Home

Position Acquisition and Control for Linear Direct Drives

1. Unregister isr void as interrupt handler this also unregisters the read handler Destroy Data FIFO Reset the register BAR2 0x02 of the PCI DASO8 Board K2 K3 Destroy Command FIFO RT Module unloaded Figure 3 17 Module cleanup 84 Linear Drive for Material Handling Figure 3 18 shows the state diagram of the Monitoring State Machine implemented in the Real Time Module This state machine is updated every second every 10 000 in terrupts The inputs and outputs of the monitoring state machine are briefly described below see also section 3 1 2 1 eE1 input active low state of the Infeed Module and of the two Monitoring Modules E1 0 signalises that the said modules are operational e E2 input active low state of contactor K1 through which the Siemens devices are supplied E2 0 signalises that K1 is closed e EM input active high state of the interface electronics power supplies EM 0 signalises the failure of at least one power supply eK2 output active high enables the On button T2 which in turn when pressed closes contactor K1 e K3 output active high drives an internal contactor of the Infeed Module which when closed supplies the DC Link The states of the Monitoring State Machine are described in Table 3 2 Initialisation MODA R320 7 DC Link Enable Command EM 0 Pu LL bo EM 0 K3 0 E2 0 Y A M6 nn M
2. monitoring h Static int El E27 EM Monitoring Si amp aitee me 23 in outputs static enum Monitoring MO M1 M9 states mon state definition void monitoring void Monitoring state machine control h include parameters h include transforms h include pi control h include modulator h include position h amp amp 6 static double u dc void ctrl init void void machine_control void Initialisation of the control parameters Machine control fun ction implements the control algorithm capture h define DATA LEN 30 Capture static double capture data DATA LEN buffer def static unsigned long capture length Capture static unsigned capture decimation parameters void data capture void Data capture function the rtf_put FIFO DATA amp capture data capt buffer sizeof capture data js written in Data FIFO Figure 5 4 Real time module detailed structure 1 3 131 Appendix e parameters h e transforms h e pi control h e modulator h modulator h include parameters h include transforms h define define define define define define define INV BRD IA INV BRD IB INV BRD IC INV BRD WRN SZT INV BRD TA INV BRD TB INV BRD TC define INV BRD IRQ START define INV BRD IRQ STOP static struct segment short wrn szt sector short ta
3. M G TIS wor f an S i L ii Vr mmu HE SIE T di Display IE reer on BBC back view Wd TIN HE TEE B TALL AL eae S Aan TE ae drew out _ Main Switch D Reserve place for 7 B Power Stacks Figure 3 10 Cabinet 12 Linear Drive for Material Handling The Power Stacks are equipped with internal current sensors but they are only used for over current protection The feedback values of the phase currents needed in control are acquired using external LEM sensors which provide better quality signals as required e g by sensorless control A diode bridge with integrated brake chopper contained in the Infeed Module supplies the DC link bus which feeds all the Power Stacks This module 6SN1145 1AA01 0AA1 has a nominal power of 10 kW and a peak power of 25 kW The Infeed Module also supplies the internal electronics IGBT drivers internal current sensors protection circuitry etc of the first seven Power Stacks while for supplying of the internal electronics of the remaining Power Stacks two Monitoring Modules 6SN1112 1AC01 0AA1 are necessary The Infeed Module and the Monitoring Modules are connected to their respective Power Stacks through a Siemens proprietary Devices Bus Ger tebus see the diagram in Figure 3 11 Each Power Stack is equipped with a plug in Power Stack Interface card developed at our department 45 This card e implements the Inverter Bus communica
4. 0 5m s 100 a tacens ee 100 T z Ije10m s s 50 IvI 10m s v increases v 0 5m s v 0 12m s Imaginary j c eo m T f Er M EE T 100 Iv 0 5m s 5000 4000 3000 2000 150 100 50 0 50 Real o Figure 3 37 Root locus of the mechanical observer observer This however should have no influence on the estimation dynamic if the decoupled EMF observer as described by Eq 3 22 is used For speeds lower than vo the mechanical observer is still stable but the dynamic depreciates with decreasing speed For speeds lower than approx 0 12 m s the dominant poles of the observer move to right half plane leading to instability It can be seen from Figure 3 37 that as the speed v approaches zero the poles of the mechanical observer converge to the origin of the s plane One must however note that in the right half plane the dominant poles remain close to the imaginary axis slow dynamic which makes possible to reverse the vehicle s moving direction using EMF based sensorless speed control this implying a passing through zero of the speed so long the reversal occurs fast enough see also the experimental results in the next subsection Up to this point only one stator segment was considered in the analysis of the mechanical observer If two consecutive stator segments m and n with no relative phase displacement are considered Eq 3 48 can still be used
5. By introducing in Eq 3 12 the electrical angular speed w the speed v of the vehicle and the EMF constant Ke respectively _ d O Tp T Q dt V C Kc Yomo Eq 3 13 P the EMF vector can be written ea _ sin 0 e Ke HE Eq 3 14 The EMF constant Ke the ratio between the modulus of the EMF vector and the speed is determined experimentally for each segment of the machine This constant depends through Ypmo on the air gap 6 which has variations especially in the curved sections of the track see Figure 3 5 For each segment the mean value of Ke will be considered The procedure used to determine the EMF constant as well as the resulting values for all 18 segments of the linear machine are given in Appendix 5 6 Assuming that the speed v varies much slower than the electrical quantities the derivative of the EMF vector can be expressed as dile cos0 T 2 Introducing Eq 3 14 in the above expression the derivative of the EMF vector yields VT all s is al vz g e Eq 3 16 Tp Eq 3 6 can be rewritten as follows MEINE Ha mp la Eq 3 17 U i dt Pig e q Based on Eq 3 16 and Eq 3 17 the electrical subsystem of one segment of the linear machine can be modelled by the following state equations 62 00 1 0 Tis 100 0 1 ER Dy O sR OU d is g o o y JO 1 0 Rus dt e pile 0 0 0 Oli Zu e 0 0 vn 0 e 00 0 0ji Tp The EMF observer will be based on the above model
6. static double CORR 1A 5001 4 static double CORR 2B 5001 4 static double CORR 3A 2501 4 static double CORR 3B 2505 4 typedef enum 1 S2D HO S2D H1A S2D H2B S2D H3A S2D H3B j t head nr typedef struct int use muxcode counter double sine cosine double x inc x abs t head data typedef struct t head data HA data t head data HB data t head nr current head short x is valid double x60 double x double v void update void t s2d data static t s2d data s2d void s2d update F void s2d_calc_head_x i void s2d_update Figure 5 6 Real time module detailed structure 3 3 EMF based sensorless position and speed estimation Position Interface Board addresses definitions Sensors period Period correction tables Head numbers definition the States of the Re construction S M Definition of one read head s data Definition of the Sensor to Digital data structure used to determine sensors position and speed including synchro nisation between read heads Function used to calc a read head s local position Sensors position and speed update function 133 Appendix 5 5 Mounting of the Optical Sensors at the Linear Drive for Material Handling The distance between the vehicle attached scale and the stationary read head is not drawn to scale Figure 5 7 Read heads mounting only one read head show
7. MC 1 Hi Scale covers head 2 USE 1 amp MC 3 USE 18MCG 2 Hi Scale covers head 3 USE 1 amp MC 4 USE 1 amp MC 3 Hj Scale covers head 4 USE 1 amp MC 5 __ USE 1 amp MC 4 H Scale covers head 5 first half USE 1 amp MC 5 USE 1 Hee Scale covers head 5 second half USE 0 USE 1 Initialisation gt H Vehicle outside sensors region Figure 2 16 Reconstruction state machine AX _X XgtXo te X X X t P 4AN P AN 2 P AN P 5AN X X Xo tjo X X X t P 4AN P AN 2 X X X t P 4AN 1 X X X ts P 3AN X X X X ts X X X t P 2AN 1 X X x t P AN X X4 P AN 2 x t X t 0 to t b t t4 t Ho Hin ig Hs x Hj EM Hs Hs Nd His Mi Hsp B His Ed Ho pum Hsp T Ha Hsp xS Ho Figure 2 17 Illustration of the principle of position reconstruction in C code 39 Position Sensing Systems for Passive Vehicles Xo to 0 Eq 2 4 Based on the position given by the current head and on current head s offset the global position x will be calculated x t Xo ty Xp t Eq 2 5 The reconstruction state machine then changes state to H 44 In the next cycles as long as USE remains 1 and USE 0 the position of head 1 A will be used to calculate the global position When Na gt AN the Sensor to Digital Board will send the position information from both head 1 A and 2 B signalled by USE 1 and USEs 1 this
8. Moreover for longer sensors Some meters the capacitances to ground increase and large current pulses will be necessary to charge all the capacitances producing electromagnetic interference EMI To avoid these problems we propose sinusoidal excitation voltages Uac t Ap cos t Up t An sin oyt Eq 2 29 with Oh 2T Eq 2 30 The excitation frequency fh is chosen e g in the 20 100 kHz range By substituting Eq 2 29 into Eq 2 25 the following expression is obtained for the charge amplifier s output Usu X AK cos a1 eos E x sin t sin E 3 Eq 2 31 Introducing the normalized position Xp 2n P x Eq 2 32 and reducing the trigonometric expression Eq 2 31 can be rewritten as Usu X AK COS apt xp Eq 2 33 The normalized position xp is the phase difference between the excitation input Uac and the output of an ideal charge amplifier Uo The simulation in Figure 2 37 illustrates the principle of phase modulation as described by Eq 2 33 The simulation was made for a very high traversing speed 20 m s only for better clarity When the incremental position of the sensor is zero the signals Uac and Us are in phase With increasing position the phase between these two signals increases proportionally for x 1 mm P 2 Ua and Uo are in phase opposition When x reaches P the phase difference reaches 360 and the two signals are again in phase By measuring the phase differences be
9. a 2 us variation of the time between the sampling instants at a mean travelling speed of approx 750 mm s as in Figure 3 33 produces a variation of the sampled position of approx 1 5 um which over the 100 us control cycle introduces a noise of ca 15 mm s in the speed signal The interrupt jitter see Figure 3 14 also contributes to the derivation noise by additionally increasing the variation of the time between the sampling instants see the spikes in Figure 3 33 Because of its high frequency the derivation noise could be almost completely eliminated by a first order low pass filter with a small time constant 300 500 us which filter would introduce only a small phase delay in the speed control loop However using such a filter has disadvantages at entering in the sensor s region Without filtering the first speed value is available with only 100 us delay with respect with the first position value as soon as the first two position values are acquired when using the low pass filter it can take up to ca 2 ms until the filtered soeed reaches steady state which translate into ca 20mm at the beginning of the sensors region where no speed signal is available considering the maximum travelling speed of 10 m s In order to eliminate the derivation noise without the use of any filter the Sensor to Digital Board s firmware must be modified by adding a counter which keeps track of the number of 2 us sampling interv
10. along with an offset correction will be called afterwards the a components two phased fixed statoric of the currents are calculated It must be noted that the current reading and the modulation information writing from to the Vehicle Control Interface Board are also necessary when the control algorithm is disabled for the correct functioning of the board s firmware 45 88 Linear Drive for Material Handling At this point the position information from the Sensor to Digital Board must be available This is checked by reading the value from ADDR STAT see section 2 2 5 As in the case of the PCI DASOS board if a communication error occurred an error flag is raised If the position information is available the sensors position and speed are calculated according to section 2 2 6 In order to proceed with the calculation of the control algorithm three conditions must now be simultaneously fulfilled e The Monitoring State Machine must be in state M4 meaning that the DC Link is powered up and there is no failure of the power supplies e There must be no communication error encountered up to this point e The user must have enabled the starting of the linear machine If these conditions are fulfilled then the control algorithm proceeds by updating the values of the control references limits and parameters this can be the case e g when passing from a stator segment to the next one which has different parameters Then it i
11. and also small angular tolerances This can be quite challenging with a linear machine where the guiding is not very stiff see section 3 2 The most important datasheet specifications of the used optical sensors are summarised in Table 2 1 Figure 2 6 shows the definitions for the positive moving direction and for the zero position When the sensor moves in positive direction the cosine signal has a phase of 90 with respect to the sine according to the trigonometric definition when the sensor moves in negative direction the cosine signal lags the sine 90 phase The periods are counted based on the zero crossings of the sine when the cosine is positive crossing from 4 to 1 quadrant increases the period counter while crossing from 1 to 4 quadrant decreases it The reference pulses are centred in the middle of the 1 quadrant and can have a width between 180 and 540 This means there is exactly one period count up or down depending on the moving direction during a reference pulse A read head Figure 2 5 LIDA 181 from Heidenhain Source 26 24 Position Sensing Systems for Passive Vehicles Specification LIDA 181 Measuring principle Imaging scanning Measuring standard Steel tape with AURODUR graduation Gap between scale read head 0 75 0 15 mm Grating period 40 um Thermal expansion coefficient 10 ppm K Accuracy grade 5 um Measuring length 220 mm Reference marks Selectable by magnet every 50 mm Maxima
