Home

Data Transmission at High Rates via Kapton Flexprints for the Mu3e

image

Contents

1. 29 To evaporate aluminum with a laser it needs to be heated above its boiling point at 2519 C 63 According to 65 the reflectivity at a wavelength of about 1 um is in the range of 0 93 and at 10 um it is approximately 0 98 Therefore the fibre laser with a wavelength of 1 06 um is used to deposit the maximal possible amount of energy in the aluminum 4 4 Structure Sizes 4 4 1 Impedance Calculations In a first step the structure sizes are calculated which are needed to match Za g 100 Q for minimal reflections In the tables below several calculations are presented based on the formulas introduced in A dielectric constant e 3 4 and an aluminum thickness of 12 um have been assumed trace width w um trace separation s um Zaig Q 10 10 328 50 50 376 100 150 468 150 150 428 Table 4 2 Calculation of the impedance for coplanar striplines on 25 um Kapton Coupling constant a 0 assumed trace width w um trace separation s um Zaig Q 35 60 100 80 100 53 100 150 38 150 150 no value Table 4 3 Calculation of the impedance for differential microstrips on 25 um Kapton Equation 3 5 not valid trace width w um trace separation s wm Zaim Q 80 100 101 100 150 91 150 150 63 Table 4 4 Calculation of the impedance for differential microstrips on 50 um Kapton 30 The comparison shows that much larger structure sizes suffice to achieve a differential impedance in the range of 100 Q with
2. J Phys IV France 12 8 3 26 2002 A P Clark Principle of Digital Data Transmission Wiley 1983 G Cariolaro Unified signal theory London Springer 2011 M Werner Signale und Systeme volume 3 Vieweg Teubner 2008 H Zimmermann OSI Reference Model The ISO Model of Architecture for Open Systems Interconnection IEEE Transactions on Communications 28 4 425 432 1980 Wikimedia Commons Low Voltage Differential Signaling http en wikipedia org wiki LVDS 2014 Online accessed 12 August 2014 http www maximintegrated com en app notes index mvp id 3662 Under standing LVDS Fail Safe Circuits 2014 Online accessed 11 August 2014 M Pauer Return to zero Coding in Optical Intersatellite Links Technische Uni versitat Wien 2003 Wikimedia Commons Leitungscode http de wikipedia org wiki Leitungscode 2014 Online accessed 12 August 2014 R Forster Manchester encoding Opposing definitions resolved Engineering Sci ence amp Education Journal 2000 D Derickson and M M ller Digital Communications Test and Measurement High Speed Physical Layer Characterization Prentice Hall December 2007 A X Widmer and P A Franaszek A DC Balanced Partitioned Block 8B 10B Transmission Code IBM Journal of Research and Development 27 5 440 1983 Wikimedia Commons Ersatzschaltbild eines kurzen St ckes einer Zweidraht leitung http de wikipedia org wiki Leitungstheor
3. versity 2012 A K Perrevoort Characterisation of High Voltage Monolithic Active Pixel Sensors for the Mu3e Experiment Master thesis Heidelberg University 2012 R Philipp Characterisation of High Voltage Monolithic Active Pixel Sensors for the Mu3e Experiment Master thesis Heidelberg University 2014 C Grzesik Fast Optical Readout for the Mu3e Experiment Bachelor thesis Hei delberg University 2014 S Corrodi Fast Optical Readout of the Muse Pixel Detector Master thesis Hei delberg University ETH Zurich 2014 S Bachmann et al The proposed trigger less TBit s readout for the Mu3e exper iment In JINST 9 C01011 doi 10 1088 1748 0221 9 01 C01011 2014 A Damyanova Development of a Scintillating Fibre Tracker Time of Flight De tector with SiPM Readout for the Mu3e Experiment at PSI Master s thesis Geneva University 2013 29 P Eckert In Preparation PhD thesis Kirchhoff Institut fiir Physik 2015 78 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 M Zimmermann Cooling with Gaseous Helium for the Mu3e Experiment Bachelor thesis Heidelberg University 2012 L Huxold Cooling of the Mu3e Pixel Detector Bachelor thesis Heidelberg Uni versity 2012 A Herkert In Preparation Master thesis Heidelberg University 2015 G S Bauer Y Dai and W Wagner SINQ layout operation applications and R amp D to high power
4. For LVDS links see Subsection 3 1 1 a pair of wires is needed When designing a PCB the trace width w the trace separation s the trace thickness t and the dielectric thickness h as well as the dielectric constant of the substrate e see Figure need to be taken into account to match the impedance to the other electric components for minimal reflections Therefore the following equation can be used for differential microstrips 50 This approximation is valid for ratios of w h between 0 1 and 3 0 174 5 98 h A Zag l 1 0 48 e7067 3 5 aara TE 2 vn zn S t h a Differential microstrips b Coplanar striplines Figure 3 5 Profiles of two kinds of microstrips Coplanar Striplines At first appearance coplanar striplines seem to be very similar to differential microstrips see Figure 8 5b but there is a crucial difference when calculating the characteristic impedance This is because in case of the latter there is a clear boundary condition for the electric field surrounding the conducting traces due to the ground plane According to 51 the impedance Zo of a coplanar stripline can be calculated by _ 1207 K ko Zo 3 6 20 where cef is an effective dielectric constant given by 1 K k K ko 2 K k K kh 1 er 1 with K e being the elliptic integral of first kind and 8 2 O a 0 s 2 w 0 0 3 8 tanh 25 ER K yi k 3 9 w s 2 tanh 7 ah For
5. i e to avoid observing the influence of different trace lengths on the adapter board 6 2 1 Influence of the HSMC Flexprint Adapter Board Although the adapter board was designed with a differential impedance of 100 Q it is likely to affect the signal quality due to imperfections at its soldering points i e the locations in which the HSMC connector and the FPC connectors are soldered onto the PCB In addition reflections certainly occur as the differential impedance of the flexprints deviate from 100 2 Comparision of Adapter Board and Loopback Card Figure depicts an eye diagram for data transmission where input and output are directly connected via the HSMC loopback card whereas the data transmission via a flexprint is shown in Figure 6 3b The comparison reveals that the shape of the signal is changed in some ways A clear distinction is visible regarding the eye height which is comprehensible due to ohmic losses along the flexprint In both eye diagrams one observes two slight dips shortly after a transition But for the setup involving the loopback card there is a strong dip nearly simultaneous with the next transition whereas the corresponding dip comes a bit later when using the adapter board and a flexprint It could be that the dip in Figure 6 3alis caused by reflections on the FPGA develop ment board Since the signal is picked up very closely to the HSMC connector in case of the loopback card but a few centimeters
6. increases the laser power in areas with a low density of graphical details A high setting for density decreases the laser power at all edges to compensate for laser lag in turning off at high speeds Table 4 6 summarizes the minimal structure sizes which could be produced such that the traces were still conductive and properly separated respectively Nevertheless it was found out that a trace width of 80 um does not conduct reliably at a length of gt 10 cm see Subsection 4 5 3 Microscopic images of the test structures are shown in Figure 4 3 One observes that the 80 100 um separations in Figures 4 3b 4 3d and have the same widths Still 31 min trace width um min trace separation um horizontal 80 100 vect 110 rast 45 100 100 vect 140 rast vertical 100 100 vect 140 rast Table 4 6 Laser settings used to produce test patterns and flexprints these separations are not always proper The laser would have to be run at a lower velocity for these cuts The reason why this can still not be used for the manufacturing of flexprints is discussed in Section 4 5 1 With this laser platform it is in principle not possible to go below a trace separation of 100 um because this is the width of a single laser cut i e the radius of the laser focus For all lines lt 100 um in a layout the laser mode is automatically changed to from rastering to vectoring Therefore the 80 100 um cuts in
7. 50 ns time frames Due to the high decay rate of 2 10 s about 100 decays per frame occur in the detector acceptance Thus a more precise timing measurement is needed and provided by the scintillating fibre detector The fibre detector consists of two or three layers of scintillating fibres with a diameter of 250 um forming ribbons The light produced by scintillation will be read out with arrays of silicon photo multipliers SiPM mounted at both ends of the ribbons 28 Figure 2 5 First mechanical prototype of the Kapton support structure for the inner two layers length 12 cm Tile Detector The third sub detector is a scintillating detector consisting of 7 5 x 7 5 x 5 mm tiles which are also read out by SiPMs It will be placed right underneath the recurl pixel double layer As this is the last measurement of the decay particles more material can be used This leads to a detection efficiency close to 100 and a very high timing accuracy below 100 ps PJ Detector Environment As mentioned above the whole detector will be placed in a solenoidal homogeneous magnetic field of 1 T to bend the electron tracks The front end electronics will be placed directly on the muon beam pipe For cooling the whole detector volume will be filled with a circulating helium atmosphere In addition channels in the Kapton support structure will be flushed with gaseous helium for cooling of the pixel detectors 2 1 4 The Readout Concept In Fig
8. Beyond the Standard Model Rev Mod Phys 73 151 202 2001 arXiv hep ph 9909265 77 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 D Nicolo MEG Collaboration The u ey experiment at PST Nucl Instrum Meth A503 287 289 2003 U Bellgardt et al SINDRUM Collaboration Search for the Decay ut etete Nucl Phys B299 1 1988 J Adam et al MEG Collaboration New Constraint on the Existence of the ut ety Decay Phys Rev Lett 110 201801 May 2013 A M Baldini et al MEG Upgrade Proposal ArXiv e prints January 2013 arXiv 1301 7225 physics ins det A Blondel et al Letter of intent for an experiment to search for the decay u eee 2012 M Kiehn Track Fitting with Broken Lines for the MU3E Experiment Diploma thesis Heidelberg University 2012 R M Djilkibaev and R V Konoplich Rare Muon Decay u eteTetvedy Phys Rev D79 073004 2009 arXiv 0812 1355 hep ph I Peri et al High voltage pixel detectors in commercial CMOS technologies for ATLAS CLIC and Mu3e experiments Nucl Instrum Meth A731 131 136 2013 N Berger et al A Tracker for the Mu3e Experiment based on High Voltage Monolithic Active Pixel Sensors Nucl Instr Meth A 732 61 65 2013 arXiv 1309 7896 physics ins det H Augustin Charakterisierung von HV MAPS Bachelor thesis Heidelberg Uni
9. Detector The Mu3e pixel detector consists of High Voltage Monolithic Active Pixel Sensors HV MAPS with a pixel size of 80 x 80 um The chips are thinned down Recurl pixel layers ia a a FB Scintillator tiles Inner pixel layers Jd gt gt gt H Beam Target gt Scintillating fibres om eT Outer pixel layers Figure 2 4 Schematic drawing of the detector design The blue and the red lines represent recurling positrons and an electron from a signal decay On the right side view along the beam axis 17 to 50 um and have a size of 1 x 2 cm in the inner layers and 2 x 2 cm in the outer and the recurl layers 21 With 275 million pixels an area of more than 1 m will be covered In contrast to the classical MAPS technology the ionization charges are collected by drift due to an applied high voltage HV This leads to a much faster charge collection compared to the diffusion process in MAPS In addition the radiation tolerance is improved Several prototype chips are tested 22H24 The chips will be glued on a self supporting Kapton structure see Figure 2 5 with a thickness of 25 wm and wire bonded to Kapton flexprint cables for power supply and readout The triggerless readout will be done by zero suppressed serial 800 Mbit s Low Voltage Differential Signaling LVDS see Subsection 3 1 1 25H27 Fibre Detector The pixel detector will be read out in
10. Ea dene De ee 12 1 Physical Properties yh ea ay ke 4 4 Structure Sizes 4 4 1 Impedance Calculations CONAN N E rFOANAAaD E 13 13 15 18 19 20 21 22 23 4 4 2 Test Structuresl 2 2 2 2 2 Hmm rn Bo Ge ce Stereo cts Dr Se ae Sy me Spt ee ea S 4 5 1 Limitations 2 2 ee ie ee ee ee ee ee ee E Dee oe a ad ene ok E dis ne Bk ee 5 Performance of BERTs 5 1 Hardware ee aa a 5 1 1 Field Programmable Gate Array 04 5 1 2 FPGA Development Board 04 5 1 3 HSMC Flexprint Adapter Board bt byes eage ee eee he eda aa A a bee bees 5 2 1 Altera Quartus lll o o e 0000 A A RE 5 3 Firmware BERT Implementation 0 5 3 1 Data Generator 2 ee BON sian Emcee ste e Sh at ae ee ee ents ee Hoey amp 0 3 3 LVDS Transmitter 5 3 4 LVDS Receiver 2 8 00 a 8 Ea aa a ee RA ew eset tay A a Messen glee ES e es eee 5 3 6 Data Checker ee B A BERT Results x 00 io de a AD al e A mm b Analysis of Eye Diagrams 6 1 Test Setup l ooo comp ua a a ee e Pa a da SM AAA 6 2 1 Influence of the HSMC Flexprint Adapter Board 6 2 2 Influence of the Cable Length 6 2 3 Influence of the Transmission Rate 6 2 4 Influence of the Pre Emphasis 6 2 5 Crosstalk between Trace Pairs 2222 o a A 6 2 7 Influence o
11. Eye Diagrams Besides the performance of BERTs it can be valuable to analyse the signal quality of a high speed transmission line with eye diagrams In contrast to BERTs which need to be run for a long time to achieve low upper limits an eye diagram is quickly measured and allows to extrapolate the behaviour of a transmission line for higher frequencies or worse input signal qualities 6 1 Test Setup Basically the same setup as for the BERTs is used In addition an active differential probe has been deployed to pick up the voltage signal from the adapter board and visualize it on an oscilloscope The setup is shown in Figure 6 1 b Closeup of the differential probe picking a Complete setup for the measurement of up the voltage signal from a via pair on the 8 eye diagrams adapter board Figure 6 1 Pictures showing the setup used for the measurement of eye diagrams 5l 6 1 1 Tektronix DPO 7254C Digital Phosphor Oscilloscope The Tektronix DPO 7254C Digital Phosphor Oscilloscope is a high end oscilloscope which is suitable to analyse fast digital data transmission due to its bandwidth of 2 5 GHz and a sampling rate of 40 GS s 79 It provides a fast acquisition FastAcq mode for the analysis of dynamic signals and the capturing of rare events such as glitches In this mode the oscilloscope records sample points of the signal in constant time steps after being triggered as depicted in Figure 6 2a In the next acquisition c
