as a heat pump When an object mounted to the top surface of a Peltier is too cool the top surface of the Peltier will become warmer and the bottom surface will become cooler When the object mounted to the top surface of the Peltier is too warm the reverse is true When the Peltiers cool the aluminum plate the heat from the bottom side of the Peltier travels down the metal post mounted to the bottom surface of the Peltier towards the floor of the chassis The heat from the post 1s distributed over the floor of the chassis which s then cooled with an external rack mounted fan that blows
59. resting The optical length of one meter of a standard optical fiber will drift by 50 60 fs deg There are approximately 2 meters of fiber along one arm of the chicane BPM optical front end so if the temperature of one arm differs from the other arm by less than 0 1 degrees Celsius the beam position measurement will be affected by less than 0 5 um If the temperature of the chassis changes by 0 1 degrees Celsius the beam arrival time measurement will change by 6 fs The active temperature stabilization system shown in Figure 7 2 4 can keep the temperature of the box stable to within 0 003 degree C rms for slowly changing external temperatures in an air conditioned laboratory setting When the box is in the accelerator tunnel the temperature changes of the plate on which the fibers rest were 6 of the 2 degree C temperature changes measured on the outside of the box even though the inner box air temperature tracked the 2 degrees of tunnel temperature change The measurement of the in tunnel temperature stability is limited by the Beckhoff ADC to 0 03 degree Celsius resolution and it takes about a 12 hours for the temperature in the box to become truly stable after the tunnel has been opened for a maintenance day It 1s anticipated that this stabilization time could be reduced if more active cooling were applied to the outside of the chassis improving the efficiency of the removal of heat from the Peltiers One other problem that limits the speed w
60. shown in Fig 7 1 1 as a sine wave with a frequency equal to the LO the difference frequency will be zero a DC signal and the sum frequency will be 20 8 GHz If one low pass filters the output of the mixer in order to remove the sum frequency changes in the phase relationship between the two mixer input signals will produce a change in the DC voltage measured at the output of the mixer When the signals have a fixed phase difference of r changes in the DC output of the mixer are proportional to changes n the amplitude of either input signal When they have a phase relationship of 7 2 the sensitivity of the mixer output to changes in the amplitude of the input signals will be minimized This is the ideal phase for measuring differences between the phases of the input signals If one approximates the output of the filters as a pulse with the shape of a single cycle of a sinusoid the output of the mixer will not be a DC voltage but a pulse This concept along with the phase relationships for measurements of a signal s phase or amplitude are illustrated in the drawing below in Fig 7 1 2 Amplitude measurement Phase measurement Amplitude measurement Figure 7 1 2 The input and output of a mixer and how an appropriate phase relationship facilitates the measurement of the phase of the input signal 99 In Fig 7 1 2 the signal from the output of the filter is depicted as a single cycle of a sinus The filter out
61. signal right Un tilted beam signal left N f P 20 l l l 25 20 15 10 5 0 5 10 15 20 25 horizontal distance mm Figure 6 6 2 Error resulting from x z tilted asymmetric charge distribution An error of half of a picosecond is seen for the tilted asymmetric tilted distribution green black compared to the asymmetric un tilted distribution red blue 93 In this simulation the zero crossings of the asymmetric charge distribution accurately give the position of the horizontally offset centroid position of the distribution but the zero crossings of the tilted and asymmetric charge distribution differ by 580 fs 6 7 Wakefields There is a valid concern that in addition to the image charge of the beam as it passes beneath the pickup the chicane BPM stripline measures fields originating from the beam s image charge from earlier in the chicane Due to the large vacuum chamber width the fields from previous positions before and in the bending magnets are detected by the pick up This would manifest as an averaging over previous beam positions While averaging over previous positions could reduce the slope of the signal produced by the pickup thereby reducing the resolution of any zero crossing sampling scheme this would be evidenced by a gradual reduction of the amplitude and slope of the signal as the beam is moved from the inside of the vacuum chamber towards the outside In oscilloscope measurements of the ampli
62. than top design but the amplitude 1s 20 smaller The different colors represent the monopole and dipole modes at the output of the pickup 2 None of the designs suffered from any distortion of the signal shape as the beam position was changed but there were big differences in the steepness of the signal slope and the amplitudes of the signals It appeared that the tapered design would have a slope at the zero crossing that was 35 steeper than the existing design When the tapered design was completed however it was apparent that unless the pickup antenna could be made light and hollow there would need to be a ring made of ceramic Alumina to support the antenna and hold it in a stable position reducing the risk that the feedthrough ceramic would crack and cause a vacuum leak Although the diameter of the ceramic was designed to minimize the impact of the impedance mismatch that it creates when this ceramic ring was added to the simulations the comparison between the existing design and new tapered design was much less dramatic Fig 5 5 4 bottom While Vespel would offer 30 better performance than Alumina it is not allowed in vacuum installations because it outgases under the influence of radiation The performance of the monitor predicted by the CST simulation cannot be measured up to 50 GHz For a bandwidth of below 8 GHz the simulation is in partial agreement with the performance measured with an 8 GHz oscilloscope Fig 5 5 5 The
63. that dynamic range is no longer a problem the first device that needs to be optimized in order to increase the bandwidth is the combiner which reduces the measurement s dependence on vertical beam tilts The current combiner is an in house built device with 6 dB insertion loss The second device that could be improved in conjunction with a switch from SMA to type K cables connecting the pickup to the EOM is the Agilent N9355C limiter which can have a bandwidth of up to 50 GHz if the type K version is used instead of the SMA version The third device which could be improved is the EOM The current EOM has a 3dB bandwidth of 10 GHz but there are more expensive EOMs that accommodate above 40 GHz The pickup is the most time intensive component to improve but if the pickup was re engineered to be hollow and suspended between the two vacuum feedthroughs the elimination of the ceramic support rings would produce a 30 improvement in the slope of the signal Fig 5 5 3 Given perfect thermal stability the device that limits the accuracy of the optical front end more than any other is the RF limiter The limiter containing a combination of Schottky diodes is used to prevent damage to the EOMs from large amounts of power Without the limiter the EOMs survive short pulses of even a few hundred Volts without immediate damage however if the beam 1s steered directly onto the pickup itself or if the pickup is sprayed with a shower of electrons in the abs
64. the input signal Simulated behavior of the mixer output around the phase for which the mixer output is minimized RF signal composed by sum of two different frequencies mixed is with the LO red the output of this mixer is shown in green Three different sample points of the mixer output when the phase of the LO is changed in simulation Distorted sinusoidal pattern that is measured at the output of the mixers when the phase of the LO is changed Mixer output when the RF signal is composed of two signals with the same frequencies but where one signal has twice the phase of the other Chicane BPM RF front end schematic Scheme for delivering a stable reference signal to the phase detection circuits using an optical signal from a length stabilized fiber link Scheme for stabilization of signal phase on an RF cable 7 2 4 1 29 1 2 6 aad 7 2 8 1 2 9 7 2 10 12 11 1 2 12 T2313 7 2 14 7 2 15 7 2 16 7 3 1 32 gt 1 3 4 7 4 1 1 2 25 7 2 4 7 2 5 8 2 1 8 2 2 8 3 1 8 3 2 8 3 3 8 3 4 8 3 5 The upper level and lower level of the HF front end chassis Resolution of the 10 4 GHz front end installed in the tunnel Resolution of the 10 4 GHz front end installed out of the tunnel Three days long measurement of the difference between the split signals Scanning the gradient of the first accelerating module and measuring the change in the position of the beam with the chicane BPM Beam position change corresponding to a
65. the mixer s Ky and characterization of the spectral noise density and drift contributed by each RF component BP 1 3GHz ZFM2000s1 P1 9 140 In Fig 8 3 4 the spectral noise density of the signal at the exit of the low noise amplifier LNA shown in Fig 8 3 1 s plotted with a calculation of the amount of timing jitter contributed by various bandwidths The net timing jitter of lt 6 5 fs is dominated by the offset frequencies above 100 kHz Since the MLO RF lock bandwidth is typically only a few kHz the noise introduced by the phase measurement will be suppressed It will not however be possible at this frequency to distinguish the real timing jitter of the laser from phase detection errors This means that the fast noise of the phase detection circuit will not limit the performance of the lock Slow drifts must be measured separately with attention to the thermal stability of the circuit Timing jitter RIN 6 468fs after mixer filter and LNA 0 237fs 0 223fs 0 191fs 0 992fs 5 895fs 2 458fs 90 6 100 gt 52 110 5 E 120 Fil 14 2 E S 430 3 op 4 140 g 29 P 150 dp 160 170 10 10 10 10 10 i Offset Frequency Hz Figure 8 3 4 Spectral noise density of signal at the ex
66. time monitor with a 100 fs phase stabilized cable Comparison of optical and RF systems phase noise detection and etc Cost estimate for 6 fs resolution optical front end and a length stabilized fiber The cost of an MO in a rack and an MLO with optical table and control hardware Comparison of energy measurements in the first bunch compressor List of Figures 1 0 1 2 0 1 2 1 1 22 1 2 3 1 2 2 23 3 2 3 4 2 4 1 2A2 3 1 1 3 1 2 3 1 3 3 2 1 3 3 1 3 4 1 4 2 1 4 2 2 4 2 3 4 2 4 4 2 5 4 2 6 4 2 7 4 3 1 4 3 2 4 3 3 Magnetic bunch compressor chicane Free electron LASer in Hamburg FLASH Simplified cross section of the RF Gun Basic structure of a klystron The phase of the accelerating RF relative to the beam determines the energy chirp of the beam Bunch profiles in the first bunch compressor for inhomogeneous compression without the third harmonic cavity Energy spread generated within chicane breaks linear achromaticity and results in an increased emittance after the chicane Single chicane a and symmetric double chicane b Undulator magnet and electron bunch producing synchrotron radiation Interaction of electron beam and photon pulse System for controlling the cavity fields of the accelerating module Simplified block diagram of the cavity regulation routines on the FPGA A desired FPGA algorithm structure incorporating reference injection and beam based information Transformation of
67. to high frequency fluctuations and it allows for easier incorporation of new information into the control algorithm In contrast the existing proportional gain feedback responds to all frequencies up to the bandwidth of the ADC and is incompatible with beam based feedback When measured with the reference tracking out of loop vector sum setup the MIMO feedback with iterative learning control a new version of the feedforward successfully reduced the pulse to pulse amplitude stability of the first accelerator section from 2e 4 to below 5e 5 and it reduced the phase stability from 0 008 down to 0 003 but it developed resonances and instabilities over the course of an hour 11 With notch filters applied to the resonance frequencies the system should be more stable The feedback is limited primarily by the resolution of the cavity measurement front ends An improvement in the resolution of the cavity front ends by a factor of 5 to meet le 5 stability 1s conceivable in the near future given the implementation of new 16 bit front ends 22 If the amplitude stability measured by the reference tracking system translated directly into beam energy stability then 5e 5 energy stability would have been measured with the MIMO controller but this was not the case The best pulse to pulse beam energy stability that was achieved with this controller was only 1 3e 4 not a major improvement over the best case 2e 4 beam energy stability produced by the proportion
68. travels straight through the RF gun and solenoid field and straight through the accelerator section thereby minimizing the effects of wakefields and dispersion The beam is transversely round in the cavities and when it reaches the middle of the chicane if it has been accelerated on the slope of the accelerating wave it is elongated in the x y plane There 1s however a curvature to the beam due to the curvature of the accelerator RF It results in inhomogeneous compression consisting of a sharp leading spike and a long trailing tail A third harmonic module has recently been installed upstream of the bunch compressor in order to remove this curvature by sending the beam through a set of cavities that operate at thrice the frequency of the accelerating cavities By selecting the phase and amplitude of the third harmonic module appropriately the bunch will be compressed homogeneously 1 The beam shape in the second bunch compressor is also impacted by the addition of the third harmonic module The slice emittance of the tail is small but due to over compression transverse tails are generated at the spike of the bunch 32 This in addition to the long longitudinal tail causes the projected emittance to be significantly larger than the slice emittance With the third harmonic module the projected emittance will be reduced hopefully enabling a larger portion of the beam to lase With a non linear energy chirp the portion of the beam that l
69. um single bunch resolution unless a new electronics concept 1s invented 54 sensitivity of 8 and 20 mm diameter buttons 0 25 gt 20 mm optimal in flat chamber 0 2 8 mm installed in flat chamber 8 mm in round chamber 0 155 20 mm in round chamber a 01 gt 5 Ss 0 05 g gt gt 0 S 0 05 N 0 1 0 15 0 2 0 25 5 0 5 10 mm Figure 5 1 5 Sensitivities of button pickups in flat chamber and round chamber configurations The sensitivity of the button configuration which was installed in the bunch compressor flat chamber is plotted in green the sensitivity of an optimal flat chamber button pickup configuration is plotted in red and the sensitivity of a standard button pickup configuration in a round chamber is shown in red Fig 5 1 5 also implies that the flat chamber button configuration that was initially installed might have worked with a resolution of about three times the 15 20 um resolution produced by the round chamber button pickups at 1 nC a surprising result given that under test the measured difference between the amplitudes of the voltages of the pickup outputs was below the signal to noise ratio of the pickup output voltage for a significant range of beam positions close to the middle of the two buttons The lesson from this is that it is not always possible to make scaling assumptions about button monitors because below a certain threshold noise dominates To
70. would be less than 30 fs relative to an optical reference to which the pump laser could be synchronized This would make it possible for the pump laser to be used to make high resolution beam arrival time measurements within a limited 30 fs dynamic range This was the goal of the FLASH optical synchronization system The development of the newer THz beam arrival time measurements relative to the pump laser does not however make the optical synchronization system for FLASH obsolete When one can control the beam arrival time with femtosecond precision one can create defined timing patterns in the bunch train enabling new sorts of experiments In addition THz radiation is not easy to transport and for femtosecond resolution the length of every hundred meters of optical path must be stable on the sub micron level While this may be possible at FLASH the distances involved at the European XFEL are prohibitive For the European XFEL 6 and for sFLASH 7 the optical synchronization system is absolutely critical for the success of the experiments The European XFEL is a much larger scale and higher energy FEL that will be commissioned in 2014 In the XFEL the THz beam line is located a kilometer away from where the other experimenters are working and it would be very difficult to transport the THz radiation over that distance to provide the corresponding beam arrival time measurement In the case of SFLASH the machine is the same as FLASH except fo
71. 1 um this tilt jitter effect could eventually limit the measurement s resolution Unless the measurements or signals from the top pickup and the bottom pickup are combined this sort of tilt will generate an error and lead a user to believe that the centroid of the beam has shifted when in fact the tilt of the beam has changed While combining signals from different pickups can dramatically reduce this effect it cannot remove it entirely and it could still become visible for very large horizontal position spreads large vertical offsets or large beam tilts 6 4 Tilted in y z plane A beam tilted in the y z plane is depicted beneath a bar representing the pickup in Fig 6 4 1 Figure 6 4 1 The beam tilted in y z plane relative to the pickups above and below the beam This effect will only make the bunch seem shorter than it really is This will not have a measurable impact on the measurement of the beam position but it could have an effect on the beam arrival time measurement 6 5 Tilted in x z plane The beam 1s tilted in the x z plane as shown n the picture of the particle distribution in the middle of the chicane Fig 6 5 1 The distribution is more complicated than just a flat distribution or a Gaussian distribution tilted in the x z plane but for the sake of simplicity in the following Green s function calculations such a simplified beam will be used 90 10 5
72. 2 1 4 1 6 change ACC1 Figure 7 2 15 1 3 GHz front end beam position measurement as a function of beam energy The 1 3 GHz front end can also measure changes in the beam arrival time Fig 7 1 16 EBPM arrival ps setpoint front end 0 2 0 4 0 6 0 8 1 1 2 1 4 1 6 change ACC1 Figure 7 2 16 1 3 GHz front end beam arrival time measurement as a function of beam energy Arrival time jitter is 400 fs rms during this measurement Instead of delivering the pickup signals over 30 meter long cables to the RF front end outside of the tunnel if the RF front end is installed in the tunnel the drifts of the cables will no longer have an impact on the resolution of the measurement and the amplitude of the signals from the pickups will be less attenuated possibly enabling the 119 removal of some amplifiers from the front end circuits The outputs of the mixers can be delivered over the long cables to the VME crates which are typically installed outside of the tunnel While the phase of a signal tends to drift on a long RF cable the amplitude does not Amplitude drifts would not therefore be a problem for the transport of the mixer output signals Noise picked up on the cable could still however be a problem Since the RF front end chassis of the chicane BPM is being used primarily as a beam position monitor and not as a beam arrival time monitor drifts of the phase of the reference signal on a long cable int
73. 5 E gt 5 0 gt Lt 3 O or sum diff slope 0 003 slope 0 050 0 5 4 2 0 2 4 6 tilt deg Figure 6 3 5 Impact of x y beam tilt on beam position measurement The position of the beam was held constant and the tilt of the beam was changed and measured on a screen The sum of the pickup signals arrival times 1s constant while the difference of the pickup signals arrival times has a dependence on the tilt of the beam The error due to a beam tilt of one degree will cause 50 um of measurement error Although the resolution of the oscilloscope measurement was approximately equal to the energy stability of the beam 5e 4 averaging over 30 pulses reduced the measurement error to a few microns The tilts were created by making orbit bumps around the first accelerator section as described in the chapter on the beam shape in the bunch compressor Ch 4 Using data from the plots in Fig 6 3 3 and Fig 4 3 6 a 4 mm orbit bump in the first accelerator section causes 2 degrees of beam tilt in the chicane and causes a 100 um position error to be measured by the BPM This means that since the tilt effect caused by dispersion downstream of the first accelerator section s a linear 89 effect the typical 100 um orbit jitter downstream of the first accelerator section could cause enough beam tilt jitter to create a 2 5 um error in the position of the beam as measured by the chicane BPM Given a target resolution of
74. 8p 0 0Ud 43 00 18 00 00 00 06 00 12 00 Figure 8 3 7 Out of loop measurement drift without temperature control and without disturbances people in room Laser amplitude drift during same time period The conclusion from this out of loop measurement is that with active control of the laser amplitude and of the temperature within the chassis the RF lock can in principle be made stable within 10 fs pkpk A DSP feedback on the laser amplitude 1s in the planning stages and a new commercial MLO OneFive has been purchased and in preliminary tests it had a good amplitude stability even without active feedback on the amplitude This stability results from the sealing of the chassis containing the laser making it insensitive to humidity changes air currents and small temperature changes All of the components in the laser based synchronization system could benefit from such packaging 144 For a frame of reference concerning the costs of these systems the cost of an MO and an MLO with optical table and control hardware s presented n table 8 3 1 The largest cost for both MLO and MO systems s for the infrastructure including racks laser table and climate control It is possible to invest millions in climate control PSI for example has made a considerable investment to stabilize the temperature of the entire 500 meter long linac within 0 1 degrees C This reduces the drift problems associated with RF components and compared to th
75. BPM in broken state energy change 0 25 Buffer number problems are apparent here 2 4 6 hours Figure 10 4 5 Optical EOM front end position measurement 10 4 GHz front end measurement photomultiplier tube position measurement time of flight measurement involving 2 BAMs and the setpoint of the gradient are plotted together over several hours 162 10 Conclusion and Outlook S x distinct methods of measuring the beam energy n and around the first bunch compressor chicane of FLASH have been presented and compared The various pickups and techniques have been described in detail The chicane BPM with the optical front end has demonstrated the highest resolution out of all of the different methods An alternative chicane BPM front end which used cheaper 10 4 GHz RF techniques also demonstrated acceptable resolution The problem with these and other high resolution systems like the PMT monitor and the BAM is that they have a limited dynamic range and require frequent mechanical delay line adjustments and calibrations in order to deliver accurate measurements To address these dynamic range limitations lower resolution and larger dynamic range measurements were developed for both the optical and RF front ends The 25 um resolution 1 3 GHz RF front end s an ideal solution for situations that require a quick installation commissioning time and no down time due to mechani
76. CD camera and when the pixels from the region of interest around the beam are summed together along the vertical axis changes in the beam position can be measured by fitting a line to the sharp rising edge of the profile and checking for changes n the horizontal position of the zero crossing of the line 146 Fig 9 1 2 This method proved to be more accurate than any method tak ng the peak or centroid of the distribution Figure 9 1 2 A picture of the beam as imaged with the synchrotron light camera A projection of the image is shown below along with a line fit to the steep rising edge Changes in the zero crossing of this line give a 10 um resolution measure of the beam position This method produces beam position measurements with 10 um resolution 52 The dynamic measurement range of the device is several millimeters and if the beam moves outside of this range a motor must be adjusted and the beam must be manually centered on the CCD The other problem with the monitor is that it does not have single bunch resolution of the entire bunch train An MCP Micro Channel Plate can enable the selection of one bunch from the bunch train for analysis but the information from the rest of the bunch train is lost 9 2 Photomultiplier tube monitors Two photomultiplier tubes placed 1 7 m away from the second bend of the fist bunch compressor can provide a beam position measurement with a resolution that is slightly worse than that of the pro
77. Chapter 3 2 3 Bunch Compression When an electron bunch travels through a magnetic bunch compressor chicane the high energy electrons travel a shorter path than the low energy electrons and therefore arrive at the exit of the chicane earlier than the low energy electrons If the electron bunch is given a longitudinal energy dependence chirp by accelerating it off of the crest of the RF wave in the cavity the electrons in the head of the bunch gain less energy than the electrons in the tail of the bunch Fig 2 3 1 The energy dependent path through the bunch compressor will make the head arrive later and the tail arrive earlier thereby shortening the bunch Figure 2 3 1 The phase of the accelerating RF relative to the beam determines the energy chirp of the beam Here the head gains less energy than the tail of the bunch Due to the curvature of the RF wave the bunch acquires a non linear energy chirp When a bunch with a non linear energy chirp travels through the chicane it is compressed inhomogeneously resulting in a sharp leading spike with a high charge density followed by a long trailing tail An RF module operating at the third harmonic of the 1 3 GHz accelerator frequency will remove the non linearity of the energy chirp and enable homogeneous compression of the bunch In Fig 2 3 2 plots of the beam distribution for non linear compression in the first chicane of FLASH were generated by a particle tracking simulation whic
78. FLASH The impacts of beam shape charge position jitter and accelerating RF properties were also investigated Measurements were undertaken to verify the predictions of these simulations and benchmark the measurements against those of an existing synchrotron light detection scheme and a beam arr val time monitor scheme 2 Free electron Lasers The facility at which the experiments were performed is the Free electron LASer in Hamburg FLASH FLASH consists of an RF _ photo injector electron source superconducting RF accelerator sections two magnetic bunch compressors and an undulator section Figure 2 1 ACC ACC 23 ACC 456 Undulators N Diag BC2 BC3 GUN Figure 2 1 Free electron LASer in Hamburg FLASH including RF gun GUN superconducting RF accelerator sections ACCI 6 two bunch compressors BC2 3 and an undulator section The electron beam is produced in the RF photo injector accelerated compressed accelerated and compressed again and finally sent through the undulator magnets producing FEL light pulses with wavelengths between 6 and 40 nm for experimenters 2 3 Experimenters use the pulses of light to among many other sorts of experiments take pictures of molecules with crystallographic and pump probe methods The pulses range in duration from 10 to 50 femtoseconds with the possibility of making them as short as a few hundred attoseconds 4 If experimenters cannot synchronize their measurement e
79. J If VN I N i IW CLV II Ni IN III SI 1 I al i I an 0 5 HRDIR IN Ill N IN I WII FE PE Mi MI CHAR IL TE PTAA UAT HAA TAT ANZ Il Hh ih Waa AY N If Al N A II AN AA A yy Ny A Ah fly IN AVA l AIM RAN I Il N A i II N I Vad y WIV IN MA N ll I D et Inn Il N Ei N N I HA 4 IN N KUREN N I AA un at Ii oh I I Wl Hii INN IN fi I IN vi if i i MATA ii IN Wi A IH N He wil WHA ty YAY Mi w 0 5 WATE WAL EA a u
80. K This is a reasonable representation of the arrival time jitter of the beam core but as the many approximations made in the derivation showed it does not describe at all what the rest of the bunch is doing for non linear compression or for operation with the 3 harmonic module It can also not be used recursively for an additional bunch compression stage 168 Higher order terms and the 3 harmonic module To see how the 3 harmonic module is affects the timing jitter more terms need to be carried in the derivation Even without the 3 harminic module the higher order terms are necessary to understand how the reference particle acts with respect to the nominal particle Starting with the energy of a particle located at z E z E z PV cos k z o Vo cos 3k Z 0 E z E z Vik sin k z V 3k sin 3k z Pro E z E z VK cos k z 9 V k cos k z 9 and the path length through the chicane 2 Da Dahn T E L Ro EEn Ta EEn nom nom l joy one BD Boo t Hse a t l T E Ms tones one can Taylor expand z z T E z T E about z 0 and Enom to get z 1 T E z Ey TE 2 and then define the compression factor C z a a T E z eny T E b a f Using these terms in the derivation of the arrival time jitter should predict the arrival time jitter of more than just the core of the bunch More than that it provides a tool for optimizing
81. Measuring the Electron Beam Energy n a Magnetic Bunch Compressor Dissertation zur Erlangerung des Doktorgrades des Department Physik der Universitaet Hamburg vorgelegt von Kirsten Hacker aus Minneapolis USA Hamburg 2010 Gutachter der Dissertation Gutachter der Disputation Datum der Disputation Vorsitzende des Pruefungsausschusses Vorsitzender des Promotionsausschusses Dekanin des Fachbereichs Physik Joerg Rossbach Bernhard Schmidt Joerg Rossbach Eckhard Elsen 16 09 10 Caren Hagner Jochen Bartels Heinrich Graener Abstract Within this thesis work was carried out in and around the first bunch compressor chicane of the FLASH Free electron LASer in Hamburg linear accelerator in which two distinct systems were developed for the measurement of an electron beam s position with sub 5 um precision over a 10 cm range One of these two systems utilized RF techniques to measure the difference between the arrival times of two broadband electrical pulses generated by the passage of the electron beam adjacent to a pickup antenna The other system measured the arrival times of the pulses from the pickup with an optical technique dependent on the delivery of laser pulses which are synchronized to the RF reference of the machine The relative advantages and disadvantages of these two techniques are explored and compared to other available approaches to measure the same beam property including a time of flight
82. Scheme for delivering a stable reference signal to the phase detection circuits using an optical signal from a length stabilized fiber link 106 While this option was not built and tested in this thesis work these sorts of schemes are under active study for other applications 49 An alternative way to improve the drift stability of the RF arr val time measurement is to use RF cable reflectrometry With the active phase stabilization method from 46 a signal of a few GHz is sent from a Master Oscillator MO over a coaxial cable to an end location at which part of the signal is reflected and sent back to the source The phase of the signal being generated and the phase of the signal returning are then be compared by RF phase detection The length of the cable is then adjusted until the returning phase matches the sent phase thereby removing the effects of cable drifts Sub 100 fs drift performance was achieved with this method Although the phase detection of the returning pulse can be made with 10 fs accuracy the reflection of the pulse at the end of the cable 1s problematic due to the temperature dependence of the mismatch that produces the reflection Any drift of the mismatch will be corrected by the feedback loop but this does not accurately represent the drift of the cable and it therefore adds an error to the cable length stabilization Another error comes from the drift of the finite directivity of the coupler Individual cou
83. The optical front end is also immune to the effect of reflections within the pickup It can be calibrated with an external reference and due to reflections in the pickup the RF measurement must be calibrated with either a scan of the accelerating gradient or of the chicane dipole current This means that the optical measurement calibration is parasitic and the RF measurement calibration is invasive and would disturb machine operation The optical front end can deliver lt 6 fs resolution beam arrival time measurements with length stabilized fiber links and under the best possible circumstances RF front end beam arrival time measurements could not be made better than 10 20 fs The cost of the RF front end is however about half of that of the optical front end Table 7 5 1 RF trombone 600 collimators 300 each Temp control 600 Beckhoff 3000 100 m cable 600 Cabling eto 5 000 Table 7 5 1 Rough cost estimate of RF front end and optical front end for the chicane BPM The cost of a length stabilized RF cable or optical link is not included 131 8 Beam Arrival Time Monitors An optical beam arrival time monitor with 6 fs resolution 20 has made a high resolution time of flight energy measurement in the bunch compressor a possibility and is the only monitor system that has the accuracy to cross check the measurements of the chicane BPM constructed in this thesis Like the optical front end of the chicane BPM this monitor require
84. This can be done at a speed of 100 kHz thereby counteracting much of the noise produced in the amplification of the reference RF signal The RF module phase jitter can then be stabilized by measuring the arrival time of the beam relative to the optical reference and then feeding back on the phase setpoint of the RF module It does not however make sense to stabilize the cavity phase in this manner without first stabilizing or measuring the laser jitter because any beam based feedback would respond to both RF module phase jitter and laser jitter 31 N Two laser Dichroic mirror BBO crystal GDD Dichroic mirror pulses Reflects second harmonic Generatea Transmits 800 nm 527 7 nm light generated group delay 527 7 nm 1550 nm Transmits when pulses a dispersive Reflects 800 and 1550nm overlap medium 800 and 1550nm Figure 3 4 1 Measurement of the arrival time of the injector laser pulse relative to a timing laser reference MLO using a two color single crystal balanced optical cross correlator From S Schulz Synchronization of Injector Laser and Master Laser Oscillator PAC 10 In Fig 3 4 1 the arrival time of the injector laser pulse 1s measured relative to the optical reference laser n an optical cross correlator In a two color optical cross correlator two laser pulses with different colors 800 nm and 1550 nm are sent through a dichroic mirror that reflects the sum frequency 527 7 nm and transmits the higher frequencies The