12. but the annulus is now centred in the origin of the xy plane There is also an improvement of the mean amplitudes of the sine and cosine signals Using period corrections Figure 2 19 c the shape of the locus of the corrected sine cosine signals comes very close to that of the unity circle which indicates the reduction of the position error caused by the systematic errors There still are some slightly thick parts in the diagram oriented after the directions of the two bisectors of the xy plane They are an indication that the sensor also has small phase errors which errors were not compensated for by the correction tables Position Sensing Systems for Passive Vehicles a Without corrections Cosine p u 0 5 0 0 5 1 Sine p u b Mean corrections 0 5 5 amp um o O O 0 5 A l 1 0 5 0 0 5 1 Sine p u c Period corrections 4 05r Of ua c Oo e tm 1 1 0 5 0 0 5 1 Sine p u Figure 2 19 XY Representation of the sine and cosine signals of head 1 whole length 43 Position Sensing Systems for Passive Vehicles The position deviation between the tested optical system and the optical sensor used as reference is shown in Figure 2 20 a The position of the tested system was acquired every 100 us as described in the previous sections together with the one of the reference sensor during a test run The numbers 1 5 in the figure represent the region c
13. e Curved sections left right curves or up downhill e Sections with high force for acceleration e Sections for high precision positioning equipped with position sensors etc Two university departments cooperate in this project The Institute for Electrical Machines Traction and Drives from Technische Universitat Braunschweig is responsible for the design of a new type of electrical machine briefly presented in Section 3 1 1 and all the mechanical constructions track vehicles Our department is responsible for hard and software of power electronics and control of the proposed system Due to its high efficiency high power density and because it allows a higher air gap the Permanent Magnet Synchronous Machine PMSM is a good choice for this application As high acceleration is mandatory lightweight passive vehicles using an active track i e the long primary configuration of the PMSM moving magnets is the best suited With passive vehicles there is no energy or information transfer from the stationary side to the vehicle In order to allow for individual motion control of several vehicles the active track is separated into many segments 5 each segment being fed by the power stack of a dedicated inverter This approach has also the benefit of reduced losses by turning off the supply of the stator segments where there is momentarily no vehicle Due to the modular construction of the machine there are gaps in the stator windin
14. electrical angle and subsequently in the generated electrical force In order to avoid this step in the electrical angle a smooth synchronisation can be realised over a time interval AT starting at tsync 2 Using a ramp function defined as ee lt 0 sync t t C Ottan SAT Eq 3 65 b feuAl R t the position used in control can be determined during the time interval AT as follows Xc t R t xs t 1 R t x t Eq 3 66 A similar expression using the same ramp function can be used for the calculation of the speed vc during AT in order to avoid a step in the reference current due to speed estimation error 3 4 1 Leaving the Processing Station In this subsection experimental results acquired at the vehicle s leaving of the processing station are discussed Figure 3 44 shows the sensed and estimated position and speed as well as the respective estimation errors at the transition between the processing station and the subsequent transport section All the quantities were acquired in the same test run The control algorithm starts at time to At this time the vehicle is inside the processing station xs 0 013 m so the sensors provide valid position and speed information When the control starts the vehicle begins to move towards a commanded 120 Linear Drive for Material Handling position outside the processing station using initially the feedback information given by the position sensors The
15. only after all the supply voltages 5V2 5V3 5V4 5V5 have all reached their operational values This ensures the simultaneous supply of all the interface circuitry independent of the power up time constants of the different power supply units this is especially critical for the analog to digital converters which have both analogue and digital power supplies These voltages are also monitored during operation and in case of a failure the K6 and K7 switches are opened By interrupting the 5V2 power supply of the Power Stack Interfaces all the Power Stacks IGBTs are also turned off The state of the interface electronics power supplies is transmitted to the Control PC through the EM signal A detailed diagram of the Supply Control Board is given in Appendix 5 3 75 Linear Drive for Material Handling 3 1 2 2 Control Software For the implementation of the control algorithm a standard PC equipped with an Intel Pentium 4 processor and 512 MB of memory was used The only hardware modification brought to the PC was the addition of the three PCI interface boards see Figure 3 11 in the previous section Using a PC for the implementation of the control provides some advantages compared with other possible architectures e g based on a DSP board among them being a fast CPU with fully integrated floating point support FPU and a large amount of primary memory RAM which can be used as a transient recorder this is especially
16. se qeueA jeg0I9 qun Ja8jeJdJequ pueuJuJo 5 qegew B 3 buisn buisse2oJd SUINO SJ9 euleJed eulupel Transfer during initialisation ul91s S9 IJ Ja pueH pee OJ14 PUEUIWO S JojoeJeu OINPOW LA San eA Jo o q Mau e UdYM payed A spJeog o2eJIo1u Figure 3 15 Software Architecture 80 Linear Drive for Material Handling The control data acquired from the Real Time Module is also saved on the hard disk of the Control PC From here it can be analysed locally offline using Matlab or sent over Ethernet to another computer for further processing For the communication between the Real Time Module and the User Interface two unidirectional real time FIFOs are used The Command FIFO character device located at dev rtf0 used to send commands together with their parameters from the User Interface to the Real Time Module The Data FIFO character device located at dev rtf1 through which the values of the status and control variables of the Real Time Module are transmitted to the User Interface Each of the two FIFOs has an associated read handler a function that is called automatically when a writing occurs in the respective FIFO see Figure 3 15 The following commands are defined for the Command FIFO eCMD SET PARAMS Set the parameters of the machine segments This command is sent from the User Interface during initialisation once for each segment of the linear
17. the Capacitive Sensor Acquisition Board s CPLD firmware saves the four sampled voltages in the CPLD registers and asserts the IRQ10 signal This interrupt request triggers the execution of the Interrupt Service Routine in the control program Here the position information from the two sensors will be read through the ISA Bus The position information of the capacitive sensor consists of the four samples of the charge amplifier s output taken in one cycle of the excitation voltages Based on them the position of the capacitive sensor can be calculated using Eq 2 28 L 72 77 77 OO RER Po a E E N N q 15 us 3V 41 7 1 2vs 280 8 V H 2 vs 1 88 V Se Figure 2 33 Capacitive sensor s output voltage measured at the input of the A D converter 55 Position Sensing Systems for Passive Vehicles The position of the reference sensor will also be used as feedback for the position and speed controllers the output of which the linear motor s reference current is written back on the ISA Bus to the Reference Position Acquisition Board s CPLD and from there itis loaded in the digital to analog converter of the board In Figure 2 31 ideal excitation voltages are depicted With a real excitation e g during the rising edge of U and the falling edge of U the sum of the two voltages may not remain constant having a variation for some nanoseconds This variation causes a spike at the input of the ch
18. useful during the development and testing of the control algorithm As operating system a Vector Linux 5 8 SOHO distribution 47 based on the Linux kernel 2 6 17 48 was chosen Included in this distribution are also tools for the software development code editor compiler etc The Linux kernel by itself is not fully pre emptive 49 i e there are certain kernel functions which cannot be interrupted so it cannot be directly used for applications which require hard real time very fast and deterministic response to real time events In order to be able to meet the real time requirements needed in control the Real Time Application Interface RTAI kernel extension 50 developed at Dipartimento di Ingegneria Aerospaziale from Politecnico di Milano was used This extension introduces a small real time kernel between the system hardware and the Linux kernel used to manage the hardware events see Figure 3 12 This real time kernel captures the hardware interrupts and realises the scheduling of the corresponding real time handlers interrupt service routines with very short delays 51 52 From the point of view of the real time kernel the Linux kernel is just another task which is run in the idle time the time when no real time task is active and which can be interrupted at any time when a hardware interrupt requires service thus the Linux kernel becomes fully pre emptive A kernel patch containing the RTAI Hardware Abst
19. when the vehicle re enters the processing station 122 Linear Drive for Material Handling time tsync 2 the observed position X is wrapped around this is necessary due to the closed shape of the track and brought as close as possible to sensors position xs but without affecting the electrical angle used in control tone tens 2Ktr withkeZ Eq 3 67 sync 2 sync 2 so that X5 teyne2 Tp lt nes s Xs teyne 2 p Eq 3 68 After the wrapping of the observed position a small difference still remains between x and xs ca 3 mm due mainly to inaccuracies in determining the phase displacements between consecutive stator segments This corresponds to a phase difference which will be eliminated in the second step during the time AT using the ramp function defined in Eq 3 65 If the position control loop is active when the vehicle passes through the zero of the track the reference position x must also be wrapped In this case no ramping is necessary the wrapping will occur as soon as the sensed position is available i e at time to in Figure 3 45 x t x t x t xs t for t2 ty Eq 3 69 In the above equation the length of linear machine is subtracted from the reference position simultaneously compensating for any position estimation error One must note that the resulting reference position must not be too close to the zero of the sensors no positioning is possible until the synchroni
20. 2 values and limits Update the control parameters Is the current control necessary sensorless posi tion and speed Calculate sensorless EMF based position and speed new X Read the S2D transfer status ADDR STAT e Figure 3 19 Diagram of the control function 1 2 87 Linear Drive for Material Handling e Synchronise sensorless and S2D position and speed new Xe Vc Update the position controller new v Update the speed controller new i Is the vehicle covering segment m Calculate the segment m electrical angle Currents a gt dq transf segment m Update the segment m current controllers new le uU Voltages dq gt a transf segment m Calculate the new seg m switching states and times SV PWM 4 Is the vehicle covering segment n Y Calculate the segment n electrical angle Currents a gt dq transf segment n Update the segment n current controllers new Un te Voltages dq gt a transf segment n Calculate the new seg n switching states and times SV PWM Q 2 Write the modulation information to VCI Start a new A D conv on PCI DASOS board Control function return point Figure 3 20 Diagram of the control function 2 2 algorithm is enabled then the current conditioning function which performs the conversion from converter units to amperes
21. 3 4 Possible orientation adjustments certain conditions to the generation of incorrect reference signals by read heads which were not covered by the scale To avoid this problem a protection against direct light was also mounted above the L profile In order to obtain the best possible signal quality sine cosine reference the read heads were mounted in such a manner as to allow for small orientation adjustments in five axes as shown in Figure 3 23 The read heads are not mounted directly on the L profile but by means of assembly brackets The oblong holes on the assembly brackets D allow for translational adjustments along the z axis as well as for rotational adjustments around the y axis whilst the similar oblong holes 2 on the read heads make possible translational adjustments along the y axis and rotational adjustments around the z axis To also allow for rotations around the x axis the supplemental set screws 3 are used These four screws are inserted in the L shaped profile close to the corners of the assembly brackets base see also Figure 3 22 They can also be used for additional adjustments translation along the y axis and rotation around the z axis All the possible adjustments of the read heads orientation at the different mounting points are summarised in Table 3 4 92 Linear Drive for Material Handling Figure 3 24 Straight section of the linear machine with the optical sensors mounted In F
22. 30 e Spacing between transmitting and modulating plates d 0 4mm e Spacing between modulating and receiving plates d 0 4mm e Thickness of each of the three plates d4 1 5mm Electrical permittivities e Absolute permittivity of air approximated to that of vacuum y 8 854187817 10 F m e Relative permittivity of epoxy FR4 35 Based on the above values the different capacitances can be calculated using the homogeneous field distribution approximation e Average value of the position dependent capacitances h A A 1 d 93 amp 3 Cavg Nmp 697 W 4 082pF 128 Appendix Variation amplitude of the position dependent capacitances P223 NE sm E 0 980pF Car E Nmp g d 3 This less than 1 pF variation of the capacitances encodes the position information Coupling capacitance between modulating and receiving electrodes N io 26 P 2A h d Position independent capacitance between transmitting and receiving electrodes N Nann amp 0 w 2A h c No New 80 w 2A h p Nm tow _7 41pF d d d Input capacitances in parallel with the voltage sources N 5 4 W 2A h MS a 59 814pF j a b c d 3 Output capacitance short circuited by the charge amplifier Cout o 3 C 29 219 pF 341 795 pF Additionally for the charge amplifier feedback following components were used Feedback capacitance Feedback resistance Re 20MQ 129 Appen