12. Figures and 4 31 show a better result than the 110 130 um separations 32 a Horizontal connection dE E e Vertical connection Vertical separation Figure 4 3 Test patterns to examine which minimal structure sizes can be achieved by rastering All trace widths and trace separations are from 80 um to 150 um in steps of 10 um from top to bottom or left to right respectively For the connection structures the gap is kept constant whereas the pitch is kept constant for the separation structures 33 4 5 Flexprint Cables Multiple flexprint cables with different characteristics have been produced examples of which are shown in Figures and Various types of flexprints have been successfully produced at lengths of 10 20 and 30 cm though broken traces were not uncommon for a trace width of 100 um see Subsection 4 5 3 Flexprints with a trace width of 100 150 wm have also been produced up to a length of 50 cm b Microscopic image of the traces and contact pads Figure 4 4 Flexprint with a trace width of 100 um a trace separation of 150 um for pairs and 650 um between pairs and a Kapton thickness of 50 um plus aluminum ground plane Length of the flexprint 10 cm The flexprint shown in Figure 4 4 consists of 17 equal trace pairs with a large sepa ration between adjacent trace pairs It has only horizontal structures so that it can be produced at a higher laser velocity This type of flexpr
13. away on the adapter card this might introduce a difference in the runtime of the reflected signal which leads to the observed shift of the dip Another reason could be crosstalk between adjacent traces on the flexprint 53 a Transmission via loopback card b Transmission via flexprint Figure 6 3 Eye diagrams to compare transmission involving the loopback card and the adapter board The flexprint has a trace width of 100 um a trace separation of 150 um between all traces a Kapton thickness of 50 um plus aluminum ground plane and a length of 30 cm Comparison of Different Channels The HSMC Adapter Board is not designed symmetrically and has traces of different lengths On the transmitter side they vary between 38 6 mm and 48 2 mm and on the receiver side between 26 2 mm and 40 0 mm respectively For the eye diagrams in Figure the signal has been picked up at channels 0 11 and 16 Since the traces on the flexprint might slightly differ in trace thickness or separation due to production tolerances for the eye diagrams in the right column the flexprint has been connected with interchanged ends see Figure 6 5 I e different channels have been compared using exactly the same differential pair on the flexprint for maximal comparability It can be seen that interchanging the ends of the flexprint makes only a very slight dif ference which could be introduced by an improper connectivity to the FPC connector On the contrary
14. be used for an errorless data transmission though the signal quality is worse in comparison to differential microstrips see Section 7 3 7 3 Eye Diagram Results The analysis of multiple eye diagrams has lead to the following results The signal deformations which can be observed in the eye diagrams are neither dependent on the transmission rate nor on the flexprint length The shape even resembles the signal picked up from the loopback card alone Therefore the signal deformations are unlikely to be caused by reflections at the FPC connectors due to improper impedance matching Consequently a differential impedance for flexprints of the type differential microstrips in the range of 40 60 Q seems sufficient to minimize the reflections to an acceptable level Comparing flexprints of different lengths has shown that the eye height decreases with increasing cable length Nevertheless also for 50 cm flexprints the signal shape allows a clear distinction between the logic states The signal shape looks much better for flexprints of the type differential mi crostrips than coplanar striplines The additional aluminum ground plane de creases crosstalk effectively and possibly also improves impedance matching More over it provides space for ground and power supply which would have to be placed on the flexprint itself in case of coplanar striplines Even though it increases the material budget the amount of scattering
15. bending the flexprint in a loop like in Figure 5 1 does not cause any damage On many of the cables a few traces were not conducting This problem increased with growing length For 10 cm cables broken traces occurred only rarely whereas at a length of 30 cm typically 2 3 of 17 channels did not work due to broken traces An attempt to produce a 50 cm cable with a trace width of 100 um yielded only 2 working channels It is likely that this problem comes up because the laser platform does not work homogeneously throughout the whole table This might be due to an imperfect cali bration of the work area so that it is slightly slanted and the laser focus passes through different vertical positions Flexprint with a trace width of gt 120 um have been successfully produced at a length of 50 cm with all channels working From this one concludes that a trace width of gt 120 um can be manufactured with a much higher reliability 4 5 4 Blackening of the Kapton The Kapton becomes darker the slower the laser is moved The question arises what happens to the Kapton during this process and whether the dielectric properties are changed The image taken from the back side of the Kapton strips shows that bubbles have formed inside the Kapton and buckled the surface This behaviour can be understood According to 66 Kapton changes its colour from dark brown to black if heated above a temperature of 500 C Additionally the gases CO and CO e
16. bertragen Die Herstellung sowie Performance Tests selbsthergestellter Kapton Flexprints sind die Schwerpunkte dieser Arbeit Es konnte gezeigt werden dass Flexprints mit Lei terbahnbreiten von 120 um und Abst nden von 110 um zuverl ssig mit einer Laser plattform an der Universit t Heidelberg hergestellt werden k nnen Zus tzlich sind Fehlerraten Tests durchgef hrt worden die Werte unter O 1071 f r 17 parallele Ka n le bei einer Datenrate von je 800 Mbit s ergeben haben Dar ber hinaus sind Augen diagramme analysiert worden um zu verstehen welche Faktoren die Signalqualit t im Wesentlichen beeinflussen Contents Theory amp Background Theoretical Background La Eye aoe So ee a ee aes Re Bee Gg Bert Ge din eee E tego a pats Ge eee ape eee ce ade eH ida 28 Alla a ad eR eh oh a oe ea G 2 1 The Muse Experiment 2 0 0 0 0 nn 2 1 1 The Signal Decay 2 2 2 2 22 Co onen 2 1 2 Background Decays 2 1 3 Experimental Concept 2 1 4 The Readout Concept 2 1 5 The Muon Beam Basics of Data Transmission 4 3 1 Signals 3 1 1 Low Voltage Differential Signaling 2 3 1 2 Data Encoding 3 2 Transmission Lines 3 2 1 The Characteristic Impedance o 3 2 2 Microstrips 3 3 Signal Quality Checks 3 3 1 Bit Error Rate Tests BERTS existir a 3 3 2 Eye Diagrams Measurements amp Results Manufacturing of Kapton Flexprints
17. on the receiver side run back to the via pair where the signal is picked up a Transmission rate 100 Mbit s x axis b Transmission rate 300 Mbit s x axis 2 5 ns div 2 5 ns div c Transmission rate 600 Mbit s x axis d Transmission rate 800 Mbit s x axis 500 ps div 500 ps div Figure 6 9 Eye diagrams of data transmission at different rates via flexprints with a trace width of 100 um a trace separation of 150 um between all traces and a Kapton thickness of 50 um plus aluminum ground plane Length 20 cm 59 The slight dips might come from a limited bandwidth of the transmitter since they appear independent of the frequency and even when using the loopback card see Fig ure 6 3a so that they are definitely not caused by the flexprint or the adapter board Besides the slew rate is observed to be equal for all frequencies which indicates that it is limited by the transmitter Also the eye height does not change with the frequency Even though the frequency could not be increased beyond 800 Mbit s due to limi tations of the transmitter the signal shape allows to assume that the flexprints would also work at higher transmission rates because the deformation of the signal does not show any frequency dependence and the bit states are always clearly distinguishable 6 2 4 Influence of the Pre Emphasis At high frequencies the slew rate of the transmitter is not fast enough to reach the full voltage level
18. wire is described as an infinite series of such elements Figure 3 4 Equivalent circuit diagram of an infinitesimally short piece of a transmis sion line with normalized inductance L and resistance R coupled to ground via G and C 45 18 3 2 1 The Characteristic Impedance If the transmission line is homogeneous along its length a single parameter is sufficient to describe its behaviour the characteristic impedance Zo It is equal to the ratio of the complex voltage and the complex current of a wave travelling along the line It can be shown see 46 that Zo is given by IR iwL Zu 1 o G iwC 31 Note that Zo is independent of the length of the transmission line For an ideal conducting material R 0 and an ideal dielectric G 0 or high frequencies R lt iwL and G lt iwC equation 3 1 reduces to L Zo a 3 2 so that the characteristic impedance is also independent of the frequency In such cases Zo is not related to the ohmic resistance of a wire which causes an attenuation of the amplitude of a transmitted signal and is self evidently dependent on the length of the transmission line The characteristic impedance is merely a parameter to describe a wire s high frequency behaviour When considering a pair of transmission lines with opposite current which is the case for LVDS another quantity is of importance the differential impedance Zag It is d
19. 00 110011 001100 D 09 01001 100101 D 25 11001 100110 D 10 01010 010101 D 26 11010 010110 D 11 01011 110100 D 27 11011 110110 001001 D 12 01100 001101 D 28 11100 001110 D 13 01101 101100 D 29 11101 101110 010001 D 14 01110 011100 D 30 11110 011110 100001 D 15 01111 01011 101000 D 31 11111 101011 010100 K 28 11100 001111 110000 Table 3 1 5b 6b encoding scheme For several 5 bit words two different patterns with disparity dp 1 exist D x mark the 2 32 possible data words K 28 is a control word 26 suffice a disparity of 0 or 2 and never have more than five equal subsequent bits 584 of the possible 1024 combinations are left One way of implementing this concept has been developed by IBM in 1983 44 The 8 bits of a word are split into two parts and are treated separately according to an 5b 6b and an 3b 4b part see Tables 3 1 and 3 2 respectively During the data transmission the running disparity is determined to control the combination of 6 bits and 4 bits being used to satisfy the condition dp 0 or dp 1 Besides the advantage of DC balancing this encoding scheme comprises some control words so called K words which do not encode data but can be used for clock recovery and phase alignment before beginning to transmit user data or to bypass phases when no actual data is transmitted to retain the synchronization The K 28 7 word is of particular importance because it does not result from a single bit
20. 5 obs k ep p e _ et 3 13 k 1 CL k 0 one arrives at 22 log CL _ 2 996 BER lt o Da ar at 95 CL 3 14 where CL is the confidence level 3 3 2 Eye Diagrams Eye diagrams allow to visualize and determine the quality of a transmitted digital signal easily and quickly 56 An eye diagram is constructed from a digital data stream by superimposing the waveform of an arbitrary bit stream in a single diagram with time on the horizontal and signal amplitude on the vertical axis Therefore it represents the average statistics of the signal An ideal waveform of the digital signal would result in a square shaped eye diagram as can be seen in Figure 3 6a Due to impairments of the signal like attenuation cross talk or noise and a limited bandwidth of the transmitter real physical signals differ from this and rather resemble the eye diagram shown in Figure 3 6b Here the unit interval Ul is defined as the time it takes to transmit one bit and corresponds to the inverse of the transmission rate UU WO UU WUV a E E T 1U l Tpit a Time Time a Ideal high speed digital signal with eye b Typical example of a real high speed diagram digital signal with eye diagram Figure 3 6 An ideal and a real arbitrary digital datastream superimposed in eye di agrams 56 From an eye diagram a number of key parameters can be determined These are shown in Figure 3 7Jand described in
21. 6 6 Schematic of the reflections at the FPC and the HSMC connectors Tgex can be calculated with equation 3 4 T ysmc is not known 6 2 2 Influence of the Cable Length To estimate the influence of the cable length on the signal quality three flexprints with different lengths are compared in Figure As can be seen the length does not affect the qualitative shape of the eye In all cases there are basically two dips after a transition This supports the theory that the reflections due to improper impedance matching of the flexprint do not dominate the total reflection In fact one would expect different signal shapes because of different runtimes of the reflected signals unless the cable lengths equal multiples of the wave length This is unlikely since with a base frequency of 400 MHz and a propagation velocity of approximately 2 3 of the speed of light the wavelength should be roughly 50 cm 56 a Length 10 cm c Length 30 cm Figure 6 7 Eye diagrams of 800 Mbit s data transmission via flexprints with a trace width of 100 um a trace separation of 150 um for pairs and 650 um between pairs and a Kapton thickness of 50 wm plus aluminum ground plane but different lengths As the signal deformation is independant of the cable length one can rather proceed on the assumption that reflections at the flexprint are not even visible in these eye diagrams However one observes that there is an attenuation of the signal wh