85. This time of flight measurement with two BAMs was unsuccessful due to the off crest Operation in the second accelerator section The large jitter seen in the two BAM measurement is due to arrival time jitter about the slope of the RF in the second accelerator section This jitter is not measured by monitors in the first bunch compressor Because the machine program was dedicated to another experiment only two sample points were taken for each step in the scan of the gradient in Fig 10 4 3 Another cross check of the measurements done by the optical front end of the chicane BPM is provided by the BAM located upstream of the chicane The arrival time upstream of the chicane tupstream In terms of the quantities measured by the chicane BPM x lehieane IS t Rss x t upstream R chicane 16 This quantity measured with the chicane BPM is compared to the beam arrival time measured with a BAM located upstream of the chicane in Fig 10 4 4 160 stripline U button en EAN 1 4 v ir A N ran Fr Ne 1 2 a f we w 1 2 z If c 0 8 j do a D Y 7 on 06 Q D 0 4 H 0 2 0 if 0 100 200 300 400 500 600 700 800 Bunch number Figure 10 4 4 The beam arrival time upstream of the chicane measured with both the transversely mounted stripline BPM installed in the chicane and with a button type pickup BAM installed upstream of the chicane The ripples observed on both the BAM
86. U Pa EN j 5 f NL N Fi 4 4 10 Yo J f J wo 1 5 E 0 2 0 1 0 0 1 02 03 04 time ns Figure 5 1 13 Ring pickup output with red and without blue limiter Simulation shown in black Undesired bump in signal changes position when beam position changes Ring pickup signal slope with and without combiner Oscilloscope 800 A 700 gt Without combiner gt E 600 6 500 gt o O f 400 o With combiner 300 FO J amp amp ab 200 Q J N 100 v 0 4 2 0 2 4 6 horizontal position mm Figure 5 1 14 Position dependence of pickup output slope with and without combiner The new button type pickup was designed to maximize the bandwidth of the output signal without sacrificing too much of the amplitude Another concern is the length of time that the signal from the pickup rings The XFEL bunch spacing is only 200 ns and the ringing from the first bunch must be gone by the time the second bunch comes While the ringing is gone within the FLASH bunch spacing of 1 us it is not gone within 200 us The limitation of the design has not however yet been completely evaluated because the in house combiner that has been used so far has a 6 dB insertion loss and creates standing waves on the cable between the pickup and the combiner A better commercial combiner will reduce this effect and reduce the ringing observed The remain
87. air up onto the bottom of the chassis The two temperature controllers are mounted to the back plane of the 3 RU chassis and their cooling blocks are mounted outside of the chassis This was a successful arrangement and could stabilize the temperature of the chassis to lt 0 003 degrees C peak to peak with the lid open in an air conditioned room with a thermal stability of 0 2 degrees C If a lid is added to the top of the chassis the temperature control system is much more unstable and difficult to manage but it can resist larger temporary temperature changes if the controller is properly set up When the gain of the controller is too high and the setpoint of the temperature controller is too low the Peltier will continuously pump heat to the bottom of the chassis thereby heating up the air within the chassis warming up the circuit and causing the Peltier to pump even more heat to the bottom of the chassis If the setpoint is higher than room temperature the system is more stable even if the controller gain is high Hence the strategy for commissioning the temperature control system with a closed lid and a high controller gain was to use a setpoint that was several degrees above room temperature and then slowly over the course of days reduce it until it was closer to room temperature This produced the most efficient operation in the air conditioned room but given large permanent changes in the room temperature of a degree or more the temperat
88. al gain controller This just serves to reinforce that stabilizing cavity field measurements does not always stabilize the beam energy A summary of the module controller benchmarks described above is given below in Table 3 1 1 Only the amplitude stability measurements have been verified with beam based measurements The jitter and drift of the phase refers to the jitter of the module RF phase relative to the 1 3 GHz reference signal phase and not relative to the beam phase a Current System Reference Injection System Identification adrift jitter drift jitter drift jitter Table 3 1 1 Module controller performance benchmarks as it now stands for the system with the addition of reference injection and for the system with the addition of the system identification algorithm It should be noted that the numbers from Table 3 1 1 represent the best measurements of the jitter performance the typical performance is 4e 4 in amplitude and 0 07 degrees in phase as measured with beam based devices 23 These beam based measurements used the beam image on a screen in the bunch compressor to measure the energy jitter and a bunch compression monitor pyrodetector to measure the phase jitter While the energy jitter measured with this method is primarily due to the amplitude jitter of the first accelerator section with a smaller contribution from injector phase jitter the phase jitter 23 measurement is limited by the injector phase j tter
89. al pattern that is measured at the output of the mixers when the phase of the LO s changed The location of the bump in the distorted signal changes when the beam position 1s changed The red signal is from the left side of the pickup and the blue is from right side of the pickup A change in the difference of these signals phases is proportional to a change in the beam position The phase of one of these signals is most accurately measured by sampling a point on a steeply rising or falling edge The bump could present a problem if it pops up on the slope which has been selected for the signal phase measurement In principle the bump can always be avoided by selecting the sampling point that is unaffected by it but this adds undesirable complexity to the 102 algorithm required to select the phase measurement sample point If the bump cannot be easily removed its origin should at least not remain a mystery Perhaps it is naive to assume that the only pulse of significance to this measurement is that of the initial beam transient There are after all rather significant reflections n the pickup If the time elapsed between the initial pulse and reflected pulse is shorter than the duration of the mixer output pulse then the mixer output pulse amplitude 1s composed by a combination of both the initial and reflected pulses Since the reflection in the pickup occurs at 300 ps after the initial beam transient pulse it would make sense that
90. amplitudes More rapidly changing charge density produces a higher amplitude signal than a slowly changing charge density This is depicted in Fig 6 2 1 for an elliptical beam shape and for a flat beam shape both are types of beams which can be generated with different injector laser parameters ai EE E i A aE Figure 6 2 1 Coupling of the beam to the pickup for an elliptical beam left and for a flat beam right is shown on the top in black The charge distribution of the beams is shown on the bottom in red The beam is directly under the pickup It is easier to imagine why the pictures above look as they do if one imagines the beams divided up into slices Each slice generates a pulse traveling to the left and a pulse traveling to the right When a slice has the same charge and vertical position as its neighboring slices the pulses that it generates will be canceled out through destructive interference with the pulses generated by the neighboring slices Fig 6 2 2 83 Figure 6 2 2 Cancellation of signals on the pickup through destructive interference for neighboring pencil like beams When the beam is no longer under the pickup the only signals remaining are the ones shown in bold While the signals on the pickup look very different directly above the beam for the two cases shown n Fig 6 2 1 by the time that the pulses have been transported to the outputs of the pickup they have very similar properties This is due to the
91. an electrical signal The incoming laser pulse is split and travels through two Lithium Niobate Li NbO3 crystals under the influence of the electric field from the electrical signal The electric field through each crystal is of opposite polarity This is not however the way n which the EOM was actually used in the optical front end of the BPM The electrical pulse in Fig 7 3 1 is very long compared to the repetition period of the laser for purely illustrative purposes In the BPM optical front end implementation the electrical pulse was long compared to the duration of the laser pulse but short compared to the period of the laser pulse train This is illustrated in the following figure Fig 7 3 2 mi beam transient 60 100 ps long pulsed laser signal 20 200 fs long laser pulse Figure 7 3 2 Mach Zehnder Electro Optical Modulator EOM used to sample the zero crossing of a beam transient pulse 122 Depending on the arr val time of the beam transient pulse with respect to the arrival time of the laser pulse the amplitude of the modulated laser pulse shown in the middle of the laser pulse train in Fig 7 3 2 will increase or decrease To convert this amplitude change into a measurement of the pickup signal s arrival time one must use a laser with a very stable repetition rate that can serve as an arrival time reference signal and the amplitude of the laser pulses must be accurately measured with a photodetector a
92. anced optical cross correlator used to measure the difference between the arrival times of pulses coming from and returning to the MLO 50 In this balanced detection arrangement when the signals at the photodetectors are balanced at about half of the maximum signal the optical length of the fiber is stabilized with an accuracy that can be well below 10 fs 20 Fiber length stabilization can also be accomplished with a scheme that detects the arrival of the MLO pulse and the returning pulse with RF phase measurements of the signals resulting from the laser signals impinged upon photodetectors This is significantly less expensive than the optical cross correlator but the best resolution that it can achieve is about 10 fs 47 In Table 8 2 2 an estimate of the cost of a 6 fs resolution optical front end and a length stabilized fiber is presented The total cost of 55 000 EUR can be compared to a total cost of 15 000 EUR for an RF front end and phase stabilized cable system Price EUR BAM front end 30 000 Link w OCC 25 000 55 000 Table 8 2 2 Cost estimate for 6 fs resolution optical front end and a length stabilized fiber Cost for MLO distribution not included 8 3 MLO RF lock All of the results quoted for the beam arrival time measurements done with the optical front end refer to the measurement of the arrival time of the electron beam relative to the laser pulses from the MLO but a beam arrival time measurement relative to th
93. and BPM arrival time measurements are due to a 50 kHz oscillation on the gun phase Both measurements were averaged over 20 shots and the more spiky texture of the BPM arrival time measurement results from problems with buffer numbers the buffer number from one side of the pickup was not equal to the buffer number from the left side of the pickup in about 20 or more of the cases It appears that the buffer number problem has since been solved The question remains which monitoring system should one believe Several of the monitors described in this and past sections are plotted together below in Fig 10 4 2 Buffer number problems aside for all of the different monitoring systems to demonstrate agreement they must all be properly calibrated and within the dynamic range of the measurement Because the dynamic range of the higher resolution measurements is only a millimeter or two these requirements are not always fulfilled and the software that is needed to diagnose whether or not the measurement is accurate becomes more complicated As the stability of the machine is improved the higher resolution measurements will become more interesting and valuable 161 0 25 Ai T It N N N A yA G A 2L INN ANAC IN Ion REN 0 N N A AN I Ir Vy iki a a Cy ae es a a yo A y 14 i IN t nh ae bi EOM BPM M ACC setpoint 0 05 Lyle Wh So N N N J PMT y H F
94. and CSR wakes in the chamber could very well impact the accuracy of the measurement 6 8 Summary The conclusion that one can take from these studies of the beam tilt is that the beam position monitor measures a combination of the centroid of the beam the tilt of the beam and the path of the beam If the beam is tilted in the x y plane the effect of the tilt can be mostly removed by combining the signals from the top and bottom pickups If the beam has an asymmetric charge distribution and is tilted in the x z plane there is a systematic error of up to several hundred femtoseconds that cannot be removed by any means In this case one measures a property of the beam that is determined by both the centroid and tilt of the beam The effect of the path of the beam appears to be so small that it is difficult to measure A table summarizing the strongest effects 1s shown below symmetrical Length mm Width mm Tilt deg l Error fs Error um 660 Table 6 8 1 Effects of beam tilts on beam position measured with transversely mounted stripline BPM Several hundred microns of measurement error can be expected from typical beam distributions and tilts 95 7 Chicane BPM Front ends Using the transversely mounted stripline pickup the beam position and arrival time is determined by measuring the arrival times of the pulses coming from opposite ends of the pickup So far we have ignored the fact that these pulse arrival time measur
95. and photon pulse The process described above s known as SASE Self Amplified Spontaneous Emission wherein the micro bunching structure develops spontaneously from shot noise and grows more distinct as saturation is achieved An alternative to SASE is to use a seed laser to initiate the micro bunching process at a desired frequency This makes the light generated in the undulators much more monochromatic and intense Such a seeding project called sFLASH is underway for commissioning during the coming year 7 A complete introduction to VUV and X ray FEL techniques is given in 18 18 3 Beam Arrival time Stabilization As we learned n the previous chapter beam energy changes upstream of a bunch compressor become arrival time changes after the bunch compressor and it is desirable for the electron beam to have a stable arrival time relative to a reference signal so that seed laser diagnostic laser and pump laser pulses arrive synchronously with the beam A feedback to control the arrival time of the FEL beam can be made no more complicated than a single monitor like a beam arr val time monitor after the chicane that tells a single klystron how to set the energy of the beam While such a feedback can produce a dramatic improvement in the arrival time jitter measured at one point in the machine it can have the flaw that it feeds back on arrival time jitter that 1s generated somewhere other than in the module it is controlling It also cons
96. and the plot looks the same if the beam is on or off crest This result implies that the 74 pickup functions appropriately over the full range of the bunch compressor vacuum chamber Beam position a 18 0 deg position cm O AE E 0 Figure 5 5 7 Beam position across the full range of the vacuum chamber as a function of the beam energy change The red stars are the beam positions as measured with the pickup and an 8GHz oscilloscope The solid blue line represents the expected position for various energy deviations as calculated with first second and third order dispersion The broken blue line is calculated with first order dispersion alone At the upper left edge of the plot in Fig 5 5 7 the beam was scraping on the edge of the beam pipe At the lower right end of the plot the gradient of the cavity could not be increased any more The higher order dispersion is also plotted it is the curved line along which the measured positions lie Because the oscilloscope samples the signal many times using an oscilloscope to measure the arrival times of the pulses ignores the effect of amplitude changes of the signal and only measures the zero crossing of the signal When one must rely on only one sample point per zero crossing measurement one must sample the signal close to the zero crossing and have a calibration measurement of the slope of the signal close to the zero crossing Of course a
97. anges In Fig 8 2 2 the arrival time of a pulse emerging from the MLO s measured relative to the arrival time of a pulse returning from a timing sensitive device This measurement is done in an optical cross correlator In an optical cross correlator two laser pulses with polarizations that are perpendicular to one another are sent through a dichroic mirror that reflects second harmonic light and transmits the fundamental mode The input laser pulses are transmitted through the mirror and are sent through a PPK TP crystal which The right side of the crystal is coated with a high reflectivity coating and an anti reflective coating in order to reflect and transmit light pulses from the fundamental and second harmonic light The left side of the crystal 1s coated with anti reflective coatings so that it transmits both the fundamental and second harmonic light When the pulses overlap in the crystal second harmonic light is generated and emitted in both forward and backward directions With the aid of dichroic mirrors these pulses each travel to a photodetector as the incoming light returns from whence it came 136 Photo detector Photo detector Two laser Dichroic mirror PPKTP crystal Dichroic mirror pulses reflects second second harmonic transmits second polarizations harmonic light light generated when harmonic light perpendicular to transmits pulses overlap reflects fundamental one another fundamental Figure 8 2 2 Bal
98. ant as it is for scans of the vector modulator phase indicates that the voltage read back of the potentiometer is not a reliable indicator of the true position of the stage in the mechanical phase shifter 115 1 6 mm V 10 4 GHz downmixed signal 0 10 20 30 40 50 60 70 80 90 100 Potentiometer mm Figure 7 2 12 Fiducializing the mechanical phase shifter potentiometer with the vector modulator If the calibration factor for the potentiometer voltage is correct the signals will look exactly the same and the slope of the signal measured by scanning the phase of the vector modulator will be equal to the slope of the signal measured by scanning the phase with the mechanical phase shifter Backlash of the motor is seen as a discontinuity between data points measured after a forward movement of the stage red compared with the data taken after a backward movement of the stage blue Further evidence of the problems with the mechanical phase shifter are seen when a scan of the beam energy is done with a feedback that uses the mechanical phase shifter to center the sampling position about the zero crossing of the signal Fig 7 2 13 The beam position should change in a linear fashion as the beam energy is changed over a small range but the beam position as a function of beam energy is curved and does not follow the setpoint of the accelerator section Despite the problems with the mechanical phase shifter whenev
99. ant to ignore n the first four modules of the accelerator section and this creates a mismatched result in the simulation With proper matching the slice emittance of the beam will be smaller and the beam will look like a curved strip in the second pair of plots In the last pair of plots the horizontal position spread has returned to the value it had before entering the compressor and the bunch has been compressed into a 10 before bunch compressor T T T T x axis mm oO z axis mm middle of bunch compressor T T x axis mm z axis mm after bunch compressor x axis mm Figure 2 3 2 06 04 02 0 02 04 06 z axis mm energy chirp before bunch compressor 124 123 122 E MeV 121 120 119 z mm 125 124 123 122 E MeV 121 120 119 energy chirp in middle of compressor 118 3 125 energy chirp after bunch compressor 124p ee e 123 122 E MeV 121 120 119 08 0 6 04 0 2 0 0 2 0 4 0 6 z mm Bunch profiles in the first bunch compressor for inhomogeneous compression without the third harmonic cavity The transverse beam size is larger than in reality because space charge effects were not taken into account after the RF photo injector sharp leading spike of charge distribution followed by a millimeter long trailing tail The energy di
100. apan M Hoffmann F Ludwig H Schlarb S Simrock Multichannel Downconverter for the Next Generation RF Field Control for VUV and X Ray Free electron Lasers Proceedings of the PAC 2007 H Schlarb et al Beam Based Measurements of RF Phase and Amplitude Stability at FLASH Proceedings of the DIPAC 2007 Venice Italy J Cawardine Results from full beam loading 9 mA experiment FLASH seminars November 2009 H Schlarb V Ayvazyan F Ludwig D Noelle B Schmidt S Simrock A Winter Next Generation Synchronization System for the VUV FEL at DESY Proceedings of the FEL 2005 H Schlarb et al Precision RF Gun Phase Monitor System for FLASH Proceedings of the EPAC 2006 Edinburgh Scotland H Schlarb Plans for beam based feedback at FLASH Internal DESY 2010 174 28 29 30 31 32 33 34 35 36 37 38 39 40 41 H Schlarb et al Beam based measurements of RF phase and amplitude stability at FLASH Proceedings of the DIPAC 2007 S Schulz et al Injector laser cross correlation Proceedings of the IPAC 2010 Kyoto Japan M Dohlus and T Limberg Bunch compression stability dependence on RF parameters Proceedings of the FEL 2005 Stanford California S Wesch Spektroskopie kohaerenter Uebergagsstrahlung zur strukturanalyse von elektronenpaketen am FLASH beschleuniger Master s thesis U
101. are filled with electromagnetic waves that have a frequency of 1 3 GHz They form standing waves in the cavity producing gradients ranging from 12 to 30 MV m The waves are produced by a klystron in 800 us long pulses with a repetition rate of 5 Hz n order to accelerate up to 800 bunches per pulse Typically however bunch trains of only 1 30 bunches have been produced for standard operation A klystron consists of a cathode from which electrons are generated and an anode toward which the electrons are accelerated with a voltage drop of many kilovolts 200kV at FLASH The electrons then enter a cavity that 1s filled with gigahertz waves by a modulator This buncher cavity gives the electrons an energy modulation which 1s transformed into a density modulation in the subsequent drift section where high energy electrons travel faster than low energy electrons The electrons are thereby bunched with a periodicity equal to the period of the wave in the buncher cavity These bunched electrons travel through a second cavity causing it to ring with the same frequency that was generated by the modulator The resonant wake fields from this cavity are transported via waveguide to the accelerating structure to accelerate the beam The phase of the klystron output is strongly influenced by changing the voltage drop of the driving electrons and the amplitude of the klystron output is controlled by the amplitude of the modulator output At FLASH a multi beam kly
102. ases is but a fraction of the sharp leading spike While the duration of the spike has been measured with a transverse deflecting cavity a sort of streak camera and is lt 60 fs FWHM 32 measurements of the bunch spectrum suggest that the fraction of the bunch that is responsible for the lasing process is closer to 25 fs in duration 33 With the addition of the third harmonic module the beam will acquire a linear energy chirp With a linear energy chirp more of the beam will acquire the charge density and emittance characteristics necessary for lasing but this is only true if the projected transverse emittance from the injector is sufficiently small a requirement that can be met with a perfectly aligned injector 4 2 Mis aligned Injector The effect of a mis aligned injector can be best observed on a beam that is accelerated on crest This is due to the fact that the minimal energy spread minimizes the effects of dispersion downstream of the injector It 1s often observed on an OTR screen in the middle of the first bunch compressor that when the beam is on crest the shape on the screen is not round but rather like ac or a boomerang The head and the tail of the bunch are offset in the y plane as shown in Fig 4 2 1 emittance growth 1 8136 emittance growth 2 0826 600 6500 7990 500 0 400 450 500 pixels pixels 600 650 700 Figure 4 2 1 The shape of the beam in the bunch compressor for on crest operation as viewed on the synchr
103. bilized on two fronts the phase and the amplitude The phase in this section determines primarily the bunch length while the amplitude strongly affects the beam arrival time and the energy Following the argument given in Sect 3 1 about arr val time stability after a chicane the desired energy stability in the first accelerator section is approximately 0 004 a factor of ten improvement over the current 0 04 stability A monitor for a feedback system should be at least two times better than the stability that it hopes to achieve and so the monitor for an energy feedback should resolve 0 002 energy changes By this logic with an Rss of 180 mm 550 ps and an R s of 345 mm an arrival time measurement of the beam s time of flight through the first chicane of FLASH should resolve 10 femtoseconds and a position measurement in the chicane should resolve 7 um 21 fs It follows that the resolution required by the position measurement is lower than that required by a time of flight arrival time measurement in proportion to the ratio between the Rig and Rss terms In the case of the first bunch compressor of FLASH this ratio is 2 1 in favor of the position measurement whereas for the XFEL the 33 ratio is 6 1 Future FLASH configurations also call for a reduction of the Rss in the first chicane by a factor of two increasing the resolution requirements of BAMs used in a time of flight energy measurement while not affecting the BPM resolution r
104. cal adjustments That is why a copy of the first prototype will be soon commissioned in the second bunch compressor of FLASH Developing the infrastructure for another optical front end 1s much more time consuming and expensive The accuracy of the chicane BPM measurements was studied with respect to likely beam shapes and thermal stability of front end systems The monitor will be sensitive to changes in the longitudinal tilt of the beam and unless the signals from the top and bottom pickups are combined the monitor will also be sensitive to transverse tilts of the beam If the chicane BPM measurement is used to measure the beam energy corrections for incoming orbit changes must be implemented based on BPM measurements from before and after the chicane Higher resolution BPMs were installed for this purpose The thermal stability of the chicane BPM front end systems has been addressed with active temperature control systems involving Peltier elements and the 163 resolution limitations due to the bandwidth limitations of the pickup and splitter have been described The impact of the AM to PM conversion effect of the RF limiter on the accuracy of the measurement was not measured but is of interest Theory and systems were described to show how the beam arrival time and position measurements from the chicane BPM could be used to simultaneously stabilize the beam arrival time and energy and to measure the energy spread of the beam The merits of the
105. cessing and signal correction RF Regulation Requirement Pkpk AE E lt 5 10 BC2 a Figure 3 3 1 Layout of synchronization sensitive components at FLASH along with desired feedback loops in an optimal configuration a and a more expedient configuration b Schematics taken from in house presentation of Holger Schlarb In Fig 3 3 1 a the arrival times of the photo injector laser pulses relative to the Master Laser Oscillator MLO pulses are measured in an optical cross correlator The arrival times of the photo injector pulses can then be adjusted with a vector modulator that controls an electro optical modulator in the actively mode locked injector laser cavity This can be done at a speed of 27 MHz with the use of a uTCA crate with an ADC a DAC and an FPGA installed within All of the RF cavities phases and amplitudes would then be influenced by a central correction algorithm operating on a single crate that collects the beam energy arrival time and compression information from monitors throughout the machine and delivers corrections to the cavities controllers The 29 super conducting cavities controllers would be sent commands from the central controller on an intra bunch train basis with a moderate bandwidth while the normal conducting RF cavity amplitude could be sent commands on a fast bunch to bunch basis Each super conducting cavity would have its own independent fast vector sum controller feedback
106. ckup through destructive interference for neighboring pencil like beams Sensitivity of the chicane BPM arrival time measurement to changes in the width of the beam The beam tilted in x y plane relative to the pickup CST simulation of the sensitivity of the chicane BPM signal amplitude to changes in y position of a pencil like beam with a charge of I nC The amplitude of the signal induced on the pickup by a slice of a beam with a flat charge distribution that is tilted in the x y plane by 5 degrees as a function of the x position of the slice within the beam Pickup outputs for tilted un tilted beams with flat charge distribution Impact of x y beam tilt on beam position measurement The beam tilted in y z plane relative to the pickups above and below the beam Particle tracking simulation of a nicely matched beam at the location of the BC2 BPM The beam tilted in x z plane relative to the pickup Illustration of the spacing of the wavelets produced by beam slices as they are transported on the pickup for a tilted beam Asymmetric horizontal charge distribution with centroid offset from center by 3 3 picoseconds Error resulting from x z tilted asymmetric charge distribution Slope at the zero crossing of pickup signal over full dynamic range of monitor Down mixing scheme to measure the relative phases of two pulses The input and output of a mixer and how an appropriate phase relationship facilitates the measurement of the phase of
107. ctangular chamber Let 2h be the height of the chamber and zero be the vertical position of the beam centered between the top and bottom surfaces located at h and h This means that a sum of image currents of peam at y 2h 4h 6h yields the magnetic field seen by a particle at location x y 41 H x ee L bean a gt 5 1 10 1 1 x nh y ger which for y h becomes H x y ny Zum l 5 1 11 2n Vx h This means that the image current can be written A FT 5 1 12 Dax h and integrating x over a pickup of width w with a horizontal offset of X gives I ee Pan I x d X w h In X tv X h 5 1 13 We can now use Equations 5 1 10 and 5 1 14 to compare the relative sensitivities of pickups installed in flat chamber and round chamber configurations Three different 53 configurations of pickups are shown below in Fig 5 1 4 The unsuccessful but expedient design which was installed in the flat chambers of the first and second bunch compressors is Shown on the left It had a button diameter of 8 mm and a distance between buttons of 55 mm A flat chamber design that would have had performance comparable to existing button pickups installed in round chambers is shown in the middle with 20 mm diameter pickups and a separation between the pickups of 21 mm Typical pickup dimensions n a 35 mm diameter round chamber are shown on the right 8 mm 20 mm gt _ 1 So h gt 55 m
108. ction coefficients and S 2 and Sz are transmission coefficients These S parameters are also the terms produced by electromagnetic simulation software such as CST A typical configuration for a test of a button monitor might involve a stretched wire running through the middle of the vacuum chamber The network analyzer would send its incoming signal down the wire representing the electron beam the outgoing parameters would be measured from the pickup output From this measurement one should be able to see if the ratio of the outgoing to the incoming waves is large enough that there will be enough voltage for the front end electronics to function for a given beam current at the frequency of interest Button pickups are used frequently in the FLASH linac for both position measurements and for arrival time measurements Position measurement BPM pickups usually require a large voltage at frequencies below a GHz whereas for arrival time measurement BAM pickups for an optical setup that will be described n chapter 8 a large bandwidth that stretches up past 40 GHz without any notches in the spectrum is desirable These properties can be adjusted by tuning the impedance of the pickup design In the spectra shown in Fig 5 1 11 the BPM button delivers more power at lower frequencies than the BAM button while the BAM button delivers a steeper signal slope 60 BPM Parameter Magnitude BAM S Parameter Magnitude Frequen
109. cy Get Time Signals 0 02114 0 01 a u 0 007907 0 23729 0 5 1 1 27 Time ns Frequency GHz Time Signals 0 02304 0 02 0 01 a u 0 0 01 0 02193 _ 0 25912 0 4 0 6 08 i 1 1118 Time ns Figure 5 1 11 Comparison of frequency and time domain simulations of two pickups Above left is the cross section of a button pickup for a BPM Above right is a button pickup that is used in a Beam Arr val time Monitor BAM 61 So far the button has been treated as a simple high pass filter without accounting for the various notches that appear in the spectrum due to resonances within the pickup and cavity Some of these resonances include e A resonant cavity between the button and the body of the vacuum pipe this resonator is usually tuned to frequencies greater than 10 GHz and is excited by short bunches e The impedance variation on the transmission line from vacuum to air will produce impedance mismatching generating reflections and standing waves To deal with unwanted resonances most BPM front ends include a low pass filter to remove the higher frequency resonances This is not possible when the pickup is used in an arrival time measurement BAM where a broadband pickup output is desired The BAM pickup that was originally developed for a 1 3 GHz RF front end was a ring supported by two SMA connector sized feedthroughs Fig 5 1 12 a This effectively delivered a signal with a voltage that was