23. De which are also listed in Table 2 3 At time tc after another delay period Tp data frame D will be seen by the PCI Interface FPGA and it will be saved at time t The remaining data frames will also be sent and saved in the PCI Interface FPGA registers at to tio t43 At t44 the Sensor to Digital has sent all the position information and stops driving the bus lines After receiving the last data frame t45 the PCI Interface does not immediately start to drive the bus a second high impedance state occurs between t 5 and tis and afterwards the PCI Interface drives all the bus lines at logic 1 It is possible to initiate a new transfer as soon as t1 The total transfer time on the Sensor Bus amounts to Tr 2 Tp 1 6 2 Tz 2 Tp which is about 3 5 us The bus will remain in the idle state all lines driven to 1 by PCI Interface until a new position request comes from the control program the time interval to tig in Figure 2 13 is the 100 us control cycle 2 2 5 PCI Interface Board The PCI Interface Board assures the data transfer between the Sensor Bus and the PCI Bus of the control PC It does not alter in any way the position requests from control program nor the position information sent by the Sensor to Digital Board acting solely as a data transfer relay It also provides galvanic separation between the control PC and the position acquisition electronics on the Sensor to Digital Board The realisation of t
24. International Power Electronics and Motion Control Conference EPE PEMC 2010 6 8 Sept 2010 Mihalachi M and Mutschler P Capacitive Sensors for Position Acquisition of Linear Drives with Passive Vehicles in Proc of the 12th International Conference on Optimization of Electrical and Electronic Equipment OPTIM 2010 pp 673 680 20 22 May 2010 Mihalachi M Leidhold R and Mutschler P Long Primary Linear Drive for Material Handling in Proc of the 12th International Conference on Electrical Machines and Systems ICEMS 2009 15 18 Nov 2009 Mihalachi M Leidhold R and Mutschler P Linear Drive System for Combined Transportation and Processing of Materials in Proc of the 35th Annual Conference of the IEEE Industrial Electronics Society IECON 2009 3 5 Nov 2009 Mihalachi M Leidhold R and Mutschler P Control of Segmented Long Primary Linear drives in Proc of the 7th International Symposium on Linear Drives for Industry Applications LDIA 2009 20 23 Sept 2009 Mihalachi M and Mutschler P Position Acquisition for Linear Drives A Compari son of Optical and Capacitive Sensors in Proc of the 34th Annual Conference of IEEE Industrial Electronics Society IECON 2008 pp 2998 3005 10 13 Nov 2008 Mihalachi M and Mutschler P Position Acquisition for Long Primary Linear Drives with Passive Vehicles in Proc of the 2008 IEEE Industry Applications Society Annual Meeting IAS 2008 p
25. Taking into C x WD Figure 2 30 Equivalent circuit of the capacitive sensor 51 Position Sensing Systems for Passive Vehicles account the equivalent circuit the output of the charge amplifier can be written sRg C x U Cp x U C x U C x U ge cud M eC out X 1 SRpCpR C x x C x C4 x OM Eques C where U is the sum of the four excitation voltages Up U U U U Eq 2 21 and s is the Laplace operator Introducing the line to line voltages hs U U er Up Ua Up os the following expression of the output voltage is obtained 1 gp N T MR a ML I sos x Jug sin 2x ESRC C 4C P P Eq 2 23 C 4Cavg 2 The term denoted with 1 contains the position information and the term denoted with 2 stands for a component which decays with t RrCr some milliseconds due to the high pass filter behaviour of the charge amplifier This component can be completely eliminated if Up t 2 0 By using the notation CC x MEL Cr Ck 4Cavg Eq 2 24 and assuming U t 0 and Rx gt the output of the charge amplifier can be written 2n 2n Usu X K Us COS EJ Uap sin 2x Eq 2 25 52 Position Sensing Systems for Passive Vehicles 2 3 3 Square wave excitation In this section the operating mode of the capacitive sensor with square wave excitation like the one used in the commercially available system will be i
26. as well as the implementation of the three observers structure at the linear drive for material handling including experimental results will be discussed in the following subsections Vehicle Inverter Bus N Vehicle Controller Interface SV PWM SV PWM it 3 NEL dq current dq curren ira SS m control SS amp speed m Vehicle Controller Position control References SUuOne3luJr EMF E Observer UCM u Observer SS m amp Observer Figure 3 34 The three observers structure integrated in the control diagram 103 Linear Drive for Material Handling 3 3 1 EMF Observers Throughout this subsection all quantities involved will refer to the same stator segment For simplicity any subscripts referring to the considered segment s number will be omitted A sinusoidal fundamental frequency model will be used for the electrical subsystem of each segment of the linear machine The voltage equation of the considered segment in the stationary a B reference frame can be written Ua R la d Ya U Zi i di LASi Eq 3 6 where Ua Up la Ip Ya Yg are the components of the segment s voltage current and flux linkage vectors respectively R is the phase resistance of the segment The flux linkage vector can be further expressed as the sum of the current dependent flux and the flux linkage generated by the permanent magnets Ta TE a Yon a l
27. be used see Figure 3 1 This Motion Coor dination PC must generate the position reference values for each vehicle For this a slower cycle time is sufficient typical values are around 1 2 ms In the same cycle time the Vehicle Controllers should send back their measured positions as well as their status for monitoring Then in each Vehicle Controller an interpolator should be integrated to generate the intermediate position reference values for the 100 us position control cycle Because the real time communication demand between the Control PCs and the Motion Coordination PC is less stringent compared e g with the demand on the Inverter Bus the Ethernet interface which is nowadays present in every PC could be used for this purpose 126 5 Appendix 5 1 Imaging Scanning Principle The LIDA linear encoders operate according to the imaging scanning principle To put it simply the imaging scanning principle functions by means of projected light signal generation two scale gratings with equal grating periods are moved relative to each other the scale and the scanning reticle The carrier material of the scanning reticle is transparent whereas the graduation on the measuring standard may be applied to a transparent or reflective surface When parallel light passes through a grating light and dark surfaces are projected at a certain distance An index grating with the same grating period is located here When the two g
28. calculated as the numerical derivative of the position aF i21 ixi Ts where Ts represents the sampling interval the time between samples i 1 and i The resulting position modulo p and speed are shown in Figure 5 9 c Finally the mean value of the ratio between the EMF module and speed over the N acquired samples was calculated Figure 5 9 d yielding the mean value of the EMF constant S feli i 1 KEmean N vii i 1 Table 5 1 summarises the measured EMF constants for all 18 stator segments Table 5 1 Measured EMF constant for all stator segments 136 Bibliography 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Hellinger R and Mnich P Linear Motor Powered Transportation History Present Status and Future Outlook Proceedings of the IEEE vol 97 no 11 pp 1892 1900 Nov 2009 Gustafsson J Vectus Intelligent Transport Proceedings of the IEEE vol 97 no 11 pp 1856 1863 Nov 2009 Bosshard C MagneTrak Paradigmenwechsel im Materialhandling Proc of 13 Deutscher Materialfluss Kongress Innovative Techniken f r die Logistik vol 1815 pp 189 200 Garching Germany 2004 St ppler G Segmentierte Langstator Linearmotoren f r die Schnellpositionierung von Werkst cktr gern in l ngsverketteten Montagelinien in SPS IPC DRIVES Electric Automation Systems and Components Nuremberg Germany 2008
29. can be detected with the help of this operation state signal The power supply units NT1 NT5 generate the supply voltages required by the interface electronics e The Power Stack Interfaces and the Vehicle Control Interface Board are fed with the 5V2 voltage for the digital circuits and with the 5V3 and 15V3 for the analogue ones e The Position Interface Board and the sensors pre processing unit Sensor to Digital are fed with 5V4 digital and with 5V5 analogue e The 24V6 generated by NT5 are used to supply the relays and switches on the Supply Control Board The Supply Control Board is connected through the PCI DASOS8 46 Interface Board from Measurement Computing company to the Control PC The functions to the realisation of which the Supply Control Board takes part are e Switching on off and monitoring of the DC link voltage through the E1 and E2 inputs and the K2 and K3 relay outputs as described earlier e Measuring of the DC link voltage The Siemens Devices Bus provides a voltage proportional to the DC link voltage This signal is routed to the Supply Control Board where is conditioned and then sent further to the PCI DASO8 Board Here the signal is digitised and made available to the control software e Switching on off and monitoring of the interface electronics power supplies The K4 relay on the board is used to control the K6 and K7 switches At power up these switches are simultaneously closed
30. capacitive sensor The second signal DATA SAVE is used to synchronise the saving of the position from the two sensors ensuring that the two positions are saved simultaneously The data from the reference sensor is saved on the hard disk drive of the control PC while the capacitive sensor data is saved in the external RAM of the DSP Board After an acquisition ends the data from the two sensors is transferred offline to an auxiliary PC where the position deviation is calculated One possibility to generate the four excitation voltages is shown in Figure 2 39 For long encoders and high excitation voltages providing improved signal to noise ratio the sinusoidal voltage may be generated in two stages First a small MOSFET H Bridge Control PC real time Linux Reference Position Acq Board Z z Current Board Bus Control GRE Reference Program Va integrated I D A u C Code E hie current control _ ee Data transfer gt p Linear Motor 2 over Ethernet A Quadrature CPLD Optical Optical Sensor 4 Pulses sensor data 7 GH 7 77 Vj FIR A F sync _ save DSP Board Auxiliary PC 3 Excitation Voltages Z e L A Control u f FE ore Fi 77 E Sianals N JTAG i Code Capacitive Sensor Cure MEY Signals i A p e B Wy ine s Interface Composer Offline pUnits i Mgpe D fen pes ee 8 External calculation Capture m RAM of position iW dure
31. control program reads from these addresses the position information as sent from the Sensor to Digital Board in the six data frames Di De Two 15 bit Sensor Bus data frames can be read in a single 32 bit PCI transfer e ADDR STAT This address gives to the control program read access to register R SENSOR BUS CONTROL DATA TR CTRL D0 15 RB DO 15 SELECT LDI 32 ADDR 32 COMM START SENSOR BUS CONTROL STATE MACHINE LOCAL CONTROL SIGNALS PCI BUS SIGNALS PCI BUFFERS SEN M ES D1 15 D2 15 D3 15 D4 15 RS 485 RECEIVERS RS 485 DRIVERS PCI INTERFACE FUNCTION LDO 32 SELECT D5 15 TO ALL REGS CLOCK RB D5 15 AND STATE RESET D6 15 MACHINES 4 77 RB D6 15 Figure 2 15 Simplified diagram of PCI Interface FPGA Firmware 37 Position Sensing Systems for Passive Vehicles STAT 4 which contains the state of the Sensor Bus Control State Machine from the PCI Interface Firmware and implicitly the Sensor Bus transfer status At the beginning of the control cycle the control program writes on the PCI bus the data frame Do at its address ADDR Do The Data Transfer Control State Machine decodes the address and consequently the data is saved in RB Do 15 from where it will be sent on the Sensor Bus Concomitantly it asserts the communication start signal COMM START signalling
32. corresponding bus registers e Sine cosine registers RB SA 12 RB CA 12 RB SB 12 RB CB 12 e Period counters RB NA 14 RB NB 14 e Multiplexers codes RB MCA 2 RB MCB 2 e Used valid head RB USEA RB USEB e Reference signals RB R1 RB R5 From these registers the position information will be sent further on the Sensor Bus The position information is updated in the Sensor to Digital FPGA every 2 us The content of the bus registers listed above is a consistent snapshot of this information taken every 100 us when it is required by the control The data stored in registers R ENA and R AN ss 14 is coming from the control program Register R ANsp 14 contains the real number of periods covered by head 5B as determined during a test run Its contents will be written only once when the control program starts A zero written in R ENA will determine the Synchronisation State Machine to ignore the reference signals thus effectively disabling all the position acquisition related firmware in the Sensor to Digital FPGA This behaviour is useful for the initialisation phase when the control must ensure that the vehicle is outside sensors region before enabling the Sensor to Digital algorithm State Ho corresponds in the CHSM to the vehicle being outside sensor s region For each of the six logical read heads 1 4 5A 5B there are two states The only difference between the two states is in the multiplex
33. d mem 0 2 0 15 0 1 0 05 0 0 05 0 1 0 15 0 2 0 25 Reference position m Figure 2 43 Position deviation using sinusoidal excitation There are several ways to reduce the errors of capacitive position sensors by software or by hardware A software method is the reduction of the harmonics by more sophisticated correction tables To generate them large amounts of sampled data are necessary to identify the characteristics amplitude and phase of some dominant harmonics The more sophisticated correction table should be loaded into a flash memory located at the sensor This could be preferably done at the manufacturer s test bed during the final production test before the sensor leaves the factory A hardware oriented way to reduce harmonics is an improved shaping of the modulating or even of the transmitting electrodes The model derived in section 2 3 2 and used in the implementation of both excitation methods is based on the assumption of homogenous field distribution But in reality the field will be quite different from this simple assumption As discussed in section 2 3 2 a large number of 3D FEM calculations may be used to optimise the shape of the modulating electrode such that the position of the slider can be extracted precisely from Uou This implies that the position dependent capaci tances contain a pure sinusoidal position dependency The optimisation will be a time consuming iterative design procedure where many 3D FE