22. Department of Physics and Astronomy University of Heidelberg Bachelor Thesis in Physics submitted by Jens Kroger born in L bbecke Germany 2014 Data Transmission at High Rates via Kapton Flexprints for the Mu3e Experiment This Bachelor Thesis has been carried out by Jens Kr ger at the Physikalisches Institut Heidelberg under the supervision of Prof Dr Andr Sch ning Abstract The Mu3e experiment is aiming to search for the neutrinoless muon decay u ee et with a sensitivity of one in 10 decays or better This decay is lepton flavour violating and strongly suppressed within the Standard Model Therefore even a single decay signal would be a clear hint for new physics The concept of the Mu8e experiment is to achieve a very good momentum and time resolution and an excellent vertex reconstruction to suppress background to a sufficient level Since the decay electrons have a low energy up to 53 MeV multiple Coulomb scattering is the dominating limiting factor for the momentum and vertex resolution Therefore the material budget inside the detector region must be kept at a minimum To achieve this pixel detectors with a thickness of only 50 um will be used which are carried by a Kapton foil support structure The pixel detectors shall be linked to front end FPGAs via ultra thin Kapton flexprints The production and performance tests of self manufactured Kapton flexprints are the main scope of this thesis It
23. High Speed Digital Signals Semiconductor Components Industries LLC AND9075 D 2014 http uniquelaserengravers tradeindia com pl6mw 1118929 html On line accessed 20 August 2014 Universal Laser Systems PLS User Guide 2012 DuPont Kapton HN polyimide film Technical Data Sheet 2011 DuPont Kapton HN Summary of Properties year unknown M Bauccio ASM Metals Reference Book ASM International 1993 http www aluinfo de index php aluminium in electrical engineering and electronics html Online accessed 20 August 2014 Los Alamos National Laboratory http periodic lanl gov 13 shtml On line accessed 20 August 2014 S Eidelman et al Passage of Particles Through Matter Physics Letters B592 2004 J Bartl and M Baranek Emissivity of Aluminum and its Importance for Radio metric Measurements Measurement Science Review 2004 Y Kaburagi and Y Hishiyama Ferromagnetism discovered on heat treating the aromatic polyimide film Kapton J Mater Res 2002 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 Altera Stratix V Device Handbook 2013 W Rogers et al FPGA based Implementation of Signal Processing Systems Wiley 2008 Altera Stratiz V Device Overview 2014 Altera Stratir V GX FPGA Development Board Reference Manual 2014 http www altera com products devkits altera kit sv gx host html Online accessed 21 A
24. Sai Ch Aa et ene EI 33 AA Blexprint type ly ss 434 wage er Re Ron eier el alee 34 4 5 Flexprint type 2 e 35 4 6 Flexprint type 3 ee 36 AA 37 pi ae pies UR 38 TT Oe ee Oe 40 5 1 Picture showing the setup comprising the FPGA development board the HSMC adapter board and the self manufactured flexprint 42 5 2 Stratix V GX FPGA Development Board 44 5 3 HSMC Flexprint Adapter Board o e e 44 5 4 Altera HSMC loopback board nn nn 45 74 5 5 Schematic of the BERT implementation 46 cagao 48 os on 52 AA 54 IE 55 6 5 Interchange of the ends of the flexprint for maximal comparability 56 6 6 Schematic of the reflections at the FPC and the HSMC connectors 56 Loe Poe A eek ee a 57 6 8 Eye diagram for a 50 cm long flexprint 58 oe kee ee Ra eS 8 59 by tk ash pe 60 6 11 Eye diagrams showing crosstalk for different Kapton thicknesses 61 pes Bo aah e ook ee ee O Ea 63 a a een 65 6 14 Interchange of the ends of the flexprint to compare different knee lengths using the same channel 2 2 Co u En 66 8 1 Schematic showing the concept of bonding on the underlying aluminum E RAGE SERRO RA apa ae aa 72 75 List of Tables 76 3 1 56 6b encodina La dos seh ars a ES die Oe ee dae 17 3 2 3b Ab encoding 3 2 4 4 ee eG EA 18 3 3 Typical parameters for microstrips on PCBs and flexprin
25. Vic o 1 2 electron neutrino lt 15 5 MeV c 0 12 t GAUGE BOSONS LEPTONS au neutrino muon neutrino Figure 1 1 Particles described by the SM 4 In spite of the great successes of the SM observations have been made which cannot be explained by this theory Several experiments have observed neutrino oscillation such as Super Kamiokande 5 SNO 6 KamLAND 7 and Daya Bay 8 To ex plain this mixing of the flavour eigenstates the neutrinos need a non vanishing mass which is not foreseen in the SM An extension by heavy right handed neutrinos called vSM yields consistent results with oscillation experiments 9 Still it cannot be ex plained why the neutrino masses are orders of magnitude smaller than those of the other particles Moreover vSM does not provide any explanation for the observed matter antimatter asymmetry the origin of dark matter or the existence three generations of particles These phenomena motivate theories beyond the Standard Model BSM such as supersymmetric theories SUSY In contrast to the SM many of these predict flavour violating processes at an observable branching ratio 1 1 2 Muon Decays Anti muons are unstable and have a mean lifetime of about 2 2 us The domina ting decay in the SM is p e vev I Considering neutrino mixing and therefore allowing lepton flavour violation the muon decay can also be realised without outgoing neutrinos see Figu
26. aluminum can be processed with a laser such that the aluminum evaporates in some areas whereas the Kapton is damaged as little as possible and conducting traces remain 4 1 The Laser Platform For manufacturing the flexprints a PLS6MW Multi Wavelength Laser Platform from Universal Laser Systems see Figure 4 1 has been used It provides a plane work area of approximately 80 x 46 cm on which a broad spectrum of materials can be processed with different wavelengths as the laser source can be changed For this thesis a CO2 laser with a wavelength of 9 3 um and a fibre laser with a wavelength of 1 06 um are available Figure 4 1 A photo of the PLS6MW Laser Platform taken from 57 For processing a material is placed on the work area and scanned with the pulsed laser For ideal results several parameters can be adjusted such as laser power laser movement velocity pulse frequency waveform and vertical focus position The laser can either treat areas called rastering by moving back and forth in x direction while scanning the y direction incrementally or move along lines called vectoring When 26 rastering the additional parameters contrast definition and density are adjustable In order to blow away evaporated material and prevent the lens from being contami nated with debris a nozzle points an air stream onto the focus of the laser Apart from compressed air every other gas can be used if not flammab
27. am control words can be used to identify bit packets words Thus the bit stream can be cut into words of a certain length For some physical layers like LVDS it is desirable to have DC balance i e to send an equal number of ones and zeros to achieve a vanishing net current By an encoding scheme like 8b 10b this can be accomplished A detection of transmission errors is possible if bit patterns are used for which a single bitflip yields an invalid word 8b 10b Encoding The disparity dp of a bit pattern is given by the difference in number of ones and zeros counting ones as 1 2 and zeros as 1 2 If considering a continuous data stream the running disparity rd is defined as the running sum over all previously trans mitted words For 8 bit words encoded in 10 bits it is possible to ensure DC balance in a long run That is because with 8 bits 2 256 words can be created whereas with 10 bits there are 21 1024 combinations Regarding the restrictions that the 10 bit patterns shall 16 Word Data dp 1 dp 1 Word Data dp 1 dp 1 D 00 00000 100111 011000 D 16 10000 011011 100100 D 01 00001 011101 100010 D 17 10000 100011 D 02 00010 101101 010010 D 18 01010 010011 D 03 00011 110001 D 19 01011 110010 D 04 00100 110101 001010 D 20 01100 001011 D 05 00101 101001 D 21 01101 101010 D 06 00110 011001 D 22 01110 011010 D 07 00111 111000 000111 D 23 10111 111010 000101 D 08 01000 110001 000110 D 24 110
28. and a base frequency of 400 Mbit s result in a runtime of approximately 2 ns But 2 ns after the transition in Figures 6 12h c no specific dip can be observed Therefore it can be concluded that the impedance matching is only of minor impor tance for the eye shape and crosstalk affects the signal quality much more a Coplanar Stripline Kapton thickness b Differential Microstrip Kapton 25 um without ground plane calculated thickness 25 um on ground plane calculated Zaif 268 Q Zaif 402 c Differential Microstrip Kapton thickness 50 um on ground plane calculated Zaig 90 Q Figure 6 12 Eye diagrams of 800 Mbit s data transmission via flexprints with a trace width of 100 um a trace separation of 150 um between all traces and a length of 20 cm but different microstrip types 63 6 2 7 Influence of the Knee Length For flexprints with a trace separation of 150 um between all traces there is a spread at both ends to meet the 0 5 mm pitch of the FPC connectors The question arises whether the length of the knees i e the sections running 45 to the main section no tably impair the signal quality because here the trace thickness and separation deviate from the main section and the edges of the traces are less smooth see Figures 4 5b and 4 6b To estimate this influence channels 0 8 and 16 i e the channel with no knee at all a mean knee length and the one with the longest knee are compared in Figur
29. be used To have an effective content of information the time evolution of the signal must be unpredictable because otherwise the receiver can foresee incoming data which makes the transmission redundant 35 Analogue Signals An analogue signal is continuous in time and amplitude In principle every classical physical observable like brightness temperature or pressure can be understood as an analogue signal Mathematically it can be described by a smooth function of time Digital Signals On the contrary a digital signal consists of a sequence of discrete values That is to say there is only a limited set of well distinguishable values which can be attained and which are furthermore only defined at periodic points in time An analogue signal can be transformed into a digital signal by quantisation and sampling in constant time intervals 36 In electronics it is common to use merely the two Boolean values 0 false and 1 true which are associated to two logic voltage levels or the transitions between those The binary numeral system provides the theoretical basis to use binary codes sequences of zeros and ones for digital information processing of all kinds 3 1 1 Low Voltage Differential Signaling Low Voltage Differential Signaling LVDS is an interface standard for high speed data transmission It describes the physical layer which means that it only comprises 13 Coupled Fields EN Fringing Sine 5mA Fi
30. before the next transition This conduces to a pattern dependent jitter because the edge of a transition comes a bit earlier if the full voltage level has not been reached yet To minimize this jitter the amplitude of the high frequency components of the out put signal can be enhanced This so called pre emphasis leads to a voltage overshoot shortly after a transition and increases the slew rate 67 a Pre emphasis disabled b Pre emphasis enabled Figure 6 10 Eye diagrams showing the effect of enabled pre emphasis The flexprint has a trace width of 100 um a trace separation of 150 um between all traces a Kapton thickness of 25 wm plus aluminum ground plane and a length of 30 cm Transmission rate 800 Mbit s 60 Figure depicts the change in the signal shape due to enabled pre emphasis The overshoot and an increased slew rate are clearly recognisable Especially in the first quarter the eye opening is notably larger Furthermore it can be seen that the amount of jitter is reduced 6 2 5 Crosstalk between Trace Pairs To examine how much noise is induced on an adjacent trace pair the firmware was slightly changed so that only every second channel was used for data transmission and the intermediate trace pairs were not assigned to FPGA pins a Signal on channel 0 b Induced signal on channel 1 Kapton thickness 25 um no ground plane Kapton thickness 25 um no ground plane c Signal on channel 0
31. ce each chip has a length of 2 cm there will be an area of 0 586 x 2 cm to spread the traces using 45 sections and place bonding pads Ground and power could be supplied by integrating holes in the Kapton to be able to bond on the underlying ground power plane as depicted in Figure For this purpose differential microstrips on 25 um Kapton would be much easier to produce than on 50 um because in case of the latter one would need to find a way to partially remove the Kapton from the aluminum to bond to the underlying metal plane These structure sizes also suffice for the outer polygonials which will have a length of 36 cm and 24 26 sides with a width of 2 cm Here the chips will have a size of 2 x 2 cm so that 18 chips will be mounted in a row These will have only 1 2 LVDS output channels so that to each side lt 18 channels need to be placed on a 2 cm wide flexprint This is at most as narrow as in the inner layers and should work as well TO 71 bonding wire Figure 8 1 Schematic showing the concept of bonding on the underlying aluminum plane Mupix chip signal plane ground power plane 8 1 2 Transmission Errors In phase I of the Mu3e experiment 180 chips with 3 LVDS output channels each will be used in the innermost double layer The outer double layer will mount 972 chips with 1 2 channels each so that the total number of channels adds up to lt 2142 10 With a BER lt 10715 at 95 CL and equation 3 12 the
32. chronization with the underlying clock can be lost 40 Another line code is the Manchester code in which falling and rising edges i e the transitions between the two logic levels represent ones and zeros respectively With this code clock recovery see below is relatively easy due to frequent transitions A disadvantage is that twice the bandwidth compared to NRZ is required to be able to send subsequent ones or zeros see Figure 42 Line codes are not only used to specify how ones and zeros are represented in a signal They also provide schemes how to encode bit words of a certain length in a bit pattern 15 000 a The non return to zero code ones and b The Manchester code ones and zeros are zeros are represented by two logic levels of the represented by falling and rising edges signal respectively Figure 3 3 The bit pattern 11011000100 represented in two commonly used line codes the NRZ and the Manchester code 41 with properties suiting the physical link 43 Furthermore additional information can be contained for the following purposes If there is no extra clock transmission line sending 0101 continuously the clock has to be recovered from the data stream clock recovery Therefore a high transition rate between the logic levels is favourable Some encoding schemes provide the possibility to send either data or a predefined control sequence of bits If transmitting a continuous bit stre