110. d of the beam is perpendicular to the beam direction Since an image current flows on the surface of the vacuum chamber to cancel out the magnetic field tangential to the metal surface the magnetic field at a distance r from the electron beam with current Ibeam 18 given by the Biot Savart law in which a vector describing the direction of the current flow of the beam is crossed with a vector pointing perpendicularly out towards the vacuum chamber walls r i dixtr I HG em CAT 5 1 5 0 Figure 5 1 3 Coordinate system for a circular vacuum chamber For a circular vacuum chamber Fig 5 1 3 with a cross section of radius r and with the beam in the center of the chamber the image line current density is then De ne ae 9 5 1 6 and for a beam that is displaced from the center of the chamber by_X and Y at a position given by D 0 the image current contained within an angular spread of can be calculated from either Laplace s equations in two dimensions yielding 43 32 J 12 j cos n y al 5 1 7 or from Biot Savart to get beam r X Y I eyes 5 1 8 m r X Y 2r Xcosp Ysing Senne p A perfect pickup electrode spanning an arc Ad integrates a fraction of Jimage Using normalized beam displacements x X r and y Y r we get the image current integrated by the pickup 2 l 2 L pickup AQ ee ta y tan 2 5 l 9 mT 1 x y The same exercise can be done for a wide re
111. deren Methoden w e zum Beispiel der Detektion der Flugzeitdifferenzen des Elektronenstrahls durch die magnetische Schikane oder der Positionsbestimmung der Elektronenpakete durch optische Synchrotronstrahlung Die Messung der transversalen Elektronenstrahlposition in einer magnetischen Schikane ist ein direktes Ma f r die Elektronenstrahlenergie und kann f r ein schnelles Regelungssystem zur Stabilisierung der Beschleunigungsgradienten genutzt werden Durch die Stabilisierung der Beschleunigergradienten kann eine Stabilisierung der Ankunftszeit des Elektronenstrahls relative zur Synchronisationsreferenz der Anlage erzielt werden Dies verbessert entscheidend Experimente die auf das seeden oder manipulieren des Elektronenstrahls durch externe Laserstrahlen angewiesen sind Contents 1 Introduction 2 Free Electron Lasers 2 1 RF Photo injector 2 2 Accelerating Section 2 3 Bunch compression 2 4 Undulator section 3 Beam Arrival time stabilization 3 1 Baseline Control 3 2 Arrival time Changes after a Bunch Compressor 3 3 Beam based Feedback Strategy 3 4 Injector Jitter 3 5 Third harmonic Module Jitter 3 6 First Accelerating Section Jitter 3 7 Second Accelerating Section Jitter 3 8 Outlook 4 Beam Shape and Orientation in and around the Bunch Compressor 4 1 Perfect Alignment 4 2 Misaligned Injector 4 3 Downstream of the Injector 5 Beam Pickups 5 1 Button Pickups 5 2 Cavity BPM 5 4 Stripline Pickups 5 3 Array of Striplines 5 5 Transversely
112. derivative of the path length will be useful later and is given by T E Rg tn 4 nom The path length through the chicane can be used to calculate the change in the longitudinal position of a particle of the beam relative to a nominal position This can be described by a transformation from a coordinate system from before the chicane z into a coordinate system after the chicane zy Fig 1 The nominal longitudinal position before and after the chicane in each coordinate system is given by z 0 This position corresponds to the nominal beam energy Enom 165 Before the chicane Dee After the chicane Z gt Z tail head Znom O a Zref Znom O0 T Figure 1 Coordinate system from before z and after z the chicane The nominal longitudinal position before and after the chicane in each coordinate system is given by z 0 This position corresponds to the nominal beam energy Enom When the position of the center of the beam before the chicane is set equal to zero the position of the center of the beam after the chicane might not be equal to zero This is especially the case when higher order dispersion terms are taken into account When this occurs it means that the center of the beam does not have the nominal energy In this derivation the center of the beam will be described by a reference particle with an energy Ere E1 z 0 A particle with position z before the chicane will be located at position zp after the chicane accordi
113. dforward control can identify patterns in the cavity signals during one klystron pulse and attempt to remove those patterns by applying a pattern of equal and opposite amplitude in a subsequent klystron pulse This control option has not however been used for day to day beam operation due to the incompleteness of its implementation The incompletely debugged failure modes of the controller have caused the superconducting cavities to quench New versions of the controller are under development 11 Feedback is different from feedforward in that it sets control parameters based on the reaction of the system to the control parameters It utilizes measurements taken at the beginning of the pulse n order to change the settings of the klystron within the pulse The current system can implement Proportional Integral and Differential PID feedback control but for typical operation only the proportional feedback is used Feedback is limited by measurement resolution and latency how long it takes for a signal to be measured interpreted and converted into a control parameter With Field Programable Gate Arrays FPGAs Analog to Digital Converters ADCs and Digital to Analog Converters DACs able to process signals at more than 100 MHz or every 10 ns and signal transport times that can be kept below 100 ns bunch to bunch feedback within the FLASH bunch spacing of 1 us and even the XFEL bunch spacing of 200 ns becomes a goal within reach The latency
114. difference between the arrival times of the incoming pickup pulses comes from the fact that the LO is common to both arrival time measurements any LO noise measured by one arm of the setup will also be measured by the other arm If one subtracts the one arm s measurement from the other arm s as in a beam position measurement the LO phase noise will cancel out The measurement of the difference between the arrival times of two pulses will only suffer from inaccuracies if the filters cables and mixers in the two different phase measurement arms drift relative to one another These thermal drift effects are counteracted by active temperature stabilization within the chassis a system described in a later sub section but first a theoretical investigation of the above circuit will be presented If the inputs of a mixer are sinusoidal voltage waves v with amplitude A frequency f and phase 6 98 v t A sin 27f t 6 the output of a mixer is the product of these signals According to the trigonometric identity sin A sin B eos 4 B cos A B we can write the output of a mixer as v t v 2 AA oos 2at f f cos 2m f f 4 6 Where K is a constant of the mixer One can see that the output of a mixer is a superposition of the sum and difference of the input frequencies The sum and difference of the phases will also govern the output of the mixer If one estimates the output of the filters
115. dispersion of the pulse as it travels along the pickup and the filtering effects of the impedance mismatches in the pickup and vacuum feedthrough The higher frequency components of the spectrum of the pulse will be more strongly suppressed than the lower frequency components such that after transport to the output of the pickup the length of a shorter pulse has increased by more than the length of a longer pulse It is still clear from the picture however that even for an elliptical beam when the beam 1s directly under the pickup there is a space between the zero crossings of the signals on the pickups That space is proportional to the width of the beam and consequently to the energy spread of the beam If an incoming beam arrival time measurement is available from before the chicane the difference between the incoming arrival time measurement and the arr val time measured with the BPM pickup n the chicane will give a measurement of the beam energy spread according to arrivalincoming arrivalgpm Ris AE E This is because incoming arrival is measured with button pickups and a non dispersed beam and the BPM arrival time in the chicane measures the arrival of the locations where the charge distribution is changing for a beam which is stretched out transversely The arrival time measured with the BPM is given by the average of the arrival time measured by both outputs of the pickup The arrival time measured by the BPM will then be late
116. duction a fixed feedforward setpoint table can be determined When the beam is then added to the system a slope on the RF pulse arises due to something called beam loading Beam loading occurs when an electron bunch enters the accelerating cavity The beam takes energy out of the accelerating field and this energy must be replaced by increasing the klystron s output If in one bunch train each bunch took a certain amount of energy out of the cavity the same thing 1s likely to happen in a subsequent bunch train as long as the beam charge or orbit is not significantly changed A fixed feedforward table may be appropriate for one set of beam parameters but as soon as the machine operator changes the setup of the machine the feedforward table will have to be manually tuned to compensate for changes in beam loading In the absence of an expert to tune the feedforward table an adaptive feedforward algorithm using Iterative Learning Control can automatically change the feedforward table in order to counteract the changes in beam loading The control decisions of the adaptive feedforward algorithm do not take place within the bunch train but after averaging over multiple bunch trains The adaptive feedforward can not only remove slopes from the bunch train it can also in principle remove ripples If ripples are periodic and appear in bunch train after bunch train with the same phase they can be removed through the feedforward An adaptive fee
117. e it was concluded that the c shape seen on the OTR screen in the dispersive section of the bunch compressor for on crest operation is due to a badly several mm mis aligned solenoid The first hint of this was that in ASTRA beam transport simulations no combination of wakefield coupler kick and dispersive effects was strong enough to cause the shape seen on the screen but if the solenoid was given a 1 cm offset from the beam axis the beam shape that would be clearly visible on the screen in the dispersive section of the first bunch compressor would be that of a C The solenoid alignment was not previously suspected as the cause of the c shape because a lot of trouble is generally taken to align the solenoid with sub mm precision 38 Measurements confirming the conclusion that the solenoid alignment was the culprit involved changing the position of the iris on the cathode This changed the position of the electron beam within both the RF module and the solenoid As the electron beam position relative to the module and solenoid was changed changes in the maximum separation of the head and tail of the beam were observed on the bunch compressor screen The maximum separation of the head and tail was determined for each iris position by scanning the current of a quadrupole in order to find the phase advance for which the head tail separation was maximized When the charge was changed between 0 5 nC and 3 nC the c shape did not change appreciably ruli
118. e RF reference of the machine can never be better than the lock of the MLO to the RF master oscillator MO that sets the reference signal for the accelerating RF If one quotes the arrival time of the electron beam relative to the MO instead of the MLO one might 137 frequently measure differences of several picoseconds in a drifting system This is not however the r ght way to look at the problem If everything of mportance lasers and arr val time feedback diagnostics are locked to the MLO then nobody cares about their phase relationship with the MO The MO signal is only important n so far as it keeps the accelerator stable and running while the MLO reference is responsible for delivering fine corrections to the stability of the accelerator and maintaining synchronization with the lasers of the FEL users It turns out however that the MLO would be a useless reference without a lock to the MO The short term phase stability of the best MOs and MLOs can be less than 4 femtoseconds 1kHz 10MHz but the long term phase stability of an MLO is really quite bad When a source with a good short term stability 1s locked with a narrow bandwidth 1kHz to a device with a good long term stability the source with the good short term stability acquires the good long term stability of the device to which it is locked The current plan to accomplish this is to lock the higher frequency MO to a temperature stabilized crystal oscillator with a lower f
119. e aperture the beam diverges with an opening angle proportional to A over distance This distance is small but some loss of photons is still a possibility The shot noise of the photo emission process is given by Ge Zu NG 9 2 6 and the photomultiplier signal is shot noise limited if it fluctuates by less than this amount To determine the shot noise limitation of the measurement we must calculate the resolution of the monitor in terms of the shot noise an outline of this derivation was provided by 57 To begin we write the normalized beam position sensitivity P f o x dx P o x dx nr SPEER EEE EEE 9 2 7 S S_ P l in terms of the probability of a photon being detected Si Np quantum efficiency The probability of a photon being detected is also equal to the probability of a single electron emitting a photon that is detected P times an integral over a portion of the bunch s charge distribution p The limits of the integrals are written such that half of the beam is detected by one detector and the other half of the beam is detected by the other The beam 149 position for which the beam is centered relative to the two photodetectors is written as xo Let us assume that there is no change in the beam profile but there is a small change of the position of the beam Ax We want to know how this affects the sensitivity of the monitor so we take the derivative of the sensitivity with respect to a small change of beam
120. e entire beam The blue line is a fit to the centroids of the slices from the central portion of the beam The orbit bumps wakefields and coupler kicks were then simulated with a combination of ASTRA in the injector and the transport matrices constructed in MATLAB for the propagation of the simulated beam through the first accelerator section and first half of the first bunch compressor The simulation results for the energy spread of the beam coming out of the injector the horizontal beam path through the accelerating module and the resulting tilt n the first bunch compressor are shown in Fig 4 3 3 for simulations with and without the energy dependent deflection of the offset trajectory through the quadrupole magnets chromatic effect The simulation was done for beams generated with and without the small influence of the energy chirp from the gun The Rosenzweig Serafini model for beam transport through an accelerator section was used without a high energy approximation 40 In the plot of the transport through the first accelerator section of a beam with an incoming angle and offset in the horizontal plane the angle goes to zero and the offset is also focused It is clear from the comparison between the simulations for which the chromatic effects of the quadrupole magnets were on and off that dispersion induced by the quadrupole field offsets is by far the largest contributor to the tilt of the beam seen in the first bunch compressor When
121. e limitation of using the chicane BPM to measure the beam energy spread will be due primarily to the limited bandwidth of the pickup itself The energy spread of the beam is related to the bunch length after the chicane and changes in several length scales of the bunch are measured with high precision are measured with high precision with a pyrodetector based single shot spectrometer or bunch length monitor Whereas at FLASH the beam is only about a centimeter wide 6 ox in the XFEL the beam may be as wide as 6 cm With wider beams comes not only a higher sensitivity of the chicane BPM arrival time measurement to the energy spread of the beam but also a higher sensitivity to tilts of the beam 85 6 3 Tilted in x y plane A beam tilted in the x y plane is depicted beneath a bar representing the pickup in Fig 6 3 1 The lines represent the waves traveling on the pickup as a result of the beam transient Where the beam is closer to the pickup the signal amplitude is larger where it is farther away the coupled signal is smaller af Figure 6 3 1 The beam tilted in x y plane relative to the pickup The side of the beam that is closer to the pickup produces a larger amplitude signal than the side that s further away One must first know the sensitivity of the monitor to changes n the y position of the beam before making an estimate of the sensitivity of the monitor to changes n y tilt This was done with the CST simulation and is
122. e monopole modes are most strongly excited but they do not contain information about the beam position That is why a slot that selectively couples out the dipole mode and not the monopole mode is desired A waveguide 1s attached to the slot In order to make a high resolution cavity monitor the dipole mode must be coupled out without picking up the monopole mode or higher order modes These requirements put strict limits on manufacturing tolerances and installation alignment While sub micron resolution has been demonstrated at multiple labs the higher the resolution that is required the smaller the dynamic range of the monitor generally becomes One major impediment to simply scaling up an existing design is that the vacuum chamber in the middle of the FLASH bunch compressor 1s flat and the standard designs of high resolution cavity monitors have round vacuum chambers This means that a good design cannot simply be scaled up Assuming that a quality large scale cavity design for a flat chamber could be made one must consider several other factors A larger cavity rings for longer than a smaller cavity and when the bunch spacing is very close as it s for FLASH and XFEL bunch trains the ringing from one bunch could overlap with that of the following bunch Slight asymmetries and deformations in the cavity production 65 impact the quality of the measurement and must be carefully investigated with respect to what is possible in manufacturing
123. e optical synchronization system represents a very different strategy to achieve good reference signal stability Table 8 3 1 The cost of an MO in a rack and an MLO with optical table and control hardware MLO numbers come from Holger Schlarb and MO numbers come from Henning Weddig There is a cheaper way to do optical synchronization using CW continuous wave lasers and one such system was built and fully commissioned at GSI by Michael Bousonville a similar concept was developed as a prototype at DESY by Matthias Felber A more complicated CW optical synchronization system has been implemented at LCLS Such systems have been shown to achieve 50 fs synchronization but not sub 10 fs synchronization They are ideal for situations that involve long distances that make RF cable costs prohibitive and where sub 10 fs synchronization is not required 145 9 Synchrotron Light Monitors Synchrotron light detectors can enable a high resolution beam position measurement after the second bend of the bunch compressor Existing synchrotron light monitor systems at FLASH are described with respect to their limitations and capabilities 9 1 Profile monitors A screen and camera positioned after the third bend of the chicane can detect the synchrotron radiation produced at the third bend of the chicane Fig 9 1 1 Mover CCD not to scale Figure 9 1 1 A synchrotron light monitor system with CCD The synchrotron light is detected with a C
124. e pickup and even though the calibrations of each side were incorrect they were incorrect in the exact same proportions and so the beam position measurement was still correct When the machine configuration changed causing the beam to sit at the far end of the pickup and the reflections to move in opposite directions temporally the calibration done by scanning the LO was two times too large on one end of the pickup and two times too small on the other end of the pickup As described in the chapter on pickup design f the pickup were made lightweight and hollow and the ceramic support rings were removed from the assembly the reflections and their attendant problems would vanish 7 2 RF Front end Execution In Fig 7 2 1 there is a drawing of the RF circuits contained in the RF front end chassis The high resolution 10 4 GHz down mixing circuit described in Fig 7 1 1 1s drawn alongside a similar 1 3 GHz down mixing circuit that delivers a lower resolution measurement but does not require any mechanical phase shifters 104 Downmixers for EBPM Left Out Coarse A X RightOut EH 30 lt x8 JES Ph hifter 2o SLP Left Out ats X lt 30 Fine 10 4 GH PatchPanel 2 1 er SS cco gt X gt Fine Clock Gen MOClock 10GHz x Instead of 10 4 GHz from x8 Ne Optional SSE Hz INS Instead of 108 MHz from MO Ne PD eons Laser Am PD j Figure 7 2 1 Chicane BPM RF front end schematic The reference freque
125. e pickup to the EOMs n the front end For the same beam conditions as in the out of tunnel case this in tunnel front end measured 7 fs arrival time resolution and 2 um position resolution The resolution is calculated by multiplying the accuracy with which the laser pulse amplitudes can be detected by a measurement of the slope of the pickup signal This measurement of the slope is done by scanning the arrival times of the laser pulses over the pickup signal zero crossing and measuring how much the amplitude of the laser pulse changes The arrival time of the pickup signal coming out of the pickup output on the inside of the chicane should change in proportion to the energy deviation times R56 2 R1 6 and on the outside of the chicane it should change in proportion to Rs5 6 2 Rj6 Facing in the direction in which the beam travels the left output of the pickup is on the outside of chicane and the right output of the pickup is on the inside of chicane The ratio of the change of the arrival time of the left side to the change of the arrival time of the right side should be equal to the ratio of Rs6 2 Ry6 Rs6 2 Ris A measurement of this ratio constitutes a check of the calibration of the arrival time measurements done on both right and left outputs of the pickup Such a measurement is shown in Fig 10 4 1 and good agreement with the expected ration is observed despite the large jitter and drift of the beam
126. e third harmonic module third are depicted in block diagram format with arrows connecting various optical cross correlators OCC chicane beam position monitors EBPM bunch length monitors EO 1D THz 1D and beam arrival time monitors BAM to the digital processing boards uTCA or SIMCON DSP with which they would be connected in a beam based feedback system that controls on a bunch to bunch basis the amplitude and phase A of a normal conducting RF cavity 3GHz NRF cavity the 1 3 GHz super conducting acceleration cavities 1 3GHz SRF ACC1 2 amp 3 and the super conducting third harmonic linearization cavity 3 9GHz SRF The reason that the system shown in Fig 3 3 1 b will be built before that shown in Fig 3 3 1 a is that the uTCA crate system shown in blue in the figures below along with the corresponding ADC and FPGA boards will not be available in 2010 VME is the crate system that has been used at FLASH since its inception but will be phased out as uTCA crates become viable In the currently available VME crate infrastructure the beam arrival time is calculated on an in house built Analog Carrier Board labeled ACB in the figure that contains ADCs delay chips and FPGAs Beam arr val time information from this board s delivered to the cavity controller via an optical Gigalink The cavity controllers reside on VME based SIMCON DSP boards that each have ADCs DSPs DACs and an FPGA 28 Accelerator Fast data pro
127. ector signal phases 47 While it has been shown that the RF technique and the more expensive cross correlation method can stabilize a several hundred meter long link to within 10 femtoseconds out of loop the resolution of the RF technique has already been pushed to its theoretical limit 47 while the cross correlator has the potential to achieve sub femtosecond accuracy 50 Drifts of the optical cross correlator have not been fully understood yet and that is why the theoretical limit has not yet been reached A fatal error was made in two recent engineered prototypes of length stabilized optical links one version was built and designed in house at DESY and the other was built by a commercial vendor This fatal error was that the correction of the length of the fiber was made for only the outgoing pulse and not for the reflected pulse While the returning and outgoing pulses were both synchronized n the cross correlator the link end was not stabilized 135 NAN RF machine reference Master RF Laser lock Oscillator Optical Cross several hundred Correlator meters long fibe Electro Photo Analog Optical detector Digital Modulator Converter Faraday rotator Beam transient Figure 8 2 1 Beam arr val time measurement with length stabilized fiber The RF pulse from the pickup and the optical pulse from the MLO meet up in an EOM The amplitude of the optical pulse exiting the EOM will change when the beam arrival time ch
128. ed as stars The dependence of the beam position on the phase of the upstream accelerator section shows less than perfect agreement between the expected position and the measured position for deviations from on crest phase which are larger than 15 degrees For these off crest phases the beam is wider but this alone should not be detected in the beam position measurement When the beam is wider the position measurement is more sensitive to tilts of the beam In Fig 5 5 11 one sees the change in position measured as the phase of the accelerating RF is changed The predicted change of beam position due to energy change is shown as the solid line while the measured positions are shown as stars 78 Beam position 29 position cm 0 20 15 10 5 0 phase deg Figure 5 5 11 Change in beam position as a function of RF phase The beam position change predicted by the change n beam energy is shown as the solid line while the measured positions are shown as stars There is poor agreement with the predicted energy change for large phases wide beams This 1s most likely due to a wide and tilted beam Wide and tilted beams will be treated in the following chapter 79 6 Impacts of beam shape and orientation A standard BPM measures the beam position by compar ng the amplitudes of s gnals from two pickups If the transverse s ze of the beam s small compared to the distance from the beam to the pickup the pos
129. ed at the expense of impedance matching 1 e the ratio between the inner and outer conductor is not the same as it is n the transmission line This mismatch however causes a low pass filter like response which will be described in the following paragraphs The time constant of the high pass filter is given by t RiCpickup and can be measured with a network analyzer but first the time constant needs to be given as a function of frequency To calculate the time constant of a button pickup for a forward traveling sinusoidal wave V t z V cos w t Re Ae 5 1 29 V phase where A is a phasor with A V and arg A 0 5 1 30 phase 59 The derivative of the phase with respect to the frequency gives the time delay through a transmission line of length z Z See e 5 1 31 00 v phase The behavior of this phasor as a function of frequency can be measured with a network analyzer Most network analyzers measure a circuit with one or two ports At each port there s an incoming wave A generated by the network analyzer and an outgoing wave B Fig 5 1 10 Ai Bj Bo Figure 5 1 10 Definition of two port S parameters in terms of incoming wave A and outgoing wave B The network analyzer measures the ratio of the outgoing wave A to the incoming wave B by delivering complex S parameters according to B 8 4 5 4 5 1 32 B S A SA The diagonal elements S and S22 are refle
130. ed by both bunch compressors C 10 C gt 2 e Arrival time jitter induced by the first accelerator section amplitude jitter RF will not be compressed in the second chicane e The Rss of the second chicane is a factor of 5 smaller than that of the first chicane so the amplitude jitter of the second accelerator section will make a smaller contribution to the total arrival time jitter than the amplitude jitter of the first accelerator section These are the reasons that when a single arrival time monitor after the second chicane was used to feed back on the amplitude jitter of the first accelerator section 1t was able to stabilize the beam arrival time to within 30 fs 20 This is possible when the second accelerator section is operated on crest but not when it is operated off crest When the beam is off crest in the second accelerator section the arrival time jitter from the first accelerator section is compressed in the second chicane making the first and second accelerator section arrival time jitter contributions more equal after the second bunch compressor Given off crest operation in the second accelerator section and a larger Rs n the second bunch compressor the amplitude stability of the second accelerator section amplitude becomes almost as important to the timing stability of the beam as that of the first accelerator section This can partially be seen by using Eq 3 2 1 recursively for the second bunch compressor but because the rat
131. electron radius and gamma is the Lorentz energy factor 15 Since the bunch length is shorter after each bend the local energy spread generated in the last bend of the chicane will be the largest Aside from the increase in emittance bunch compressors can also cause something called a micro bunching instability when a density modulation created by the impedance of geometric wakefields longitudinal space charge or CSR get caught in a feedback loop in which these energy modulations are coupled into density modulations via the momentum compaction of the chicane For example the CSR creates an energy modulation of the bunch and the dispersion chops the beam up into slices with lengths that are comparable to the coherent synchrotron radiation wavelengths These micro bunches then interact with one another experiencing resonant oscillatory motion that can break the macro beam apart Although it was initially suspected that CSR would be the primary driver of the microbunching instability the longitudinal space charge effects in the injector are now the primary focus of concern The microbunching instability can be avoided by increasing the residual energy spread of the bunch with a laser heater that imposes a periodic energy modulation over the bunch and then smears it out longitudinally via dispersion 16 At FLASH a single chicane is used as the first bunch compressor and a symmetric double chicane is used as the second bunch compressor Fig 2 3 4
132. ements must be made relative to a reference signal In the equation for the beam position if the same reference is used for both left and right pickup outputs the influence of the phase stability of the reference cancels out beam _ position lt arrival _left reference arrival _ right reference where c is the speed of light and reference is the phase of the reference signal against which the arrival times of the pulses are measured If however the arrival time of the beam is measured from the same pickup signals that are used to perform the beam position measurement the phase of the reference does not cancel out beam _arrival arrival _left reference arrival _ right reference From this we can conclude that while a front end for a chicane beam position monitor and for a beam arrival time monitor must both be able to measure the arrival times of the beam transient pulses emerging from the pickup with femtosecond precision n order to meet the resolution requirements described in previous chapters the beam position measurement has much looser tolerances on the stability of the reference signal Both types of measurements must however measure the amplitudes of the pickup signals 96 around their zero crossings n order to avoid measuring changes n the charge or vertical position of the beam These problems can be approached with either RF or optical methods The RF front end and the optical front end f
133. ence of an RF limiter the EOM crystals become opaque within a matter of minutes This was directly observed during a machine studies day when the beam orbit was dramatically off center The pickup in this instance was a ring type pickup and had therefore a large surface area open to beam spray Using pickups with a smaller surface area exposed to the beam could reduce the possibility for beam spray damage of the EOM and therefore remove the requirement of 130 using the RF limiter This might be desirable because an RF limiter produces an appreciable yet difficult to measure amount of AM PM conversion This means that the limiter can convert changes in amplitude into changes in phase Since the phase is the quantity that we want to measure if the AM PM conversion is large enough it can limit the accuracy of the measurement Reducing the amplitude of the signal entering the limiter can also reduce this effect but unless the bandwidth of the signal is preserved when the amplitude is reduced this will reduce the resolution of the measurement Measuring this AM PM effect requires high amplitude 60 Volts RF signals with a frequency of several GHz 7 5 Front end Costs The RF front end and the optical front end can deliver comparable 5 um resolution in their present configurations but the optical measurement has the potential to reach lt 1 um resolution while the RF front end can only achieve 3 um resolution under best case circumstances
134. eory for the Design of Beam Transport and Charged Particle Spectrometers SLAC Report 75 1982 J Aurthur et al Linac Coherent Light Source LCLS Conceptual Design Report Tech Report SLAC R 593 SLAC 2002 173 14 15 16 17 18 19 20 21 22 23 24 25 26 27 T O Raubenheimer et al Chicane and Wiggler Based Bunch Compressors for Future Linear Colliders SLAC PUB 6119 May 1993 Ya S Derbenev et al Microbunch Radiative Tail Head Interaction DESY Sep 1995 E L Saldin E A Schneidmiller and M V Yurkov Nucl Instrum and Methods A 528 2005 355 P Emma and R Brinkmann Emittance dilution through energy spread generation in bending systems Proceedings of the PAC 1997 P Schmueser M Dohlus and J Rossbach ntroduction to Ultraviolet and X Ray Free electron Lasers Springer to be published T Schilcher Vector Sum Control of Pulsed Accelerating Fields in Lorentz Force Detuned Superconducting Cavities Ph D Thesis Department of Physics of the University of Hamburg 1998 F Loehl Optical Synchronization of a Free Electron Laser with Femtosecond Precision Ph D Thesis Department of Physics of the University of Hamburg 2008 F Ludwig C Gerth K Hacker M Hoffmann W Jalmunzna G Moeller P Morozov C Schmidt Drift Calbration Techniques for Future FELs Proceedings of the IPAC 2010 Kyoto J