34. direction of the sensors system was chosen along z axis see also Figure 5 7 in Appendix 5 5 because as specified by the machine s manufacturer this axis presents a higher guiding stiffness The distance between the optical read heads and the scale which is mounted on the underside of the load carrying plate on the vehicle must be kept in the 0 6 0 9 mm range for the adequate operation of the optical system 26 In order to avoid an impact between the scale and the read heads a collision protection system was also installed for each read head see also Figure 5 8 During the testing of the optical system it was noticed that due to the mounting position of the read heads variations of the ambient light s intensity could lead under Protection against direct light Load carrying plate y sae KET je u 33 x 3 Supplemental head orientation gt adjustment Carrying roller uono oejoJd uoisi Jo L profile Yj Gk Wi NN Ya Figure 3 22 Mounting of the optical sensors system Linear Drive for Material Handling T T Assembly Z V4 gY Bracket mounting Adjustments Points Tz Translation along z axis Ry Rotation around y axis in the xz plane Ty Translation along y axis Rz Rotation around z axis in the xy plane Ty Translation along y axis Rz Rotation around z axis in the xy plane Rx Rotation around x axis in the yz plane Table
35. e H H lt O Transition Read head 2 Read head 3 Sp S2 Cg C2 o Sa 3 Ca C3 0 4 0 3 0 2 0 1 0 0 1 lt x iim nm f H Middle of Read head 3 0 5 1 1 8 2 ss S Cas C3 Time ms g H End of Read head 3 Sp S3 Cg C3 0 2 0 0 2 0 4 Time ms Figure 3 27 Measured sine cosine signals continuation positive zero crossings of the sine signals can be noticed exactly one positive zero crossing of the sine signal while the corresponding reference signal is high compare also with Figure 2 6 The phase differences between successive read heads can be seen in Figure 3 27 c and e A decrease of the amplitudes of the sine cosine signals as the vehicle travels towards the end of the sensors region in positive direction can also be observed in Figure 3 27 This is also visible in the xy representation of all the periods of the four logical read heads from Figure 3 28 on the left column 96 Before correction Ca p u 1 05 0 05 1 CoB p u 1 05 0 09 1 Cota p u 1 0 5 0 0 5 1 1 Sa A p u 1 05 0 05 1 S50 P U 0 5 Linear Drive for Material Handling After correction 05 0 05 1 V Alcor p u 2 B corr p u 05 0 05 1 S2 B corr p u 3 A corr p u 05 0 05 1 S a A corr p u 3 B corr p u 0 05 1 S 3 B corr p u Figure 3 28 XY Representation of the sine cosine sign
36. edge of SB ACK it is copied to the bus registers On the falling edge of SB ACK t4 the transition to Hiao occurs CNTRg is disabled and loaded with Ng 1 the B position information is invalidated USEg 0 and the position evaluation will continue based only on the A values f Leaving sensors region moving in negative direction The vehicle is close to the left limit of the region covered by head 1 moving in negative direction CHSM is in state H44 4 and CNTR counts down At t the rising edge of R4 occurs but it has no effect At time to the zero position of the sensors system is crossed towards negative values At ti Na becomes less than 1 so at t4 one FPGA clock later CHSM changes state to Ho and another clock later 1 is stored in CNTRa The system is now again in its initial state waiting for one of the two reference signals 34 Position Sensing Systems for Passive Vehicles 2 2 4 Sensor Bus Communication The Sensor Bus connects the Sensor to Digital Board with the PCI Interface Board It is a 16 bit parallel bus implementing the RS 485 differential signalling standard Line 15 the most significant bit of the bus is used as control signal whilst the other 15 lines 14 0 are used as data lines So in one transfer frame 15 information bits can be sent simultaneously through the bus At one given time one of the two communication parties drives all the lines of the bus control and data This signall
37. etc eAs long as the User Interface is still active the shell script hangs waiting for its termination eWhen the user closes the User Interface program the shell script resumes unloading all the previously loaded kernel modules in reverse order In Figure 3 14 a measurement of the interrupt latency interrupt jitter and control algorithm processing time is depicted The measurement was done while the complete control algorithm including simultaneous current control for two stator segments position acquisition from the optical sensors and sensorless EMF based speed control was running on the Control PC under RTAI Linux The signal in the upper part of the figure is the control interrupt generated by the Vehicle Control Interface Board it was measured directly on the PCI bus The second signal is one output bit of the Control PC s parallel port which is set at the beginning of the Interrupt Service Routine and reset at its end thus this signal remains high as long as the control algorithm is executing The interrupt latency is defined as the time interval elapsed between the beginning of an interrupt request the falling edge of the interrupt signal and the beginning of the execution of the associated Interrupt Service Routine the rising edge of the parallel port output signal in this case The latency is mainly determined by delays in the interrupt controller hardware the time needed for the interrupt to be transm
38. heads 1 and 2 Scale covers head 2 first half Scale covers head 2 second half Synchronisation between heads 2 and 3 Scale covers head 3 first half Scale covers head 3 second half Synchronisation between heads 3 and 4 Scale covers head 4 first half Scale covers head 4 second half Synchronisation between heads 4 and 5 Scale covers head 5 first quarter Scale covers head 5 second quarter Synchronisation between heads 5A and 5B Scale covers head 5 third quarter Scale covers head 5 fourth quarter Vehicle outside sensor region initial state Reset gt Ho gt R 1 N lt 1 4 Hia L4 N gt AN 2 N lt AN 2 T Hia m n a SB ACK 1 N gt AN H a28 I ni Hostia N gt 4 SB ACK t Y N 1 R Hasa m v N gt AN 2 N lt AN 2 4 j EE Hp 82 aa N gt AN A SB ACK 1 N gt AN Hg f m H3 25 N gt 1 A SB ACK t N lt 1 May gt N gt AN 2 N lt AN 2 4 EE Hza aa N gt AN A SB ACK 1 N gt AN H5 48 zz i Hg N gt 1 a SB ACK t Y N 1 d Hasa m x N gt AN 2 N lt AN 2 4 E Has a ur a SB ACK 1 N gt AN Hjs sa m IE Hz 45 a gt 1 a SB ACK t Y N lt 1 Ei Hz m gt N gt AN 4 N lt AN 4 4 E Asa aa N gt AN 2 SB ACK 1 N gt AN 2 Hz sp zZ BE Hss sa N gt 1 A SB ACK t Y N 1 E Ase pm v N gt AN 4 N lt AN 4 4 H52 Na gt AN R 1 Reset gt Ho
39. is time t1 in Figure 2 17 The global position x will still be calculated based on the head 1 position information Now the local position of head 2 xs t can also be calculated in a similar manner to that of head 1 Knee t atan2 S t C t if S t 0 Enos Xincg t XincB t P Xp t P N t Kine B t Subsequently the offset of head 2 will be determined and stored in variable xo Xp t x t Xp t Eq 2 7 The state of the reconstruction state machine changes to H 2 and in the following cycles between t4 and t2 the global position x will be calculated based on the position given by head 2 B and on the offset calculated at time t4 XL tst ext ext etset Eq 2 8 At t Ng gt AN a new transition between heads 2 and 3 occurs and the offset of the new head 3 xo t2 will again be calculated in a similar manner During state H 34 a reversal of the travelling direction is assumed The global position continues to be calculated based on the A position information as previously described until t4 when the Sensor to Digital data USEs 1 MCs 2 indicates that the vehicle is back again in the region covered by head 2 It must be noted that the position x tz transition between heads 2 and 3 in positive direction and the position at x t3 transition between heads 3 and 2 in negative direction are not equal There are two reasons for this firstly the phase difference that ma
40. of the position signal the more complex period corrections will be used the amplitudes and offsets of the sine cosine signals will be determined for each 40 um period of each read head and stored in correction tables These tables will then be used by the control program for online corrections To calculate the period correction tables the Sensor to Digital Board will be used for the acquisition of the sensors signals A high enough number of samples e g 100 in each period of each read head is necessary In order to be able to acquire this high number of data points modifications were made in the Sensor Bus transfer protocol as well as in the Sensor to Digital and PCI Interface firmware The Sensor to Digital firmware will sample the sine cosine signals at a higher rate 41 Position Sensing Systems for Passive Vehicles 1 2 us and then send the data directly through the Sensor Bus in a so called Burst Mode In this mode a transfer rate of 20 Mbit s is achieved on the Sensor Bus The PCI Interface FPGA will temporarily store the incoming data in a firmware implemented FIFO buffer from where it will then be read from the control program and transferred on the hard disk of the control PC for subsequent offline processing Because the position is not available during the acquisition of the correction data no field oriented control can be used the vehicle must be moved using voltage frequency control or by hand The speed of the vehicle c
41. of the vehicle spacing between each two segments the synchronisation function also determines the addresses m and n of the Power Stacks which correspond to the stator segment s where the vehicle is located Due to the closed shape of the track of the machine used in our experimental setup a wrapping of the position xc must also be implemented Because the estimated position is not absolute and only one processing station exists in our setup the zero position of the processing station i e the zero position of the optical sensors system was chosen as zero position of the entire track More details about the position wrapping are given in subsection 3 4 2 Figure 3 43 shows schematically the implemented synchronisation procedure between the sensed and the estimated position and speed Let us assume that initially the vehicle is moving inside the processing station using feedback from the position sensors and approaching one of the processing station s Xc Xs position and speed used in control R t u tsync 2 AT Xc R 1 R X Vo Rv 1 R Xe Xs Ve Vs AN Leaving the processing station Speed used in contro Xc X AX IN Ve V Figure 3 43 Synchronisation between sensed and the estimated position and speed 119 Linear Drive for Material Handling extremities i e the point where the sensors will provide no more valid position information Following equation is valid
42. our experimental setup has an oval track with the dimensions of ca 3m x 4 m It is composed of 18 independently fed stator segments which are grouped in 5 sections The main data for the segments of each section is listed in Figure 3 5 where en number of turns of the coils e Np number of stators connected in parallel ens number of elementary machines connected in series e F nominal thrust force e nominal current e 0 nominal air gap ed actual air gap Section 1 containing SS1 and SS2 is a straight section The two segments are double sided p 2 and each side of each segment is formed by the series connection of two elementary machines k 2 The optical sensors system described in section 2 2 of this work is implemented on this section of the linear machine see chapter 3 2 All the other sections of the machine are single sided p 1 The nominal values of the currents from Figure 3 5 were used for the dimension 68 Linear Drive for Material Handling ee Elementary Machine Figure 3 5 Segments data ing of the inverters for each section of the track see 3 1 2 1 The nominal values of the electrical force F can be obtained with the nominal values of the current when the air gaps of the different sections are at their nominal values 6 However it was not possible to set the air gaps to their nominal values due to the high magn
43. plate is much shorter than the other two plates there are constant capacitances Cao Cp Ceco Cao from the transmitting electrodes to the receiving electrode These four capacitances are approximately equal and their common value will be referred to as Co Cao Co Coo Cao Co Eq 2 9 Np Nmp s9 w 2A h ui a 0 TEETH Eq 2 10 The position independent capacitance between coupling and receiving electrode is c _ Nmp so P 2A h rr Eq 2 11 Nmp is the number of periods at the modulating electrode 30 periods eo is the electrical permittivity of air and s3 is the relative permittivity of FR4 35 A h P w d4 da and d3are dimensions of the capacitive sensor as shown in Figure 2 26 Figure 2 29 The values of the geometrical dimensions of the sensor as well as the values of the resulting capacitances calculated under the assumption of an homogenous electric field are given in Appendix 5 2 Capacitances Cj in a b c d and Cout see Figure 2 27 and Figure 2 30 can be neglected in the analysis The first ones are in parallel with the voltage sources U Ua between the transmitting electrodes and ground whilst Cout is short circuited by the charge amplifier see below Information on the position of the slider is provided by position dependent capacitances Figure 2 28 shows the areas where transmitting electrodes a d face the ki modulating electrode The area F4 depends on the position
44. segments Segm k A xx 1 Segm k AO8 amp 4 Segm k Abk 167 36 96 70 360 00 166 55 321 81 104 87 0 00 164 97 43 75 84 01 133 86 305 20 Table 3 5 Measured phase differences between all stator segments 117 Linear Drive for Material Handling m h es tt 40 NO e NO Observed EMF in a axis V O Observed Speed m s 0 0 1 0 2 0 3 0 4 0 5 0 6 Time s Figure 3 42 The a components of the observed EMF and the observed speed SS4 SS5 and SS6 Because the mechanical observer described in Section 3 3 2 only uses the phase information contained in the observed EMF vectors these air gap changes are not reflected in the estimated speed 118 Linear Drive for Material Handling 3 4 Sensor Sensorless Transition As previously described there are two sources for position and speed information eWhen the vehicle is inside the processing station the sensor provided position and speed xs Vs are used in control e When the vehicle travels through the transport region the estimated position and speed X V are used as feedback by the control algorithm The position and speed synchronisation function as depicted in Figure 3 20 see also the block diagram in Figure 3 34 determines which position and velocity will be used in control Xc vc Based on the position xc and on the linear machine s parameters length of each stator segment length