33. corresponds to the width of the laser focus It is problematic to use small trace separations in a design because the laser changes its mode from rastering to vectoring for a separation of lt 100 um and an offset or imprecision introduced by the mechanics of the laser platform leads to inaccurate traces This means that the production of flexprints is limited to be done with a single laser setting for rastering Vectoring is only suited to cut out the flexprint It has also been shown that differential microstrips can be realized by gluing a flex print and an unprocessed piece of Kapton laminated with aluminum of the same shape together This can be done with both sides of the Kapton aluminum foil so that differ ential microstrips with a Kapton thickness of 25 um plus adhesive layer or 50 um plus adhesive layer can be obtained The influence of different Kapton thicknesses on the impedance matching and crosstalk was examined and measurements were performed 68 Possible reasons for the blackening of the Kapton due to the processing with the laser have been discussed According to the interpretation given the laser treatment does not significantly change the dielectric properties of the Kapton to the detriment of the experiment 7 2 BERT Results BERTs have been performed that have shown that data transmission works flawlessly for a variety of flexprints Even coplanar striplines without a ground plane can
34. ctron from pair electron positron pair from internal production conversion with a Michel positron Figure 2 3 Possible combinatorial background combination of an electron and a positron from the internal conversion process u ete e v v with a Michel positron see Figure 2 3b 2 1 3 Experimental Concept As discussed above for an efficient background suppression high rate capabilities as well as excellent spatial time and momentum resolution are crucial The basic concept of the Mu3e experiment is to measure the momenta of the muon decay electrons in a solenoidal magnetic field of 1 T with a silicon pixel detector Because multiple Coulomb scattering in the detector material is the limiting factor for the momentum resolution minimizing the material budget below 1 o of radiation length per layer in the active detector region is essential In F igure 2 4 a schematic view of the detector design is shown The incoming muon beam will be stopped in a hollow double cone made of aluminum to decay at rest Four layers of very thin pixel detectors are arranged in two double layers black to track the decay electrons Furthermore scintillating fibres grey are used for precise timing measurements As the experiment is placed in a magnetic field the electrons are curled back and detected again by another double layer of pixel sensors Finally they are stopped in scintillating tiles which again yield precise timing information Pixel
35. d Induced signal on channel 1 Kapton thickness 50 wm plus ground plane Kapton thickness 50 wm plus ground plane Figure 6 11 Eye diagrams showing crosstalk for different Kapton thicknesses Data transmission at 800 Mbit s via flexprints with a trace width of 100 um a trace separation of 150 um between all traces and a length of 20 cm 61 For the eye diagrams in the right column of Figure the signal was picked up at an unassigned trace pair between two properly working channels One observes that crosstalk is present for both types of flexprints i e for coplanar striplines as well as for differential microstrips But in case of the coplanar striplines there is a clear periodic structure in the induced signal whereas for the differential microstrips it is mostly flat and thus interpreted as noise Moreover the amplitude of the induced signal is much larger for the coplanar striplines This means that the aluminum ground plane really effectively shields the adjacent chan nels and decreases the amount of crosstalk 6 2 6 Influence of the Microstrip Type In Figure 6 12 eye diagrams for different microstrip types can be seen The comparison shows that the eye diagrams clearly differ regarding the band thicknesses The bands in the signal for the coplanar striplines Figure are notably widened and also the amount of jitter is increased On the other hand the two signals for the differ ential microstrips with different Kapton thick
36. decay is indistinguishable from the signal decay The only difference is a fraction of momentum and energy carried away by the neutrinos which cannot be detected This clarifies the need for a very high momentum resolution If both neutrinos have a vanishing momentum this decay looks exactly as the signal decay This is the only irreducible background In Figure the branching ratio for the internal conversion as a function of the missing energy is plotted The missing energy corresponds to the difference of m e and the energy carried away by the electrons which is measured To suppress this background sufficiently an energy resolution below 1 MeV is needed 10 10r 10 10 10 F U3e 105 107 1078 LAA CUA EN ER 2 3 4 5 6 m E MeV tot Branching Rat Figure 2 1 The internal conversion decay Figure 2 2 Branching ratio for the ut etevetvev 13 internal conversion as a function of the missing energy 19 Random Combinatorial Background Due to a limited spatial resolution two nearby vertices cannot be properly separated if they are too close together As can be seen in Figure 2 3a two positrons from the dominant Michel decay u gt e vev see Figure can be accidentally combined with an electron for exam ple produced by photon conversion or Bhabha scattering Another possibility is the e y 14 a Accidental combination of two Michel b Accidental combination of an positrons with an ele
37. detail below e The zero level and the one level are the mean values of the lower and upper logic levels respectively In NRZ coding they correspond to a 0 and a 1 23 24 The eye amplitude is the difference between the two logic levels This quantity is used by the receiver to determine whether a 0 or a 1 has been transmitted The eye height describes the vertical opening of an eye diagram Ideally it would be equal to the amplitude but in reality it is smaller due to noise or saturation effects The level at which rising and falling edges cross is called eye crossing percent age and should be at 50 of the amplitude The eye width corresponds to the vertical opening of an eye diagram and should ideally be equal to the unit interval Jitter refers to variations in the transmission frequency so that rising and falling edges are slightly shifted and occur too early or too late Bit Period One Level Eye Crossing Percentage Zero Level Fall Time Rise Time Figure 3 7 A typical eye diagram with key parameters 56 modified Part II Measurements amp Results 4 Manufacturing of Kapton Flexprints A flexprint is a bendable plastic film carrying conducting traces which consist of copper or aluminum To manufacture flexprints usually lithographic processes are used which resemble those for PCB production For this thesis it was examined in what way Kapton foils laminated with
38. differential microstrips on 50 um Kapton than with thinner Kapton For the coplanar striplines it is not possible to find structure sizes yielding a Zgig close to 100 Q 4 4 2 Test Structures The second step was to examine down to which scales the laser platform works properly and to find optimal settings Therefore a series of test structures has been produced see Figure 4 3 All of the test strutures and flexprints have been produced with the settings listed in Table 4 5 The Kapton foil was placed on a sticky rubber mat with the aluminum layer on top In addition the vertical position not in the table had to be adjusted to the thickness of the rubber mat and the Kapton Since the calibration was not very stable and changed from day to day it was easier to produce a quick test structure and try out different settings for the vertical position than recalibrating the laser system again power speed freq MHz waveform contrast def density rastering 100 65 30 0 20 10 80 vectoring 100 14 30 0 20 10 80 Table 4 5 Laser settings used to produce test patterns and flexprints Besides the power the speed and the frequency of the pulsed laser it is possible to set a value of 0 to 5 for the waveform In 58 the different waveforms are not specified but only 0 yields proper results A high contrast increases the laser power at edges in areas with a high density of graphical details Definition on the contrary
39. e Again the flexprint is connected with interchanged ends in the right column to dis tinguish the influence of the knee from the intrinsic differences of the channels see Figure 6 14 The eye diagrams for channel 8 should and do look exactly the same in both cases since 8 is in the middle of 17 and the direction of the data transmission should make no difference due to the symmetric structure of the cable Also the other eye diagrams for no knee and the maximal knee length look fine The comparison of Figures 6 13a and 6 13b shows that there is only a very slight difference in the height of the wiggles whereas the basic signal shape is conserved The amount of jitter does not notably differ The same is observed in Figures 6 13e and 6 131 Consequently the influence of the knee length can be considered to be of minor importance and the use of 45 sections in a flexprint does not significantly impair the signal quality 64 a No knee channel 0 b Maximal knee length channel 0 Flexprint connected with interchanged ends c Medium knee length channel 8 d Medium knee length channel 8 Flexprint connected with interchanged ends e Maximal knee length channel 16 No knee channel 16 Flexprint connected with interchanged ends Figure 6 13 Eye diagrams for different knee lengths 800 Mbit s transmission via a flexprint with a trace width of 100 um a trace separation of 150 um betw
40. e branching ratio BR u etete lt 1071 could be set at 90 confidence level CL H In the experiment which was placed inside a solenoid magnetic field of 0 33 T muons with a momentum of about 28 MeV c were stopped in a hollow double cone target The decay products were measured by five tracking layers of multiwire proportional chambers and an array of scintillators for triggering and timing measurements MEG To search for the LFV decay u ety the MEG experiment uses a drift chamber to detect the positron and a liquid xenon calorimeter for the photon It has been running at PSI since 2008 and is currently being upgraded to MEG II 16 The current upper limit is BR u ety lt 5 7 107 at 90 CL 15 104 v 5 o 10 a l Y 2 10 S E u ey 10 o u 3e A uN eN ee SINDRUM A t 3u SINDRUM II 10 MEG plan Mu3e Phase O 107 Muse Phase IIC 1940 1960 1980 2000 2020 Figure 1 4 The history of LFV muon decay experiments adapted from 11 2 The Mu3e Experiment 2 1 The Mu3e Experiment Mu3e is an experiment to search for the lepton flavour violating decay ut gt etete I7 It is aiming to be sensitive for better than one signal decay in 10 muon decays This would increase the sensitivity by four orders of magnitude compared to the current upper limit given by SINDRUM To perform the experiment on a reasonable time scale a very high muon stopping rate of O 10 s is needed Consequent
41. e first generation contains the up quark u and the down quark d with charges of 2 3 and 1 3 in units of the elementary charge the electron e7 and the electrically neutral electron neutrino ve The second and the third generation are also made up of a pair of quarks a lepton and a neutrino These are the charm quark c and the strange quark s together with the muon u and the muon neutrino v in the second generation and the top quark t the bottom quark b together with the tau 7 and the tau neutrino v in the third generation All quarks and leptons are spin 1 2 particles i e they are fermions The interaction between them is mediated by the gauge bosons which have spin 1 The eight gluons are responsible for the strong interaction the photon y mediates the electromagnetic force and the Z W and W go with the weak interaction In the SM the neutrinos are considered to be massless The so called lepton flavour number i e the number of anti leptons subtracted from the number of leptons of one generation is a conserved quantity mass gt 2 3 MeV c 1 275 GeV c 173 07 GeV c 0 charge gt 2 3 po 2 3 r 2 3 4 0 spin gt 1 2 1 2 1 2 1 up top gt Higgs gluon bogon charm 4 8 MeV c 95 MeV c 4 18 GeVic q 1 3 g 1 3 g 1 3 de BE lu 12 v2 dy lt lt down strange bottom LE S P 105 7 MeV c 4 1 2 0 511 MeV c 1 777 GeVic A 12 electron lt 2 2 e
42. edefined functional entities like transceivers or encoders They can easily be configured with a graphical interface to meet the user s individual requirements Moreover the input and output pins of the logic circuit can be assigned to the real physical pins of the FPGA and further options like a pre emphasis see Subsection 6 2 4 or differential termination can be enabled or disabled Also the compilation of the design is done within Quartus II This process consists of multiple steps After analysing the code for syntactic errors it is synthesized and fit to the logic blocks on the FPGA considering timing constraints as well as available resources such as memory blocks registers or clocks Finally a netlist is written which serves to be loaded on the FPGA by the Device Programmer To do this the FPGA development board can be connected to the computer via USB 45 5 2 2 ModelSim ModelSim is a software by Mentor Graphics for the simulation of the logic and timing behaviour of an HDL design 76 All external signals such as the position of switches or the duration of pushing a button can be specified by the user Therefore ModelSim is a powerful tool to verify the logic validity of a code especially because besides the output signals the software also allows to display internal signals which cannot be physically measured in practice 5 3 Firmware BERT Implementation The schematic in Figure shows the architecture of the fi