135. equirements As argued in Sect 3 2 while we know that the 200 fs injector timing jitter is compressed in the chicanes it is not compressed enough to make it negligible especially if the compression factor of the first chicane is reduced from 10 to 5 as is planned for upcoming operation with the third harmonic module The injector jitter must therefore be measured and incorporated in any energy arrival time feedback scheme While the arrival time is stabilized using the accelerating amplitude as an actuator in a feedback the bunch length is stabilized using the accelerating phase as an actuator in a feedback The bunch length stability is less of an issue of the total bunch length but more of an issue of the lasing bunch length namely the tiny slice of the bunch that lases in the undulator section The length of this is 15 um and no existing monitor can truly resolve it but one n particular can in practice stabilize it An array of pyrodetectors located after a diffraction screen can be arranged with various filters in order to produce single shot spectra of all of the bunches in the bunch train 32 Previous spectrometers required multiple shots as a delay stage moved over several centimeters The single shot measurement detects fewer frequencies than the scanning measurement and cannot reconstruct the longitudinal profile as well as the scanning measurement but this is offset by the advantage of having an unchanging beam emittance pos
136. er of slices into which the beam is divided and the amount of charge contained within each slice In Fig 6 3 3 the pulse amplitudes that slices of a tilted beam with a flat charge distribution would produce on the pickup are shown The amplitudes from Fig 6 3 3 can be added together for pulses traveling to the left and to the right in order to produce Fig 6 3 4 This gives an estimate of the shape of the signals produced by a flat beam that is tilted in the x y plane compared to the same beam that is not tilted The beam length was 1 mm the beam width was 10 mm and the tilt was 5 degrees The difference in the arrival time of the tilted signal s zero crossing at the exit of the pickup compared to the non tilted case is 1 ps If the charge distribution is not flat but Gaussian with a length of 1 mm a width of 4 mm FWHM and a tilt of 5 degrees the difference between the tilted and non tilted cases is 0 3 ps 200 um These differences due to x y tilt constitute errors in the measurement of the beam centroid 87 Vertical tilt sensitivity 5 degrees Signal amplitude V x position mm Figure 6 3 3 The amplitude of the signal induced on the pickup by a slice of a beam with a flat charge distribution that is tilted in the x y plane by 5 degrees as a function of the x position of the slice within the beam 20 Beam length 1 mm a Beam width 10 mm 15 N 2 5 Beam tilt 5 degrees ff f O Error due to tilt 1 ps m ff f
137. er the beam position changes by more than a millimeter it 1s necessary to use it to keep the system sampling the zero crossing of the signal In the long term measurements which will be presented in Chapter 10 the mechanical phase shifter feedback was active 116 3000 ae ACC1 Setpoint ae lt BC2 BPM 2500 u 2000 1500 1000 EBPM and PMT positions um 500 0 A 01 02 03 04 05 06 07 08 09 1 change ACC1 Figure 7 2 13 Curvature of the BC2 BPM measurement results from the problems with the mechanical phase shifter BC2 PMT is another chicane beam position monitoring system which will be introduced in Chapter 9 Due to the unreliability of the mechanical phase shifter position measurement whenever the mechanical phase shifter moves the accuracy of the measurement is reduced This 1s not a concern for the alternative to this scheme the optical front end because a high resolution linear encoder was mounted to the mechanical delay stage of the optical measurement If a high resolution linear encoder were mounted to the mechanical phase shifter of the RF front end the concerns about the accuracy of the measurement could be removed Alternatively a 12 GHz vector modulator that is in development through a collaboration between DESY and PSI might be available in the future An alternative that eliminates the need for any phase shifter at all is to down mix not to base band but to some intermediate fre
138. er to take into account the dependence of the kick on the phase of the RF The main effect is a time varying dipole kick for a situation in which the forward power and reflected power have reached a steady state reflected 0 The predictions made in 35 were incorporated into a beam transport simulation in order to check if it 1s possible to measure the kicks with existing diagnostics In this simulation the cumulative effect from each coupler of 20 30 urad tilt was so weak that it could not be measured with existing diagnostics The separation between the head and tail due to the coupler kicks alone is only 17 um after the first accelerator section F Er I DESY i A y 32mm y y 22mm F th F 1 6 iin 2 LO 7 7 2 j54 41 j9 7 124559 1 9 j54 7 5 j50 75073228 38 j6 6 x mm x 372mm 4 14j53 21 431 9 Figure 4 2 3 Coupler geometry with pickups top and voltage of kick bottom V and V are proportional to the accelerating voltage Vj The voltage is given in complex notation in order to take into account the dependence of the kick on the phase of the RF The main effect is a time varying dipole kick for a situation in which the forward power and reflected power have reached a steady state reflected 0 From M Dohlus Field Asymmetries and Kicks 35 A beam that travels diagonally through an accelerator section will acquire a tilt due to the difference in the orientation of the field axis re
139. er which is forward pumped This amplifies both the pulses coming from the link and the pulses which are reflected by the Faraday rotator and sent back to the fiber link stabilization unit The 60 mW traveling in the direction of the Faraday rotator is split such that 80 goes towards the BAM measurement and 20 goes to the chassis containing the chicane BPM front end The fiber links must not only deliver a pulse that is stable in arrival time but the pulses must also have an appropriate length when they reach the EOMs and when they return to the optical cross correlator To this end a length of LLWBDK dispersion compensating fiber from OFS is incorporated into the fiber link stabilization chassis The length of dispersion compensating fiber in the present case must make the pulses short when they reach the EOMs n both the BAM chassis and chicane BPM chassis Because the connection between the BAM chassis and the EOM chassis requires about 8 meters of fiber an additional 8 meters of fiber is wound up inside the BAM chassis after the splitter but before the Faraday rotator This makes the distance between the EOMs and the laser the same for both the BAM EOMs and the chicane BPM EOMs This additional 8 meters of fiber will not have an impact on the length stability of the BAM fiber link because all fiber from the MLO up to the Faraday rotator 1s length stabilized If however the 8 meters of fiber connecting the BAM chassis to the chicane BPM chas
140. ered signals This chassis has been used to provide the RF phase measurement for the MLO RF lock and in parallel to measure the out of loop performance of this lock 138 RF Lock Box for FLASH MLO Synchronization MO INJ3 lQ N 1 3 GHz 108 MHz 11dBm 216MHz 25dBm 1 3GHz PD eT P 20 40dB 1 3GHz Lock1 SO ZFM2000 SLP1 9 LNA MO MLO2 N LEMO Power atchPanel 2 1 216MHz AD8302 TIER Looki LEMO 216MHz Lock2 Q 1 3GHz A fais ZFM2000 SLP1 9 wa LNA ug MLO2 diag MO Power2 216MHz E rooe i LEMO216MHz SLP250 Lock2 Q 28dBm Figure 8 3 1 Schematic of MLO MO laser RF lock Inputs include 1 3 GHz and 108 MHz from the MO and photodetector signals from two different MLOs Components include vector modulators VM1 VM2 for shifting MO phases in calibration routines bandpass filters and amplifiers for removing and amplifying the 1 3 GHz frequency component from the photodetector outputs mixers for generating a baseband signal from the MO and MLO signals phase and amplitude detectors AD8302 for a coarse determination of the MO and MLO phases with a 216 GHz measurement and couplers for online troubleshooting of the circuit The strategy of the circuit shown in Fig 8 3 1 is to filter out one frequency from the frequency comb produced by impinging the pulsed MLO laser signal on a photodetector and then measuring the phase of the filtered signal relative to the MO
141. file monitor Because the photomultiplier signals are short in duration and are sampled with an 81 MHz ADC they can be used to produce single bunch resolution across the whole bunch train while the profile monitor cannot The Hamamatsu R5900U 00 M4 photomultipliers used in the monitor come from a HERA B experiment and they consist of 4 photomultiplier tubes packed together in one square package with dimensions of 16x16 mm Each individual tube is 8x8 mm and there is 10 20 cross talk between the tubes meaning that photons that are measured on one side will also be partially registered by the other side The signal from each tube is amplified and filtered with a Gaussian filter in order to give the signal a more rounded peak and increase its width This reduces the impact of the clock jitter from the ADC used to sample the signal 53 The synchrotron light is centered on a pair of the four tubes such that each gets half of the light If the beam moves to one side or the other the signal from one side will increase and the signal from the other will decrease If the beam 147 is moved away from the center of the two detectors a motorized stage will move the detectors to where the beam is Fig 9 2 1 mover Photomultiplier detectors Figure 9 2 1 Two Photomultipliers used to measure the beam position in the chicane In this application if the measurement resolution is limited by the number of photons that each tube detects it is s
142. g 7 1 3 or 7 1 4 is sampled at a given time for an array of LO phases the amplitude measured at this sample point will change in a sinusoidal pattern as is shown in Fig 7 1 5 Even though there is an asymmetrical pulse emerging from the output of the mixer the dependency of the output pulse amplitude on the phase of the LO is still sinusoidal regardless of the sample point This is not however what was measured Fig 7 1 6 There is a bump in the measured mixer output signal that changes its position within the signal when the position of the beam changes This is not a desired effect because the measurement of the beam s position is given by the difference in the phases of the two signals shown in the plot red and blue The difference in the phases is measured by adjusting the various delay lines until the ADC sampling time falls about the zero crossing on the falling or rising slopes of both signals 101 amplitude of mixer output as LO phase is changed 0 8 0 6 0 4 0 2 oF a u 0 2 0 4 0 6 0 1 00 200 300 400 500 600 700 800 degrees Figure 7 1 5 Three different sample points of the mixer output when the phase of the LO is changed in simulation The RF signal is composed of two signals with different frequencies but the same phases 10 4 GHz downmixed signal Sampling location 7 Vector modulator LO phase _ gt Figure 7 1 6 Distorted sinusoid
143. ghted in blue Measurement System Position resolution Energy resolution In loop Vector Sum 25 um to 70 um 17e 5 to 2e 4 le 2 10 cm drifts Out of loop Vector Sum 25umto 70um _ 7e 5 to 2e 4 10 cm drift free Photomultiplier Tube Monitorj15 um to 30 um Ae 5 to 9e 5 BC2 BPM 10 4 GHz front 6 um to 10 um 2e 5 2mm end time of flight with 2 BAMs 9 fs 1 5e 5 anticipated Table 10 0 1 Comparison of beam energy measurements in the first bunch compressor 152 The highest energy resolution AE E 6e 6 is provided by the optical front end of the chicane BPM but it also has the smallest dynamic range In the following sections the measurements from each of the monitors listed above will be compared with respect to their reliability and agreement 10 1 RF BPM Measurements Calibration scans of the 1 3 GHz coarse and 10 4 GHz fine front ends were presented in Chapter 7 and long term measurements are presented below in Figs 10 1 1 and 10 1 2 They took place over the course of several hours during relatively quiet SASE operation shifts In Fig 10 1 1 the setpoint of the first accelerating module is shown in black while the coarse measurement is shown in green and the fine measurement is shown in blue The y units are in percent energy change and the x units are in hours The measurements of the BPM are converted into percent energy change by multiplying the beam position times the R s of the chicane in accordance with Eq 1 10
144. gnals from the pickups n the first bunch compressor sent over 30 meter long RF cables The patch panel signals were connected to the chassis with meter long cables 109 The resolution of the 10 4 GHz front end was evaluated by splitting a signal from the beam pickup and measuring the difference between the arrival times of the pulses If the front end measurement had perfect resolution the difference measured would be a constant value regardless of horizontal beam position changes The difference between the split signals is plotted below in Fig 7 2 5 difference of split signals arrival times resolution 60 I I I I I std 3 3296 58 Correlated with a it i vertical 54 I Fill eee al pii 7 Sy agii If WN position I Hh II il I WI change LE STA PAA 2 5 ll EET AE EEL WAST FI 5 II Um yy INN E I WW Pe
145. gy changes If the electron beam after a chicane is used to generate a photon beam as in the case of a Free electron Laser FEL the arrival time stability of the photon beam will be directly affected by the stability of the beam energy a quantity which can be measured through the position or arrival time of the beam in or after a dispersive section To detect the position of an electron beam one could look at the optical transition radiation that is produced as the particles travel through a metal film but the beam is significantly disturbed by the film and it cannot be used parasitically Instead one could look at the synchrotron light that is produced by the beam as its trajectory is changed by the magnetic field The resolution of the synchrotron light based beam position measurement would then depend on how much light there is and how well it can be detected If instead the position of the beam is detected with a metal antenna in which a current is induced as the beam passes close to it the resolution would depend on how well the beam couples to the antenna and how well the properties of the current pulses generated n the antenna can be measured In this thesis the design and characterization of a pickup antenna and pulse measurement system was completed n order to measure the position of an electron beam over a range of 10 cm and with a resolution below 2 um Expanding the range even to the 40 cm required by the XFEL chicanes should
146. h stabilized fiber Balanced optical cross correlator used to measure the difference between the arrival times of pulses coming from and returning to the MLO Schematic of MLO MO laser RF lock One frequency is filtered out of the frequency comb of pulsed laser signal on photodetector Setup for measurement of the mixer s Ko and characterization of the spectral noise density and drift contributed by each RF component Spectral noise density of signal at the exit of the LNA shown in Fig 8 3 1 RF phase measurement drift with temperature control with and without disturbances people in room 8 3 6 8 3 7 9 1 1 9 1 2 9 2 1 10 1 1 10 2 1 10 2 2 10 2 3 10 3 1 10 4 1 10 4 2 10 4 3 10 4 4 10 4 5 RF phase measurement drift w thout temperature control without disturbances people in room Out of loop measurement drift w thout temperature control and without disturbances people in room A synchrotron light monitor system w th CCD screen A picture of the beam as imaged with the synchrotron light camera Two Photomultipliers used to measure the beam position in the chicane Measurements of energy stability in the chicane taken by the coarse and fine HF front ends of the chicane BPM plotted with energy setpoint values from the upstream accelerating module Correlation between the measurements of the beam position in the chicane taken by the chicane BPM labeled EBPM and the photomultiplier tube monitor PMT
147. h used a particle distribution from the gun which was generated with ASTRA together with transport matrices for the first accelerator section and for the bunch compressor The evolution of the transverse distribution is poorly described by this simulation because the lattice 1s designed for a space charge limited beam and space charge effects were not taken into account after the gun This results in a simulated transverse beam size which is much larger than it would be in reality The plots however serve to demonstrate the gymnastics of longitudinal bunch compression In the first pair of plots in Fig 2 3 2 the energy chirp is shown on the right and the particle distribution in the horizontal and longitudinal planes is shown on the left The beam is spread out over 7 mm of longitudinal space and it has a longitudinal energy chirp which is curved due to the curvature of the accelerating RF In the second pair of plots the higher energy particles have begun to overtake the lower energy particles and the bunch has become shorter longitudinally and wider in the horizontal plane of the bunch compressor The slice emittance increase in the tail of the bunch is due to a mismatched lattice f a quadrupole strength is changed slightly the emittance increase appears in the head of the bunch and not the tail This is partially an artifact of the particular simulation used Space charge forces were only used up to the exit of the RF gun but they too signific
148. hange To do this let us first write down the energy of an electron subject to an accelerating module with an acceleration voltage of U and a phase of o k Ast o with k 27 2 4 in terms of the wavelength of the accelerating RF and o equal to the phase for which the longitudinal position is equal to that of the reference trajectory As 0 E E eUcosp 2 3 5 where is the initial energy of the particle and Ey is the energy after the accelerating module We can describe the energy chirp produced by the accelerating RF by doing a Taylor expansion of 6 about a small longitudinal position change As 5 s 5 As s As As 6 As 5 As 2 3 6 Res initial Res AS Ress As 7 The first term describes the initial energy spread over the position change the second term describes the linear chirp acquired over the position spread and the third term describes the quadratic chirp acquired The indices of the R coefficients describe the coordinates of the values in the beam transport matrix The first index coordinate equal to six corresponds to energy deviations and the second index coordinate equal to five 12 corresponds to terms that are linear within As The third index coordinate equal to 5 is for terms that are quadratic within As The coefficients are given by eU sin E Lel E ky _ E E and 2 3 7 E eU icc COSp 2 if Ey Res Ress For a quick calculation of the change in beam properties
149. harge Distribution Tilted When the charge distribution is asymmetric and the beam is tilted in the x z plane a systematic error is generated that cannot be removed by any available methods For an asymmetric Gaussian charge distribution shown in Fig 6 6 1 the width is 4 5 mm the length is 4 mm and the centroid is offset horizontally from the center by 3 3 ps 92 x 10 Horizontal Charge Distribution centroid 3 302ps 3 5 3 7 eS fo y 2 5 a gt 4 5 mm 4 rf c 1 5 a O f N f ib 7 y 0 5 Fi p ze Z N 0 8 6 4 2 0 2 4 6 8 X position mm Figure 6 6 1 Asymmetric horizontal charge distribution with centroid offset from center by 3 3 picoseconds When a bunch with an asymmetric Gaussian charge distribution like the one shown in Fig 6 6 1 1s tilted by 45 degrees like the beam shown in Fig 6 4 2 a 580 fs systematic error compared to a non tilted beam results Fig 6 6 2 This 1s called an error because if one is trying to measure the beam centroid one does not want to measure the tilt as well Beam length 4 mm 20 p f ven Beam width 4 5 mm se fi Beam tilt 45 degrees jf ff Centroid 3 3 ps 10 L jj ff Error due to tilt 580 fs j i A 5 S f y i E A A 2 Fi Ah Das S oF NN i g 2 x f ff Of i f 40 Tilted beam signal right Tilted beam signal left j j iy 15 Un tilted beam
150. hat the beam experiences when it passes the RF couplers ports from which the accelerating cavities are filled up with RF waves or from which higher order modes are removed The kick arises because the coupler introduces a field asymmetry into the structure There are two HOM couplers per module one power coupler and eight cavities per module The HOM couplers are oriented in such a way that the kicks of pairs of upstream and downstream couplers should partially cancel one another out making the power coupler the coupler of concern Fig 4 2 2 With hole Without hole PA cavity field axis exaggerated HOM coupler pick up HOM coupler Figure 4 2 2 Coupler kick concept A hole in a cavity changes the axis of the cavity field There are two sets of couplers per accelerating module They are oriented in such a way that the kicks of pairs of HOM couplers should partially cancel one another The power coupler produces the strongest kick and it scales with the gradient in the cavity Estimates for the strength of the coupler kicks n a steady state situation have been made using software that calculates the fields in cavities of arbitrary shape 35 The coupler 39 geometry and results from this simulation are shown in Fig 4 2 3 One can use the voltages V and V along with the accelerating voltage V in order to predict the strength of the kicks for different gradients The voltage is given in complex notation in ord
151. he final energy is close to the nominal energy E1 Enom one can multiply the result by a small change in accelerator voltage AV to write AV Rss AN 12 OV CO V Arrival time jitter due to phase changes Using Eq 5 together with t z c the sensitivity of the arrival time to the accelerator phase is Ot Deren 13 OP Cy 09 amp Using Eq 1 to solve for s n k z 7 we can write 167 ER AE ar Bz Eg OP C ky 14 Using the approximation that lt lt E and multiplying the sensitivity by a small phase change gives EN ay T E z E z A 15 Pp i z 1 z p op Cok Now solving for T E in terms of the compression factor Eq 8 gives Ot u L N l S 40 16 a0 Cok C Arrival time jitter due to arrival time changes prior to the accelerator section z Using Eq 5 and Eq 8 we can immediately write down the sensitivity of the arrival time after the chicane to a small change of the incoming arrival time Az 17 This shows that the incoming arr val time jitter is compressed by the compression factor of the chicane C Sum of all contributions Since we assume that the incoming arrival time changes voltage changes and phase changes are statistically independent they can be added in quadrature to find the net arrival time jitter after the chicane Eq 3 x x 2 2 2 g 2 5e Z Re oy BB gt Og Jk 5 5 i 1 8 a Co V C p C
152. hey have an intrinsically high failure rate when the beam conditions are unstable If any beam parameter changes significantly the monitor can be pushed out of range and require mechanical movement and recalibration all of which can take several seconds to complete Such a beam based monitor can only be used to add small corrections to the more robust cavity field measurement based regulation 36 4 Beam Shape in the Bunch Compressor Because the beam position monitors of interest for this thesis are located in the middle sections of the chicanes where the dispersion is at a maximum it is important to understand the likeliest shapes of the electron bunches at these locations for various machine configurations The impacts of various beam shapes on the chicane BPM performance will be described in chapter 6 The transport of the beam has been simulated with a code called ASTRA 34 and with the use of transfer matrices 12 ASTRA was developed for the space charge dominated beams that one finds in the injector and will in the following simulations be used to generate a charge distribution at the exit of the RF gun This charge distribution will then be transported to the middle of the chicane by multiplying together transfer matrices with Matlab Simulation results will be compared to measurements of the beam done with the beam image from the synchrotron light monitor screen 4 1 Perfect Alignment Ideally the beam is centered on the cathode
153. high enough for the 1 3 GHz processing electronics When a front end with a higher resolution was used the shape of the signal produced by the pickup was more clearly resolved and it was noticed that a bump in the signal at the sampling location desired for the arrival time measurement moved from side to side when the beam position changed in the horizontal plane Since the monitor was supposed to measure beam arrival time and not position this was undesirable and a new pickup was constructed with button like pickups combined with external cables Fig 5 1 12 b Figure 5 1 12 Cross sections of the old a and new b beam arr val time pickups The beam position dependent bump observed in the time domain signal from the old ring pickup was the result of a notch in the frequency spectrum at 5 GHz with a corresponding wavelength of 6cm This length scale was close to the length of half of the circumference of the ring The output of the ring pickup is shown below in Fig 5 1 13 as given by CST simulation and measurements with an oscilloscope done with and without an RF limiter The use of the combiner to reduce the beam position dependence is shown below in Fig 5 1 14 62 arrival time monitor with and without limiter 15 Ar no limiter limiter 10 NI simulation f Dr Li o wo yA Io N Me N Y a a AG PEN Bu ff y N i 1 ATS L 0 R i A Q
154. hot noise limited If a measurement is shot noise limited no matter how much a signal is amplified the signal to noise ratio will not improve To estimate the resolution of this measurement one must first know how many photons are intercepted by the monitor Synchrotron radiation is emitted over a wide range of frequencies Different frequencies are emitted with different angular distributions but most of the power is present in an opening angle of 1 y To accurately calculate the radiation produced in the frequency range to which the photo detector is sensitive and within the angular spread determined by the aperture of the optics leading up to the photodetector one must integrate over the number of photons radiated per unit frequency per unit solid angle The unit solid angle can be written in terms of its components the emission angle and the bending angle One can use the following formula to find the number of photons N emitted over an emission angle 6 and a bending angle w 36 dN _ d dy Ao C QE z gE OA T peis C yO ke 9 2 1 1 r 0 K3 3 u Where Co 1 3273e16 photons sec mrad GeV A K is a Bessel function Q is the beam charge E is the beam energy y is the Lorentz factor is the angular frequency of the radiation and the critical frequency 1s EP _ C 411 6nm 9 2 2 p S Il where C 3 37e18 m sec GeV and p lep siny is the bending radius Above the critical freq
155. how the measurement differs from the actual center of mass beam position and arrival time The position and arrival time of the beam are measured according to Al and ore t_ 6 1 13 N meas with U t 0 at the zero crossings of the signals The real center of mass position and arr val time of the beam are given by 82 Kaum 2 TAN xdtdx thea A 126 1 tdtdx 6 1 14 Using either this Green s function method or the numerical simulations from CST one can predict that the measured beam arrival time will be different from the real center of mass arrival time when the beam width changes Using the Green s function method alone one can predict that the measured beam position changes compared to the center of mass beam position when a wide asymmetrical charge distribution 1s tilted in the x y or x z planes This approach has the advantage that Green s functions for 2 D and 3 D transient cases can be found through multiplication of 1 D cases The solution takes the form of the superposition sum of several integrals The magnitude of these tilt and width effects will be detailed below 6 2 Beam Width Changes When the phase of the upstream accelerator section is changed the width of the beam will change The charge density changes when the beam width changes Since the coupling of the beam to the pickup occurs at locations of changing charge density different charge densities will produce signals with different
156. ile the RF field provides some longitudinal focusing In order to minimize the detrimental effects of wakefields and dispersion the beam must travel directly through the middle of the beam pipe where the focusing is the most effective and where the distorting effects are the weakest Another aspect of producing a low emittance electron beam in the photoinjector is the quality of the laser pulse If it is unstable then the electron beam will be unstable If it is badly longitudinally shaped the emittance produced by the photoinjector will be too large for lasing to occur at the end of the machine The laser pulses originate in an actively mode locked Nd YLF laser operating at a wavelength of 1047 nm 10 These pulses are amplified and then frequency quadrupled in order to produce a beam charge of around nC Changes in the amplification of the laser result in changes in the charge produced at the cathode In order to reduce the effect of the laser s pointing jitter and to make the intensity of the laser spot more uniform the laser pulse s sent through an iris before it 1s impinged upon the cathode The synchronization of the laser pulses and the RF in the cavity to the reference of the machine are critical to keeping the electron beam properties stable and maintaining synchronization to other devices including the downstream accelerator section 2 2 Accelerator section The accelerator sections consist of superconducting niobium cavities which
157. ilt so far but when pushing toward sub femtosecond resolution such techniques may be employed After the first delay line the fiber is split into two arms so that half of the light goes to one end of an ODL and the other half goes to the other end of the ODL When the ODL stage moves the path length of one arm gets longer while the other gets shorter This ODL stage would need to move whenever the position of the beam is changed by more than a millimeter or so At the exits of the two arms on this ODL the fibers are split again Two of the four fibers will be used to sample electrical signals coming from the right side of the pickup and the two of the fibers will be used to sample the signals from the left side Of the sets of two one fiber will be used to perform the fine measurement of the arrival time of the pickup signal and the other fiber will be split with 90 going towards the coarse measure of the pickup signal arrival time and 10 being used to generate the clock for the ADC which will be used to sample the amplitudes of the signals emerging from the EOMs 127 Because fibers and EOMs are sensitive to temperature changes the temperature of the plate upon which most of the fibers rest is actively controlled with Peltier elements Peltier elements were chosen instead of a heating mat because they can cool as well as heat and they react more quickly to changes n the control voltage The 4 Peltier elements are mounted underneath the alum
158. image is in the x y plane Plot by C Gerth The images seen on the monitor are projections of the beam streaked out in a longitudinal direction The x axis of the monitor corresponds to a combination of the z axis and x axis of the beam The y axis of the monitor corresponds to the y axis of the beam Each beam projection 1s composed of beam slices that are tilted in the x y plane and streaked and tilted in the x z plane Fig 4 3 2 This means that if one makes a fit to the entire beam image there will be an error associated with the x y tilt of the beam when the quantity that is sought is the x z tilt of the beam This error can be avoided by fitting only to the central portion of the beam and ignoring the head and tail sections This is shown through the difference between the red line which 1s fitted to the entire beam and the blue line which is a fit to the central portion of the beam One can see that the central portion is not affected by the x y tilt of the individual slices while the end portions are In practice this is done by cropping the image of the beam so that the head and tail of the beam are cut off The colors of the pixels are assigned weights according to the intensity of the detected synchrotron light A line is then fitted to the weighted distribution 45 Figure 4 3 2 Images on the screen contain projections of the beam streaked out in a longitudinal direction The red line 1s a fit to the centroids of the slices of th
159. ing ringing will likely be caused by the cavity created by the channels opening into the beam pipe and the sensitivity of type N size pickups Fig 5 1 15 a to resonances with wavelengths on length scales corresponding to the diameter of the coaxial channel and the distance between the feedthrough ceramic and the beam pipe An SMA feedthrough option was simulated Fig 5 1 15 b and t does not suffer from all of these resonances but it would take longer to manufacture and have a higher risk of breakage at the ceramic in the feedthrough a b Figure 5 1 15 BAM pickup designs Design a provides an expedient production and installation process and is the design that has been installed at FLASH for BAM applications Design b would be very complicated to realize but t would have better performance than design a It was decided that the type N design Fig 5 1 15a would be produced for FLASH because of expediency of manufacture installation and low risk of breakage At ELETTRA a similar design was installed but t used an SMA sized pin and channel instead of a type N sized pin and channel Out of all of the designs the SMA sized pin delivers the best ratio of the amplitude of the signal to the signal slope This 1s ideal for avoiding problems related to AM amplitude modulation to PM phase modulation conversion problems but because the amplitude of the signal it generates is so much smaller than that of any other picku
160. input laser pulses are then sent through a BBO crystal When the pulses overlap in the crystal new pulses with the sum frequency are generated and emitted n both forward and backward directions With the aid of dichroic mirrors and a group delay generated in a dispersive medium the pulses generated in the crystal each travel to a photomultiplier tube PMT The incoming laser pulses return from whence they came In this balanced detection arrangement the measurement of the relative arrival times of the two input pulses is insensitive to laser noise and is background and drift free 3 5 Third harmonic Module Jitter The third harmonic module is comprised of 4 cell cavities which are filled with 3 9 GHz It is used to linearize the energy chirp of the beam In the first bunch compressor the energy of an electron at position z in the bunch 1s E E V cos k z V cos 3k z 3 5 1 where V V3 3 are the RF amplitudes and phases of the first accelerator section and the third harmonic module As in Eq 1 7 the path length through the chicane as a function of the energy of a given particle is written by 32 2 E E E E L E Ly Rs e e z2 3 5 2 0 0 where Ey 1s the energy of the bunch center z 0 The first and second order chirps of the beam energy as a function of the compression factor C are 30 E R a pA d prea Rew ER ass Rz C Rz Ey Taken all together we have three equations with four free para