45. speed travelling This super elevation requires a precise three dimensional bending of the two guiding tubes which is difficult to realise In order to compensate for unavoidable variations from the nominal value of the distance between the tubes the outer guiding rollers the ones that are located towards the outside of the track i e opposed to a single sided stator of the articulated vehicle are mounted on springs This lets the vehicle balance the variations in the distance between the tubes but with the disadvantage of a highly increased friction force between the vehicle and the track Measurements of the static friction conducted at different points along the carriage way have shown it to be very high up to 200 N at some points even higher than the nominal electrical force Due to the modular construction of the linear machine and to the tolerances in the dimensions of the SMC stator elements there are gaps in the machine s winding between consecutive stator segments which lead to a phase displacement between the two corresponding EMF vectors The influence of these gaps on the control algorithm will be discussed in detail in section 3 3 Figure 3 8 Articulated vehicle without magnets and load carrying plate Source 43 NArtieulateckVehicle gt I pl Anm TAIAN en 1 E f l S Rk ID x A al ASSA x f gt Ss Y Figure 3 9 Entire linear machine 71 Linear Driv
46. the amplitude and offset errors In Figure 2 36 the deviation between reference sensor and capacitive sensor is shown The blue line shows the result when no correction is used i e the signals from Figure 2 34 are used directly for atan2 calculation Eq 2 28 In the zoom it is clearly seen that the deviation ca 50 um has the wavelength of the pitch of the capacitive sensor 2 mm due to the offset and amplitude errors In addition to this high frequent deviation there is a long term integral deviation ca 120 um over the entire measuring length which cannot be corrected without using c c U ut U ut2 out 1 out 2 0 5 0 05 1 c c out 2 out 3 1 1 0 5 0 05 1 U _u out 3 out 4 1 0 5 0 05 1 1 0 5 0 05 1 Figure 2 35 XY Representation of the sampled voltages before and after period corrections 57 Position Sensing Systems for Passive Vehicles Position deviation over the entire measuring length mi n ul IW ui j Without corrections Period corrections Period and pitch corrections Position deviation um ii Ves M x ye i 0 2 0 15 0 1 0 05 0 0 05 0 1 0 15 0 2 0 25 Reference position m Figure 2 36 Position deviation using square wave excitation the reference sensor when generating the correction table The result of applying a period correction table generated without the use of the reference sensor
47. this position acquisition system will be presented as well as experimental results regarding the performance of the optical system Section 2 3 presents the evaluation of the capacitive sensor Figure 2 3 Firstly an electrical model of the sensor will be derived Then two evaluation methods developed for the capacitive sensor based on rectangular and respectively on sinusoidal excitations will be presented Experimental results for both excitation methods will also be shown and compared 22 Position Sensing Systems for Passive Vehicles 2 2 Optical Sensors System 2 2 1 General Description of the System In this section the implementation of an optical sensors system for use in linear drives with passive vehicles will be described Optical sensors are the industry standard for position measuring when high precision positioning is required The used sensor has an incremental scale with a grading pitch of 40 um and in order to further increase the resolution the read heads generate two analogue signals sine and cosine with a cycle length equal to the pitch Fine interpolation using arctangent function will provide a position resolution but not accuracy in the nanometers range The high resolution is necessary in order to generate a smooth speed signal for the speed control loop as the numeric derivative of position within a short sampling interval and without delay by filtering The sensors also provide a reference signal use
48. to the Sensor Bus Control State Machine to initiate a new transfer on the Sensor Bus The Sensor Bus Control State Machine handles the communication protocol on the Sensor Bus and saves the position information coming from the Sensor to Digital board in registers RB D4 15 RB De 15 The control program can check at any time the status of the transfer by reading the contents of R STAT 4 When the Sensor Bus Control State Machine is again in idle state meaning that the position transfer is completed the control program can read the position information in three 32 bit PCI accesses and based on this information it will determine a new value of the position 2 2 6 Position Reconstruction in C Code The position of the vehicle is reconstructed from the information sent through the Sensor Bus in the position calculation function of the control program This function is called every 100 us when new values of USE MCA Sa Ca Na and of USEg MCg Ss Cg Ns are available A software implemented state machine Figure 2 16 uses USEA USEs MC and MG to keep track of the current head in parallel with the Current Head State Machine CHSM from the Sensor to Digital firmware One state of the software state machine is allocated for each of the six logical read heads plus the state H o when the vehicle is outside the sensors region and no position information is available Figure 2 17 depicts the principle of the state changes in the positi
49. two counters will be blocked holding the values Na 1 and Ng ANss respectively Multiplexer code register R MC 2 holds the value MCA 1 so that Sa Si Ca Ci and R MCa 2 the value MC 5 Sg Ss Cg Cs Registers R USE and R USEs indicate whether position information from heads A or B respectively is valid and should be used for position calculation Initially they are both reset neither of the active heads 1 and 5B is providing valid signals The Conversion Timer in the Synchronisation 28 ADC CONTROL ADC BUS 12 C55 BS S p R TMP S 12 SIN COS REGISTERS R R R TMP S 12 gt R TMP C 12 A UPDTATE SC REG Position Sensing Systems for Passive Vehicles S 12 l C12 R TMP C 12 4 J S 12 gt sa Fa SENSOR BUS CONTROL RB S 12 RB C 12 RB S 12 RB C 12 FPGA FPGA CLOCK RESET UY TO ALL REGISTERS AND STATE MACHINES RD S MC 2 jcsx s 6 W dz COUNTER A CONTROL MC 2 E ES a RD C EG DECODE mag CS 5 9 z y LD 3 T mss gh B la lt N 14 lt PH 2 AN Bj CNTR 14 RB N 14 b E CS S H CURRENT 2 ME BS C AN 2 HEAD CONTROL mm
50. u P Ang rm i atan2 Usut3 Yout2 n Eq 2 28 Bla Ala fla P X34 2n atan2 Usut 3 Uout 4 u A new position value might be calculated each 3 75 us one fourth of the excitation period if the data transfer and the controller are fast enough In a practical application an average value may be calculated using values from the past in a user defined time window thus providing over sampling and filtering as desired Figure 2 32 shows the experimental setup used for the evaluation of the capacitive sensor using rectangular excitation Two identical ISA boards Reference Position Acquisition Board and Capacitive Sensor Acquisition Board located in the control PC are used for the evaluation of the signals from the two sensors Each board is equipped with a Complex Programmable Logic Device CPLD as well as with analog to digital and Control PC Reference Position Acq Board Current Board Bus Converter with Reference ZI integrated 4 D A Converter current control A Linear Motor gruen Buffers CPLD IRQ11 Quadrature Pulses M Optical Sensor VELL er r Excitation Voltages Signals Capacitive Sensor Acq Board Capacitive Sensor Board Bus ISA Bus A D Converter Sensor Output Buffers IRQ10 Figure 2 32 Overview of the setup used for evaluation with rectangular excitation 54 P
51. x by Xe A 2n n F x h w 2 1 sin y dy al gt Fr 3 X Eq 2 12 Xx Solving the above integral we get F x w hA P75 sin SE cos 2x CL A qt P P Eq 2 13 Constant uS c E Constant The area Gais complementary to Fa G x 2 w h 2 A E x Eq 2 14 and together they define the position dependent capacitance x Nmp g F3 x Nmp g G x d d 3 Eq 2 15 amp 3 which can also be written 27 C x Cavg Cvar COS E i Eq 2 16 50 Position Sensing Systems for Passive Vehicles The constant values Cay and Cvar are h A A Cava No E0 W 9 d EM C c NS PA PA G ee a In a similar way the position dependent capacitances between the other three electrode structures b c and d and the modulating plate can be written All the position dependent capacitances are given below Eq 2 17 CR Cag 6 4 8605 d C X Ca COS mij MN q C X Ca Cvar COS xen 2 3 C4 X Cava Cvar COS Te These four capacitances consist of a constant part Cayvg on which sinusoidal position dependent variations of amplitude Cvar and 90 phase shifted are superimposed Based on the above determined capacitances the equivalent circuit from Figure 2 30 is obtained for the capacitive sensor In the following it will be assumed that at the output receiving electrode of the sensor a charge amplifier is connected
52. 1 where a sonic strain pulse is induced in a specially designed magnetostrictive waveguide by the momentary interaction of two magnetic fields One field comes from a movable permanent magnet fixed at the vehicle as it passes along the outside of the sensor tube the other field comes from a current pulse interrogation pulse applied along the waveguide The interaction of the two magnetic fields produces a strain pulse which travels at sonic speed along the waveguide until it is detected at the head of the sensor The position of the magnet vehicle is determined by measuring the elapsed time between the application of the interrogation pulse and the arrival of the resulting strain pulse 22 These sensors have many interesting properties such as absolute measurement possibility to simultaneously measure the position of several vehicles and to measure also along curved tracks But due to the travelling time of the sonic stain pulse the above mentioned requirements 4 and 5 are not fulfilled The measuring cycle is 500 us for a 1 2 m measuring distance and 1000 us for 2 4 m 21 23 Therefore these types of sensors are interesting for other applications where such a delay in the control loop can be accepted The rest of this chapter is organised as follows In Section 2 2 a position sensing system for passive vehicles based on optical sensors as shown in Figure 2 2 will be described The hardware firmware and software developed for
53. 2 N OS vr K3 0 K3 1 At 1s Y EM 0 Ll M3 K2 0 else Kay E1 0 EM 0 Y N M K2 0 else El 1 amp Kar At gt 2s ra D x DC Link Ei 1 amp E2 1 Disable E2 0 Command v M7 M5 M8 K2 0 K2 0 K2 0 K3 0 K3 0 K3 0 co CO Co Figure 3 18 Monitoring State Machine 85 Linear Drive for Material Handling State Description Initial state K2 and K3 are both open and the state machine is waiting for aDC Link power up enable command from the User Interface Wait for T2 to be pressed The DC Link power up was enabled from User Interface the relay K2 is now closed and in order to feed the Siemens electronics the user must press the On button T2 on the cabinet s door The transition to next state occurs when input E2 is activated E2 0 i e Siemens electronics is supplied Close K3 In this state the feeding of the DC Link is activated K3 1 E2 is already active so K2 can now be deactivated K2 0 The transition to next state occurs automatically after 1 s Wait for E1 Wait for the Infeed Module and the two Monitoring Modules to become operational E1 0 Active state The DC Link is fed and the machine control can operate The Power Stack Interfaces can be addressed only if the Monitoring State Machine is in this state Normal stop The user has commanded the power down of the DC Link from the User Interface this being possible only after stopping the machine control algorithm T
54. 5 being controlled oO i i N a zZ x PLD m i i gt XE j 4 E S9 lt 2 7 fo eo i i poa M HH wo Figure 3 2 Inverter Bus Protocol Source 42 66 Linear Drive for Material Handling Figure 3 2 the Bus Master calls the first Vehicle Controller VC1 Each VC keeps track of the Stator Segment where its corresponding vehicle currently is i e of the Power Stack whose currents must be controlled in this case PSn After being called VC1 sends the modulation information to PSn 4 x 12 bits the first 12 bits encode the address of the Power Stack 6 bit and the commanded state of each inverter phase upper IGBT on or lower IGBT on or both off 6 bit The next 3 x 12 bits are the switching times in the three inverter phases After receiving the modulation information the addressed PSn answers to VC1 by sending the actual values of the three phase currents 3 x 12 bits which currents have been sampled in the middle of the zero vector interval The communication between VC1 and PSn takes about 5 us the half of the 10 us time slot accorded to VC1 A special situation appears when a vehicle passes from one segment to the next one and for a time occupies two stator segments as V3 in Figure 3 1 In this case VC3 has to control the currents of both PS n 3 and PS n 2 The necessary communication utilises fully the 10 us alloc
55. A FPU FR4 HAL IGBT IRQ ISA JTAG MOSFET MSB Abbreviations Three Dimensional Analog to Digital Analog to Digital Converter Address Advanced Power Management Application Specific Integrated Circuit Four quadrant Arctangent Function Basic Input Output System Current Head State Machine Period Counter Comparator Complex Programmable Logic Device Direct Current Direct Memory Access Driver Digital Signal Processor Electromotive Force Electromagnetic Interference Elektromotorische Kraft Enable Electrically Programmable Read Only Memory Finite Elements Method First In First Out Field Programmable Gate Array Floating Point Unit Flame Retardant woven glass reinforced epoxy resin Hardware Abstraction Layer Insulated Gate Bipolar Transistor Interrupt Request Industry Standard Architecture Joint Test Action Group Metal Oxide Semiconductor Field Effect Transistor Most Significant Bit Abbreviations MUX PC PCB PCI PI PLL PMS L M PS PSI PWM RAM RCV r m s RTAI RTM RTW S2D SB SMC SS TTL UI USB VC VCI Multiplexer Personal Computer Printed Circuit Board Peripheral Component Interconnect Proportional Integrative Controller Phase Locked Loop Permanent Magnet Synchronous Linear Machine Power Stack Power Stack Interface Pulse Width Modulation Random Access Memory Receiver Root Mean Square Real Time Application Interface Real Time Module Real Time Workshop Sensor to Di
56. As error function correction term of the observer the difference between the measured and the estimated current dependent flux vectors is used bid Ei Pia L F LM x eit E a E 3 19 Ey Pi Yip Ip Pip i 105 Linear Drive for Material Handling Using this error function and considering the model of the permanent magnet synchronous machine the EMF observer is implemented by the following state equations 9 DO d 0 Y 1000 1 9 10 R 0 4 Gy 0 d Vip Y 5 01 0 Riu 0 G feya ae ep eser cn Eq 3 20 dt Tp 6 000 0 fi Gs 0 Ey s 00 o Le 199 0 Odi 0 G Tp Pa Pips and s denote the estimated components of the current dependent flux and of the EMF vectors respectively while v is the estimated speed Because the measured values of the machine s phase voltages are not available their reference values u and uj will be used instead Gy and G are the gains of the EMF observer A block diagram representation of the EMF observer is given in Figure 3 35 Because of the speed dependent cross coupling of the observed EMF components the dynamics of the EMF observer cannot be analysed independently of the mechanical observer which calculates the estimated speed v based on the estimated EMF unless it is assumed that the estimated speed follows exactly the real speed v v It is however possible to eliminate the speed dependent cross coupling from the EMF observer which will allow the indepen