43. een all traces and a Kapton thickness of 50 ym plus aluminum ground plane Length 10 cm 65 FPC FPC 0 channel 16 Transmitter side Interchange ends Receiver side Figure 6 14 Interchange of the ends of the flexprint to compare different knee lengths using the same channel 66 Part Ill Conclusion amp Outlook 7 Conclusion For this thesis Kapton flexprints have been produced with a laser system at the Hei delberg University for the Mu3e experiment Bit Error Rate Tests BERTs at a trans mission rate of 800 Mbit s have been performed which resulted in upper limits for the bit error rates below O 107 Furthermore eye diagrams have been analyzed to understand which factors mainly affect the signal quality 7 1 Manufacturing of Kapton Flexprints It has been shown that flexprints with a horizontal trace width and a trace separation of gt 120 um can be manufactured with a high reliability i e without any broken traces These can also be combined with 45 sections and vertical sections of traces with a width of gt 150 um The vertical sections should rather be short and can be used as pads to connect the flexprints to the FPC connectors Also flexprints with smaller horizontal trace widths down to 100 um can be manu factured but show broken traces mostly 1 3 in 34 sometimes more With the used laser platform it is not possible go below a trace separation of 100 um which
44. efined as the ratio of the differential voltage Vaig and the current J on one line 47 Due to an inductive coupling between the lines a current J in one trace will cause a current a J in the second trace with a 0 1 being the coupling constant A calculation which can be found in 48 yields Zaif 2Zo 1 a 3 3 The differential impedance is of particular importance for reflections which occur if wires or other electric components with different characteristic impedances are con nected The reflection coefficient I is given by r Zaif Zaiff a 3 4 Zaift a Zaitt b where Zaift a and Zaig y correspond to two arbitrary connected components and Zair R in case of an ohmic terminating resistor T is in the range of 1 to 1 and negative val ues correspond to a reflection with a phase shift by 7 Obviously the characteristic 19 impedance of all components of an electric circuit should be matched as good as possi ble to ensure proper signal propagation and minimize reflections 49 3 2 2 Microstrips Microstrips are a certain type of transmission lines which are commonly used on printed circuit boards PCBs and flexprint cables see Chapter 4 They are suited for signals in the microwave range 300 MHz to 300 GHz Differential Microstips A microstrip consists of a conducting trace which is separated from a ground or other constant potential plane by a dielectric layer called substrate
45. elds TRS S Cross Section of Differential Pair Driver Current Source Receiver 4 Figure 3 1 Schematic drawing of a basic LVDS circuit Also the field coupling of the differential pair is shown 38 the mechanical and electronic components i e the hardware of a physical link but no encoding schemes or protocols 37 Figure shows the architecture of an LVDS link A current source is injecting a constant current of 3 5 mA into the circuit Transistors work as switches to control the direction of the current towards the receiver At the receiver side there is a 100 Q terminating resistor at which a voltage of 350 mV drops according to Ohm s law The receiver which is a comparator with a transition threshold of about 50 mV senses the polarity of the signal to determine the logic state being transmitted LVDS has the advantage of generating very low electromagnetic noise due to the closely coupled wires As they carry the same current in opposite directions most of the radiation is cancelled For the same reason an LVDS link is relatively insensitive to external electromagnetic noise because the noise will nearly equally affect both wires and cancel out This means that there is only very little cross talk between adjacent wire pairs even if they are close to each other Another considerable advantage of LVDS is the low power consumption because of the low voltage and little radiative losses compared to other conce
46. esults for two or more errors in a row But since a BER far below 10 10 is headed for it is appropriate to assume that there are never two subsequent error bits Din Dexp Dait Errcount Odee O0000000pin Odee 000000006n 00000000 0 37d bit flips 334ec 00100001 pin laec 00000001 pin 00100000 1 2dec OOO000010pin 34dec 00100010b n 00100000 2 3dec 00000011pin 3dec 00000011 pin 00000000 2 Table 5 2 Example for an error being counted twice 5 4 BERT Results Before transmitting data via the self manufactured flexprints the correctness of the BERT implementation was tested by directly connecting transmitter and receiver with the help of the HSMC loopback board Unless manually injected by using a push but ton not a single error was detected independent of the frequency 100 MHz 200 MHz 800 MHz After that multiple flexprints with different trace densities lengths and knees have been tested The BERs can be seen in Table The results were obtained for a data transmission at a rate of 800 Mbit s and disabled pre emphasis see Subsection 6 2 4 In all shown cases not a single error has been detected Consequently only upper limits are given using equations and Broken channels i e non conducting traces which were not carrying a proper signal but rather random noise have been ignored for the BER calculations 49 trace width w um 100 150 100 100 trace width w um 100 100 100 100 100 trace sepa
47. f the Knee Length Ut Conclusion amp Outlook Conclusion 7 1 Manufacturing of Kapton Flexprints 12 BERT Results s soent o soroek a a a oa ee 7 3 Eye Diagram Results a 7 4 Recommendations ooo a a a nn 8 1 Consequences for the Mude Experiment 8 1 1 Spatial Constraints o o a 0000 2 eee eee 8 1 2 Transmission Errors 2 2 2 2 Ho oo nn 8 2 Further Work List of Figures 71 71 71 72 73 74 75 76 80 Part Theory amp Background 1 Theoretical Background 1 1 The Standard Model 1 1 1 The Elementary Particles The Standard Model of particle physics SM I is a quantum field theory that com prises the description of the smallest constituents of matter the elementary particles as well as the electromagnetic the weak and the strong interaction Only gravity cannot be described The SM was developed throughout the second half of the 20th century and has passed countless experimental tests until today Especially the recent discovery of a new particle which is likely to be the long predicted Higgs boson has given further support to this theory 2 3 According to the SM the fundamental particles comprise six quarks six leptons and their corresponding anti particles Furthermore there are four types of gauge bosons and the Higgs boson see Figure 1 1 The quarks and leptons form three generations Th
48. flip in the data stream 17 Word Data dp 1 dp 1 K Word Data dp 1 dp 1 D x 0 000 1011 0100 K x 0 000 1011 0100 D x 1 001 1001 K x 1 001 0110 1001 D x 2 010 0101 K x 2 010 1010 0101 D x 3 011 1100 0011 Kx 3 011 1100 0011 D x 4 100 1101 0010 K x 4 100 1101 0010 D x 5 101 1010 K x 5 101 0101 1010 D x 6 110 0110 K x 6 110 1001 0110 D x P7 111 1110 0001 K x 7 111 0111 1000 D x A7 111 0111 1000 Table 3 2 3b 4b encoding scheme For several 2 bit words two different patterns with disparity dp 1 exist For D x 7 either P7 or A7 is used to ensure that in the resulting 10 bit pattern never more than five equal bits occur The K x y words can be combined with the K 28 word of Table to form a control word 26 3 2 Transmission Lines Only for direct and very low frequency alternating currents electric wires can be char acterized solely by their ohmic resistance If the wavelength of a signal is in the order of the length of the link this simple model is not sufficient any more Instead a description as a transmission line is appropriate This comprises not only the resistance of the wire but also its capacitance and inductance see Figure 3 4 In place of the total resistance R the inductivity L and the capacitance C their normalized i e per length dx counterparts R L and C are used G represents the conductance of the dielectric material between the transmission line and ground Within this model a real
49. for the Mu3e experiment because of the multiple scattering dominated elec tron interactions with matter According to 64 multiple scattering can be quantified by Oms being the root mean square RMS of the central 98 of the planar scattering contribution It is given by T T 1 i log 4 1 Oms x x 0 038 08 0 where x is the material thickness d multiplied with its mass density p x pd and Xo the radiation length which can be approximated by 64 716 4 g cm A4 Z Z 1 log 287 VZ where A is the mass number of the nucleus The radiation length characterizes a 4 2 0 material with regard to the energy loss of electromagnetically interacting particles With the values from Table 4 1 the ratio of ms for aluminum and copper comes out to be Omsai d 12 um Oms cu d 5 um 0 59 4 3 Consequently an aluminum thickness of 12 um leads to a clearly decreased amount of multiple scattering compared to commercial leading edge flexprints e g offered by Dyconex which provide a minimal copper thickness of 5 um see Table 3 3 Even repeating the calculation with a double layer of aluminum gives Omsai d 24 um Omus cu d 5 um 0 87 4 4 thickness d mass density p atomic number Z mass number A rad length Xo Al 12 um 2 699 g cm 13 26 98 24 01 g cm Cu 5 um 8 960 g cm 29 63 55 12 86 g cm Table 4 1 Some physical properties of aluminum and copper 1
50. has been shown that flexprints with a trace width of 120 um and a trace separation of 110 um can be manufactured reliably with a laser platform available at the Heidelberg University In addition bit error rate tests have been performed which resulted in bit error rates below O 10 1 for 17 parallel channels at a transmission rate of 800 Mbit s each Moreover eye diagrams have been analyzed to understand which factors mainly affect the signal quality Zusammenfassung Das Mu3e Experiment hat zum Ziel den neutrinolosen Myon Zerfall ete e mit einer Genauigkeit von einem in 10 Zerf llen oder besser zu suchen Dieser Zerfall ist leptonzahlverletzend und damit im Standardmodell so stark unterdr ckt dass selbst ein einziger gemessener Zerfall ein klarer Hinweis auf neue Physik ware Das Konzept des Mu3e Experiments ist es eine sehr gute Impuls und Zeitaufl sung und Vertexrekonstruktion zu erreichen um den Untergrund ausreichend stark zu unter drticken Da die Zerfallselektronen eine niedrige Energie von bis zu 53 MeV haben ist haupts chlich Coulomb Vielfachstreuung der limitierende Faktor f r die Impuls und Vertexaufl sung Deshalb soll innerhalb des Detektors so wenig Material wie m glich verbaut werden Um dies zu erreichen sollen Pixeldetektoren mit einer Dicke von ledig lich 50 um verwendet werden die von einer Struktur aus Kaptonfolie getragen werden Die Daten werden ber ultrad nne Kaptonflexprints zu Front End FPGAs
51. ich grows with the cable length This meets the expectation since the ohmic resistance of the aluminum traces rises with their lengths In addition the jitter is enhanced for the longer cables Nevertheless the basic eye shape is conserved and the signal transmission works properly 57 Figure 6 8 Eye diagram of an 800 Mbit s data transmission via a flexprint with a trace width and a trace separation of 150 um a Kapton thickness of 50 um plus aluminum ground plane and a length of 50 cm Another eye diagram which is shown in Figure has been recorded for a 50 cm flexprint In contrast to the flexprints discussed above it had a trace width and a trace separation of 150 um Compared to the eye diagrams in F igure 6 7 one observes a further decrease of the eye height and less distinct dips This meets the expectations due to higher ohmic losses caused by the length Surprisingly the amout of jitter is reduced towards the 30 cm flexprint in Figure but a reason for this observation could not be found 58 6 2 3 Influence of the Transmission Rate In Figure eye diagrams for different transmission rates are depicted In all four cases one observes two slight dips before a strong dip appears approximately 1 7 ns after a transition followed by further slight dips Since the eye shape does not depend on the cable length as seen in Subsection 6 2 2 it is likely that this strong dip occurs because reflections at the HSMC connector
52. ie mediaviewer Datei Leitungsbelag svg 2014 Online accessed 11 August 2014 M Johnson and W Graham High Speed Signal Propagation Prentice Hall PTR 2003 E Bogatin Training for signal integrity and interconnect design http www ewh ieee org r5 denver rockymountainemc archive 2000 diffimp pdf 2000 79 48 49 50 ol 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 80 D Brooks Differential Impedance What s the Difference Printed Circuit De sign 1998 P Vielhauer Theorie der bertragung auf elektrischen Leitungen Verlag Technik 1970 D Brooks Printed Circuit Board Design Prentice Hall PTR 2003 S Gevorgian and H Berg Line Capacitance and Impedance of Coplanar Strip Waveguides on Substrate with Multiple Dielectric Layers IEEE Xplore year un known Wirth Elektronik Basic Design Guide 2012 http www wedirekt de web live media 16_wedirekt spezifikationen leiterplatten_daten_ downloads WEdirektBasicDesignGuide pdf Dyconex Design Guide 2013 External 2013 Available on request via e mail T Hebbeker Calculating Upper Limits with Poisson Statistics http web physik rwth aachen de hebbeker 13note 2633 pdf 2001 I Narsky Estimation of Upper Limits Using a Poisson Statistic arXiv hep ex 9904025 page 24 1999 ON Semiconductor Understanding Data Eye Diagram Methodology for Analysing
53. ik with whom I have spent much time in the lab and at the laser platform I also profited a lot from his experience Qinhua Huang who taught me much about VHDL and supported me on FPGA programming Dirk Wiedner who gave me valuable of support on technical and electronic issues and helped me to design my HSMC Flexprint Adapter Board Niklaus Berger who is an expert in FPGA programming and was able to solve every problem Qinhua and I were facing Ann Kathrin Perrevoort and all the others who were always there to answer questions proof read my thesis with great commitment and gave a lot of sugges tions Finally the whole Mu3e group for providing a very pleasant and cordial work atmosphere and sharing a time full of new experiences 83 Erklarung Ich versichere dass ich diese Arbeit selbstst ndig verfasst und keine anderen als die angegebenen Quellen und Hilfsmittel benutzt habe Heidelberg den 15 09 2014