161. inum plate shown n the side view of the chassis Fig 7 2 4 They are incorporated into a fast control loop with a Wavelength PTC temperature controller that uses a single temperature sensor mounted on the top of the metal plate in order to deliver control voltages to the Peltiers If the Peltier is cooling the plate heat will travel down a conducting channel towards the outside of the box If it is heating the plate the opposite will occur The outside of the box is insulated from the inside of the box with a centimeter thick layer of neoprene While the conductivity of neoprene is higher than that of air it is a better insulator because it prevents convection from transferring as much heat from the outer box to the inner A fan is used to cool the outer box Figure 7 2 3 The layout of the fibers in the top layer of the optical front end chassis for the chicane BPM The stage on the left will move when the beam position changes The stage on the right will move when the arrival time of the beam changes The four EOMs are in the middle of the drawing on an actively temperature stabilized plate 128 Peltier element on metal foot Pump diode is mounted to metal block and Thermal contact to external box Thermally connected but electrically isolated Insulation Figure 7 2 4 The side view of the optical front end chassis The two layer design was made in order to facilitate the control of the temperature of the plate on which the fibers were
162. io of the incoming energy chirp to the outgoing energy chirp and the ratio of the incoming energy to the outgoing energy are not small as assumed in the derivation of the equation the prediction of the equation will be wrong by up to 30 depending on the machine configuration In general because a beam that arrives earlier or later on the falling slope of the RF wave will gain a lesser or greater amount of energy whenever the accelerator section upstream of a bunch compressor is operated off crest it is not advisable to use the energy measurement from in the chicane or the arrival time measurement from after the bunch compressor to directly feed back on the upstream accelerator section without first taking into account the effect of incoming arrival time jitter on the quantity that 1s measured Although for large compression factors the incoming arrival time jitter may be compressed enough in the chicane that it can be ignored this is not always the case It 1s 21 not possible to disentangle which energy change was caused by incoming arr val time jitter and which energy change was caused by accelerating gradient and phase jitter unless the arrival time jitter generated upstream of the accelerator section has been measured There are two different strategies to deal with this problem One could use an accelerator section upstream of each chicane in order to stabilize the arrival time after each chicane regardless of how much the beam energy
163. is fully compressed near the end of the machine A bunch compressor at the end of the machine can also compress some of the timing jitter generated up stream and can possibly provide a compensation effect for some emittance increases generated in the first compressor 17 A horizontal emittance increase will occur if a significant energy spread is generated within the chicane This emittance increase occurs because the energy spread breaks the linear achromaticity of the chicane In a linear achromat the particles will have the same transverse position after the achromat that they had before they entered the achromat When the achromaticity is broken the off energy particles no longer follow the same orbit as the on energy particles and they will emerge at a different transverse position than the on energy particles This is depicted in Fig 2 3 3 AE generated in dispersive section x Z Ax D AE Nig X E Chromatic orbit nai Ax D AE Achromatic orbi E Figure 2 3 3 Energy spread generated within chicane breaks linear achromaticity and results in an increased emittance after the chicane Within the chicane dipoles incoherent and coherent synchrotron radiation ISR and CSR generate an energy spread Collective effects and space charge effects can also generate an energy spread in the chicane Because these energy spreads are generated in a dispersive section they will result in emittance growth Emittance growth will not of cour
164. is then calculated from a gain setting and the difference of the measured signals from the setpoint This correction is multiplied by calibration factors appropriate for each cavity and sent to a DAC which generates a 250 kHz signal The 250 kHz signal goes to a vector modulator labeled VM in the diagram which shifts the phase and amplitude of the 1 3 GHz that is sent to the klystron Module fe nn k Cavity field probes SSeS controller Detectors n n TunnunuuuundensunmunnunnununnnEnEnEEEnBEnEEEEENENEENEENEEENERENEEENEEREENENENEEENEERI ESS creer rec er eres rer rr res rrrrecir HEHEHHNNENHHNENNNENEENNENENEENNENENENENENEEENENENENENENEEENENENENENENEEENENENENENENEEENENENEEENENEEENENENRENERENEEERNENENE NnENEEENENENENENEN calib 1 Q ADC in detection Setpoint gain feed fwd Figure 3 1 1 System for controlling the cavity fields of the accelerating module The phase and amplitude of the fields are detected from cavity pickups the difference from the setpoint is calculated and a correction is sent to the klystron 11 The boxes labled I Q detection calibration in vector sum setpoint gain feedforward calibration out comprise the routines that take place on an FPGA This routine is depicted in more detail below in Fig 3 1 2 In Fig 3 1 2 the phases and amplitudes detected from each cavity are added together in a vector sum and then subtracted from the setpoints generated by the feed forward table The feed forward
165. it of the LNA shown in Fig 8 3 1 With temperature control of the RF circuit shown in Fig 8 3 1 and amplitude control of the laser it is conceivable that the RF lock could be stable over the long term to below 10 fs pkpk without disturbances people in the room and 30 fs pkpk with disturbances This would be the case if the lock were limited by the RF phase measurement alone Fig 8 3 5 141 MLO RF Lock Drift 6 g nt 21 ll tl nA ERITI pfa nr m aid oa NM wi nam I N A 0 11 12 13 14 15 16 Time hours S 0 01 2 0 005 Bot A LUT Nie pa a A gt TE N M i lj An IN ily ANN 7 ia An ee u Mal Mi i ai ar a N i vf 2 0 005 5 Baur 11 12 13 14 15 16 u Time hours MLO RE Lock Drift 3 j i PN u in room E ML Hi Li if vi un l ur igi i an N Hy Hl ih if y 4 As N IH Pe amp 10 HN za 2 4 e i l 8 10 12 o 5 2 4 6 ze 10 12 Time hours Figure 8 3 5 RF phase measurement drift with temperature control with and without disturbances people in room RF signal source came from a signal generator and not from the laser Without temperature control the phase measurement drift jumps to 20 fs pkpk Fig 8 3 6 without disturbances and several
166. ith which the temperature regulation loop can function is the latency between the time that a setpoint change command has been given and when the resulting temperature change 1s measured by the sensor Because the controller reacts too quickly to the setpoint change by the time the thermistor measures the resulting temperature change the system has already over shot the target setpoint by a large margin In the case of a small 0 1 degree C setpoint increase the system will heat up by more than half of a degree before t converges back to the new setpoint This problem can be addressed by adjusting the feedback parameters of the temperature controller and by locating the feedback thermistor as close as possible to the peltier 129 Celsius Temperatur EOM1 31 93 31 92 31 91 T 31 89 l 31 88 31 31 34 Figure 7 2 5 Effectiveness of active temperature control in the tunnel While the thermal stability of the system limits the long term stability of the measurement the resolution 1s limited by the bandwidth of the signal that is transmitted to the EOM Maximum bandwidth can frequently not be used because the machine is so unstable that the beam is constantly jumping out of the measurement s dynamic range While the measurement can achieve sub micron resolution for large signal slopes the measurement was only stable for signals that were attenuated so that they produced 2 4 um resolution If the machine becomes so stable
167. itial planning stages of the chicane BPM 2005 was to take an array of 10 cm long striplines and mount them above and below the flat vacuum chamber of the bunch compressor 1 3 GHz would be filtered out from the pickup signal and the down conversion scheme used by the FLASH LLRF system would be used to sample the amplitudes of the signals The beam centroid would be determined by processing the multiple channels of data with the LLRF system FPGA thereby providing bunch to bunch beam position measurements that could be incorporated into an intra train beam based feedback The difficulties with this technique come from drifts and noise of the filters amplifiers and down conversion electronics As described in Sect 3 1 the individual LLRF amplitude measurements suffer from drifts of 2e 3 deg C and rms jitter of Se 4 This would not be sufficient for the lt 5e 5 resolution desired for the chicane BPM While a reference injection scheme has recently reduced these drifts to within the resolution of the measurement and a new down conversion front end has recently been developed with le 4 resolution this would still not be sufficient to meet the desired BPM resolution Nevertheless using stripline pickups with digital down conversion and reference injection has been successfully used by LCLS for their stripline BPM system The LCLS stripline BPM 140 MHz front end addresses the problems of filter and amplifier drift with the injection of a reference signal to
168. ition and profile over the course of the spectrum measurement Moreover it 1s not mathematically possible to accurately reconstruct the asymmetrical longitudinal profile of the beam with either method This is also not important because when the beam 1s lasing certain frequency components measured with the single shot spectrometer become stronger and if they are maintained at a constant level with a feedback on the accelerating phase then the beam has been shown to lase at a more constant level 20 There is a complication that has not yet been mentioned namely the the energy jitter that is caused by accelerator phase jitter When the Rss is small in Eq 3 2 1 the contribution of the amplitude jitter of the accelerating section to the overall timing jitter becomes smaller relative to the contribution of the phase jitter to the timing jitter Using information from the bunch compression monitor it 1s then necessary to disentangle which energy changes are caused by accelerating amplitude jitter and which are caused by accelerating phase jitter To do this one must consider the transformation 2 es hd 3 6 1 O z0 M O The first line of this transformation should be familiar from Eq 3 2 1 and the second line might be derived along the lines of Eq 2 3 9 but in practice the quantity M must be measured To use this transformation to generate feedback commands for the upstream accelerating section the inverse of the matrix must be found a
169. ition of the beam s center of mass s measured and the particular transverse distribution can be neglected This is not the case for the chicane BPM pickup The measurements of the transversely mounted stripline pickup use the arrival times of the pulses at the ends of the pickup and due to the large dispersion n the chicane and energy spreads of up to 1 the transverse beam size may influence the measurement If the beam is tilted in the x y plane or has a longitudinally asymmetric charge distribution this will also affect the signal produced by the transversely mounted stripline pickup First however the way in which a wide beam couples to the pickup will be described 6 1 Pickup Signals from a Wide Beam In a Green s function approach to the description of the pickup signals generated by a wide beam traveling under a transversely mounted stripline pickup we first take the voltage output of the stripline resulting from a pencil like beam passing beneath the middle of the stripline This voltage can be taken from the CST simulations presented in the previous chapter and will be given the name Uyft We can use it to determine the voltage output for various charge distributions by summing together the results from various arrangements of pencil like beams There is a Green s function G associated with the differential operator L from the linear differential equation 80 L x U Axt 6 1 1 where A is a known homogeneous li
170. itter of the RF phase relative to the phase of the laser was most accurately measured in 28 In this measurement the phase of the laser was changed with a vector modulator that acted on the master oscillator signal feeding the EOM such that the beam arrived on the falling slope of the cavity RF signal At this phase the changes n the RF phase relative to the laser phase produced a change in beam charge that could be measured with a downstream toroid By scanning the phase of the laser relative to the RF phase a calibration of the beam charge dependence on the phase relationship between the RF and the laser could be determined Multiplying this calibration by the charge fluctuations measured at the toroid gave a measurement of the phase jitter between the laser and the cavity RF This jitter was larger than 0 5 degrees from pulse to pulse a quantity that requires significant improvement Nevertheless it is of little use to the machine if the laser and cavity are locked together if they are drifting or jittering relative to a downstream reference Relative to the optical timing reference the injector laser timing jitter can be measured by optically cross correlating the injector laser pulse with a pulse from the optical reference Fig 3 4 1 29 This measurement can then be used to feed back on the phase of the RF signal sent to the EOM in the injector laser cavity and thereby stabilize the injector laser timing relative to the optical reference
171. jector 43 or X mm 2 5 0 08 Y mm 7 al 0 X a 0 05 N P IN RMS beam size N we Centroid position OON 0 2 0 5 o wA es nn 0 25 0 0 3 0 2 4 6 8 10 12 14 2 4 6 8 10 12 14 Z m Z m Figure 4 2 6 A beam offset in the injector is magnified as it travels through the first accelerator section After the RF gun the RMS beam size does not change significantly This mismatch cannot of course be avoided in the injector because the extra focusing strength is needed in order to counteract the space charge forces The only way to avoid the exacerbation of asymmetries that it promotes is align the injector with sub mm precision 4 3 Downstream of the Injector When the beam is accelerated off crest in the first accelerator section a much larger head tail separation is observed on the OTR or synchrontron light monitor in the bunch compressor than for the on crest case described n the previous section This off crest tilt is the result of the effects of a mis aligned injector described in the previous section plus the effect of dispersion in and after the accelerator section The tilt produced by a mis aligned injector is however insignificant compared to the tilt created by dispersion from beam offsets in dipole fields downstream of the first accelerator section To measure the strength of these effects the beam was placed off crest n the first accelerator section and then give
172. jitter is increased One would then need to stabilize the beam energy using an accelerator section located after the chicanes Alternatively one could use a single accelerator section gradient setpoint to simultaneously stabilize the beam energy and arrival time after each chicane The end result of both schemes would in principle be the same For the sake of machine stability the author believes that the latter option is better a feedback on the first accelerator section should not respond to changes in the injector jitter and a feedback on the second accelerator section should not respond to changes in the first accelerator section To simultaneously stabilize beam energy and arrival time after a bunch compressor one could execute a combination of the following e use measurements of the arrival time jitter upstream of a bunch compressor to keep the energy arrival time feedback from responding to the energy arrival time jitter that t creates e stabilize the arrival time jitter upstream of the bunch compressor before correcting the energy arrival time jitter downstream 3 3 Beam Based Feedback Strategy Two schematics of the synchronization sensitive components in the machine are shown in Fig 3 3 1 27 An optimal feedback setup is depicted in Fig 3 3 1 a and a more quickly realizable architecture 1s depicted in Fig 3 3 1 b The injector laser Laser the injector RF Gun the super conducting accelerator sections ACC1 7 and th
173. k signal is the beam transient pulse from the pickup The red dot is the sampling location The red signal is the laser pulse The optimal sampling location for a measurement is at the zero crossing of the signal middle drawing Whenever the beam arrival time changes significantly the laser pulse arrival time must be adjusted so that t samples the signal at the zero crossing and the measurement must be re calibrated Whenever beam properties change by enough to influence the amplitude of the signal the measurement must be re calibrated This sampling concept was the idea of Holger Schlarb and was first implemented in 20 7 4 Optical front end execution In the following figure Fig 7 4 1 a fiber splicing plan for the optical front end of the chicane BPM is shown The goal of this splicing plan is to steal a small portion of the light from a length stabilized fiber link which has been delivered to a nearby BAM chassis transport it over an 8 meter long fiber patch cable and adjust its polarization amplify it and then adjust its arrival time at two pairs of EOMs The tap off point within the BAM chassis is shown in the top portion of the drawing while the splicing plan for the devices residing within the chicane BPM chassis is shown n the lower portion of the drawing The lengths of the fibers within the BPM chassis are written above the splicing plan and the optical power levels at and insertion losses of the various components are writte
174. l py 3fi eBl p eB 1 8 d cBL 2 R 2p l 3 1 eBl p eB I1 eBl py which by substituting cosa wherever the square root of 1 eBley p arises can be written more simply in terms of the bend angle and the effective dipole length tang 4 R 2d as cosa sing 21 er p er 1 tana a 1 9 cos ga sing cosg Both Rss and R76 increase when the length of the chicane increases and when the bending angle increases With a rough small angle approximation R s 2 R Although the results for the single chicane are the same as those presented here a different and more complete treatment of dispersion and momentum compaction in different types of chicanes is given in 1 Because the particles in the chicanes are relativistic we can use the approximation Ap p AE E and will use the term relative momentum changes synonymously with relative energy changes This allows us to write the useful formulas for the change in path length Al and change in horizontal position 4x that occur as a result of momentum compaction and dispersion 1 10 E Al Rss Ax Rie Because the momentum compaction and dispersion in a chicane are non zero changes in the x position of the beam within the chicane correlate with changes in the energy of the beam Likewise the path length changes in and after the chicane will result in beam arrival time changes which are correlated with beam ener
175. lack of a normal conducting cavity limits the speed with which the beam energy can be changed As a result the first 20 bunches in the train will have a different energy than the stabilized bunches that follow Because the scheme depicted in 3 3 1 a will take 2 3 years to be realized the scheme shown n 3 3 1 b to be commissioned in the coming months will be the focus of the following sections 3 4 Injector Jitter The injector needs to be stabilized on two fronts the laser timing jitter and the cavity RF phase jitter The present infrastructure synchronizes both devices to an RF reference and the delivery of this reference 1s subject to drifts and noise The infrastructure described in the previous section synchronizes both devices to an optical timing reference for which drifts have been actively compensated The injector laser 1s actively mode locked and has an electro optical device within the laser cavity that regulates the arrival time of the laser pulses at the cathode These electro optical modulators EOMs are driven with the 1 3 GHz reference from the master oscillator Temperature changes and noise picked up by the cable and amplifier that bring the master oscillator signal to the EOM have of course an impact on the phase stability and drift of the laser Temperature changes and noise picked up on the cables involved in cavity field regulation will likewise have an impact on the phase stability of the field in the cavity The j
176. lative to the beam path The vectors contributing to this tilt are shown in Fig 4 2 4 and are defined below in terms of Vay the accelerating voltage of the cavity Vy the portion of the accelerating voltage in the beam direction and V7 the portion of the accelerating voltage acting on the beam in the plane of the module s tilt Vro Vea SIN O cav 40 Vr Vro cos k Z Ur cos d k zsing 1 k z cos g Ay V z E Figure 4 2 4 Voltages acting on a beam as it travels in z through a cavity tilted in the y z plane where z is the beam direction in the internal coordinates of the beam y is the direction of the tilt O is the angle of the tilt amp is the phase of the cavity RF field and ky is the wavenumber of the cavity RF field Using this description of the cavity we can estimate that for a 1 mrad cavity tilt at 25 MeV where the beam length is 3 degrees of the cavity wave at a phase of 10 degrees off crest the transverse accelerating voltage is 25 keV This would produce a 227 V potential for one sigma of the beam dimension corresponding to bunch tilt of 10urad This is even smaller than the effect of coupler kicks Wakefields tend to magnify any particle offsets driving the head and tail even further apart and they are best described by the so called wake function which is the Fourier transform of the coupling impedance The wake function is an integral of the electric field of a particle over an accelera
177. lectrical signal When the crystal 1s under the influence of an electric field 1t becomes birefringent and causes a phase shift of the light that is transmitted Each crystal experiences the opposite polarity of the electric field and will shift the phase of the light in opposite directions When the laser pulses are recombined constructive or destructive interference between the pulses will result in a laser pulse amplitude that changes when the amplitude of the electrical signal changes A less symbolic drawing of the EOM would depict the laser pulse and RF wave co propagating along a stripline such that the group velocity of the laser pulse 1s equal to the phase velocity of the RF wave Usually this relationship cannot be maintained above a certain frequency The EOMs are presently available with bandwidths of 10 20 and 40 GHz The 40 GHz EOM is 3 times as expensive as the 10 GHz EOM and until other limiters of the bandwidth are addressed it will not be used In Fig 7 3 1 the electrical signal black shown above the EOM has a wavelength that 1s long compared to the repetition period of the laser pulses red shown below the EOM Looking at the amplitude of the electrical signal and the amplitude of the laser pulses it s clear that the amplitude of the electrical wave modulates the amplitude of the laser pulses 121 200 1000 fs lo laser pulse Figure 7 3 1 Mach Zehnder Electro Optical Modulator EOM used to sample the amplitude of
178. ler kicks Table 4 3 1 Contributions to x z beam tilt for an off crest bunch observed in x y beam images taken in the first bunch compressor In the first two cases a flat energy distribution coming out of the gun was used and in the last case the energy chirp that comes from the gun was used 47 In the first two cases of Table 4 3 1 a flat energy distribution coming out of the gun was used and in the last case the energy chirp that comes from the gun was used Dispersion contributes most to the tilt while the impact of wakefields was barely measurable by changing the charge of the beam especially given a tilted module in which the same tilts could be generated by different paths through the module Coupler kicks simulated with equations from 38 are so small that they are not possible to measure with either an on or off crest beam When the beam is on crest there is no dispersion contribution after the injector and the primary effect that should be visible on the beam are the coupler kicks and injector mis alignement effects The coupler kicks have such a small impact that it would not be possible to measure them with the screen in the dispersive section of the chicane The beam tilts predicted with the simulation matched the measured results with high accuracy when offsets at the exit of the gun were taken into account Fig 4 3 4 SR Camera BC2 Bunch tilt measurement of slices 19 Fit range 5 15
179. lter and it is seen on an oscilloscope measurement as well as n simulation The higher frequency components beat with the lower frequency components to cause the asymmetrical signal shape for LO phases that minimize the amplitude of the signal This is easier to visualize when the pulsed situation from Fig 7 1 3 is simulated in continuous wave format Fig 7 1 4 In Fig 7 1 4 two different frequencies with the same phases but different amplitudes are added together to make a single RF signal The RF signal is mixed with the LO As the phase of the LO is changed a beating behavior becomes apparent between the lower and higher frequency components that make up the RF signal The relative 100 strengths of the peaks seen in Fig 7 1 4 change as the LO phase is changed This is the same effect that is seen in the asymmetric behavior of the mixer output in the pulsed signal simulations of 7 1 3 1 57 1 5 05 17 l i i ae eee mn PERA IT LII SCROLL CIDA ll
180. m 21 mm _1I1 LE Ze ee Installed Optimal Standard reference Figure 5 1 4 Dimensions of pickups in three different configurations the unsuccessful but expedient design which was installed in the tunnel left a flat chamber design that would have had performance comparable to existing button pickups installed in round chambers middle standard pickup dimensions in a round chamber right In Fig 5 1 5 the sensitivities of the three different pickup configurations shown previously to changes in the beam position are shown below The sensitivity of the expedient configuration that was installed in the bunch compressors is shown in green This configuration has roughly half of the sensitivity of the same 8 mm diameter buttons installed in a typical 35 mm diameter round chamber 20 mm diameter button pickups installed with a spacing of 21 mm would have delivered resolution that is comparable to the resolution of 8 mm button pickups in a 35 mm round chamber Both 8 mm and 17 mm buttons in a round chamber when coupled with typical DESY front end electronics get 15 20 um resolution over the central linear portion of their 10 mm dynamic range at InC This means that it is possible to install large closely spaced buttons that have a sensitivity that is comparable to existing button installations that get 15 20 um resolution with existing electronics It is not however possible to find a design with buttons that can get the required sub 5
181. m left to right is that the incoming arrival time jitter X is compressed in the chicane The second thing to notice is that at FLASH the amplitude stability o 4 A of the klystron will be more critical than the phase stability O cok 7 for most typical values of o4 A and 6 cok This becomes more apparent when typical values from the first bunch compressor of FLASH BC2 are inserted into the equation giving Vane o 0 2ps 0 1ps ps T 0 04 5 5ps er 0 01 2ps deg case With a compression factor of 10 n the first bunch compressor the 200 fs injector jitter would be compressed to 20 fs arrival time jitter after the bunch compressor The 0 04 gradient stability of the first accelerator section would limit the arrival time jitter downstream of BC2 to 220 fs The 0 01 deg phase stability of the first accelerator section would limit the arrival time jitter downstream of the first bunch compressor BC2 to 20 fs Given no additional contributions to arrival time jitter from the second accelerator section amplitude or phase and further compression of the injector jitter by a factor of 2 in the second bunch compressor the arrival time stability at the end of the machine would be limited by the first accelerator section to 220 fs An improvement by a factor of 10 in the gradient stability would make the arrival time jitter contributions of the first accelerator section amplitude and phase approximately equal The arrival time stability at the end
182. meters V7 V3 3 in Eq 3 5 1 To optimize this system the second order energy chirp will be compensated by an appropriate phase and amplitude setting in the third harmonic module Under the current plan for the operating parameters the gradient of the third harmonic cavity is only a 9 of the gradient of the downstream accelerator sections so the amplitude jitter contribution will be only a 9 of that of the downstream accelerator sections This means that it doesn t make sense to stabilize the gradient of the third harmonic module until the down stream accelerator sections gradient stabilities are improved by a factor of nine Because this has yet to be accomplished the third harmonic module is a lesser worry Schemes that attempt to compensate for the third order energy chirp call for large amplitudes in the third harmonic module and this would begin to have an impact on the arrival time stability The jitter contribution of the phase of the third harmonic module depends much more dramatically on the setpoint of the module Some settings make the module the dominant contributor to bunch length jitter while others make it the weakest contributor It can be argued that the ideal setting for timing jitter and peak current considerations is one for which the phase jitter contributions of the 3 9 and 1 3 GHz modules are approximately equal 31 3 6 First Accelerator Section Jitter The first accelerator section needs to be sta
183. n Monitoring AIP conference proceedings on Accelerator Instrumentation AIP Upton NY 1989 pp 26 55 David M Pozar Microwave Engineering Wiley amp Sons January 2004 F Marcellini M Serio M Zobov DAPHNE Broadband Button Electrodes Frascati January 16 1996 Note CD 6 T Smith SLAC Personal Communication J Zemella Driftfreier Detector zur Messung des Zeitversatzes zweier verschiedener Laserpulszuege Diploma Thesis University of Hamburg 2008 B Lorbeer et al Noise and drift characterization of critical components for the laser based synchronization system at FLASH Proceedings of the DIPAC 2007 Venice Italy F Ludwig et al Noise and drift characterization of direct laser to RF conversion scheme for the laser based synchronization system for FLASH at DESY Proceedings of the PAC 2007 Albequerque New Mexico USA J Kim et al Long term femtosecond timing link stabilization using a single crystal balanced cross correlator Opt Lett 2007 no 9 1044 1046 K Hacker et al RF lock for Master Laser Oscillator Proceedings of the EPAC 2009 Edinburgh Scotland C Gerth Synchrotron Radiation Monitor for Energy Spectrum Measurements n the Bunch Compressor FLASH Proceedings of the DIPAC 2007 Venice Italy A Wilhelm Installation and Characterization of a Diagnostic System for Bunch Resolved Beam Energy Measurements at FLASH Diploma Thesis De
184. n a closed orbit bump with several correctors The tilt of the beam was maximized by adjusting the phase advance at the screen and then measured by analyzing the images taken with the synchrotron light monitor screen The measurements were done together with C Gerth and E Pratt 39 In Fig 4 3 1 the closed orbit bumps are depicted on the left and the resulting tilts are depicted on the right 44 8 deg off crest in ACC1 eA Tog GUN acci horizontal plane o 298 0 181 v 1 691 HOM RAM 1 923 1 554 cun HL Gi a W M F 0 184 1 542 4 408 8 129 0 000 ad dd ash aaa Aa d d 3 2501 an 1 036 70 009 2 501 A 0 208 A 1 036 0 009 A vertical plane Dec horizontal plane i k ii 1 59 Hon RAW 2 277 i EN Bi if Hh ni wre 1 10 F 0 004 BS z 1 000 z 9 207 z 1 036 0 015 1 001 A 0 208 A 1 036 0 015 A vertical plane DEC ACEL M TTUN acci UBC2 BC2 0 765 0 970 0 026 A a A A aa a A aaa a Aha giai GR HEHE zu L 1 50 0 625 HOM RAW 2 548 0 672 APPs O Bern teed 2 3 71 F 0 004 B 1 000 70 207 0 264 0 008 1 001 A 0 208 A 0 260 0 008 A vertical plane DEC 20 40 60 80 100 120 140 Figure 4 3 1 The tilt of the beam for various closed orbit bumps through the first accelerator section Screen shots taken from a synchrotron light monitor located after the 3 dipole of the bunch compressor Bunch is streaked out longitudinally but the
185. n be adjusted until the returning phase matches the sent phase thereby removing the effects of cable drifts Although the phase detection of the returning pulse can be made with 10 fs accuracy 20 or more fs of noise is picked up over the long cable and the reflection of the pulse at the end of the cable s problematic due to the temperature dependence of the mismatch that produces the reflection Any drift of the mismatch will be corrected by the feedback loop but this does not accurately represent the drift of the cable and it therefore adds an error to the cable length stabilization The directivity of the coupler is perhaps the fatal flaw in the method because any changes in the directivity will directly impact the comparison of outgoing and returning signal phases This reflectrometry method has not been tried extensively and the absolute limitations are not completely clear Nevertheless it does not look very promising It also becomes more difficult for cables that are longer than 100 meters due to the attenuation of the signal on the cable If one had access to an RF reference signal with a stable arrival time one could measure the arrival time of the electron bunch relative to the reference signal by taking a signal from a pickup in the beam pipe filtering out a frequency and measuring its phase my mixing the filtered signal with the MO signal Beam arrival time measurements using such a technique were presented in the previous chap