57. Board 27 Position Sensing Systems for Passive Vehicles generated on the board using the DC DC converters 19 The distances covered by each read head are shown in Figure 2 9 The scale has a length of 250 mm and the read heads are mounted along the track at approximately every 210 mm Considering the grating period Pitch P 40 um of the sensor each of the heads 1 4 will cover a number AN 5250 periods Head 5A covers AN 2 2625 periods whilst head 5B covers AN s periods If the heads would be mounted exactly AN periods apart from each other the relation ANsg AN 2 would hold true but in reality a small difference exists on the experimental setup a difference of 5 periods was identified ANsg can be determined during a test run in the positive direction as the value of the B period counter on the falling edge of the second pulse of the reference signal Rs see below If instead of the real value ANsp the ideal value AN 2 is used a hysteresis of the determined position occurs between runs in positive and negative directions RAL TLR AN AN AN AN OLANA ANg 4 pd gt gt lt i gt H H i H i H Hs Hss Zero Position Maximal Position Figure 2 9 Distances between read heads The values AN AN 2 and ANsg will be used by the FPGA to decide when the transition between one read head and the following one must be made The firmware in the Sensor to Digital FPGA a simplified diagram of which is sho
58. Cy from the PWM units of the DSP are subtracted by a differential input amplifier in order to obtain Uac The operational amplifier used also allows for biasing of the output which was used to correct the offset of Unc Filtering is realised by integrated filter circuits 40 Due to restrictions of the Qf product of these circuits the excitation frequency was limited to 20 kHz A variable gain amplifier is used to correct the gain errors of Unacr and a rail to rail differential output operational amplifier generates the two complementary outputs U and U The output amplifiers power supply is 15 V which allows for a swing of 30 V in the line to line voltage Uac For generating the other two voltages Uy and Ug an identical circuit was used fed by Sp and Sy which are 90 shifted with respect to Cp and Cy The output of the charge amplifier feeds a comparator which detects the positive zero crossings of the signal Usu A Capture unit of the DSP uses the output of the comparator in order to determine the phase of U ut The timer used by the Capture unit is hardware synchronised with the timers used to generate the control signals A second comparator could be used to determine the negative zero crossings of Us reducing thus to half the acquisition latency In Eq 2 33 it was assumed that the generation of the sinusoidal excitation voltages and the charge amplifier introduce no phase shift between the fundamental of Uac and 62 Positi
59. G fe a RD C CUP CDN l MSB S QUADRANT A E d R MC 2 P RB MC 2 v STATE MACHINE E E d a Ea MSB C V T R MC 2 RB MC 2 d L I Q i c o o E CL wu S iS A Im USE N E gt F MSB S y E R USE a RB USE aE OQ oO E MSB C QUADRANT B 3 E o o 280 g C p STATE MACHINE use US ple use 3 SE A we lt CUP CDN Ri Rs e 2 5 E 1 5 A m uU Z AN u CNTR 14 RB N 14 a 5 gt DS z T Z5 R g jos P RB R d e s SYNCHRONISATION A DECODE LD 3 a zu STATE MACHINE 5 RB R OF COUNTER B CONTROL AN 14 R AN 14 5B SELECT ENA R ENA SB ACK SB REQ oe SM Figure 2 10 Simplified diagram of the pre processing unit FPGA code Block is stopped and no analog to digital conversions take place The system remains in this state until one of the comparators outputs R4 or Rs is asserted signalling the entering of the vehicle into the sensors region from left or from right Let us assume that the vehicle comes from left thus travelling in the positive direction The synchronisation logic detects the rising edge of R4 and asserts the signal ADC START and simultaneously starts the 2 us Conversion Timer It then waits for the first conversion to end When the ADC START signal is asserted the ADC Control State Machine immediately asserts the four conversion start signals CS S4 CS Ca CS Sp CS C
60. M calculations have to be performed for different shapes of the electrodes 64 3 Linear Drive for Material Handling In this chapter the position acquisition and control of a linear drive for material handling currently under development at our department is discussed The chapter is or ganised as follows First in section 3 1 the system architecture is presented After an overview of the power electronics and control topology the linear machine used in our experimental setup is briefly introduced Then details regarding the power electronics and control cabinet and the control software are given The implementation at the linear drive of the optical sensors system described in section 2 2 is presented in section 3 2 In section 3 3 the implementation of the EMF based sensorless speed control at the transport section of the drive is discussed Finally section 3 4 presents the synchronisa tion between the position given by the sensors and the estimated one when the vehicle leaves or re enters a processing station 3 1 System Architecture In order to allow for individual motion control of several vehicles the active track of the linear machine is separated into many segments each segment being fed by the power stack of a dedicated inverter For the application discussed in this thesis a relatively small number of vehicles is intended so a system with one controller assigned to one vehicle is best suited as it involves the lowest p
61. Multiplexers A B Sensor Bus Request Acknowledged Sensor Bus Data Input Sensor Bus Data Output Sensor Bus Read Enable Sensor Bus Data Request Sensor Bus Write Enable Delay Period Sensor Bus Communication Pulse Period Sensor Bus Communication Total Transmission Time Sensor Bus Communication High Impedance State Duration Sensor Bus Communication Specify if the information from read heads A B is to be used in the calculation of the position 11 Symbols x Xo XA XB Xinc A Xinc B Position given by the optical sensors system Position offset of the current read head Positions calculated based on the information provided by optical read heads Ha Hg Incremental positions calculated based on the information provided by optical read heads Ha Hs Capacitive Sensor Section 2 3 A An Ga Ga Cao m Cao Cain ee Cain Cawg Ck Cout CR Cvar di d2 d3 Fa Height of the upper and lower sinusoidal copper patterns on the slider Amplitude of the sinusoidal excitation voltages Position dependent capacitances Position independent capacitances between transmitting and receiving electrodes Input capacitances Average value of the position dependent capacitances Coupling capacitance between modulating and receiving electrodes Output capacitance Charge amplifier feedback capacitance Amplitude of the variation of the position dependent capacitances Spacing between transmitting and modulating plate
62. Mutschler P Comparison of topologies for linear drives in industrial material handling and processing applications in 7 International Conference on Power Electronics 2007 ICPE 07 pp 1027 1032 Oct 2007 Nyce D S Linear Position Sensors Theory and Application ISBN 978 0 471 23326 8 John Wiley amp Sons 2004 Reininger T Welker F and von Zeppelin M Sensors in position control applica tions for industrial automation Sens Actuators A Phys vol 129 no 1 2 pp 270 274 May 2006 B hr A Speed Acquisition Methods for High Bandwidth Servo Drives Dissertation TU Darmstadt 2004 Canders W R Mutschler P Mosebach H Shi Z Weigel J and Lamsahel H Neue Funktionalit ten von Linearantrieben Forschungsvereinigung Antriebs technik e V Forschungsheft 692 Jan 2003 Rashed M MacConnell P F A Stronach A F and Acarnley P Sensorless indirect rotor field orientation speed control of a permanent magnet synchronous motor with stator resistance estimation IEEE Trans Ind Electron vol 54 no 3 pp 1664 1675 Jun 2007 Seok J K Lee J K and Lee D C Sensorless speed control of nonsalient permanent magnet synchronous motor using rotor position tracking PI controller IEEE Trans Ind Electron vol 53 no 2 pp 399 405 Apr 2006 De Angelo C Bossio G Solsona J Garcia G and Valla M I A rotor position and speed observer for permanent magnet motor with non sinusoid
63. Position Acquisition and Control for Linear Direct Drives with Passive Vehicles Vom Fachbereich Elektrotechnik und Informationstechnik der Technischen Universit t Darmstadt zur Erlangung des akademischen Grades eines Doktor Ingenieurs Dr Ing genehmigte Dissertation von Dipl Ing Marius Alexandru Mihalachi Geboren am 29 08 1981 in C mpina Rum nien Referent Prof Dr Ing Peter Mutschler Korreferent Prof Dr Ing Wolf R diger Canders Tag der Einreichung 1 7 2010 Tag der m ndlichen Pr fung 23 11 2010 D17 Darmstadt 2011 Abstract For combined processing and transportation of materials in industrial production lines a long primary linear synchronous drive with passive lightweight vehicles is being designed and experimentally tested This thesis concentrates on position acquisition and motion control of the proposed system In order to allow a high degree of independency in the movement of the vehicles the stator primary of the linear machine is divided into many segments Each segment of the track is fed by a dedicated power stack and control information is exchanged between all power stacks and all vehicle controllers via an Inverter Bus A number of processing stations are spread along the track of the linear drive being connected by transport sections Inside the processing stations high quality speed and position control of the vehicles is required For this precise and fast position measurement is ne
64. Position Sensing Systems for Passive Vehicles am Ww Ui s5899 A B H a2s Has n mimm 2359 A B Hos Ea J A J A agg MD 3 B Figure 2 11 Current Head State Machine CHSM USE 0 USE 0 LD 001 LD 100 N 1 N AN USE 1 USE 0 LD 000 LD 100 N changes N AN USE 1 USE 0 LD 000 LD 001 N changes N 1 USE 1 USE 1 LD 000 LD 000 N changes N changes USE 0 USE 1 LD 010 LD 000 N AN N changes USE 0 USE 1 LD 001 LD 000 N 1 N changes USE 1 USE 1 LD 000 LD 000 N changes N changes USE 1 USE 0 LD 000 LD 010 N changes N AN USE 1 USE 0 LD 000 LD 001 N changes N 1 USE 1 USE 1 LD 000 LD 000 N changes N changes USE 0 USE 1 LD 010 LD 000 N AN N changes USE 0 USE 1 LD 001 LD 000 N 1 N changes USE 1 USE 1 LD 000 LD 000 N changes N changes USE 1 USE 0 LD 000 LD 010 N changes N AN USE 1 USE 0 LD 000 LD 001 N changes N 1 USE 1 USE 1 LD 000 LD 000 N changes N changes USE 0 USE 1 LD 100 LD 000 N AN 2 N changes USE 0 USE 1 LD 001 LD 000 N 1 N changes USE 0 USE 0 LD 001 LD 100 N 1 N AN 31 Position Sensing Systems for Passive Vehicles On the rising edge of SB ACK the content of the position information registers is transferred into a set of
65. S e N EAST N SAO 2 CAO Cum Sn GAS o EIT ZAS A t an TLN N yous uleW N m 5l 3d 2 E 2 a HDI vor o Cd 40 q pup dea J0y9e U09 d g 0 14U0J TL n amp ne SIN A ddn do gt B quon z l WLU LUL l S uo TH OH N LE TA CA CX gI juawainseaw DAN EIN HE X NINN lt NINN 9 0 gt S c eJ JERE maer BER J 5 EZ Sp IM U sng Ajddns 13 sng A ddns 3 sng A i vor so2iMeg ewag so2ieg ewou saa aaq ii EE f m gt i l eS PS Pj zo einpow 25 PS Pj TO a npow es z ux ES y Mei u u 9 Gunopuow u u o Bunoyuow u u u S EE TAC E Sa TNC 009W 009W Fd EE TT opewog LETT e EIT 009d 009d e Seus TTT DOAN uonoas xung L 8 J9MOd jeuuequr E AO um um mi 009d Avc d oH jeuoneisdo jeuoneisdo jeuoneisdo oo ENG 8TSd STSd VESd 8Sd T Sd ISd emnpow N eoe SIN IN peojup NIS aasjoje di3 i Tu i mp iI at i dp ilk Ra ilk mud rer rem VLLL youlqey 81SS a SISS vISS 8SS ZSS TSS Figure 3 11 Cabinet Diagram 74 Linear Drive for Material Handling After the Control PC has closed the contact K2 it will monitor the input E2 This input is connected to an auxiliary switch of the contactor K1 and signalises the state of the contactor During the running time of the control algorithm the Control PC also monitors the state of the Infeed Module and of the two Monitoring Modules through the input E1 Faults like phase failure or under voltage
66. The compiled code is downloaded through the Host Interface to the control DSP and the ControlDesk software is used for simultaneously acquiring the two sensors position The results position differences between the tested and the reference for three Converter with integrated current control Current Reference Controller Board DS1102 DSP 12b D A Converter TMS320C31 T 4 1 Encoder Input T mte RL a Host 2 4 Encoder Input Interface L Host Simulink Control Desk System RTW Runtime Original processing Figure 2 24 Overview of the setup used in initial tests 47 Position Sensing Systems for Passive Vehicles 250 n l INN Position Deviation um oo Wi m psn MI M IM i 0 0 05 0 1 0 15 0 2 0 25 0 3 0 35 0 4 Reference Position m O1 o Figure 2 25 Test results using the original signal processing test runs at relatively slow speeds are shown in Figure 2 25 The position error due to the processing delays amounts to more than 120 um at a speed of 0 5 m s For high speeds e g 10 m s this error would be unacceptable for high dynamic position control 2 3 2 Electrical Model of the Capacitive Sensor In order to develop an improved processing algorithm a model of the capacitive sensor must first be derived The sensing arrangement itself consists of three printed circuit boards PCBs as shown in Figure 2 26 1 A stationary Tra
67. Units gt deviation men Pr Capacitive sensor data Figure 2 38 Overview of the setup used for evaluation with sinusoidal excitation 60 Position Sensing Systems for Passive Vehicles Transformer To the Capacitive Sensor MOSFET H Bridge Y Analog U Power Ground Figure 2 39 Possible implementation of the sinusoidal voltages generation generates the PWM voltage Unc as shown in Figure 2 39 and Figure 2 40 Then a passive L C filter with a high quality factor Q suppresses the harmonics of the PWM voltage yielding Unc The line to line voltage Uac Ua Uc An cos unt is generated by a three winding transformer which also separates the Power Ground of the H Bridge switching and the Analog Ground charge amplifier By appropriately choosing the winding numbers N and N2 of the transformer high excitation voltages can be obtained even with a small voltage Upc The second line to line voltage Uab Ug Ub An sin wt is generated by a similar circuit not shown in Figure 2 39 The gate signals Cp and Cy for the H Bridge are produced by the PWM circuitry of the DSP They are 180 shifted and have a pulse width of 120 which eliminates the 3 harmonic from the voltage Uac All the frequencies contained in Uac must be eliminated by filtering except the fundamental component frequency fh This would normally be achieved by a band pass filter with the band pass frequen