54. int has also been produced with a trace width from 100 150 um and trace separations between 110 150 um In Figure one can see a flexprint on which all traces have the same separation of 150 wm To meet the 0 5 mm pitch of the Flexible Printed Circuit FPC connec 34 tor which was used to clamp the flexprints it was necessary to introduce 45 sections Therefore this type of flexprint had to be manufactured with a laser velocity of 65 b Microscopic image of the transition between main section the 45 section and the contact pads Figure 4 5 Flexprint with a trace width of 100 um a trace separation of 150 um between all traces in the horizontal section and trace width and separation of 175 um each in the 45 section plus aluminum ground plane Length of the flexprint 10 cm A third type of flexprints has been produced to examine the feasibility of designing flexprint cables with an FPC connector rotated by 90 Such a flexprint can be seen in Figure Also this flexprint type had to be manufactured with a laser velocity of 65 This arrangement of the FPC connector is of particular interest for the assembly Mu3e experiment because of the spatial constraints 4 5 1 Limitations Since the laser focus has a radius of approximately 100 um it was expected that the minimal trace separation which can be achieved corresponds to this value As described in Section 4 4 2 the laser mode changes to vectoring for all lines
55. is notably reduced in comparison to leading edge flexprints applying copper see Section 4 3 An enabled pre emphasis enhances the signal quality and increases the slew rate of the signal Nevertheless also without this option enabled a flawless data transmis sion is possible at 800 Mbit s Although pre emphasis involves further electronics 69 namely an additional LVDS driver which leads to a higher power consumption this option is foreseen on the final Mupix chip 82 7 4 Recommendations In the present work it could be shown that it is possible to produce Kapton flexprints with the PLS6MW Multi Wavelength Laser Platform and transmit data at a rate of 800 Mbit s with a BER below O 1071 at 95 CL A proper trace separation works down to 110 wm and the trace width can be decreased down to 100 wm However it has been observed that this minimal trace width leads to broken traces and is therefore not suitable for mass production with the used laser platform whereas for trace widths gt 120 um the connectivity is reliable In addition it has been seen that a ground or other constant potential plane notably improves the signal quality Based on these results the following recommendations see also Table 7 1 can be made Since a reliable connectivity of the traces is crucial a trace width of 120 um is recommended This can be combined with a trace separation of 110 um For traces which are routed with an angle of 45 to
56. its n 9 10 In principle a BERT can be realized by sending a deterministic datastream and com paring the incoming bits to the expected pattern at the receiver side 25 The exact architecture of the BERT used for this thesis is described in Chapter 5 Mathematically a BERT can be understood as a classical counting experiment As for every transmitted bit there are two discrete results false or correct it obeys a binomial distribution For data transmission at high rates n becomes large within seconds and the probability p for an error bit should be very small p lt 1 due to a good physical link Therefore the binomial distribution converges to a Poisson distribution if u dim p n gt 0 and the probability P k to find k error bits in a large number of transmitted bits is given by P k a eh 3 11 For large n the Poisson distribution converges to a normal distribution so that the standard deviation is given by o Vk Thus for the BER one gets k k per 4 VE 3 12 n n Upper BER Limit If no error is detected i e k 0 an estimation for an upper limit of the BER must be found In the following a perfectly working data checker is assumed so that no error counts occur accidentally i e there is no background and no error is overlooked According to 54 and 55 a Bayesian approach with a flat prior distribution is used Assuming a Poisson distribution and not observing any error bits kobs 0 from 5
57. le As light materials such as Kapton foils can ripple or might even be blown away by the air stream they should be adhered to a sticky rubber mat Layouts must be designed as vector graphics with a third party graphic software such as CorelDRAW and different laser settings can be assigned to the RGB encoded colors black red green yellow blue magenta cyan and orange The laser platform is then addressed by the Windows Print System like an arbitrary printer 58 4 2 Kapton Kapton is a multi purpose polyimide film developed by DuPont It has an excellent balance of electrical thermal mechanical physical and chemical properties and was used in applications with a wide temperature range from 269 C up to 400 C 59 Furthermore it can be laminated with a thin metal layer or glued to another film of Kapton For the Mu3e experiment 25 wm Kapton Type HN is planned to be used for flexprints and the mechanical support structure to minimize the material budget inside the active detector region 4 2 1 Physical Properties Kapton is mechanically stable though flexible and retains its physical properties over a wide temperature range It does not melt or burn and has the highest UL 94 flamma bility rating V 0 60 Between 360 C and 410 C a second order phase transition occurs which is assumed to be a glass transition According to the manufacturer different measurement techniques result in different transition tempe
58. lt 100 um in a layout 35 a Photo of the full flexprint b Microscopic image of the transition between main section and 45 section and the contact pads Figure 4 6 Flexprint with a trace width of 100 um a trace separation of 150 um between all traces in the horizontal section and trace width of 140 um and a trace separation of 150 um in the 45 section Length of the flexprint 10 cm When using the laser it always rasters first before it vectors Therefore the idea was to produce flexprints first with one thick trace instead of a pair and then to cut these with a single laser movement However it was observed that the mechanical laser positioning system is not precise enough to separate the thick trace exactly in the middle see Figure 4 7 Due to this imprecision one rather obtains two highly asymmetric traces or the laser even cuts so close to the edge that it simply reduces the width of the single trace For the same reason it is not possible to process the Kapton foils with multiple laser settings successively Trying to correct the offset in the layout itself does not solve this problem because overlapping colours encoding different laser settings are not converted correctly 36 Figure 4 7 Attempt to cut a single trace into two thin traces 4 5 2 Obtaining Different Types of Microstrips The aluminum traces on the processed Kapton strips correspond to coplanar striplines as they are mounted
59. ly the main challenges for the detector design are to handle high data rates and to have a very efficient accidental background suppression For the latter a very precise vertex fitting lt 200 um as well as a momentum resolution below 0 5 MeV c and an excellent time resolution lt 100 ps are needed The main limiting factors for the momentum and vertex resolution is multiple Coulomb scattering of the decay electrons in the detector material Thus the material budget of the detector must be as low as possible 2 1 1 The Signal Decay The muons will be stopped in a target to decay at rest Thus the total momentum of the decay electrons is vanishing 3 Prot Y i 0 2 1 1 The decay at rest constrains the total energy to be equal to the rest mass of the muon 3 Esos Y Ej my 105 7 MeV 2 2 i 1 In summary the signal decay is given by three electrons with a vanishing total mo mentum and an energy between mec 0 5 MeV and 1 2 my 53 MeV coming from a common vertex and being coincident in time 2 1 2 Background Decays Any background is due to fake signals that can be divided into two groups internal conversion and random combinatorial background Internal Conversion Internal conversion is the decay pt eteTetvev see Figure 2 1 which has a branch ing ratio of 3 4 10 I Here the emitted photon immediately converts into an electron positron pair Considering vertex and timing this
60. m which is much more than that of produced the flexprints 37 um for coplanar striplines 74 um for differential microstrips For this reason and for mechanical stability i e to be able to push the flexprints into the FPC connector without bending them it is necessary to thicken their ends For this purpose double sided adhesive tape was used As shown in Figure it was glued onto the back side of the flexprint with one protective foil left differential microstrips or a small piece of a slightly thicker plastic foil coplanar striplines Thus a thickness of approximately 350 400 um was obtained which even exceeds the required 300 um and worked well In case of the coplanar striplines this procedure leads to an increased thickness of the dielectric at the ends of the flexprint The dielectric constant of neither the adhesive 38 tape nor the plastic foil are known but a generic calculation with an assumed dielectric thickness of pure Kapton of 350 um yields Zo 174 Q instead of 234 Q In case of the differential microstrips the adhesive tape does not affect the dielectric properties of the flexprint since it is placed below the shielding aluminum plane Fragility Is has been observed that traces break very easily if the flexprints are unintentionally folded while pushing them into the FPC connector Consequently they need to be handled carefully and the adhesive tape is absolutely necessary for mechanical stability However
61. mean time 7 for an error to occur can be estimated errors 1 BER 8 1 total bits 800 Mbit s channels 7 en gt T gt 86 min 8 2 Note that the true value for 7 might be much larger because no error bit has been observed Since the pixel detectors will be read out in time frames of 50 ns see Subsection 2 1 3 this means that only a very small fraction of lt 107 of the readout frames will contain an error bit if at all whereas it is very unlikely that two error bits occur in the same frame The firmware or software should be able to detect a single error bit so that such an unemployable readout frame can be discarded Due to the very low rate of frames containing error this would not harm the experiment 72 8 2 Further Work Several tasks are left which must be focused on in subsequent works These are The flexprints which have been produced so far did only act as test cables to show that data transmission at 800 Mbit s works in principle and to find out suitable trace widths and trace separations which can be manufactured reliably Further investigations have to be done on flexprint designs which really fit the Mupix chips and the spatial constraints in the detector It has been seen that an aluminum ground plane notably increases the signal quality since it suppresses crosstalk between adjacent trace pairs and presumably also enhances impedance matching Still it has the dra
62. merge at the decomposition of the polyimide film Since aluminum evaporates at 2519 C the Kapton is certainly heated above 500 C for a short time Considering also the manufacturer information that a second order phase transition is 39 a Processed front side b Back side Figure 4 9 Microscopic images of flexprints showing the blackening of the processed Kapton assumed to occur between 360 C and 410 C 60 it is likely that mainly two things happen e CO and CO emerge inside the Kapton which starts decomposing due to the high temperature e The Kapton becomes expansible and soft as the thermal energy of the polyimide molecules exceed the binding energy of the hydrogen bonds As a consequence the gas bubbles can easily expand inside the Kapton and lead to buckles at the surface which are stronger on the back side of the flexprint as the air stream of the laser platform could push them down on the front side The front side of the Kapton which has been processed by the laser and carries the aluminum traces is observed to be darker than the back side One possibility is that this is due to burned residues of the glue between the Kapton and the aluminum foil An additional effect could be that the aluminum might not completely be evaporated by the laser but red hot particles fly away and possibly burn holes into the front side of the Kapton To find out whether this effect is enhanced by blowing the compressed ai
63. nd for clock recovery a fixed number of K 28 5 words see Subsection is sent before the actual user data transmission begins Once synchronization is achieved actual user data is transmitted To check the incoming bit stream for errors it is compared to the expected pattern As the data consists of a simple counter this can be easily implemented Figure shows a flow diagram of the BERT procedure The incoming data Din n at a clock cycle n is incremented by 1 and compared to Din n 1 the data received in clock cycle n 1 by using the logic operation XOR The number of ones in the resulting 8 bit string Daig n 1 corresponds to the number of wrongly transmitted bits in Din n 1 presumed that Din n was correct If an error bit occurs and is detected correctly Dexp n 2 will be wrongly calculated for the incoming data in the next clock cycle Consequently every error bit is counted Dexp n 1 D n n 1 Daipp n 1 Dexp n 1 XOR Din n 1 ETT count ETT count count ones Daisp n 1 Corr count COTT count 8 count_ones D gig n 1 1 ETT tot 2 ETT count COTT tot COTT count Figure 5 6 Schematic of the functionality of the data checker 48 twice see Table 5 2 Therefore the total number of error counts Errcount is divided by two to get the true error number Error It should be kept in mind that this test concept only works for rare errors i e it does not yield reasonable r
64. nesses Figures and do not significantly deviate from each other This is comprehensible when regarding the results from Subsection In case of the coplanar striplines the amount of crosstalk is much larger than for the differential microstrips since the aluminum ground plane effectively shields adjacent trace pairs and minimizes crosstalk In contrast the basis eye shape does not differ In all three cases there are two slight dips after a transition and for those bands remaining at a logic level for longer than one unit interval there is another stronger dip somewhat after the next transition Furthermore all dips are nearly of equal intensity In Subsection it has been seen that the signal deformation for differential striplines on 50 wm Kapton with ground plane are not dependent on the length except for the attenuation of the amplitude This led to the conclusion that reflections due to improper impedance matching are not observable in the eye diagram Since the intensity of the dips does not notably differ for the different microstrip types it is likely that the effect of impedance matching is negligible in all three cases To support this theory the runtime of the reflected signals are estimated The dis tance the signal has to run comprises twice the cable length of 20 cm and a trace length of 4 mm from the FPC connector to the via pair where the signal is picked up 62 A propagation velocity of 2 3 of the speed of light