186. n below the splicing plan FC APC connectors are shown as thick black marks optical delay lines are labeled as ODLs the amplifier 1s represented as a triangle with 4um inside representing that 4 um gain fiber was used The 3 picosecond length of the laser pulses when they arrive in the amplifier is written above the amplifier and the transition from single mode fiber SM to polarization maintaining fiber PM is written underneath the first ODL 124 20mW 0 BAM 60mW Erfiber 60mW Link 8m 2m 2m 9mW 3ps pulse arrives from BAM Acrobat Polarization Controller 90 10 200 mW 100 mW 50 mW 5 mW clock 1dB 3dB 1dB 3dB 1dB SdB 2dB 3 mW at ADC required Figure 7 4 1 Chicane BPM optical front end schematic Laser pulses from a near by length stabilized link are tapped off and delivered over an 8 meter long optical patch cord to the chicane BPM chassis Within the chassis the polarization of the incoming link light is controlled the light is amplified and the arrival time of the pulses at the 4 EOMs is adjusted with the Optical Delay Lines ODLs The length stabilized fiber link entering the schematic in the upper right corner is a long stretch of SMF optical fiber connecting the Master Laser Oscillator MLO to a timing sensitive device in this case the chicane BPM The pulses coming from the MLO are reflected by a Faraday rotator in the timing sensitive device and sent back to the room containing
187. ncy of 1 3 GHz is used as an LO for a mixing scheme with signals from the BPM pickup The reference frequency of 108 MHz is used as the clock for an ADC that samples the outputs of this schematic An optional scheme for generating the reference frequencies from a master laser signal is sketched in the bottom left corner The figure above contains a symbol labeled MO INJ3 This refers to the signal from the master RF oscillator in the injector racks The MO signals were delivered over a few meters long cable to the patch panel labeled PatchPanel2 1 on the left side of the drawing Signals from the BPM pickup were delivered over 30 meter long cables to the same patch panel From the patch panel the signals are sent over 2 meter long RF cables to the chassis depicted by the blue square Within the chassis the 108 MHz signal from the MO is sent through a clock generation circuit The clock generation circuit turns the sinusoidal 108 MHz signal into a square wave that is appropriate for the clock input of the ADC that will be used to sample the mixer outputs The 1 3 GHz signal from the MO enters the chassis and is split into three signals two of the signals are mixed with 1 3 GHz waves coming from a filtered pickup signal and one of the signals is multiplied by 8 in a Hittite frequency multiplier After the frequency multiplier the 10 4 GHz signals are split and then mixed with 10 4 GHz waves coming from a filtered pickup signal In one of the tw
188. nd an ADC that is clocked with a signal that is generated by the 216 MHz repetition rate laser pulses themselves Fig 7 3 3 detector gt 200 MHz Photo detector gt 200 MHz 216 MHz Figure 7 3 3 Measuring the amplitude of the laser pulses with an ADC that is clocked with a signal that is generated by the laser pulses themselves A band pass filter is used to extract a clock signal from a higher bandwidth photodetector In Fig 7 3 3 a bandpass filter with a center frequency of 216 MHz filters the signal from the photodetector The 216 MHz signal is conditioned to make a square wave that is appropriate for the clock of the ADC In initial experiments an AD9510 clock divider evaluation board and an RF bias adjustment were used to make the clock signal In later versions a more compact printed circuit board design was used Within the in house ADC FPGA board digital clock dividers and shifters were employed to optimize the ADC sampling time Once the amplitude of the laser pulse is measured one must scan the arrival time of the laser signal about the slope of the steeply falling edge of the beam transient pulse This provides a calibration factor for the pulse arrival time measurements Fig 7 3 4 123 Figure 7 3 4 Calibrating the arrival time measurement requires scanning the arrival time of the laser pulse about the zero crossing of the beam transient pulse Such a scan is shown in the drawings from left to right The blac
189. nd used in the following way ala AA io eb Ft 3 6 2 0 M o o p 34 3 7 Second Accelerator Section Jitter When operated on crest the amplitude jitter of the second accelerator section affects the arrival time stability in a way that is similar to that of the first accelerator section only the effect is smaller because the Rss of the second bunch compressor at FLASH has typically been five times smaller than that of the first bunch compressor This means that the conversion of energy jitter into timing jitter has been five times less dramatic This also reduces the required sensitivity of the second bunch compressor monitors by about a factor of 5 compared to that of the monitors in the first bunch compressor chicane In future machine configurations however the compression factors of both compressors will be more equal In these future machine configurations the Rs6s will be reduced thereby increasing the resolution requirements of beam arrival time measurements for both the first and second accelerator section energy jitter measurements In future compression schemes the second accelerating section will be operated off crest with a larger R56 in the following bunch compressor making the trouble with arrival time jitter upstream of the chicanes the same as it is in the case of the first accelerator section As in the case of the first accelerator section if the arrival time jitter upstream of the second chicane is not small
190. near charge distribution from a pencil like beam and where the inverse of the differential operator is defined in terms of the Green s function by L x DAX t Gx t A dt 6 1 2 The solution to Eq 6 1 1 can then be written in terms of the Green s function U Naar 6 1 3 u Let xo be the center of the stripline and when the pencil like beam position is altered by Ax Ax is much less than xp so that the signal at the exit of the pickup will not change in shape but only be delayed by Ax co This critical assumption was verified with both CST simulations and a mockup of the pickup and electron beam which could be moved with a micrometer We will also assume that the pulse shape will not change if the beam arrival time changes In terms of the Green s function these assumptions are written for the left and right pickup outputs as G x Ax t G x t At Ax c 6 1 4 meaning that the function is invariant under translation in space and time and can therefore be used as a convolution operator Let us define the charge distribution of the thin pencil beam as Gaussian in x and tty x Xy Nad 2 2 ne Fe 6 1 5 21 0 0 xt Which is normalized to the bunch charge Q according to A t x dtdx O 6 1 6 Integrating in over slices of the beam defined by the pencil like beams Eq 6 1 3 we get the pickup output voltage that would result from a wide beam U t aet 0At ded 6
191. ng a pickup geometry that keeps Zo constant Fig 5 1 7 Issues related to the resistance of the transmission line have not been covered ceramic Alumina ral Pout vacuum air metal Figure 5 1 7 Button geometry that keeps the impedance constant will keep the ratio between the inner and outer conductor constant A button pickup can be modeled by the equivalent circuit shown in Fig 5 1 3 in which the beam current beam in parallel with the transfer impedance Z makes a voltage a source that is experienced by the high pass filter composed by capacitance Cyickup and R the load impedance Coiekup V pickup Figure 5 1 8 Button pickup equivalent circuit pcam 15 the current of the beam Z is the transfer impedance Cyickup 18 the electrode capacitance and R is the load impedance The transfer impedance Z is the quantity that we are looking for and to get it we start by writing down the voltage induced at the button with no load lo o Vi are E dt 5 l 24 pickup o If the bunch is longer than the button radius and the beam s n the middle of the pipe we use an approximation depicted in Fig 5 1 9 C button lt gt At g beam current Figure 5 1 9 Approximation that bunch is longer than button radius allows for integration over the beam current in steps of Ar It is valid for frequencies with wavelengths longer than the button radius along with Eq 5 1 10 to get V L anh d
192. ng out the influence of wakefields If the gradient of the first accelerator section was changed the c shape was unaffected ruling out the influence of the coupler kicks The c shape was present both before and after a major injector upgrade In fact it was more pronounced after the upgrade This suggests that the screen in the bunch compressor should be used as a diagnostic to verify proper solenoid alignment With transport through even a perfectly aligned accelerator section and lattice any offsets gained n the injector are magnified This s clear from the simulation shown in Fig 4 2 6 in which the offset of a beam is tracked with ASTRA from the cathode through the first accelerator section An offset of 100 um in the horizontal plane X at the cathode becomes 140 um in Y and 40 um in X at the exit of the first accelerator section As the beam travels through the accelerating module the focusing effect of the module is evident through the curvature of the beam s path But before the beam enters the accelerating module the offset particles are over focused in the injector This 1s because the injector lattice 1s optimized for a space charge dominated beam while the centroid orbit is not affected by the space charge effect A mismatched beam line is a group of focusing fields that are either too weak or too strong for the beam that is being transported The mismatch in the injector magnifies any position chirps that are generated in the in
193. ng to z T E z T Epon 5 nom ay Adding and subtracting the energy of the reference particle gives z 2 IE z T E TE T Enon 6 where the second term in brackets is a constant which vanishes if the reference energy of the bunch is equal to the nominal energy of the chicane A Taylor expansion about z 0 of the first term in brackets then gives z T E g T Eqom A TENz 7 om This result can be used to write the compression factor of a bunch which is linearly compressed a Ge Ley 8 O 166 Solving Eq 8 for Ez provides a relationship between the bunch energy chirp and the compression factor We now have the longitudinal position of a particle in terms of five free parameters Zp Z ZV PoE Ly 9 These results can be used to determine the sensitivity of the longitudinal position to each of the free parameters Arrival time jitter due to voltage changes Using Eq 5 together with t z c the sensitivity of the arrival time to the accelerator voltage is Ot 1 l f f T IE z cos k z 10 OV C v C z 42 p R en and uses Eq 4 to write T 1 nom When one solves Eq 1 for cos k z Eq 10 can then be written ot l Rs Ed ee 11 OV a E V a nom Making the approximations that the initial energy is much smaller than the final energy Eo lt lt E and that t
194. niversity of Hamburg 2008 M Roehrs Investigation of the Phase Space Distribution of Electron Bunches at the FLASH Linac Using a Transverse Deflecting Structure Ph D thesis University of Hamburg 2008 C Gahl et al Nature Photonics 2 165 169 2008 K Floettmann Astra user manual http www desy de mpy astra documentation M Dohlus Field Asymmetries and Kicks http tesla desy de fla publications talks seminar FLA seminar 090904 pdf H Wiedemann Particle Accelerator Physics II Springer Verlag Berlin Heidelberg 1999 I Zagorodnov Wakefield Effects of new ILC cavity shapes European Particle Accelerator Conference Edinburgh Scotland 2006 M Krasilnikov et al Beam based Procedures for RF Guns Particle Accelerator Conference Knoxville Tennessee 2005 Work done together with E Prat and C Gerth 2008 J Rosenzweig and L Serafini Transverse Particle Motion in Radio frequency Linear Accelerators Phys Rev E 49 2 1599 1602 Feb1994 A W Chao Physics of Collective Beam Instabilities in High Energy Accelerators John Wiley amp Sons Inc 1993 175 42 43 44 45 46 47 48 49 50 51 52 53 54 E Medvedko R Johnson S Smith R Akre D Anderson J Olsen T Straumann A Young Stripline Beam Position Monitors for LCLS Beam Instrumentation Workshop 2008 R E Schafer Beam Positio
195. o arms of the 10 4 GHz down mixing circuit there is a box labeled phase shifter This mechanical motorized trombone phase shifter from the company ATM is used to synchronize the arrival times of the signals coming from the pickup If the phase shifter is adjusted appropriately the signals from the right and left sides of the pickup 105 will arrive at the mixers at exactly the same time The other type of phase shifter is the vector modulator There are two of them and they are labeled VM in the drawing One shifts the phase of the 108 MHz clock and is used to adjust the sampling time of the ADC and the other will shift the common LO for all of the different phase measurement circuits When the beam arrival time is changed the phase of the 1 3 GHz signal coming from the MO will need to be adjusted in order to maintain the optimal LO phase An optional feature that was not built into the tested design is the generation of the LO signals from the optical reference produced by the master laser oscillator This is shown in the lower portion of Fig 7 2 1 and again in more detail in Fig 7 2 2 This scheme would be advantageous if the front end were relied upon to generate beam arrival time measurements as well as beam position measurements In a beam position measurement the drift of the LO is irrelevant because the position depends on the difference between the arrival times of the signals The beam arrival time however is measured by the
196. o scale An existing DESY design for quadrupole mounted striplines left A new design with tapering to the feedthroughs right The existing design had been required due to space constraints in other installations The figures of merit for measuring the performance of the monitor include the steep slope of the signal at the zero crossing low amplitude and the absence of distortions in the signal that occur when the position of the beam is changed The older and newer designs 71 were compared with the aid of CST software and it appeared that the performance of the tapered design Fig 5 5 4 middle would offer a significant improvement over the existing design Fig 5 5 4 top re Tr ae u m 2 S Par reter Magrini un Il Ab itt au Wa 0 05 z L 14215 Tere rs Parameter Moore Alt 101 ariel oahi Sieh 0 01 0 02 D0251 i i l 0 42322 3 1 0237 10 0 a zi Time re Frequency CHE Time Signals Parameter Magniude 2 0918 of 1 U1 i eea 0 42322 05 0 8 i 1 2 1425 Time fins Frequency GHz Figure 5 5 4 Comparison of time and frequency domain simulations of three different stripline designs older stripline design top tapered without ceramic support middle tapered with ceramic support bottom Slope of time domain signal of middle design is 35 steeper than top design Slope of bottom design is only 5 steeper
197. o the tunnel do not present a problem Of course if the 1 3 GHz reference signal had been delivered to the in tunnel chassis with a length stabilized RF cable Fig 7 2 3 or with a length stabilized optical fiber Fig 7 2 2 the in tunnel RF front end could be used for precise beam arrival time measurements as well as beam position measurements This in conjunction with an improved pickup reduced reflections could make the arrival time resolution and stability of the RF front end competitive with the optical front end that will be described in the next section Under the best circumstances the RF arrival time measurement would not be able to easily achieve the lt 6 fs few minute resolution lt 30 fs few hour stability demonstrated by the optical system but 10 20 fs resolution and comparable stability for this RF front end would be possible Given that the 10 4 GHz RF front end is much more robust than the optical front end and need not rely on a complicated optical synchronization infrastructure it 1s an interesting option to consider in situations that require a quickly built and inexpensive system that doesn t require high precision synchronization of various laser systems A cost comparison of the considerably cheaper RF front end and the optical front end will be made in the last section of this chapter To summarize the beam position resolution achieved with the 1 3 GHz down mixing scheme was 25 um The beam position resolution that co
198. oes to Peter Schmueser Holger Schlarb Bernhard Schmidt and Joerg Rossbach for pre reading sections of this document to DESY for extending my contract whenever I had another baby and to Bernhard Schmidt for hosting me in his group and allowing me access to all of the labs and equipment and people therein Thanks to Eduard Prat and Christopher Gerth for investigating chicane beam tilts with me Thank you to Jan Hauschildt for the quick technical drawings manufacture and installation of the BPM pickup to Silke Vilcins for the drawings manufacture and installation of the new BAM pickups to Dirk Noelle and others for re arranging the BPMs around the chicane to Bernd Beyer for the technical drawings and manufacture of the mechanics for the first and second versions of the optical BPM and thank you to Matthias Hoffmann and Albert Schleiermacher for so quickly and professionally assembling the 2 layer RF chassis the Beckhoff boxes and the last version of the optical BPM chassis 178
199. of present systems s currently limited to 3 us but in certain locations where cable lengths can be minimized faster performance could be realized with future hardware While the delivery of control decisions to the klystron between bunches is possible making large changes in the field of the accelerating structure at that rate is physically limited due to the large quality factor or small bandwidth of the cavity Large 20 changes n the amplitude of the klystron produce only small changes n the cavity gradient When the corrections demanded by the feedback loop become small enough then the large quality factor of the cavity 1s no longer a problem and the cavity phase and amplitude can in principle reach a stability determined solely by the resolution and drift of the monitoring system used in the feedback In Fig 3 1 1 the current digital processing architecture is depicted 11 The cavity field probe signals 1 through n are sent to the field detectors wherein they are down mixed from 1 3 GHz to 54 MHz and sampled with an 81 MHz ADC The digitized signals from the ADC are sent into a digital phase and amplitude I amp Q detection algorithm This calculation requires a multiplication by a calibration constant The phase and amplitude information from each individual cavity are then added together in a vector sum The vector sum is compared to the setpoint values generated from a feed forward table A correction to the cavity fields
200. on the contact to the feedthrough When the contact was cleaned much better agreement with the simulation Fig 5 5 6 right was achieved i Parameter Magie E E nn nme en mn pm Bang en nenne en nein a OES N un nn nn nn nn gu nn nn mn pe ee eee eee ee ee eee k i 500 1000 1500 2000 2500 MHz Frequency Getz Figure 5 5 6 The pickup network analyzer measurement of two of the stripline outputs left and the simulation of a single pickup output right The network analyzer measurement shows poor agreement with the simulation at low frequencies due to oxidation on the pickup contacts The ringing in the measured signals s due to ceramic supports which were not included in the simulation shown here Agreement with simulation only really serves to engender a sense of confidence in the simulation The only test that really matters 1s to change the position of the beam over the full range and observe the corresponding changes in the zero crossings of the signals The zero crossings of the pickup signals were tracked as the beam was moved across the full range of the bunch compressor vacuum chamber and the resulting beam position measurements Eq 5 5 1 were plotted for both on and off crest beams as a function of a change in beam energy Fig 5 5 7 There are no anomalous distortions in the signal shape that disturb the position measurement over the full length of the pickup
201. on to give a function of the change in image current per change in beam position Let us call that function AI and use equations 5 4 1 and 5 1 10 to write Z AI 2 AV z z 8 z 2L 5 4 2 which can be Fourier transformed with ide AV ae az 5 4 3 20 to get ZA AV k e e sin kL 5 4 4 IT This value has a maximum at frequencies where the length is an odd multiple of quarter wavelengths 67 kL zen 1 o f pen 1 5 4 5 where n 1 2 3 For n 1 and a stripline length of 10 cm the central frequency would then be 750 MHz with a bandwidth of 1 9 GHz given by the 3 dB points fim I fan 3 So 5 4 6 and the time resolution is given by the inverse of the bandwidth BW times rt T l m BW yielding a signal width of 170 ps for the 10 cm long stripline case The beam position resolution achieved with FLASH 30 cm long striplines has been around 5 um using a front end method that combines the pickup outputs with a delay between them and sends them through a single filter and amplifier thereby removing the impact of filter and amplifier drifts LCLS gets 2 3 um resolution with 10 cm long striplines for beam charges ranging from 0 2 to 8 nC They send each pickup signal through its own filter and amplifier and calibrate away filter and amplifier drifts by injecting a reference signal prior to the arrival of each bunch 5 4 Array of Striplines One idea that was popular in the in
202. or the chicane BPM are compared and contrasted with respect to their limitations cost and performance In general given certain modifications they can deliver comparable performance but the optical measurement has a much lower potential for making systematic errors The optical measurement is considerably more expensive than the RF measurement and requires a complicated infrastructure to implement 7 1 RF Front end Concept The RF front end of the chicane BPM contains circuits that operate at 10 4 GHz and 1 3 GHz delivering two distinct measurements of the beam position The lower frequency and lower resolution measurement gives the information required to set the position of a mechanical phase shifter for the higher frequency and higher resolution measurement Both measurements utilize the same down mixing to baseband principle they take the outputs from the pickups filter out a certain frequency from the spectrum and mix that with the same frequency generated by the machine reference signal from the nearby master oscillator A simplified schematic illustrating the down mixing concept for the higher frequency is depicted in Fig 7 1 1 60V Center 10 4 GHz 200ps BW 400 MHz LP 150 MHz RY HR Noise on LO is common mode clock Figure 7 1 1 Down mixing scheme to measure the relative phases of two pulses In Fig 7 1 1 the output signals from the pickup are depicted on the left as bipolar pulses that are about 200 ps long and 60 V in
203. ort was also available for the front end that was to accompany it 5 5 Transversely Mounted Stripline A transversely mounted stripline BPM pickup is depicted in yellow above a wide rectangular vacuum chamber Fig 5 5 1 The pickup rests in a coaxial shaped channel which is open to the vacuum chamber below The pickup is tapered to an SMA sized vacuum feedthrough The beam path 1s depicted in green below the pickup Tapering Channel Beam Path SMA Vacuum Feedthrough Figure 5 5 1 3 D transparent representation of the upper half of the chicane BPM pickup not to scale The beam green travels under the pickup yellow If a standard stripline pickup is rotated so that it 1s perpendicular to the beam direction the current pulses induced as the beam passes beneath it will travel to each end of the pickup Fig 5 5 2 This is an idea that was proposed in early 2005 at DESY by Manfred Wendt The average difference in the arrival times of the current pulses multiplied by the speed of light in a coaxial cable gives the position of the beam beam _ position lt arrival 5 left arrival B right l 5 5 1 Alternatively the average of the arrival times of the current pulses multiplied by the speed of light gives the arrival time of the beam beam _arrival arrival _left arrival 5 right 5 5 2 This requires a measurement of the pickup signals phases and it is distinct from the typical stripline BPMs that measure
204. otron light monitor in the first bunch compressor As the phase of the upstream accelerator section is moved slightly off crest relative to the longitudinal center of the bunch the head of the bunch is visible sticking out above left When the phase is changed slightly in the opposite direction the tail 1s seen sticking out below right This head tail separation of 400 um accounts for a projected emittance growth of 1 8 38 There has been some debate over the cause of th s shape but not much interest n removing it Because only a slice of the bunch lases in the present machine without the third harmonic module the shape of the rest of the bunch has been irrelevant This situation while true now will not be true when the third harmonic module is commissioned and the projected emittance becomes almost as important as the slice emittance Some wondered if the c shape was the direct result of often simulated but never measured coupler kicks in the first accelerator section Others believed it was from oft simulated but never measured wakefields due to the mirror in the injector The contribution of linear dispersion induced downstream of the RF gun but before the first accelerator section would be too small to create this effect all by itself as would the effect of a tilted first accelerator section To put these effects in a frame of reference the strength of each effect will be quantified below Coupler kicks are caused by the forces t
205. ould be common to both cables It is not of course fair to make claims about the stability of a system without showing data that has been taken over several days to that end three days worth of data is plotted below in Fig 7 2 7 111 Mixer output for split signal resolution ee __ nn Te um 30 40 50 hours 70 80 Figure 7 2 7 Three days long measurement of the difference between the split signals Temperature disturbances 1 degree n the rack are marked in green and mornings are marked in red The temperature of the hall in which the cables reside slowly increased by one degree over the course of the three days In Fig 7 2 7 the changes n the temperature of the room in which the chassis is located have a larger impact on the stability of the
206. p the resolution of any low charge BAM scheme would be severely limited by this pickup That is why it was not selected It would however be an ideal pickup for high charge applications gt 1nC An SMA design with a tapered button Fig 5 1 15b would be preferable for the XFEL due to the shorter distances between electron bunches and the necessity to limit the amount of time that the pickup rings It is however much more complicated to build For low charge applications where 100 pC beams are expected one would prefer a pickup that s 10 times more sensitive than either of the pickups in Fig 5 1 15 In order to accomplish this the diameter of the beam pipe needs to be smaller so that the pickup is closer to the beam 64 5 2 Cavity BPM Cavity BPMs Fig 5 2 1 have been built that produce beam position measurements with sub micron precision and the same sort of monitor could n principle be scaled up to accommodate larger apertures Alternatively a small scale cavity BPM could be put on movers so that it slides from side to side along the flat chamber of the dispersive section of a chicane Both concepts however present significant challenges given the 20 cm of longitudinal space allocated for the chicane BPM Dee ge wee newer cree cre nccccenacccenenseceussssesesseeneses EEE wall currents Pree ee Ore CS Oe OOO rer rer rr rrr rer rr rrr reer rrr rere ere errr monopole dipole Figure 5 2 1 Side view cavity BPM Th
207. partment of Physics of the University of Hamburg August 2009 H Wiedemann Particle Accelerator Physics I Springer Verlag Berlin Heidelberg 1999 176 55 56 57 58 C Gerth personal communication H Schlarb personal communication A Wilhelm and C Gerth Synchrotron Radiation Monitor for Bunch Resolved Beam Energy Measurements at FLASH DIPAC 09 K Hacker et al Demonstration of BPM with sub 5 um resolution over a 10 cm range Proceedings of the EPAC 2009 Edinburgh Scotland 177 Acknowledgements Many thanks must be given to all of the people on the optical synchronization team for building up the infrastructure lasers and optical links that made the optical BPM measurements possible Holger Schlarb was the mastermind of the system Axel Winter started the upgrade of the system so that it could serve more users like myself Florian Loehl built first prototypes of many devices and ordered countless components for the optical front ends Sebastian Schulz built the laser and aligned the distribution section Marie Bock and Holger Schlarb got the link up and running on time Patrick Gessler got all of the ADCs and BAM servers going and Matthias Felber took care of countless issues related to piezo drivers and whatnot Especial thanks go to Holger Schlarb for always being open to answer my questions and for explaining just about everything that is written n this thesis to me A big thank you g
208. pect that as in the x z tilt case the tilt in the vertical plane s also generated primarily upstream of the first bunch compressor The monitor developed in this thesis can be made highly sensitive to the tilt of the beam and that is why it was given so much attention here The response of the measurements of the beam position in the chicane to beam tilts is described in Chapter 6 49 5 Beam Pickups Several types of beam position monitors have been used in accelerators and could be appropriate for use in the dispersive section of a bunch compressor Of these the button pickup the cavity monitor and various forms of stripline pickups will be described with respect to their relative merits for this application The primary focus will be on the design that was actually built and installed the transversely mounted stripline CST Microwave Studio simulations of some of the pickups were conducted and matched the measured performance of the monitors with high accuracy 5 1 Button Pickups Before the commencement of this thesis an array of button pickups was the first method attempted to meet the challenge of measuring the beam position over a 10 cm aperture The measurement principle relies on determining the normalized beam position Xnormalized AS a function of the voltages of the pickup outputs Vien Vn Lead ea 5 1 1 Viet at V asn where Vieg and V ign are the voltages from the buttons on the right and left sides of the beam and C
209. ple of operation Cross section of striplines in a round chamber and in a flat chamber Simulation of sensitivity of 25 mm diameter striplines in a round chamber configuration and in a flat chamber configuration 3 D transparent representation of the upper half of the chicane BPM pickup Cross section of a transversely mounted stripline pickup with tapering to vacuum feedthroughs Stripline feedthrough cross sections Comparison of time and frequency domain simulations of three different stripline designs The simulated blue and measured red performance of the pickup below 8 GHz left and below 50GHz right The pickup network analyzer measurement left and the simulation right Beam position across the full range of the vacuum chamber as a function of the beam energy change Impact of charge change on single sample point which resides 100 ps away from the zero crossing of the pickup signal Measurement top and simulation bottom of chicane BPM pickup signal amplitude response to changes in y position 5 5 10 5 5 11 6 2 1 6 2 2 6 2 3 6 3 1 6 3 2 6 3 3 6 3 4 6 3 5 6 4 1 6 5 1 6 5 2 6 5 3 6 6 1 6 6 2 6 7 1 Teka 212 7 2 1 22 W223 Dependence of the slope of pickup signal on the phase of the upstream accelerating section Change in beam position as a function of RF phase Coupling of the beam to the pickup for an elliptical beam and for a flat beam Cancellation of signals on the pi
210. plers have to be measured to determine their drift sensitivity but typical high power RF couplers drift by 40dB This RF reflectrometry method was not employed for the chicane BPM beam RF front end because beam arrival time measurements relative to a pulsed optical reference had already demonstrated 10 times better accuracy and precision than beam arrival time measurements relative to an RF reference could ever hope to achieve What was built and tested is shown below in Fig 7 2 4 The emphasis will be placed on the robustness of the beam position measurements and details about beam arrival time measurements are reserved for the following chapter The RF front end shown in Fig 7 2 4 was constructed in two layers within a 3 rack units high chassis The lower layer contains circuits that are not sensitive to temperature changes and the upper layer contains circuits which are sensitive to temperature changes The temperature of the upper layer was actively stabilized with two Wavelength HTC temperature controllers which determined the heating and cooling action of two Peltier elements Because the circuit element that was most sensitive to temperature changes was the band pass filter and because the stability of the high resolution measurement was a key goal of this setup the Peltier elements were installed as close as possible to the 10 4 GHz band pass filters Although the Peltier elements could have been directly mounted on the filters in order to
211. position dF ok ee a p x Ax dx p dx 2p x Ar 9 2 8 If p is a Gaussian distribution we x Ax e 9 2 9 p 0 J210 and then C a 9 2 10 dAx 270 Since the horizontal width of the beam is dominated by the energy spread this means that the sensitivity function decreases for increased energy spread regardless of the detector size The question that we want to answer s how big is the uncertainty of the sensitivity function when the beam position is constant 1s it limited by the shot noise Because S is statistically independent of S 9 2 11 Is the uncertainty of the sensitivity function and at the point of highest resolution S Si 8 so l E a 9 2 12 6S IS S S S_ Substituting Eq 9 2 12 into Eq 9 2 11 we get los _ N2 Os o 2 9 2 13 AS 2 S l 150 Finally we can use this together with Eq 9 2 10 to write the uncertainty of the position in terms of the uncertainty of the sensitivity for a Gaussian beam l y2 Oo 27 0 Os Ou T Opn ee ne ey Ot 9 2 14 dAx Now since o JS and S N the number of photons detected by one detector 1 e half of the total number of photons Np times the quantum efficiency of the detector s mo Wal e 2N 2 JN 9 2 17 where l N N quantum _ efficiency 3 10 Substituting N into Eq 9 2 17 the uncertainty of the position is given by the beam width times 0 0015 Taking into account the