68. Us of the charge amplifier may be calculated Such a procedure will be helpful to optimise the shape of the modulating electrode such that the position of the slider can be extracted precisely from Uout 34 But in this work the commercially available arrangement of electrodes is used and we concentrate on extracting the position information from Uou For that purpose a simplified analytical model of the capacitive sensor was developed based on the assumption of homogeneous field distribution which permits the approximation of the electric field through concentrated capacitances as shown in Figure 2 30 48 Position Sensing Systems for Passive Vehicles Transmitting Electrodes Receiving Electrode Modulating Electrodes Grounded Electrode Grounded Electrode Coupling ER qM TETE TEN Plate slider Transmitting iuum d Receiving Plate fixed fixed Figure 2 26 Exploded view of the capacitive sensor Transmitting Modulating Receiving Plate Plate Plate Uout Figure 2 27 Cross section of the sensor Figure 2 28 Position dependent overlaping areas j L N P 280 2mm 560mm E Figure 2 29 Transmitting electrodes P 2mm 49 Position Sensing Systems for Passive Vehicles Besides the position dependent capacitances there are several position independent ones some of which can be neglected in the analysis As the modulating
69. a Yon a L f Eq 3 7 with L being the inductance matrix E L cos20 L sin20 2 L sin20 L Lg 4 L cos20 Eq 3 8 In the above equation Lo L4 and L2 represent the mean value of the self inductance the amplitude of the inductance variation and the mean value of the mutual inductance respectively This last term L2 does not appear in rotating motors it is specific to linear machines being caused by end effects 56 57 0 is the electrical angle of the vehicle within the considered pole pitch and is related to the local position x by A _y ox Eq 3 9 Up tp being the machine s pole pitch 24 mm It was experimentally determined that due to the relatively large air gap the linear machine presents only small saliencies the amplitude L4 of the inductance variation as well as the end effects determined inductance L2 are considerably smaller than the mean value Lo of the self inductance ca 1 2 Therefore L and L gt will be subsequently neglected and the inductance matrix L will be approximated as L 0 L o Ll Eq 3 10 0 The following expression will be considered for the permanent magnets flux linkage Pome WV COS g Pomp PMO sing Eq 3 11 Using the above equation the vector of the voltage induced in one stator segment by one vehicle can be expressed as e d Ppma Y sin 0 de e dt Yomg PMO cos 0 dt Eq 3 12 104 Linear Drive for Material Handling
70. al EMF waveform IEEE Trans Ind Electron vol 52 no 3 pp 807 813 Jun 2005 Robeischl E and Schroedl M Optimized INFORM measurement sequence for sensorless PM synchronous motor drives with respect to minimum current distortion IEEE Trans Ind Appl vol 40 no 2 pp 591 598 Mar Apr 2004 Briz F Degner M W Garcia P and Lorenz R D Comparison of saliency based sensorless control techniques for AC machines IEEE Trans Ind Appl vol 40 no 4 pp 1107 1115 Jul Aug 2004 137 Bibliography 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 138 Consoli A Scarcella G and Testa A Industry application of zero speed sensor less control techniques for PM synchronous motors IEEE Trans Ind Appl vol 37 no 2 pp 513 521 Mar Apr 2001 K Patel S D D Arco A Monti D Patterson and R Dougal Design and testing of a modular permanent magnets brushless linear drive in Proc 20 Annual IEEE Applied Power Electronics Conference and Exposition 2005 APEC 2005 vol 3 pp 1883 1888 March 2005 Perreault B M Optimizing Operation of Segmented Stator Linear Synchronous Motors Proceedings of the IEEE vol 97 pp 1777 1785 Nov 2009 Yoshida K Takami H and Fujii A Smooth section crossing of controlled repulsive PM LSM vehicle by DTC method based on new concept of fictitious section IEEE Trans I
71. al conditioning part 6 the sensors signals are converted to single ended their gain is adjusted and the high frequency noise is filtered The sine and cosine signals are than routed through the multiplexers 7 to the four analog to digital converters 8 whilst the analogue reference signals from heads 1 and 5 are digitalized using the two comparators 9 The threshold values of the comparators are adjustable so that the reference pulses can be kept in the limits defined by Figure 2 6 The outputs of the comparators as well as the ones from the analog to digital converters are brought to the FPGA 10 EPROM 11 is used to store the FPGA firmware when the power is off and to automatically program the FPGA at power up The FPGA can also be programmed using the connectors 13 or 14 An 80 MHz oscillator 12 generates the clock signal Auxiliary input output port 15 can be used for debugging In the area 16 are the RS 485 drivers and receivers for the Sensor Bus and 17 is the connector for the Sensor Bus cable The board requires 5 V and 5 V power supplies for the analogue circuitry and a separate 5 V supply for the digital one These are supplied to the board using the connector 18 Additionally the FPGA also needs 3 3 V and 1 5 V supplies which are e 1a mi ni Flle ie is 1i a i a E Ti u E z E ia E Figure 2 8 Realisation of the preprocessing unit Sensor to Digital
72. als rigid vehicle 97 Linear Drive for Material Handling Before correction After correction 1 05 5 Q E lt 0 8 Oo 0 5 O l i i ree 1 05 0 05 1 1 0 5 0 05 1 Siqa PU 3 A corr PY 1 Bu 05 5 a amp E N m O 0 5 0 Pipes 0 5 0 05 1 1 05 0 05 414 S B p u S2 B corr p u 1 _ 05 5 a amp E lt EE e lt O 0 5 0 3 1 SEE Lad 0 5 0 0 5 1 1 0 5 0 0 5 1 Sy p u S3 a corr p u 1 _ 05 3 5 a amp E 9 8 Oo 0 5 0 1 1 05 0 05 1 1 Sy PU S 0 5 05 1 3 B corr p u Figure 3 29 XY Representation of the sine cosine signals articulated vehicle 98 Linear Drive for Material Handling The cause of this amplitude decrease is an increase of the distance between the scale and the read heads along the travelling length Using the sensors guiding construction shown in Figure 3 25 and Figure 3 26 reduces this variation as shown in the left column of Figure 3 29 an xy representation of the sine cosine signals with the scale mounted on the articulated vehicle and with sensor guiding On the right sides of Figure 3 28 and Figure 3 29 the corrected sine cosine signals are depicted the individual period correction method described in section 2 2 7 was used In Figure 3 30 the data acquired for the calculation of the correction tables for the articulated vehicle is shown For this acquisition th
73. als elapsed between two successive position readings coming from the control program Then the Sensor Bus transfer protocol must also be changed to send the value of this counter together with the other position information back to control to be used in the calculation of the numerical derivative of the position 102 Linear Drive for Material Handling 3 3 EMF Based Sensorless Speed Control In the present implementation of the control algorithm for our application it is assumed that a vehicle will start stop only inside a processing station using feedback from the position sensors This makes the use of the EMF based estimation method adequate for the transport sections outside the processing stations even though this method does not work at zero speed Expanding the functionality of this linear drive system e g by implementing vehicle buffering before entering a processing station requires future work on sensorless position acquisition methods using signal injection For the sensorless EMF based determination of position and speed a structure consisting of three observers will be used e Two EMF observers one for each of the controlled stator segments e One mechanical observer which uses the two observed EMF vectors in order to estimate the position and speed The integration of this three observers structure in the algorithm of one vehicle controller is depicted in Figure 3 34 The EMF observers the mechanical observer
74. an change during the acquisition of the correction tables data as long as it remains low enough so that a sufficiently large number of samples per period is acquired The amplitudes and offsets of the sine cosine signals for each period of each read head are calculated offline using the acquired data The resulting correction values more than 100 000 will be then stored on the control PC and used by the position calculation function in the control program Based on the corrected values of sine and cosine a corrected incremental position will be calculated Because the counters in the Sensor to Digital FPGA firmware use the uncorrected values of sine and cosine the period numbers Na Ng read from Sensor to Digital must also be corrected in the control program elf the uncorrected sine and cosine determine a point in the 4th quadrant and the corrected ones a point in the 1st quadrant a unit is added to the period counter elf the uncorrected sine and cosine determine a point in the 1st quadrant and the corrected ones a point in the 4th a unit is subtracted from the period counter e Otherwise the period counter remains unchanged After the correction of sines cosines and period counters the corrected position will be calculated in the same manner as the uncorrected one Figure 2 18 Experimental setup used to test the position acquisition 42 2 2 8 Experimental Results The optical sensors system described in the previous s
75. arge amplifier An ideal charge amplifier would reflect this spike at its output according the last term of Eq 2 23 note the differentiating s in the nominator After the spike a real charge amplifier needs some time to settle as measured in Figure 2 33 therefore the sampling is done at the end of each 90 interval Figure 2 34 shows a large number of samples taken as in Figure 2 33 which generate the four sinusoidal position dependent signals Uo Uout 4 note the time scale difference in the two figures The offset and amplitude errors of the four signals are very large so their correction is necessary For the correction of the four position dependent One sample taken like in Figure 2 33 lt EE ee a T U ut 3 V D ik 2 V U ut 4 V Time ms Figure 2 34 Sampled voltages 56 Position Sensing Systems for Passive Vehicles signals a parametric table will be used to correct the amplitude and offset of each period of the four sinusoidal position dependent signals similar to the period corrections described in section 2 2 7 The results of the amplitude and offset correction are shown in Figure 2 35 Here the position dependent signals Uput1 Uout4 are normalised to the input range of the analog to digital converter After applying the period corrections the shape of the locus of the corrected signals comes close to that of the unity circle indicating the reduction of the position error caused by
76. ate a negative value of the position b Transition between heads H and H2 moving in positive direction The vehicle is travelling the region covered by head 1 When it approaches the end of this region the scale will also cover head 2 CHSM is in state H i45 and CNTRa is counting The sine cosine signals from head 2 are coming through MUXs but CNTRg is blocked at 1 not counting When Na gt AN at time ti CHSM switches to state H1 28 and CNTRa is enabled At time t the quadrant defined by Sg and Cg changes from 4 to 1 and CNTRg counts up yielding Ng 0 The two counters are now both counting in parallel until t5 when the rising edge of SB ACK arrives The position information is saved into the bus registers and on the falling edge of SB ACK time t4 CHSM changes state to Hog CNTR3 continues counting but CNTRA is blocked holding Na AN State H1 28 can last up to 100 us the interval between two successive readings on the Sensor Bus which translates in up to 25 position periods 1 mm at the maximum speed of 10 m s C Leaving sensors region moving in positive direction The scale covers head 5B CHSM is in state Hsgo and CNTRg is counting At time t1 the reference signal Rs comes but this has no influence on the Synchronisation block The position acquisition algorithm in the Sensor to Digital FPGA will ensure that the transition to state Ho will occur after the falling edge of R based solely on the value of CNTRa I
77. ated to VC3 t 20 30 us in Figure 3 2 After receiving actual currents values a Vehicle Controller can proceed with the calculation of the algorithms concerning current speed and position control for the corresponding control cycle More than one VC can be calculated by one industrial PC within each 100 us under Linux RTAl operating system 3 1 1 Linear Machine Preface As mentioned previously the design of a new type of electrical machine and all the mechanical constructions was done by Institute for Electrical Machines Trac tion and Drives at Technische Universit t Braunschweig Our department is responsible for hard and software of power electronics and control of the system In order to understand the operation of the control system some information on the electrical machine and the mechanical construction is included here although this is not part of the author s work As a high thrust force density in combination with a rather large air gap and a curvilinear track is necessary for the application a Permanent Magnet Synchronous Linear Machine PMSLM in the long primary moving magnets configuration corresponds best to the requirements The linear machine designed for this project has three phased single layer concentrated coils tooth windings which simplifies the modular construction of stator sections of different lengths In order to further increase the modularity the stator is constructed from individual stator el
78. atrix State Space Representation Friction coefficient of the vehicle Output Matrix State Space Representation Components of the EMF vector of a stator segment in the stationary a B reference frame Components of the estimated EMF vector in the stationary a P reference frame Components of the EMF estimation error vector in the stationary a reference frame Transfer function associated to the the estimation error dynamics Electrical force Reference value of the electrical force Load force Estimated load force Load force estimation error Gains of the mechanical observer Gains of the EMF observer Identity matrices of 3 respectively 4 order Components of the current vector of a stator segment in the stationary a B reference frame Symbols igs i Ke KF Kp L Lo L4 Lo Mo M7 My D1 P2 P3 R S Ti Ts u Qo Us Components of the current vector of a stator segment in the rotor oriented d q reference frame Reference value of the quadrature q current EMF constant Current Force coefficient Proportional gain PI Controller Inductance matrix Mean value of the self inductance of a stator segment Amplitude of the inductance variation of a stator segment Mean value of the mutual inductance of a stator segment States of the Monitoring State Machine Mass of the vehicle Poles of a characteristic polynomial Ramp function Laplace Operator Integral time constant Pl Cont