65. om Altera from modified of the flexprints The FPC connectors from Molex have 34 gold plated contacts with a pitch of 0 5 mm and require an FPC thickness of 0 3 mm For the data input and output of the development board the HSMC port A see Figure suits best because it provides 17 differential LVDS pairs suitable for trans mission rates up to 800 Mbit s 70 The pin assignment of the Stratix V allows to locate the transmitter and receiver pins in opposite rows so that a simple loopback card see Figure 5 4 can be used to send the data directly back and test the setup even without any flexprint a Bottom side of the adapter board with an b Top side of the adapter board with two HSMC connector FPC connectors Figure 5 3 Pictures of the HSMC Flexprint Adapter Board 44 gt TP4 2007 ALTERA CORP ia sv i e a HA a AAA 4 2 Figure 5 4 The Altera HSMC loopback board front side and back side view 5 2 Software 5 2 1 Altera Quartus II Quartus Il is a software package providing the complete environment needed for the design of programmable logic devices in particular FPGAs and can be used for the synthesis and analysis of HDL hardware description language designs from scratch 73 Therefore the logic needs to be described in VDHL 74 or Verilog HDL 75 or assembled by using a graphical interface It is also possible to use the MegaWizard Plug in Manager to input megafunctions which are pr
66. on a dielectric without a ground plane To obtain differential microstrips an extra aluminum layer had to be added For this purpose another strip of aluminum Kapton foil of the same size has been glued onto the back side of the flexprint Gluing the aluminum layer directly to the back side yields a dielectric thickness of 25 um whereas gluing Kapton on Kapton yields 50 um For gluing a two component adhesive was used which is also used to build the Kap ton support structure for the pixel tracker After the adhesive was dripped onto the Kapton with the help of a syringe the two strips were pressed together for 20 24 hours for curing Thus a very homogeneous distribution of the adhesive was obtained 37 Figure 4 8 Photo showing the thickened ends of two flexprints Top adhesive tape plus plastic foil Bottom adhesive tape with one protection foil left 4 5 3 Mechanical Properties Connectivity and Interfaces To achieve a proper connectivity of the flexprints to the FPC connector they have to be pushed in and adjusted carefully since the contact pins have a width of only 200 um This works best with the help of a microscope Especially newly produced flexprints often had small aluminum filaments at their ends which occasionally shorted two traces or contacts and thus prevented proper signal propagation These filaments could be easily removed by hand or with compressed air The FPC connectors require a flexprint thickness of 300 u
67. ous LVDS bitstream The signals were conducted via the flexprints and fed back to the FPGA where they were checked for bit errors Figure 5 1 Picture showing the setup used for the BERTs comprising the FPGA development board the HSMC adapter board and the self manufactured flexprint The hardware components needed are described in detail below followed by a pre sentation of the necessary software packages After that the explicit architecture of the firmware which implements the BERT is presented followed by a discussion of the results 42 5 1 Hardware 5 1 1 Field Programmable Gate Array A field programmable gate array FPGA is an integrated circuit IC in which a logic design can be implemented For this purpose it consists of multiple electronic components such as logic array blocks adaptive logic modules and embedded mem ory blocks which can be arbitrarily interconnected 67 68 The interconnections are realized by electrical switches i e transistors in the FPGA so that the device needs to be reprogrammed after every power off In comparison to application specific integrated circuits ASICS FPGAs have a lower logic density a higher power consumption and lower clock frequencies But once ASICs are produced their logic cannot be changed anymore whereas FPGAs can easily be reconfigured This advantage compensates the above named drawbacks for many applications 68 For this thesis a Stratix V by Al
68. practical purposes a 0 and therefore Zar 2Z is assumed for coplanar striplines within the scope of this thesis Typically PCBs and flexprints with differential microstrips are produced in litho graphic processes consisting of many steps in which multiple substrate and copper layers are assembled An overview of typical dimensions of microstrips in standard and leading edge processes is given in Table 3 3 copper thickness t trace width w trace separation s dielectric thickness h Multilayer PCB standard process 52 2 50 pm gt 100 um gt 100 um gt 60 um Multilayer PCB advanced a approx 25 30 um gt 75 um gt 75 um gt 60 um Flexprint gt 5 um gt 20 um gt 25 um gt 25 um leading edge 53 Table 3 3 Typical parameters for microstrips on PCBs and flexprints 3 3 Signal Quality Checks Once a physical link has been realized it needs to be tested to ascertain how reliable data can be transmitted There are mainly two concepts which will be introduced in the following 21 3 3 1 Bit Error Rate Tests BERTs A bit error rate test BERT is a method to determine the quality of a transmission link for digital data transmission The bit error rate BER is the ratio of the number of wrongly transmitted bits k and the total number of transmitted bits n and represents an estimation for the probability p for an error to occur if one bit is transmitted error bits k BER p al RER total b
69. pts of signaling The power consumption is particularly low if it can be ensured by the transmitter side that there is no disparity i e the number of ones equals the number of zeros transmitted because in this case there is not even a net current averaged over time see 8b 10b encoding below 14 1 Datastream Vem 1 25V Figure 3 2 A binary datastream and the corresponding voltage dropping across the terminating resistor 39 An LVDS connection is a serial link which means that the data is transmitted bit after bit and not in parallel except when using several LVDS links in parallel In Figure an arbitrary binary datastream and the corresponding voltage levels are shown 3 1 2 Data Encoding Encoding is an injective mapping to transform data i e a series of logic states into a pattern suitable for transmission The purpose of so called line codes is to specify how the bit states 0 and 1 are represented in a signal Because different transmission media have different properties there are several schemes to match these characteristics The simplest possible code is the non return to zero code NRZ see Figure 3 3a Ones and zeros are represented by two distinct logic levels of the signal and form a con tinuous sequence Using this code can cause problems if a long sequence of subsequent ones or zeros is transmitted because in this case the logic level remains constant over a long period so that the syn
70. r at the red hot particles and thus supplying oxygen the compressed air was exchanged by Argon No observable difference has been noticed Consequently the oxygen does not seem to be the reason for the blackening of the Kapton 40 Change of the Dielectric Properties The influence of the blackening on the dielectric properties of the Kapton is hard to es timate One possible assumption is that due to the gas bubbles which have originated inside the Kapton the dielectric constant becomes inhomogeneous and is somewhat decreased in average To get an impression of the order of magnitude a decline of 20 is assumed which is certainly overestimated Then Equation 3 5 yields a differential impedance of 98 Q instead of 91 Q in case of differential microstrips with a trace thick ness of 100 um and a trace separation of 150 um on 50 um Kapton Two important things can be observed e Zas is only weakly dependent on er i e a change of 20 of e implies a change of only 8 of Zaig e From Zaim x 1 y41 14 er it can be concluded that the differential impedance grows with a declining dielectric constant Consequently the influence of the partial decomposition is likely to have only a small impact on the differential impedance 41 5 Performance of BERTs In order to perform BERTs an appropriate test setup has been developed which is shown in Figure The concept was to implement a data generator in an FPGA to output a continu
71. ration s um 150 150 150 150 distance between pairs um 150 650 150 650 length of working cm 10 10 20 30 channels 14 17 15 16 Table 5 3 BERT results for various coplanar striplines All upper limits are given at 95 CL trace separation s um 150 150 150 150 150 distance between pairs um 150 150 150 150 150 length of working cm 10 20 30 30 30 channels 15 14 10 12 16 BER lt 4 57 lt 2 04 lt 1 88 lt 2 09 10715 10713 10715 105 BER lt 1 19 lt 3 83 lt 1 33 lt 7 09 lt 2 25 Table 5 4 BERT results for various differential microstrips on 25 um Kapton All upper limits are given at 95 CL trace width w um 100 100 100 110 100 110 120 130 140 150 trace separation s um 150 150 150 110 150 150 150 150 150 150 distance between pairs um 650 150 650 650 150 650 650 650 650 650 length of working cm 10 20 30 30 50 50 50 50 50 50 channels 17 15 14 16 2 16 17 17 17 17 10 15 To 10744 10714 107 BER lt 2 76 lt 3 68 lt 5 05 lt 3 28 lt 9 44 lt 3 07 lt 4 14 lt 6 12 lt 3 02 lt 8 31 Table 5 5 BERT results for various differential microstrips on 50 um Kapton All upper limits are given at 95 CL 50 10715 10 15 10715 10713 102 10713 10 10 13 10 13 10 16 6 Analysis of
72. ratures 60 4 2 2 Electrical Properties The dielectric constant e of Kapton depends on the relative humidity of the air as well as on the temperature and the frequency In Figure 4 2 e is plotted versus these quantities Assuming normal conditions and a relative humidity around 50 er is approximately 3 4 27 Dielectric Constant 4 0 se 38 bas u 36 e 32 8 2 5 Esa 3 3 0 o v a a 3 2 2 8 3 0 2 6 0 20 40 60 80 100 100 0 100 200 300 Relative Humidity Temperature C a e vs rel humidity at room temperature b e vs temperature for two frequencies type HN film 25 um type HN 25 um 3 6 0 C 35 296 s 40 C 3 3 100 C E 32 E 31 8 30 200 C soc A 2 8 300 C I 27 02 103 404 40 107 10 10 101 10 Frequency Hz c e vs frequency at various temperatures type HN 25 um Frequency Hz d e vs frequency at 25 C and 45 rel humidity type HN 125 um B corresponds to the same measurement as A after conditioning the film at 100 C for 48 h Figure 4 2 Dependence of the dielectric constant e on various quantities from modified 28 4 3 Aluminum Though aluminum has only 63 of the electrical conductivity of copper and a lower heat conductivity 61 it is widely used in electronic applications due to its low mass density and easy processing 62 63 It also has a much lower atomic number Z than copper which makes it particularly interesting
73. re 1 2b et q H Va m a Dominant SM muon decay the Michel b Strongly suppressed decay u gt etete decay with neutrino oscillation Figure 1 2 Feynman diagrams of possible SM muon decays 10 As the W mass of 80 4 GeV c is much higher than the neutrino mass differences lt 2 eV I the decay u ete e is strongly suppressed with a branching ratio BR below 10754 and thus unobservable There are BSM theories that predict a much higher BR for this decay introducing new tree couplings see Figure 1 3a or loop contributions with new particles see Figure 1 3b Therefore any observation of this process would be a clear hint for new physics A yr Acn rz E r m ALS e Nr PO Ce a Tree diagram involving an b Penguin diagram with a SUSY unknown particle X loop Figure 1 3 Feynman diagrams of possible BMS muon decays 0 1 2 Experimental Situation Since 1953 experiments are performed to search for lepton flavour violation in muon decays see Figure 11413 Until today only upper limits for the branching ratios were found The current upper limit for the y etete decay is set by the SIN DRUM experiment 14 whereas the best measurement for zt ety was performed by the MEG experiment 15 SINDRUM From 1983 86 the SINDRUM experiment was in operation at the Paul Scherrer Institut PSI in Villigen Switzerland Because no signal event was detected an upper limit for th
74. rmware implementing the BERT Though not shown in the figure 17 parallel channels have been realized in the design to be able to examine crosstalk at a high trace density In the following the depicted components will be described part by part 8b 10b LVDS transmitter encoder data generator 8b 10b decoder data checker LVDS receiver Figure 5 5 Schematic of the BERT implementation 5 3 1 Data Generator As 8b 10b encoding is used it is reasonable not to exceed word lengths of 8 bits Hence the output data is chosen to be a simple counter from 0 to 28 1 255 which increments by 1 every clock cycle and restarts from 0 when reaching the limit To make sure that adjacent channels do not transmit equal data neighbouring data generators start counting from different values see Table 5 1 Otherwise crosstalk might possibly not lead to errors because of constructive interference 46 channel no start value 0 4 8 12 16 00000000 pin Odec 1 5 9 13 01000000 pin 64aec 2 6 10 14 10000000pin 128dec 3 7 11 15 11000000pin 192dec Table 5 1 Start values of the 17 data generators 5 3 2 8b 10b Encoder The 8b 10b encoder maps 8 bit words onto 10 bit patterns as described in Subsection 3 1 2 to achieve DC balance and ensure a high rate of transitions which is necessary for clock recovery The encoded bitstream can be regarded as random on a short time scale i e if sending less than 256 words It wa
75. s not necessary to write an own implementation of the encoder as a block of open source code from Critia Computer Inc could be inserted into the project 77 5 3 3 LVDS Transmitter For the transmitter the Altera megafunction ALTLVDS_TX was used It serializes the incoming 10 parallel bits from the encoder and outputs them as a differential signal The output pins of the transmitter block are directly assigned to the physical LVDS ports of the FPGA The transmitter is the only entity on the transmitter side with an external clock input Its output clock is fed back to all other entities on the transmitter side namely the data generator and the encoder to guarantee synchronization 5 3 4 LVDS Receiver For the LVDS receiver Altera provides the ALTLVDS_RX megafunction which is the counterpart to the transmitter presented above It performs clock recovery and deserializes the incoming bit stream The recovered clock is used for all entities on the receiver side 5 3 5 8b 10b Decoder To recover the data from the encoded bit stream an open source code from Critia Computer Inc was downloaded and integrated into the project just like for the encoder 47 5 3 6 Data Checker Directly after booting or resetting the system one can not immediately send user data because the transmitter and receiver are not synchronized i e the receiver does not know where a 10 bit word ends and thus where to cut the bit stream For this purpose a