212. ptical signals are immune to electro magnetic interference e Optical signals can contain a large bandwidth enabling high precision arrival time measurements of both optical pulses and electrical pulses A beam arrival time measurement that uses the optical system to measure the arrival time of an electrical pulse is depicted below in Fig 8 2 1 It is essentially the same as the system that was described in the previous chapter for the optical front end of the chicane BPM except that in the case of the BPM the emphasis was on measuring the difference in the arr val times of two beam transient pulses whereas the beam arr val time measurement is concerned with the arrival time of a single beam transient pulse relative to an optical reference When two different beam arrival time monitor systems separated by 60 meters were measured against one another over few minute time scales they had a resolution of 6 fs Over longer time scales they differed by more than 30 fs This was due to the sensitivity of the measurements to small changes in the beam shape 20 Within the BAM front end a portion of the incoming laser light is reflected backwards along the fiber which delivered the light The arrival time of the returning pulse is measured relative to the arrival time of a pulse from the MLO with an optical cross correlator Optical fiber links can also be stabilized with much less expense with an RF technique that utilizes balanced detection of photodet
213. put is not just one frequency it has a bandwidth determined by the design of the filter The more poles and the more bandwidth that are used n the filter the flatter and broader the group delay of the pass band becomes and the wider the mixer output pulse becomes The wider the output pulse becomes the easier it is to sample it with an ADC Given a filter with 4 poles and a bandwidth of 150 MHz at the 3dB attenuation point the output of the mixer has a pulsed characteristic shown in Fig 7 1 3 for five different phase relationships centered about the phase for which the mixer output pulse amplitude is a minimum Each line plotted represents a 1 degree change from the adjacent line Fig 7 1 3 was produced with a simulation of the circuit shown in Fig 7 1 1 using ORCAD software and it is an accurate representation of what is visible on an oscilloscope when the mixer output is measured for different LO phases Mixer output around zero crossing of pickup signal Sampling times Volts oO IN 0 05 0 1 gt N 0 15 0 2 0 5 10 15 20 25 30 35 Time ns Figure 7 1 3 Simulated behavior of the mixer output around the phase for which the mixer output is minimized There is an asymmetry in the behavior but regardless of which sampling time you choose the amplitude changes n a sinusoidal fashion as a function of the LO phase The asymmetrical behavior seen in Fig 7 1 3 arises due to the bandwidth of the fi
214. quency The phase of the intermediate frequency could be measured through digital down conversion This would be recommended if the front end is developed in conjunction with a larger project that also demands low ADC clock j tter but it is too complicated to implement as a novelty system 117 The main concern about the long term supportability of this down mixing to baseband scheme is its uniqueness While the RF front end is cheaper and easier to set up than the optical front end it is the only front end of its kind at DESY Since the optical front end uses a format that is repeated throughout FLASH in the form of the beam arr val monitor it has a higher likelihood of being supported long term If it were not for the robustness and dynamic range afforded by the lower resolution 1 3 GHz phase measurement contained together with the higher resolution 10 4 GHz phase measurement in the RF front end chassis the chicane BPM RF front end would probably not be supported long term Because it has no moving parts the 1 3 GHz phase measurement s more stable reliable and easy to commission than any of the other higher resolution measurements The same cannot be said of the 10 4 GHz front end While the 10 4 GHz front end is more robust than the optical front end it essentially delivers the same information as the optical front end and like the optical front end t requires the precise adjustment of a mechanical stage in order to deliver any measu
215. quipment to the beam within these time scales they cannot directly make use of the short pulses n for example pump probe experiments In pump probe experiments one laser the pump excites some behavior in a sample and a second laser the probe records the behavior of the sample in for example a picture called a diffraction pattern In the THz beamline of FLASH it is possible to simply measure the arrival times of the electron bunches relative to the pump laser pulses when they arrive and then use those measurements to make sense of the data 5 This is like filming a movie with all of the frames taken at random and then later sorting the frames to make a sensible sequence This has been done with 5 fs resolution over a range of 500 fs with the possibility to deal with beam arrival time drifts of several picoseconds For such a measurement the timing jitter of the electron beam must only be kept within the dynamic range of the measurement a requirement which 1s already fulfilled by the present machine Less than a year ago this measurement of the arrival time of the electron beam relative to the pump laser was unprecedented It was previously anticipated that t would not be possible to make such a high resolution measurement with such a large dynamic range It was believed that the entire accelerator would need to be actively stabilized with longitudinal intra bunch train feedbacks so that the arrival time jitter and drift of the FEL pulse
216. r for narrow beams and it will be earlier for wide beams A simulation of this principle for different beam widths is shown in the plot of Fig 6 2 3 84 BPM beam width dependence p x 8 10 16 5 ps cm 15 O 2 N 25 30 0 1 beam width cm Figure 6 2 3 Sensitivity of the chicane BPM arrival time measurement to changes in the width of the beam Based on this plot generated with CST simulations of the average pulse arrival times at the stripline outputs for various widths of the elliptical beam shown on the left hand side of Fig 6 2 1 the dependence of the beam arrival time as measured with the chicane BPM on the width of the beam is 1 65 ps mm for beams that are more than 20 mm wide where the width is given by 30 This is consistent with the 3 3 ps mm conversion factor of a pulse traveling at the speed of light When the beam is less than a centimeter wide this sensitivity drops to lt 0 1 ps mm This is due to the fact that when the length or height of the beam is comparable to the width there is no longer a significant sensitivity of the arrival time to the width of the beam The FWHM bandwidth of the simulation used for this plot was 20 GHz and this corresponds to a beam length of 15 mm In the figure above when the beam width s greater than 15 mm the sensitivity of the arrival time measurement to the beam width starts to become significant The actual length of the beam is much shorter than 15 mm so th
217. r the addition of a seeding laser In sFLASH a short seeding laser pulse needs to overlap with the short electron bunch in order to stimulate the FEL process with a defined wavelength If these two sources are not synchronized the seeding process will not be effective This s why the electron bunch arrival time must be measured and kept under control throughout both the XFEL and sFLASH It is possible to stabilize the beam arrival time at the expense of the energy stability of the beam but this would be unacceptable because the wavelength of the light generated by an FEL depends on the energy of the electron beam If the energy of the beam is not precisely controlled seeding schemes like sFLASH will not work To simultaneously stabilize both the beam energy and arrival time the stability of the beam arr val time and energy prior to a bunch compressor must be measured and the nd v dual sub systems must be controlled The individual sub systems are described n more detail in the following sections 2 1 RF Photo injector The RF photo injector 8 generates the electrons for the machine by shooting laser pulses onto a cesium telluride cathode and then accelerating the electrons that are ejected with a strong electric field in an RF cavity A solenoid around the beam pipe provides additional transverse focusing to counteract the strong space charge forces that push the beam apart Figure 2 1 1 laser cathode Iris F ze IN
218. ravel across the upstream gap as a displacement current which will give rise to a voltage pulse on the upstream end of the stripline Because the stripline 1s terminated on both ends with the same impedance the pulse will split into two equal pulses and travel to each end of the pickup The fraction of the image current continues to travel downstream until it encounters the downstream gap where another pulse is created that is equal in amplitude to the upstream pulse but opposite in polarity This pulse also splits into two pluses traveling to each end of the stripline This is depicted in the lower of the two pictures in Fig 5 3 1 If the velocity of the beam and the phase velocity of the pulse are equal the pulse that was created at the upstream gap will arrive at the downstream gap at the same time that the image current induces the downstream voltage pulse Since the pulses have opposite polarity they will cancel one another out and no energy will be dissipated in the downstream termination 66 Figure 5 3 1 Long tudinally oriented stripline BPM principle of operation The output consists of a bipolar signal with peaks separated by twice the length of the stripline In order to calculate the frequency response of this type of pickup let us model the bunch as a Dirac impulse I yeam O Z 5 4 1 This s an appropriate approximation for the short bunches of an FEL We can use equation 5 1 10 for a stripline as well as for a butt
219. re changes because it is carried out within a single temperature stabilized enclosure but it is possibly more sensitive to beam profile changes than the BAM Such sensitivity to beam profile changes can be a good thing if it is related to the stability of the peak current but it can be a bad thing if it is unrelated Simulations of these effects are presented in Chapter 6 but when more BAMs are commissioned a true comparison can be made If the chicane BPM is used to measure beam energy changes the incoming orbit jitter must be measured using BPMs 35 from before and after the chicane It is important that these BPMs have resolution that is comparable to that of the chicane BPM and that there are not any quadrupoles between these BPMs and the first or last dipole of the chicane BPMs with lt 5 um resolution were installed before and after the first bunch compressor for this very purpose Initial attempts to benchmark the available beam energy measurements are made in Chapter 10 of this thesis but most measurement devices of interest were only commissioned for a couple weeks during the course of these studies In general the measurements that deliver the highest resolution have a limited dynamic range and require a mechanical delay line to accommodate larger changes The designs of these monitors will be presented in Chapters 7 8 and 9 Experience with these monitors gained in the course of producing the measurements in Chapter 10 showed that t
220. red to remote locations without loosing its phase stability This is done with a Master Laser Oscillator MLO that sends pulses along fiber links to end stations At each end station a portion of each incoming pulse is reflected and sent back to the source At the source the arrival times of the reflected pulses can be compared to the arrival times of the new pulses coming from the MLO The length of the fiber is then adjusted until the returning pulse timing matches the sent pulse timing thereby removing the effects of temperature induced timing drifts At this point the principle of the optical synchronization system sounds identical to that of a purely RF synchronization system with only the word MO replaced with MLO and the word cable 134 replaced with fiber The advantage of the optical system becomes apparent when the accuracy with which the detection and transport of an optical signal can be carried out 1s compared to that of an RF signal Table 8 2 1 In terms of phase shifting phase detection EMI and vibration the optical system has performance that is about an order of magnitude better than that of an RF system po Optica RF source phase noise 1kHz 10MHz phase shifting phase detection EMI vibration Table 8 2 1 Comparison of optical and RF systems phase noise detection and etc To summarize the advantages of the optical system e Attenuation s not an issue when transporting an optical signal over fibers e O
221. reement between the RF chicane BPM and the PMT BPM on a shot to shot basis was observed until multiple new ADCs were installed n the crate in which the chicane BPM ADC is installed These new ADCs sent large volumes of data over the crate BUS and caused all of the devices in the crate to suffer from buffer number problems This made it impossible to correlate any data from this crate with other devices in the machine This was not yet a problem when the data from Fig 10 2 1 was taken but for all subsequent data it was a problem When the PMT BPM and RF chicane BPM measurements don t agree the reason is frequently that both measurements have a limited few millimeter dynamic range and when it is exceeded a motor must be moved and the monitor must be re calibrated Whenever the calibration constants for these monitors are not correct the measurements do not agree The results from several day measurements of the 1 3 GHz coarse and 10 4 GHz fine BPM front ends are presented below in Fig s 10 2 2 and 10 2 3 In the first plot good agreement is observed between the BPMs and with the gradient setpoint In the second plot poor agreement is observed and was not identified by the control software Benchmarking RF chicane BPM against PMT BPM 1000 A 500 N Mer BC2 PMT S tk AC We a Ho ulm ACCI setpoint Q H oO BC2 BPM ge A a 500 a oO m u 1000 1500
222. rement at all The 1 3 GHz front end in contrast must be set up once and re calibrated periodically but if the beam moves by several centimeters it will still deliver a measurement and the operator would not have to wait for a minute for the high resolution measurement to scan a mechanical stage in order to find the correct sampling position This can be seen by comparing the dynamic range of the 1 3 GHz phase measurement to that of the 10 4 GHz phase measurement in Fig 7 2 14 There you see less than 5 mm of dynamic range in the fine measurement before a mechanical phase shifter must be adjusted The dynamic range of the coarse measurement is 80 mm with no need for mechanical elements Mixer output scan of vector modulator phase 0 5 a EN ei right fine 2 lt omm range gt 0 _ gt 0 5 0 20 40 60 80 100 120 140 mm 1 2 I coarse 80 mm range 1 0 200 400 600 800 1000 1200 mm Figure 7 2 14 Comparison of mixer outputs for 10 4 GHz phase measurement top and 1 3 GHz phase measurement bottom The 1 3 GHz measurement has a much larger dynamic range than the 10 4 GHz measurement A 1 3 GHz front end measurement of the beam position change resulting from a change in beam energy 1s shown below in Fig 7 2 15 118 EBPM and PM positions mm PMT monitor ACCI setpoint 1 3 GHz front end 0 6 0 8 1 1
223. requency This crystal oscillator would be locked to a GPS standard frequency 56 When the MLO is locked to the MO it will gain the long term phase stability of the GPS standard This is the reason that it is desirable to have an MO MLO lock that does not drift When it doesn t drift then one can be certain that the long term phase stability of the optical reference is as good as the long term stable RF references The setup to accomplish this lock is shown in Fig 8 3 1 The schematic in Fig 8 3 1 depicts one chassis containing two identical circuits the top half and the bottom half are identical The concept was developed by several people from the optical synchronization team but the construction and characterization of the lock was done by the author Each circuit provides two distinct measurements of the relative phases of the MO and the MLO One is a fine measurement mixing 1 3 GHz from the MO with 1 3 GHz generated from an MLO based photodetector signal The other 1s a coarse measurement using 216 MHz that keeps track of which bucket the fine measurement is measuring These measurements of the relative phases of the MO and MLO are used in an ADC DSP DAC feedback loop n order to set the voltage of a piezo fiber stretcher that adjusts the round trip time of a pulse in the laser This adjusts the phase of the MLO relative to the MO Additional signals are provided by the chassis for the monitoring of laser power and amplification of the filt
224. resulting from a chicane the following formulas make use of the above transfer matrix parameters in order to relate a beam energy chirp given by the difference in the energy of the particles in the head of the bunch and the energy of the particles in the tail of the bunch to a change in bunch length o or a change in position spread in the middle of the chicane ox To find the beam width in the middle of the chicane one can use the linear transformation oh E to write x gt zs x En ye x3 2R x080 5 where the middle term is equal to zero because we assume that there is no dispersion upstream of the bunch compressor and therefore no correlation between xo and do We can also write x x R 0 with an assumption that the beam is centered about zero Now with the definition of standard deviation o x x we can describe the beam width in the middle of the chicane o Xo R 2 3 8 Likewise to calculate the bunch length after an accelerator section and a bunch compressor one can use the linear transformation H Resse Roses i Res Res do Og SA a 1 Rss Res T On Ri Ris 2 3 9 This result can be used to write the compression factor C 0 0 0 a term that is used frequently n the following chapter The ranges of values that the Rs6 and R s assume for FLASH and XFEL are listed in Table 2 3 1 along with the expected range of beam positions X and widths DX that a moni
225. rom one end and measure the gain of the amplifier After cutting off a few centimeters at a time the gain curve will begin to become clear For the best noise performance the best place to stop cutting off lengths of gain fiber is when the gain curve starts to become linear Of course the length for the best noise performance may not have the gain that is needed so a compromise has to be reached After the amplifier the laser pulses enter an Optical Delay Line ODL This ODL will move whenever the arrival time of the beam in the chicane changes The fiber entering the ODL is a single mode fiber SMF and the fiber exiting the ODL through the collimator attached to the mobile portion of the stage 1s polarization maintaining PM It is better to put the PM portion on the moving part of the stage because if the SMF is moved the polarization controller will have to be adjusted While the polarization controller actuator is fast enough to react to fast changes caused by vibrating and moving fibers this makes the task of the polarization controller harder than it needs to be For fast polarization changes a micro controller or DSP can be used to adjust the polarization controller based on a measurement with a photo detector of the un modulated EOM output or based on an actual measurement of the polarization of the pulse using an in line polarimeter General Photonics A necessity for such fast adjustments has not yet been seen for any of the systems bu
226. rotron light with a fundamental wavelength of 18 A i BA with K Bol 2y 2 Figure 2 4 1 Undulator magnet and electron bunch producing synchrotron radiation Based on the presence of the Lorentz factor in the above formula it is clear that if the beam energy changes then the wavelength of the fundamental mode of the light will change 17 One of the keys to making the light generated by the undulators coherent as in a laser is to maintain a sustained interaction between the light pulse and the electron bunch over a distance known as the gain length After several gain lengths the sustained interaction creates an energy transfer from the light pulse to the electron bunch causing the electron bunch to break up into microbunches with a periodicity equal to the wavelength of the synchrotron light These microbunches begin to radiate coherently the light generated by each microbunch adds coherently to the light generated by the other microbunches This increases the intensity of the coherent light pulse and increases the energy transfer from the electron bunch to the light pulse furthering the microbunching process When this sustained interaction is maintained over a sufficient number 20 of gain lengths and high intensity synchrotron light s generated the machine is said to be in saturation or lasing Photon pulse WWW gt teren gt eQQ LULL b0e Micro bunching Figure 2 4 2 Interaction of electron beam
227. rrent ol OV Oe 5 1 18 OZ Ot where L and C are the inductance and capacitance per unit length The voltage and current can then be written down n terms of forward and backward propagating waves Z Z V V t V t phase phase 5 1 20 pa 9 04 9 Zo phase 0 V nase where Vphase 18 the phase velocity of the waves given by 56 l phase JIC Finally the characteristic impedance of the transmission line is defined by the ratio between the wave voltage and the wave current and can be written in terms of the line inductance L and capacitance C as L Z J 5 1 22 where for a coaxial line the inductance and capacitance per unit length L and C can be written n terms of the radius of the inner conductor to the radius of the outer conductor g 5 1 21 Me and Zei r 5 1 23 IN ou Fn 2m This solution of equations 5 1 17 and 5 1 18 is very general and does not predict any distortion of the pulse as it travels down the transmission line when amp and u are constants It is instructive to substitute Eqs 5 1 23 into Eq 5 1 21 to see that the phase velocity of a pulse traveling in an in vacuum coaxial line is equal to the speed of light While this is not very important for button pickups it is very useful for the stripline pickups that will be described in the following sections For the button pickups described here Eqs 5 1 22 and 5 1 23 are absolutely necessary when designi
228. rs have the resolution to make such a high resolution cross check It is envisioned to use beam based feedback to provide a small correction to the work that the cavity feedback is already doing Under typical operating conditions the proportional gain feedback is on and the adaptive feedforward is off But in all tests of beam based feedback to date the proportional gain module feedback was always off and the adaptive feedforward was on 20 the beam based corrections were too large to be implemented without the adaptive feed forward and the system was unstable with the proportional gain feedback on This sort of beam based feedback architecture however 24 creates a single point of failure system that is not robust enough for long term operation especially since the high resolution monitors have such a limited dynamic measurement range Ideally multiple systems should be used simultaneously so that if one measurement is out of range another system can step in An architecture that takes beam based and reference tracking information into account in the cavity controller 1s depicted below in Fig 3 1 3 Setpoint FPGA From ADC Cavl Vector Feed forward Matrix Cav2 Injection MO BLM BAM Figure 3 1 3 A desired FPGA algorithm structure incorporating reference injection and beam based information So far there has only been a description of what the module controller can do with respect to amplitude and phase stability b
229. rticle from a straight ahead trajectory l cosa Xofre ly ag 1 4 We can use equations 1 3 and 1 4 to describe the situation that one finds in a magnetic bunch compressor chicane where not one but four magnetic fields are applied two of which cause the beam to deviate from its straight ahead trajectory and two of which bring the beam back to its original trajectory Fig 1 1 Reference orbit in BC2 for 15 18 and 21 deg 0 6 0 5 d gt gt 0 4 x m 0 2 335 0 5 1 1 5 2 25 3 3 5 4 45 z m Figure 1 1 Magnetic bunch compressor chicane High energy particles travel a shorter path than low energy particles The path length of an electron traveling through the symmetric chicane shown above can be expressed by lo 4ra 2 2 d 1 5 cosa and the x position in the middle of the chicane is l l cosa Xzc 2 rsina d tana 1 6 sin a where d is the drift space between the first and second dipoles and dz is the drift space between the second and th rd dipoles If these equations are rewritten n terms of the magnetic field and the momentum of the electrons Eqs 1 1 and 1 3 we have l eB 1 p 4p weil i 2 eb p leB py and 1 6 l eBd 1 eB p Eu p ll I eB py Because high momentum particles
230. s synchronized to the Master Laser Oscillator MLO in the same manner depicted in Fig 3 3 1 a but that is where the similarity ends The injector RF phase s stabilized with a cavity controller that incorporates beam based feedback from a beam arrival time monitor downstream of the first accelerator section but upstream of the first chicane The first accelerator section amplitude is stabilized with a cavity controller that incorporated beam based feedback from an arrival time monitor located directly after the first chicane The first accelerator section phase is stabilized with a bunch length monitor located after the second chicane The second accelerator section amplitude and phase are stabilized in a similar fashion with an arrival time monitor and a bunch length monitor after the second chicane The system shown in Fig 3 3 1 b has the following disadvantages e If an upstream feedback loop fails to deliver acceptable beam stability the downstream loop will start to feed back on noise that is generated some where other than in the cavity it is controlling e The downstream feedback loop has to be slower than the upstream loop in order to avoid instabilities e Itis not able to make use of cross checks monitors that measure the same quantities in different ways e The energy changes due to arrival time jitter from upstream of a bunch compressor are not subtracted from the energy changes measured after the bunch compressor 30 e The
231. s the delivery of laser pulses from the optical synchronization system But before describing the advantages of the newer technology optical system the advantages and limitations of the older technology RF system should be clearly defined 8 1 RF Front end RF synchronization of accelerator facilities has been used with great success since their inception achieving picosecond synchronization between locations separated by several kilometers without the aid of active feedbacks to compensate for cable length changes With the addition of an active feedback on the cable length in loop measurements of synchronization that is better than 100 fs have been made 46 It 1s important to note that this method and any other active point to point synchronization schemes can only synchronize one location with another any locations in between these two points are not necessarily synchronized This is because the waves may travel at different speeds in different segments of the cable There are several problems that may be evident in an out of loop measurement of the performance of an RF cable reflectrometry set up With an active RF cable feedback method a signal of a few GHz is sent from a Master Oscillator MO over a coaxial cable to an end location at which part of the signal is reflected and sent back to the source where the phase of the signal being generated and the phase of the signal returning can be 132 compared The length of the cable can the
232. s the signal amplitude decreases and the slope of the signal decreases the resolution of the one sample point measurement will decrease Changes of the amplitude of the signal will make changes in the slope of the signal and consequently after any change in the amplitude of the signal the measurement will need to be quickly re calibrated or there will be an error in the measurement The amplitude of the signal changes when the beam width changes and when the charge or y position changes The 3 ps error produced by sampling the beam pickup signal 100 ps away from the zero crossing for a 3 change in the charge of the beam is shown below in Fig 5 5 8 75 Half of falling edge 3 charge change Or ae Ideal Case 0 005 gt 0 0 1 5 0 015 0 02 0 025 100 DS IN 265 2 eg 0 47 0 475 0 48 0 485 0 49 time ns Figure 5 5 8 Impact of charge change on single sample point which resides 100 ps away from the zero crossing of the pickup signal Ideally the signal would be sampled at the zero crossing but a 3 ps error 1s incurred by sampling 100 ps away from the zero crossing when the beam charge changes by 3 The influence of these pickup signal amplitude changes on the accuracy of the measurement can be removed through a routine that automatically re calibrates the monitor on a regular basis or through a calibration constant that is updated based on a measurement of the phase y position
233. se result from an energy spread generated in a non dispersive section like the accelerator section When the path of an electron beam is influenced by a magnetic field the beam emits synchrotron light With wavelengths corresponding to the length scales longer than the bunch length the bunch radiates coherently and for length scales shorter than the bunch length it radiates incoherently ISR is generated through a random process and cannot therefore be corrected It increases for higher energy beams 14 making it less significant in the first bunch compressor and more significant in the second bunch compressor Unlike most other emittance increases this acts solely on the slice emittance The CSR is much more powerful than the ISR it is correlated along the bunch and acts strongly on the projected emittance The power of the CSR increases in proportion to the bunch length raised to the power 1 3 meaning that the power of the CSR increases for shorter bunches As the CSR and the electron bunch co propagate in the bends of the chicane the CSR can catch 15 up with and then interact with the electron bunch giving it an energy spread which is correlated along the bunch For an rms bunch length oz dipole length Lg and dipole bend radius R the CSR induced rms relative energy spread per dipole for a Gaussian bunch under steady state conditions 1s Nr L AE pcp ORG 7 2 3 4 Where N is the number of electrons re is the classical
234. shown in Fig 6 3 2 Over a few millimeter range it was verified with oscilloscope measurements Using an estimation of the macro beam as a collection of pencil like beams that can be arranged in the x y plane the signals produced by each beam slice can be summed together as described in the Greens function integration of Sect 6 1 This provides an estimate of what the pickup output would look like for a charge distribution which is tilted in the x y plane In the first step of this integration process the y position sensitivity data from Fig 6 3 2 was used to calculate the amplitudes of the signals that would be induced on the pickup for each slice of a 1 cm wide 30 beam which 1s tilted by 5 degrees n the x y plane From Fig 6 3 2 it is clear that the further the beam is away from y 0 the stronger the effect of an x y tilt will become 86 iad Vertical position sensitivity 90 80 705 60 Bottom pickup Top pickup X x EN 4 0 he A 30 N Signal amplitude V y position mm Figure 6 3 2 CST simulation of the sensitivity of the chicane BPM signal amplitude to changes in y position of a pencil like beam with a charge of 1 nC For y 0 the beam is in the middle of the vacuum chamber In Fig 6 3 3 the position of the beam slice is shown on the x axis and the signal amplitude induced on the pickup by the beam slice is shown on the y axis The scaling of the amplitude varies according to the numb
235. significantly attenuated by transmission over long fibers RF signals sent over long cables are e While optical signals are immune to electro magnetic interference from noisy devices RF signals are not The prevailing opinion at the present time 1s that pushing RF components below the 10 fs level would create a cost explosion as expensive cables and couplers are required The alternative is to develop the optical technology in order to easily achieve 10 fs performance and have the possibility to later push to the sub fs level If the optical system is made widely available to other FEL facilities the costs could decrease due to economies of scale This argument does not apply however if the target synchronization accuracy is above 20 fs and the distances involved are much less than a kilometer This is a regime for which RF arrival time monitors could still be considered Table 8 1 2 shows a cost estimate of an RF arrival time monitor with a phase stabilized cable Engineered enclosure 600 100mcable 600 Table 8 1 2 Rough cost estimate for a 20 fs resolution RF arrival time monitor with an RF reflectrometry setup 8 2 Optical Front end The optical synchronization system was developed in order to lock the beam to an optical reference signal which can be delivered to remote locations with sub 10 fs accuracy 2 To this end the beam arrival time relative to the optical reference must be measured and the optical reference must be delive
236. sis drifts relative to the 8 meters of fiber which is wound up inside of the BAM the arrival time of the MLO pulses at the chicane BPM front end will not be stable and the arrival time measurement will no longer be usable The optical length of one meter of a standard optical fiber will drift by 60 fs deg 8 meters will drift by 480 fs deg When the chicane BPM front end is used purely as a beam position monitor this drift would have no impact but if the BPM front end is also used as a beam arrival time monitor this drift will cause systematic errors n the measurement The drift can be avoided by using a special and more expensive type of fiber called PSOF Phase Stabilized Optical Fiber At the moment it s only available from one firm Furukawa Compared to the standard SMF 28 fiber the optical length of PSOF does not change significantly with temperature While using kilometers of this sort of fiber would be prohibited by the 60 150 EUR meter cost 8 meters of PSOF is considerably less expensive than building an additional 25 000 EUR actively length stabilized optical cross correlator fiber link Upon entering the chicane BPM chassis depicted below Fig 7 2 3 the polarization is adjusted with a polarization controller from BATI labeled PC in Fig 7 2 3 and acrobat polarization controller in Fig 7 2 1 While the polarization of the light from the fiber link can be adjusted in the fiber link stabilization chassis the polari