76. tera has been used This FPGA is optimized for high bandwidth applications 69 and is planned to be employed on the readout boards in the Mu3e detector see Subsection 2 1 4 5 1 2 FPGA Development Board The Stratix V is mounted on a PCB which provides a hardware platform for the develop ment and prototyping of high performance designs see Figure B 2 Besides supplying power for the FPGA it provides a variety of communication ports for several interface standards In addition it bears three push buttons and eight DIP switches which the user can integrate in his design to communicate with the Stratix Moreover output data from the FPGA can be sent to 16 LEDs and an LCD display 70 5 1 3 HSMC Flexprint Adapter Board The Altera High Speed Mezzanine Card HSMC standard specifies a high performance interface for the connection of secondary PCBs named mezzanine cards to the host board of an FPGA and allows for fast differential signaling on multiple parallel chan nels 72 In order to create an interface between the FPGA Development Board and the flex print cable an adapter board had to be designed as part of this thesis see Figurel5 3 On the bottom side it hosts an HSMC connector to be plugged into the HSMC port of the development board and on top two FPC connectors are mounted for the connection 43 HSMC Port A LEDs push buttons DIP switches LCD display FPGA Figure 5 2 The Stratix V GX FPGA Development Board fr
77. the main section an increased trace width and trace separation of each 150 175 um should be used Also structures perpendicular to the main direction can be included in the design but should rather be used as pads for the connection of FPC plugs or the like An additional aluminum plane improves the signal quality and delivers more space for power supply and ground so that it should be used for the Mu3e experiment even though it slightly increases the material budget trace width um trace separation um horizontal 120 110 45 gt 150 gt 150 vertical gt 150 gt 150 Table 7 1 Recommended structure sizes for the Mu3e experiment 70 8 Outlook 8 1 Consequences for the Mu3e Experiment 8 1 1 Spatial Constraints When regarding the spatial constraints of the Mu3e detector the structure sizes shown in Table are sufficient for flexprints to connect the Mupix chips to the front end FPGAs The inner double layer will have a polygonial shape with an active length of 12 cm consisting of 12 and 18 sides respectively Since the final chip size of the Mupix chips will be 1 x 2 cm in the inner region 6 chips will be used on every side of which 3 will be connected upstream and 3 downstream Each of these chips will have 3 LVDS output ports Consequently 9 parallel LVDS channels need to be placed on one flexprint A trace width of 120 um and a trace separation of 110 um lead to a total width of 0 414 cm for the horizontal section Sin
78. there is a strong disparity when comparing the different channels For the flexprint used in this comparison it can be assumed that the reflections at the HSMC connector dominate the total reflections on the receiver side because these are of first order whereas reflections caused by an improper impedance matching of the flexprint are of second order or higher see Figure 6 6 Therefore it is likely that the differences between the eye diagrams seen in Figure are primarily caused by different runtimes of the reflections due to different trace lengths on the receiver side of the PCB and possibly also on the FPGA development board 54 a Channel 0 PCB trace length at b Channel 0 Flexprint connected with transmitter receiver side 46 1 mm 26 2 mm interchanged ends c Channel 11 PCB trace length at d Channel 11 Flexprint connected with transmitter receiver side 38 7 mm 35 54 mm interchanged ends e Channel 16 PCB trace length at f Channel 16 Flexprint connected with transmitter receiver side 40 7 mm 40 0 mm interchanged ends Figure 6 4 Eye diagrams picked up at different channels 800 Mbit s transmission via a flexprint with a trace width of 100 um a trace separation of 150 um for pairs and 650 um between pairs and a Kapton thickness of 50 um plus aluminum ground plane Length 10 cm do FPC FPC Transmitter side Interchange ends Receiver side Transmitter signal picked up here Figure
79. tons which hit a carbon target where pions are created as secondary particles Slow pions decay immediately into muons which are collected in the 7E5 beamline so that a low momentum muon beam with a rate of 2 10 1 s is provided for a first phase of the Mu3e experiment In a second phase the Mu3e experiment aims to improve the sensitivity by another order of magnitude Therefore a more intense muon beam is needed Such a beam could be provided by the planned High intensity Muon Beamline HiMB for which the protons from the cyclotron mentioned above will be shot on the Swiss Spallation Neutron Source SINQ target 33 There a high number of muons will be created as a by product and could be collected by the HiMB to provide a muon rate of 2 10 1 s 11 4860 Pixel Sensors 4000 Fibres 7000 Tiles JOE 142 FPGAs FPGA 48 FPGAs y i up to 56 800 Mbit s links 1 6 Gbit s link each 12 10 Gbit s links per RO Board Subfarm D Gbit Ethernet Data Mass Collection Storage Server Figure 2 6 Schematic of the Mu3e readout chain consisting of three stages 12 3 Basics of Data Transmission Data transmission is the transport of information from one point to another whereby the data is represented by a physical signal 34 3 1 Signals A signal is the time dependent magnitude of an observable Electrical signals can be represented by a voltage or current Also electromagnetic waves optical or radio can
80. ts 21 4 1 Some physical properties of aluminum and copper 29 4 2 Calculation of the impedance for coplanar striplines on 25 um Kapton 30 4 3 Calculation of the impedance for differential microstrips on 25 um Kapton 30 4 4 Calculation of the impedance for differential microstrips on 50 um Kapton 30 4 5 Laser settings used to produce test patterns and flexprints 31 4 6 Laser settings used to produce test patterns and flexprints 32 5 1 Start values of the data generators 22222222 47 5 2 Example for an error being counted twice 2 2 2 2 2 49 5 3 BERT results for various coplanar striplines 50 5 4 BERT results for various differential microstrips on 25 um Kapton 50 5 5 BERT results for differential microstrips on 50 um 50 7 1 Recommended structures sizes for the Mude experiment 70 Bibliography U 2 10 11 12 J Beringer et al Particle Data Group Review of Particle Physics RPP Phys Rev D86 010001 2012 G Aad et al ATLAS Collaboration Observation of a new particle in the search for the Standard Model Higgs boson with the ATLAS detector at the LHC Hyper fine Interact 214 1 3 5 11 2012 arXiv 1207 7214 hep ex S Chatrchyan et al CMS Collaboration Observation of a new boson at a mass of 125 GeV with the CMS experiment at the LHC Phys Lett B 2012 arXiv 1207 7235 hep e
81. ugust 2014 Altera High Speed Mezzanine Card HSMC Specification 2009 Altera Quartus IT Handbook Volume 1 Design and Synthesis 2014 V A Pedroni Circuit Design with VHDL MIT Press 2004 S Palnitkar Verilog HDL A Guide to Digital Design and Synthesis Prentice Hall PTR 2003 ModelSim SE User s Manual 2002 Ken Boyette http read pudn com downloads104 sourcecode embed 425959 8b_10b 8b10_enc vhd__ htm 2006 Online accessed 25 April 2014 Ken Boyette http read pudn com downloads104 sourcecode embed 425959 8b_10b 8b10_dec vhd__ htm 2006 Online accessed 25 April 2014 Tektronix Digital Phosphor Oszilloskope DPO7000C Serie Datenblatt 2013 Altera DPO7000 Series Digital Phosphor Oscilloscopes Quick Start User Man ual Tektronix Differential Probes TDP1500 and TDP8500 Datasheet 2013 I Peric Personal Contact September 2014 81 82 Acknowledgements At the end of my thesis I would like to thank everyone who supported me to carry out this thesis First of all I would like to express my appreciation to Prof Dr Andr Sch ning who gave me the opportunity to work in the Mu8e group He helped me especially by critically questioning my strategies and supposing new approaches I would also like to thank Prof Dr Ulrich Uwer for being my second examiner Moreover I would like to express my gratitude to the following persons in no specific order Carsten Grzes
82. ure a schematic of the Mu3e readout chain is shown The pixel sensors as well as the fibre and tile detectors will be connected to so called front end FPGAs field programmable gate arrays see Subsection 5 1 1 These FPGAs will be located directly on the beam pipe inside the detector 10 The connection between the pixel detectors and the front end FPGAs will be realized by serial 800 Mbit s LVDS links low voltage differential signalling see Subsection 3 1 1 which will consist of tiny aluminum traces on Kapton The manufacturing and perfor mance tests of these so called flexprints are the main scope of this work see Chapter 4 From the front end FPGAs the data will be sent out from the detector to FPGA driven readout boards via high speed optical links Thus a galvanic separation is guaranteed The data from the readout boards will then be transmitted via high speed optical links to a GPU graphical processing unit filter farm where an online track and event reconstruction will be performed Events of interest will then be sent to a data collection server which stores them in a mass storage system 2 1 5 The Muon Beam As an extremely high number of muons needs to be stopped to decay a very intense muon beam is required Therefore the Mu3e experiment is supposed to be run at the Paul Scherrer Institut PSI in Switzerland which operates the world s most intense muon source The PSI operates a cyclotron to accelerate pro
83. wback of an increased material budget which leads to stronger multiple scattering It should be tested whether a thinner aluminum thickness of lt 12 um can function as a ground or power plane Research has to be done on the question how to bond the Mupix chips to the Kapton flexprints So far no statement could be made about the lifetime of the Kapton flexprints The aluminum might possibly suffer from oxidation which could decrease the conductivity to the FPC connectors or make it more vulnerable to bending since aluminum oxide is a brittle material 73 List of Figures A 3 AA 4 pa o pad q 4 as e a oad 5 1 The internal conversion decay ppt efeTetvave A aa ates E pao e ls E oe a a a e 7 2 3 Possible combinatorial background o o 8 2 4 Schematic drawing of the detector design 2 2 2 2 2 9 2 5 First mechanical prototype of the Kapton support structure for the inner len A og a ote e did GA AOR DR hud ce CRE Spee Ge did 10 2 6 Schematic of the Mude readout chain 12 3 1 Basic LVDS circuit 2 2 Co oo oo rn 14 Peer ere 15 VOTE a Re ee 16 3 4 Equivalent circuit diagram of a short piece ofa transmission Ime 18 are 20 23 24 4 1 PLS6MW Laser Platform 0 0 00000008 8 26 4 2 Dependence of the dielectric constant e on various quantities 28 4 3 Test structures to examine which minimal structure sizes can be achieved eee
84. x Wikimedia Commons Standard Model of Elementary Particles http en wikipedia org wiki File Standard_Model_of_Elementary_Particles svg 2014 Online accessed 10 Februar 2014 Y Fukuda et al Super Kamiokande Collaboration Evidence for Oscil lation of Atmospheric Neutrinos Phys Rev Lett 81 1562 1567 1998 arXiv hep ex 9807003 Q R Ahmad et al SNO Collaboration Measurement of the charged current interactions produced by B 8 solar neutrinos at the Sudbury Neutrino Observatory Phys Rev Lett 87 071301 2001 arXiv nucl ex 0106015 K Eguchi et al KamLAND Collaboration First results from KamLAND Evi dence for reactor anti neutrino disappearance Phys Rev Lett 90 021802 2003 arXiv hep ex 0212021 F P Anetal Daya Bay Collaboration Observation of electron antineutrino dis appearance at Daya Bay Phys Rev Lett 108 171803 2012 arXiv 1203 1669 hep ex L Canetti and M Shaposhnikov The v MSM and muon to electron conversion experiments Hyperfine Interact DOI 10 1007 s10751 013 0796 7 2013 A Blondel et al Research Proposal for an Experiment to Search for the De cay u eee ArXiv e prints 1301 6113 January 2013 arXiv 1301 6113 physics ins det W J Marciano T Mori and J M Roney Charged Lepton Flavor Violation Experiments Ann Rev Nucl Part Sci 58 315 341 2008 Y Kuno and Y Okada Muon Decay and Physics
85. ycle it does the same with a small time shift so that after multiple acquisition cycles a complete waveform record with a very high sample density is obtained see Figure 6 2b By this the sampling rate can be increased far beyond the real time sampling rate which is limited by the shortest period between two acquisitions 80 Record points JL To No 1st acquisition cycle II N 2nd acquisition cycle A ML nn 3rd acquisition cycle SN nth acquisition cycle b Repetitive scanning of a waveform to obtain a very high sample density Figure 6 2 Schematic showing the FastAcq concept both taken from 80 6 1 2 Tektronix TDP3500 Differential Probe The Tektronix TDP3500 Differential Probe is an active probe which can be connected to an oscilloscope via the TekVPI probe interface It features very good common mode rejection and a high bandwidth of 3 5 GHz With a very low differential input capaci tance of lt 0 3 pF and a high input resistance of 100 kQ it provides a high fidelity signal reproduction for high frequency measurements 52 6 2 Eye Diagram Results In the following several parameters have been varied and compared to estimate their influence on the signal quality If multiple eye diagrams are shown side by side all settings and in particular the scales are equal unless explicitly stated In addition the signals have been picked up from the same channel except in 6 2 1 to achieve maximal comparability

Download Pdf Manuals

image

Related Search

Related Contents

Mode d`emploi  Manual de instrucciones COCINAS  Manual paineis linha GT  A propos de terroristes algériens à la tête de groupes  Baxi Potterton Promax SL User's Manual  Sequence Detection Systems Chemistry Guide  触れるほどに、拡がっていく航海の自由  

Copyright © All rights reserved.
Failed to retrieve file