237. slices Each slice will couple to the pickup and a pulse will travel to the left and to the right If you detect the arrival times of each slice on one side of the pickup the time elapsed between the slices will be larger on one end of the pickup than the time elapsed between the slices on the other end of the pickup If all of these slices are added together on each end of the pickup one finds that the periods of the signals on the opposite ends of the pickup are different but the arrival time of the zero crossing is the same as it would have been for a non tilted beam Pictorially this is represented in Fig 6 5 3 VN NW NN X Figure 6 5 3 Illustration of the spacing of the wavelets produced by beam slices as they are transported on the pickup for a tilted beam In the numerical simulation of this effect Eqs 6 1 6 6 1 7 the beam charge is broken up into many small slices Each slice induces a wavelet on the pickup The wavelets are added together to generate a transient for the whole beam When the zero crossings of the transients for the tilted beam case are compared to the non tilted case it is apparent that although the amplitudes of the beam transients are affected by the beam s tilt the phases of the pulses are not both measurements return the same value for the beam position and there would be no systematic error from this effect The story changes however when the charge distribution is asymmetric 6 6 Asymmetric C
238. small energy change Beam arr val time change corresponding to a small energy change Measurements of the beam arrival time changes resulting from scans of the RF GUN and laser phases in the photo injector Fiducializing the mechanical phase shifter potentiometer with the vector modulator The curvature of the BC2 BPM measurement results from the problems with the mechanical phase shifter Comparison of mixer outputs for 10 4 GHz phase measurement top and 1 3 GHz phase measurement bottom 1 3 GHz front end beam position measurement as a function of beam energy 1 3 GHz front end beam arrival time measurement as a function of beam energy Mach Zehnder Electro Optical Modulator EOM used to sample the amplitude of an electrical signal Mach Zehnder Electro Optical Modulator EOM used to sample the slope of a beam transient pulse Measuring the amplitude of the laser pulses with an ADC that is clocked with a signal that is generated by the laser pulses themselves Calibrating the arrival time measurement requires scanning the arrival time of the laser pulse about the zero crossing of the beam transient pulse Chicane BPM optical front end schematic Length stabilized fiber link concept The layout of the fibers in the top layer of the optical front end chassis for the chicane BPM The side view of the optical front end chassis Effectiveness of active temperature control in the tunnel Beam arr val time measurement with lengt
239. so better agreement between the PMT monitor and BPM for the first setpoint change which took place just prior to the 20 hour In the 3 setpoint change at the 40 hour it appears that the chicane BPM has exceeded its dynamic range and there is better agreement between the out of loop vector sum and the PMT monitor energy change 0 3 0 10 15 20 25 30 3 40 nours Figure 10 3 1 Fine HF front ends position measurement and photomultiplier tube position measurement 10 4 Optical BPM Measurements The optical front end was only in operation for a few days over the course of the last machine run and it was possible to calibrate it and check the resolution of the measurement There were two separate optical front ends constructed One had an older version of the optical delay lines that have a limited lifetime It did not have an active temperature control system and it was installed outside of the tunnel after a 30 meter long 157 cable connecting the pickup to the EOMs in the front end It measured 12 fs arr val time resolution and 4 um position resolution The other front end had more robust delay lines and an active temperature control system It was installed n a lead shielded box with 2 meter cables connecting th
240. stribution of the bunch has folded over on top of the bunch for this 10 degrees off crest simulation When the head of the bunch overtakes the tail of the bunch in the chicane this is frequently referred to as over compression Over compression is typically employed for beams with a non linear energy chirp The end result of this process is that the bunch is longitudinally compressed the energy chirp increases and the energy spread remains constant 11 Bunch compressors are frequently designed to accommodate a range of energies and compression schemes which are defined to first order by parameters denoted Rss and R s after their location in a six dimensional transfer matrix used to calculate the beam transport 12 In the introduction analytic formulas for these parameters were derived for a symmetric single chicane Physically they relate the change in position to the change in energy deviation 6 according to OZ Ox GF Rus T 2 3 2 They are related to one another by S R Ra ls Jas 2 3 3 o TCs where the integral is along the reference trajectory s and r s is the bending radius of the magnets From Eq 2 3 3 we can see that for smaller bending radii the R56 increases and for drift spaces where r is infinite the Rss vanishes It 1s useful to be able to make a quick estimation of the energy chirp of an electron bunch after an accelerator section and calculate the resulting bunch length change or x position spread c
241. stron from Thomson Tubes Electronics 1s used to generate the 1 3 GHz for the acceleration modules A simplified depiction of the basic structure of a klystron is shown in Fig 2 2 1 Buncher Catcher cavity cavity Density of Electrons Cathode Anode Collector L Te A Er Uo Modulator Output to input Accelerating t l structure Figure 2 2 1 Basic structure of a klystron Electrons travel from the cathode to the collector and along the way they are bunched and used to excite waves that are sent to the accelerating structure Large amounts of effort have been invested into actively controlling the amplitude and phase of the RF generated by the klystron with digital signal processing feedbacks and feedforward loops on Field Programmable Gate Arrays FPGAs 11 RF pickups inside of the couplers of the modules detect the 1 3 GHz field and the signals are converted to a lower frequency signal that can be sampled with Analog to Digital Converters ADCs FPGAs execute a real time algorithm to determine the phase and amplitude of the signals from the ADC outputs and use feedback and adaptive feedforward loops to send signals to actuators that adjust the phase and amplitude of the RF n the cavity The goal of this is to make it possible that every bunch in every bunch train experiences the same accelerating field gains the same amount of energy and behaves in the same manner in the bunch compressors These efforts are described in more detail in
242. t It was suspected that a significant portion of this ringing comes from the splitter combiner that was used in the 73 distribution of the signal but when it was replaced with a splitter combiner with a much lower insertion loss the ringing was the same suggesting that the ringing is primarily generated in the pickup For the construction of the XFEL pickup it 1s recommended that extra time be allotted for the development of a hollow pickup antenna that can be suspended between the two vacuum feed throughs without the need for the ceramic support rings When measured with the 10 GHz optical front end setup a slope of 1 5 V ps is measured This is less than the 6 V ps predicted in the 50 GHz simulation and this is due to the bandwidth limitations of the RF components that were used in the distribution of the signals Nevertheless at this stage it is not desirable to further increase the slope of the signal because of the limitation that high signal slopes impose on the dynamic range of the measurement Bypassing these limitations in order to achieve the highest resolution possible will be discussed in the context of the front end measurement setup in Chapter 7 Good agreement was also observed between a simulation of the pickup s frequency domain response and a measurement with a network analyzer Although in the network analyzer plot Fig 5 5 6 left the blue curve is lower than the green curve the poor performance s due to oxidation
243. table is generated from a setpoint given by a user and an older feed forward table 11 21 Setpoint I phase amp Vector Feed amplitude sum forward detection Matrix cal bration new phase amp amplitude detection number of cavitites Figure 3 1 2 Simplified block diagram of the cavity regulation routines on the FPGA The limitations of the current digital processing system are primarily due to cavity amplitude and phase measurement resolution and drifts the field measurements inside the cavities are not completely accurate When an inaccurate cavity measurement is used to regulate the cavity parameters the measurement drifts and fluctuations can turn into cavity drifts and fluctuations which in turn become beam drifts and fluctuations The drift problem of the cavity amplitude and phase measurements has recently been approached through reference injection and reference tracking techniques 21 Reference tracking refers to the use of a reference signal to measure the out of loop performance of a regulation loop Reference injection refers to the use of an out of loop measurement to actively improve the drift performance of the loop In the system under consideration the MO signal was the reference used to measure the out of loop performance of the vector sum of the phases and amplitudes of all of the cavities This sum drifts due to the influence of temperature changes on the RF cables and circuits that make the cavity
244. ter With the RF front end described in the previous chapter the measurement of the phase of the MO relative the arrival of the pickup signal delivered over a 30 meter long cable could be done with a resolution of 20 fs rms without taking into account any drift or noise on the MO signal The arrival time resolution can be halved by using a short cable from the pickup to the front end chassis halving the noise picked up over the 30 meter cable As good as these numbers may sound neither address the problem of delivering a stable MO signal over a large distance They also do not address the question of costs RF components are frequently cheaper than optical components but low drift RF cables can be much more expensive than optical fibers and will have much higher attenuation Table 8 1 1 cost meter drift meter attenuation km degreeC 1 3GHz Phase Stable Optical Fiber PSOF 25 EUR 0 02 dB Single Mode Fiber SMF 0 20 EUR 50 60 fs 0 02 dB Table 8 1 1 Costs and performance of RF cables and optical fibers Data from Henning Weddig When contained in a bundle of 6 fibers e While there are RF cables with better drift properties than SMF fiber amplifiers will frequently be required to compensate for their attenuation and amplifiers 133 tend to make significant contributions to drifts as well Couplers are another source of drifts and can become very expensive when tolerances are tight e While optical signals are not
245. the MLO Once there the arrival times of the returning pulses are measured relative to the arrival times of pulses that have just been generated by the MLO Fig 7 4 2 This measurement is used to adjust the path length of the light with an Optical Delay Line ODL and with a piezo fiber stretcher Piezo stretcher lt gt long fiber Loe ODL y feedback Figure 7 4 2 Length stabilized fiber link concept The arrival times of pulses from the MLO are compared to the arrival times of pulses that are reflected within the timing sensitive device This information is used to adjust the length of the fiber with a piezo fiber stretcher and an optical delay line ODL 125 For beam arrival time measurements it is important that the arrival times of the pulses from the MLO are constant For beam position measurements this is not important Stabilizing the arrival times of MLO pulses requires the stabilization of the optical length of the fiber over which the pulses are transported to the beam arrival time measurement BAM front end While the chicane BPM can be used as a special sort of BAM that is sensitive to the width energy spread of the beam the BAM application of the optical front end will be reserved for the next chapter At the end of the link upper right corner of Fig 7 4 1 within a nearby BAM chassis containing the link end the 5 10 mW pulses from the fiber link are amplified in a 60 cm long stretch of 8 um Erbium doped fib
246. the chromatic effect of the quadrupole is applied to an offset beam the resulting tilt 1t 3 38 degrees in the middle of the bunch compressor When the chromatic effect is off the tilt is only 0 2 degrees 46 Energy Chirp Off of Gun X transport through ACC 1 0 03 4 75 0 025 0 02 4 73 q 0 015 4 72 0 01 Z 4 71 S 0 005 4 7 4 697 0 a 4 68 0 005 4 67 l 0 01 ae 2 0 2 4 0 2 4 6 8 z mm cavity number Chromatic Effect of Quads slope 3 3851deg No Chromatic Effect of Quads slope 0 18971deg 6 5 4 L E 2 2i 2 D x oS 0 a gt 2 4 L 5 8 6 4 2 0 2 4 8 6 4 2 0 2 4 x axis mm X axis mm Figure 4 3 3 Simulation results for energy spread of the beam coming out of the injector top left the horizontal beam path through the accelerating module top right and the resulting tilt n the first bunch compressor with bottom left and without bottom right the chromatic effect of the quadrupoles The beam was 8 degrees off crest in the first accelerator section Table 4 3 1 summarizes the contributions of the various effects to the tilt stimulated in the bunch compressor ACCI entrance ACCI entrance ACCI entrance No offset 3 mm offset gun energy dist added No angle 15 mrad lt same orbit degrees degrees degrees Dispersion after inj 0 00 2 38 2009 Wakefields Inc 0 00 2340022 Coup
247. the measurement shown n Fig 8 3 7 no one was in the room and the temperature of the chassis changed by about 0 1 degrees over the cours of several hours This amount of temperature change typically causes 10 fs of phase measurement drift based on measurements shown in Fig 8 3 5 At least 60 fs out of the 77 fs drift seen in the out of loop measurement n Fig 8 3 7 can be accounted for by drifts of the laser amplitude The laser amplitude drift was 0 3 pkpk over the course of the measurement and this would cause 60 fs of phase change in the signal emerging from a photodetector according to measurements performed in 47 Some of the drifts observed in the out of loop measurement can also be accounted for by 0 03 degree change of the photodetector temperature If the temperature of the photodetector changes by 1 degree the phase emerging from the photodetector will change by 340 fs 48 Based on the result from 48 0 03 degrees photodetector temperature change would cause 10 fs of phase drift 143 100 Fast timing changes wo drifto 16 2 fs slow drifts pkpk 77 fS 710 05 Out of loop RF drift measurement fs 9 yoe uoneuen Sunjesadwe 0 05 Fast temp changes wo drifto_ 0 01 deg slow drifts pkpk 0 11 deg l i l 12 00 18 00 00 00 06 00 12 00 Correlation between Out of Loop measurement and the Laser Amplitude Change 75 Out of loop RF drift measurement fs aap amod 10 998
248. the phase dependencies of both the incident and reflected pickup pulses are contained within the 100 ns FWHM pulse that emerges from the mixer This however could not create the bump seen in the measured mixer output signal In a simulation where the RF signal is composed by two signals with different phases when the LO phase is scanned the mixer output behavior is still sinusoidal without any distortions The only mechanism that could create this bump phenomenon is if there are two LO signals with different rates of phase change One LO for example has a phase that changes at twice the rate of the other LO This could happen if a harmonic of the RF and LO is in the mix A continuous wave simulation of this concept 1s shown below in Fig 7 1 7 In the simulation the RF signal is composed by two waves that have the same frequencies but one signal has twice the phase of the other When the amplitude of the mixer output is sampled at a single point in time a distorted sinusoidal pattern will be measured at this sample point as a function of the phase of the LO In Fig 7 1 7 the LO phase dependence of the mixer output is plotted for three different sample times amplitude of mixer output as LO phase is changed au f f 3 bo J Sj j J J 0 5 f a X fs 3 N f BEN J x x O 1 00 200 300 400 500 600 700 800 degrees Figure 7 1 7 Mixer output when the RF signal is composed of t
249. the setpoints of the first accelerating section and the third harmonic module so that the timing jitter 1s minimized Presently this sort of optimization 1s carried out with particle tracking code but analytic solutions offer a more global picture and more flexibility in terms of their predictions for different machine configurations Additional bunch compression stage If the first bunch compressor is followed by an additional accelerator section and bunch compressor a few modifications need to be made to Eq 18 n order to describe this 169 situation First the equations for the energy and energy chirp after the second accelerator stage are written E z Z V cos k Z E z E 2 Vk sin k Z p Case 1 E lt lt E and Er lt lt E This is the case for present LCLS operation If the second bunch compressor is located at a higher energy than the first one and the energy chirp for the second chicane is much larger than for the first one then the approximation from Eq 18 can be applied iteratively to the second bunch compression stage just as it had been applied to the first bunch compression stage Case 2 2 0 This was the case for operation of FLASH without the 3 harmonic cavity For this case one can make the approximation E z E z V E z E 2 Neglecting the non linearity of the compression process the path length through the second chicane is written E le T E L Rs6
250. titutes a single point of failure an unfortunate design flaw for a system that requires a high level of robustness A feedback architecture that prevents these conditions is described here along with a description of the relative contributions various machine sections bring to beam arrival time jitter 3 1 Baseline Control Proportional gain feedbacks and adaptive feedforward loops that utilize measurements of the cavity fields in order to stabilize the cavity fields have been the workhorses of FLASH beam energy stabilization since its inception 19 while beam based feedbacks feedbacks that utilize measurements of beam parameters to control the cavity fields have only been tested briefly 20 Additional improvements in the feedback architecture involving an enhanced low frequency gain profile and reference injection to reduce drifts have also only recently been tested 11 21 The relative merits of these systems will be described below along with the limitations of what is currently available 19 Feedforward s a term describing an element n a control system that delivers commands in a pre defined way without responding to how the system reacts A fixed setpoint table that takes into account various calibrations done at an earlier time is an example of feedforward After the effects of Lorentz force detuning cavity detuning and imbalances in the actuator chain have been calibrated away over weeks of studies in the absence of beam pro
251. tor in the middle of the chicane would have to measure to find 13 Chicane Rss mm R c mm X mm DX mm range 3 sigma FLASH BC2 140 228 284 358 FLASH BC3 14 84 100 250 0 150 XFEL BCI 500 600 0 400 XFEL BC2 200 300 0 400 Table 2 3 1 Rss and R s values for FLASH and XFEL the corresponding dynamic apertures of the chicanes X range and the position spreads of the beam within the chicanes DX The X range starts at zero for all of the bunch compressors because it is desirable to allow for operation with the compressors off Whereas at LCLS Linac Coherent Light Source at SLAC 13 another FEL facility the entire bunch compressor beam pipe was placed on motorized movers in order to accommodate different compression modes the FLASH and XFEL bunch compressors have wide and flat vacuum chambers that do not move The reasons for building chicanes that have adjustable properties are twofold it is sometimes desirable to turn the chicane off for a different mode of machine operation and it may be desirable to have the freedom to independently adjust the bunch length and energy chirp emerging from the chicane The limitation of the adjustable range of a bunch compressor is not determined by the feasibility of constructing large aperture diagnostics which are the subject of this thesis but instead by the feasibility of constructing dipole magnets with a high field quality over the entire dynamic aperture Rectangular bends are
252. tor segment of length L For longitudinal wakefields 36 w fE s ds where E is the electric field parallel to the beam direction For transverse wakefields per unit transverse offset Au one can see the dependence of the wake on the offset of the beam from the center of the chamber wi JE reo Bla Convolving the wake functions with the longitudinal charge distribution g s will give the wake potential 41 o0 V s ws s q s ds o Integrating over all slices ds gives the total energy change of the bunch due to the wake L s 7 anas ves 00 This energy change amounts to a transverse kick or a longitudinal energy loss In general the transverse wakes are stronger for long bunches and the longitudinal wakes are stronger for short bunches The primary wakefield effect that one would see in the injector comes from the geometric wakes of the accelerating cavities 37 w s 344 exp s5 S V pC module w s 1000h i SS Jexp y s S vipc m module where sg 1 74e 3 m and s 0 93e 3 m For a beam that travels directly down the center of the acceleration module these effects are negligible but when large orbit deviations are present the effect can become significant Nevertheless without any other effects geometric wakes in the first accelerator section could only produce a tenth of the tilt that it is observed on the screen in the bunch compressor for a large 4 mm orbit
253. tream accelerating module 153 10 2 Photomultiplier Tube Monitor The photomultiplier tube monitor PMT provides measurements of the beam position that have a resolution that two to six times that of the 10 4 GHz front end of the chicane BPM In Fig 10 2 1 the chicane BPM is labeled as EBPM Energy BPM The arrival time is plotted in blue and the beam position is plotted in red 100 I mean 214 2887 rms 42 5809 300 a 400 0 10 20 30 40 50 60 70 80 90 100 seconds 800 mean 612 3209 rms 55 4429 400 0 10 20 30 40 50 60 70 80 90 100 seconds 0 I I I I mean 360 2359 rms 73 6827 200 gt goo V V JY vn IN Ly Val V A 600 0 10 20 30 40 50 60 70 80 90 100 seconds PMT 73 um RMS EBPM 55 um 750 e ee e L 700 ee e 9 A 650 e 009e o 1 e gt E we ce 3 3 A J ns a 600 ose 8 8 gt e Sa A Ru e o o m LU e oe 550 o e amp 500 e 450 550 500 450 400 350 300 250 200 150 Figure 10 2 1 put BC2 um EBPM arrival EBPM position PMT position C Gerth Correlation between the measurements of the beam position in the chicane taken by the chicane BPM labeled EBPM and the photomultiplier tube monitor PMT 154 Good ag
254. tude of the signal over the full range of the pickup the amplitude and signal slope remained constant over the full range of the pickup Fig 6 7 1 This implies that this averaging problem will be difficult to detect unless one s measurement integrates over hundreds of nanoseconds instead of over a picosecond as in the zero crossing sampling scheme which will be described in the following chapter BPM slope 1 6 1 5 2 gt L 1 4 i O S fi o 1 37 Ff Ry a ae 4 i 2 ae m O x TP r i tre 7 O 412 SA5 Tk o D 2 a gt an a f a le o Er Yn 41 0 9 12 13 14 15 16 gradient MV m Figure 6 7 1 Slope at the zero crossing of pickup signal over full dynamic range of monitor For low gradients the beam was beginning to scrape on the edge of the vacuum chamber 94 The problem that could ar se due to th s averaging effect s that the average of the positions might change when the second dipole current changes or when the beam position or angle in the beam pipe is altered While a true simulation of this sort of effect would require high frequency analysis of the entire chicane vacuum chamber one can be relatively certain that for measurements that integrate over a picosecond this averaging wake effect should be negligible For measurement techniques that integrate over hundreds of nanoseconds modes
255. uency the intensity of the radiation falls off exponentially If one integrates over the full emission angle 0 1 y and bending angle w 18 one 1s left with the number of photons emitted by a bunch as a function of frequency o 148 N vYC gr Ks dx 9 2 3 ST olo where Cy 3 967e16 photons sec mrad A GeV 36 The Hamamatsu R5900U 00 M4 photodetectors have a quantum efficiency of greater than 10 percent in the frequency range between 300 and 500 nm Between 250 and 550 nm the efficiency is greater than 5 percent This is centered about the critical frequency of the synchrotron radiation of 411 nm Using Eq 9 2 3 with charge O le 9 C and energy E 0 130 GeV the total number of photons in the frequency range to which the photodetector 1s sensitive is then N hv e 250nm gt 550nm 9 105 9 2 4 This does not take into account the number of photons which are cut away by the aperture of the optics leading up to the photodetector To find that number one must integrate over a smaller solid angle For an aperture of 13 6 mm at a distance of 1 6 mm from the source the number of photons that reaches the detector is 7e6 between 250 and 550 nm and 5e6 between 300 and 500 nm One can then estimate that the maximum number of photons that could be detected with a quantum efficiency of 10 is N hv e 250nm gt 550nm 6 10 9 2 5 A cross check of th s result was provided by code from 55 After th
256. uld be achieved with the 10 4 GHz down mixing scheme was 6 um with a 30 meter cable and 4 um with a 2 meter cable delivering the pickup signals This was measured by splitting the signal from one pickup output and measuring the difference between the outputs of the two arms of the measurement These numbers ignore the drifts n the measurement which can be 15 um over a day and 50 um over a week These drifts are due to cable length changes and would be eliminated if the thermally stabilized chassis is installed in the tunnel Due to reflections in the pickup the monitor must be calibrated by scanning the position of the beam and not by scanning the LO reference phase Corrections to the calibration factors need to be calculated based on measurements of the vertical beam position and of the beam charge 7 3 Optical front end concept The fundamental principle of the optical front end is the same as that of the RF front end in that for both types of front ends a low resolution phase measurement is used to set the position of a phase shifter that keeps a high resolution phase measurement in range As in the RF front end four different phase measurements are done in this optical front end two of which have a lower resolution and two of which have a higher resolution The resolution of each measurement is adjusted through the manual application of attenuators limiters or filters to the signals from the pickups If a large attenuator or a low pass filter
257. ure controller of the chassis could become unstable and need to be re commissioned Peltier elements were used instead of a heating mat because they can cool as well as heat and have a faster reaction time to changes in the control voltage While it is possible to stabilize the temperature of a chassis by simply making it warmer than everything in the room this technique was not chosen because if too many warm devices are installed in an air conditioned room the temperature stability of the room will be more difficult to manage Not everything in the chassis is so sensitive to temperature changes These less sensitive elements are installed on the lower level of the chassis Aside from a voltage regulator and a circuit to control the DC motor for the mechanical phase shifter the lower level contains a circuit that takes the sinusoidal 108 MHz reference signal and generates a 108 MHz NIM level square wave clock signal for the ADC This was done with an AD9510 clock divider evaluation board The clock divider generated a TTL level square wave The offset voltage of this signal was adjustable with an RF bias voltage circuit element and a potentiometer Lowering the bias by a couple of Volts made the TTL signal appropriate for the NIM level clock input of the Struck 100 MHz 14 bit ADC The RF front end chassis was installed in a rack that was adjacent to a patch panel containing signals sent from the master oscillator MO over 5 meter long cables and si
258. used and as the field quality deteriorates at the outer limits of the dynamic aperture chromaticity and higher order dispersion may have to be corrected due to the quadrupole and sextupole field errors contained therein With a chicane on movers the field tolerances on the dipole magnets are reduced and it 1s easier to add quadrupole magnets to the chicane in order to correct the chromaticity of the beam The energy spread of a bunch is unaffected by the bunch compression process and can be used to further compress the bunch in multiple bunch compressor stages until a minimum bunch length is reached The minimum bunch length for a given energy spread is taken from the second term of Eq 2 3 9 Eo Ras es O min E f It is not however always desirable to reach this value in one bunch compressor alone Bunch compression is not typically done in only one stage because of non linear energy spread wakefield and space charge issues At FLASH the bunch is shortened two times once at 130 MeV and again at about 460 MeV When the bunch is highly compressed the space charge forces become stronger and force the beam to expand When the beam is not longitudinally compressed enough the non linearity of the energy 14 spread will be larger and the transverse wakefields will be stronger Transverse wakefields are the reason that the bunch is partially compressed early in the machine and the issue of space charge is one of the reasons that the bunch
259. ut we have not given a description of what it must do in terms of beam arrival time stability This is determined by the relation of the accelerator RF parameters to the bunch compressor parameters and this is described in the following section 3 2 Arrival time Changes after a Bunch Compressor The equation 3 2 1 gives a good representation of how an incoming arrival time jitter dio is altered by transport through an accelerating module followed by a bunch compressor Fig 3 2 1 2 2 2 gt PED Rss Og 9 3 2 1 C Cy A C Cok 25 O 4 Re Aao 8 Ma A Ea 2 2 Figure 3 2 1 Transformation of arrival time jitter with an accelerator section followed by a bunch compressor In Eq 3 2 1 Rss is the longitudinal dispersion constant of the chicane C is the compression factor of the bunch defined in Sect 2 3 A is the amplitude of the upstream accelerating voltage is the phase of the accelerating gradient and do is the arrival time jitter upstream of the accelerator section It makes several approximations the bunch is short relative to the wavelength of the RF the initial energy is small compared to the energy after the accelerator section the incoming energy chirp is small compared to the energy chirp gained in the first accelerator section and the jitter is statistically uncorrelated It was first published n 25 and a derivation is written in the Appendix A The first thing to notice about the equation reading it fro
260. wo signals with the same frequencies but where one signal has twice the phase of the other The mixer output amplitude for three different sample points are plotted Harmonics do emerge from both the LO frequency multiplier and from the band pass filter output The frequency multiplier data sheet warns of harmonic content and band pass filters typically open at harmonics of the pass band When the beam position 1s 103 changed the phase of the harmonic coming from the band pass filter w ll change at twice the rate of the fundamental This is consistent with the bump s beam position dependence The harmonic could be removed with a low pass filter with a cut off starting around 15 GHz While such filters exist they are fairly specialized and one had not been purchased in time for the experiments presented in this thesis Consequently in subsequent sections the distortion is simply ignored and a sampling point is found that does not reside on or near the signal distortions Frequent calibrations measurements of the signal slope about the zero crossing also negate any errors this effect could occasion Bumps and harmonics aside the question of the accuracy of the calibration should be approached from a theoretical standpoint Because the bandwidths of the filters used in the scheme are up to 400 MHz wide reflections within the pickup are mixed with the LO along with the initial beam transient The down mixing technique then measures the sum
261. y 0 3 and the energy measured with the chicane BPM changed by a comparable amount In Fig 10 4 1 the beam energy measurements done with the optical front end of the chicane BPM are compared to those done with the PMT monitor and for a time of flight energy measurement done with two BAMs one before the first bunch compressor and one after the last bunch compressor Energy change over bunch train measured with optical front end of chicane BPM Energy deviation 0 100 200 300 400 500 600 700 800 Bunch number Figure 10 4 2 The beam energy was changed by 0 3 with the accelerator gradient setpoint and the beam energy measured by the chicane BPM changed by a comparable amount The beam was outside of the measurement range of the BPM at the end of the bunch train 159 reality check PMT setpoint oe EOM BPM d 2 BAMs i 0 1 p rq gt pr i i p Ta ae ji i oe _ i e pu re Eo a 4 a ole 0 05 l l 131 3 131 32 131 34 131 36 131 38 131 4 131 42 131 44 131 46 131 48 ACC1 gradient Figure 10 4 3 Optical EOM front end chicane BPM measurement and photomultiplier tube PMT BPM measurement along with a time of flight measurement involving two BAMs and a line showing how the setpoint of the accelerating gradient predicted a beam energy change of 0 1 The other measurements showed an energy change of 0 15
262. zation that is appropriate for the BAM EOMs will not necessarily be appropriate for the chicane BPM EOMs This is the reason for the in chassis polarization controller It 126 can be placed anywhere prior to the transition from single mode fiber to polarization maintaining fiber Because there is an ample power level coming into the chassis the polarization controller was placed before the amplifier instead of after it The amplifier is a 60 cm long stretch of 4 um diameter Erbium doped gain fiber which is pumped from both directions It amplifies the 3 picosecond long 6 mW pulse up to 200 mW This sort of large amplification would not have been advisable if the pulse length had been shorter than a picosecond When a short pulse is amplified above 80 mW unstable distortions in the pulse shape will be created that could have an impact on the accuracy of the sampling scheme In order to achieve the best gain in the fiber it is advisable to wind it over the space with a large radius of curvature wrapping it around a cylinder with a small radius can degrade the amplifier s performance Another consideration in amplifier design is noise To get the highest gain with the least amount of noise one should try to minimize the length of the gain fiber while maximizing the output The rules governing this optimization vary depending on the type of gain fiber that is used so if time allows it 1s good to start with a longer stretch of gain fiber pump it f

Measuring the Electron Beam Energy in a Magnetic

Contents

Download Pdf Manuals

Related Search

Related Contents