79. beitungsstationen werden entlang des Fahrwegs verteilt und durch Transportabschnitte verbunden Innerhalb der Bearbeitungsstationen ist eine hochgenaue Geschwindigkeits und Positionsregelung der Fahrzeuge erforderlich Eine pr zise und schnelle Positionsmessung ist dazu unerl sslich weshalb Positionssensoren eingesetzt werden m ssen Passive Fahrzeuge stellen zus tzliche Herausforderungen f r das Positionserfassungssystem dar da weder Energie noch Informationen an die beweglichen Teile bertragen werden k nnen Die Auswertung zweier Positionserfassungssysteme die die 0 9 Anforderungen erf llen wird in dieser Arbeit dargestellt Das erste System basiert auf einem optischen Encoder mit hoher Aufl sung F r die gegebene Anwendung wird die Mafiverk rperung des Sensors am Fahrzeug angebracht und mehrere aktive Abtastk pfe entlang der Fahrbahn montiert so dass in allen Positionen die Ma verk rperung mindestens einen Abtastkopf berdeckt Beim bergang der Ma verk rperung von einem Abtastkopf auf den n chsten m ssen die Signale von beiden Abtastk pfe gleichzeitig abgetastet und synchronisiert werden damit ein kontinuierliches Positionssignal ber die gesamte Messl nge erzeugt wird Das zweite Positionserfassungssystem verwendet einen kapazitiven Sensor mit einer relativ niedrigeren Aufl sung welches eine einfachere und kosteng nstigere Alternative zu dem optischen System darstellt Die Arbeitsweise des kapazitiven Sensors
80. ble in a practical application Compared to optical sensors the capacitive sensors achieve a lower accuracy but the hardware and the signal processing for capacitive sensors are extremely simple especially when the evaluation method based on phase measurement is used and can be produced at very low cost so the capacitive sensors can be an attractive alternative to the optical ones for applications where their accuracy can be accepted The optical sensors system was implemented at the experimental linear drive des cribed in Chapter 3 of this dissertation For the power electronics and control of the drive a centralised architecture is used A dedicated Power Stack feeds each of the 18 segments of the drive and a controller is assigned to each vehicle for the time being only one vehicle is available at the linear drive All Power Stacks as well as the Control PC where the Vehicle Controller is implemented are located in a central cabinet Control information is exchanged between the Vehicle Controller and the Power Stacks via an Inverter Bus In the range outside of the processing station transport section EMF based sen sorless position control was implemented The construction of the linear machine must allow for gaps between consecutive stator segments which bring phase differences in the EMFs of the segments These phase differences were addressed in the implementation of the observer structure used in sensorless control At th
81. capacitive sensor well suited for use in linear drives with passive vehicles The sensor is commercially available from the Netzer Sick Stegmann company 31 It was first tested as provided by manufacturer with the original electronics and processing algorithm 32 The measurements during the initial tests see subsection 2 3 1 confirmed what was to be expected from datasheet the original electronics generates the output signal with a time delay which delay causes an increasing position error as the travelling speed increases This makes the sensor unsuitable for position control of high dynamic linear drives Therefore we developed the hardware and software for two new signal processing methods which use from the original system only the arrangement of electrodes The first method uses the same square wave excitation scheme as the commercial system and the second method uses sinusoidal excitation The new methods presented in subsections 2 3 3 and 2 3 4 respectively avoid the signal delay and velocity dependent errors Figure 2 23 shows the linear drive used for the evaluation of the capacitive sensor with the original as well as with the improved processing algorithms The linear motor 1 with active mover 2 is supplied via a drag chain 3 from the converter 4 The converter has integrated current control so only the speed and position control loops will be implemented in the control software An optical sensor LIF 171R from Hei
82. cessary SO position sensors must be used The passive vehicles impose additional challenges for the position acquisition system as neither energy nor information must be transmitted to the moving parts The evaluation of two position acquisition systems which comply with this require ment is presented in this thesis The first system is based on a high resolution optical encoder For this application the scale of the optical sensor is mounted at the vehicle and several active read heads are installed along the track such that at each position the scale covers at least one read head When the scale is passing from one read head to the next one the position information from both read heads must be evaluated simultaneously and synchronised so that a continuous position signal will result for the entire measuring length The second position acquisition system uses a comparatively lower resolution Capacitive sensor and is intended as a simpler and cost effective alternative to the optical system The principle of operation of a capacitive sensor is first analysed and a model is determined Then based on this model two methods of extracting the position information are presented one uses instantaneous sampling based demodulation while the other is based on phase measurement In the transport sections of the linear drive the requirements concerning the accuracy and dynamic of the position measurement are less demanding than in the proces
83. correcting the sine cosine signals The other one has a wavelength comparable with the entire measuring range and is not influenced by corrections The scale of the tested system does not cause this low frequent component because in that case a similar variation should be observed for each of the five heads In a test the first head was mounted in the place where the second one normally is The same position deviation as the one of head 2 in Figure 2 20 was observed so the low frequent component of the position variation also does not depend on the read heads themselves or on the signal processing in the position acquisition algorithm It can only be attributed to the sensor used as reference which is possible to present large variations of the accuracy in different regions due to its mechanical mounting and many years utilisation By using period corrections the position deviation along the entire measuring range was reduced from ca 5 15 um to approx 5 um The component of position deviation due to systematic errors amounts to ca 1 um peak to peak The reduction of the position error is important not only for accurate positioning but also for the calculation of the speed 30 Figure 2 21 shows the speed signals as obtained by numerically deriving in the 100 us control interval the uncorrected mean corrected and period corrected positions of the test system compared with the speed obtained from the reference sensor No filterin
84. cy fi 100 kHz The Bode diagram of such a filter 2 order with a quality factor Q 20 is shown in Figure 2 41 The phase response of the band pass filter has the steepest slope for f 100 kHz N max 0 6 Ts LZ 2173 5146 T Figure 2 40 Signal diagram Sinusoidal voltages generation 61 Position Sensing Systems for Passive Vehicles 40 Low pass 70 kHz Band pass 100 kHz pi Harmonic Amplitude dB Phase deg Frequency Hz Figure 2 41 Filter Bode diagram which can have a negative influence on the phase measurement used to determine the sensor s position This is why a 2 order low pass filter will be used instead with the same quality factor and a corner frequency of 70 kHz see Figure 2 41 The amplification of the low pass filter at 100 kHz is very close to unity 0 95 and the phase at the said frequency is almost flat 178 The highest harmonic which will appear in the filtered signal is the 5 Before filtering it has an amplitude of 20 of the fundamental and is attenuated by the filter to 296 This means that the 5 harmonic will be present in the filtered signal with an amplitude of 4 o from the fundamental of Uac For the experimental tests a simplified method was used to generate the sinusoidal excitations as schematically shown in Figure 2 42 The signals Cp and
85. d Signal Handling Algorithm Diploma Thesis Nr 1316 Institut f r Stromrichtertechnik und Antriebsregelung TU Darmstadt 2004 Li X De Jong G and Meijer G C M The influence of electric field bending on the nonlinearity of capacitive sensors IEEE Transactions on Instrumentation and Measurement vol 49 no 2 pp 256 259 April 2000 VDE VDI Gesellschaft Mikroelektronik Mikro und Feinwerktechnik GMM VDE VDI Schulungsbl tter f r die Leiterplattenfertigung 3711 Blatt 2 Mai 1999 Netzer Y Capacitive Displacement Encoder U S Patent No 6 492 911 Dec 2002 http www freepatentsonline com 6492911 html Link retrieved in June 2010 Texas Instruments OPA627 OPA637 Precision High Speed Difet Operational Amplifiers Datasheet Literature Number SBOS165 March 1998 hitp focus ti com docs prod folders print opa627 html Link retrieved in June 2010 Baxter L K Capacitive Sensors IEEE Press New York ISBN 0 7803 1130 2 1997 pp 190 191 Texas Instruments TMS320F2812 Digital Signal Processors Data Manual Literature Number SPRS174R April 2001 http focus ti com docs prod folders print tms320f2812 html Link retrieved in June 2010 Texas Instruments UAF42 Universal Active Filter Datasheet Literature Number SBFS002A Nov 2007 http focus ti com docs prod folders print uaf42 html Link retrieved in June 2010 Mathworks Inc filtfilt Zer
86. d X 3 0 Positive direction gt Figure 2 6 Definition of the positive direction and of the zero position 25 Position Sensing Systems for Passive Vehicles H H SB WR Nr o DRV cc H o E m RCV SB RD H H Figure 2 7 Block diagram of the pre processing unit converters and multiplexers The essential function of the signal processing that occurs in the FPGA is to ensure that when the scale passes from one read head to the next one no position information loss occurs To achieve this the four analogue signals of the neighbouring heads Sa Ca Sg Cp will be sampled simultaneously in order to correctly calculate their relative phase difference The sampling rate of the analog to digital converters is set to 2 us This ensures that up to a maximal speed of 10 m s there are at least 2 samples of the sine cosine signals in each 40 um grating period of the sensor and the position acquisition algorithm can function correctly Reference signals from the first and from the last read heads R and Rs are also needed in order to detect when the vehicle is entering the sensors region They are brought to the FPGA through two comparators COMP R and COMP Rs At the beginning of every 100 us control cycle the control algorithm requests updated position information through the 16 bit Sensor Bus This bus is connected to the board s FPGA through RS 485 drivers and receivers Detailed descript
87. d for the zero position detection For a more detailed description of the sensors see section 2 2 2 Normally the passive scale of the sensors has the same length as the maximal travelling distance of the linear machine and is mounted on the stationary side while the active read head is on the moving part In the case of a passive mover the mounting must be reversed which means that the length of the sensor s scale is limited to the length of the mover If a travelling distance longer than the mobile part is required more than one read head will be necessary In the proposed system a number of five read heads are spread along the carriageway ca 210 mm apart from each other so that the 250 mm long scale covers at any given position at least one read head Because it is not feasible to mount the heads at distances that are exact multiples of the 40 um pitch one must calculate and compensate for the difference between the phases of two neighbouring heads as determined with the arctangent function When the scale passes from one read head to the next one the signals from both heads will be simultaneously evaluated once and the initial position of the incoming read head can be calculated Future position calculations are based on this initial value Thus a monotone variation of the calculated position along the entire measuring length can be ensured Figure 2 4 shows the block diagram of the proposed position acquisition system The signal
88. denhain company is used for position feedback in the control loop as well as for position reference The reference sensor has a pitch of 8 um and provides at its output TTL quadrature signals with a resolution of 0 8 um The scale of the reference sensor 6 is mounted along the track of the liner machine whilst the active read head is fixed on the mobile arm 5 attached to the machine s mover On the same mobile arm the slider of Fixed part mounted on the track Mobile part slider attached to the vehicle Figure 2 22 Capacitive sensor LE C 050025 1 A5 from Netzer Sick Stegmann 46 Position Sensing Systems for Passive Vehicles s E uBreeg oneg v mn Figure 2 23 Linear drive for testing the capacitive sensor the evaluated capacitive sensor 7 is also mounted 2 3 1 Initial Tests An overview of the setup used in the initial evaluation 33 is given in Figure 2 24 For controlling the linear motor and acquiring the capacitive sensor s position a DS1102 Controller Board equipped with a TMS320C31 Digital Signal Processor DSP was used The original electronics of the capacitive sensor output incremental quadrature signals RS 422 standard with a resolution of 4096 counts per pitch 2 mm The current reference for the converter is given through one of the Controller Board s digital to analog converters The control code was written using Matlab Simulink and complied with the Real Time Workshop RTW tool
89. dent parameterisation of the EMF and of the mechanical observers gt Figure 3 35 Block diagram of an EMF observer 106 Linear Drive for Material Handling By assuming die O0 dtl LOJ Eq 3 21 the resulting state equations of the EMF observer will be written as ia 00 1 ol amp 10 R Ou G 0 d Viy 0 0 0 1 Pis 0 1 0 R Iu 0 Gy bel dt 0000 6 00 0 Of G O yg Eq 3 22 000 0 amp loo o oji 0 G Eq 3 22 describes a so called disturbance observer 58 which assumes that the EMF term is a disturbance in the voltage equation see Eq 3 17 The consequences of this assumption on the estimation error will be analysed later on Introducing the estimation error vector Pis Tig Tig Sus e Tip Tip l Eq 3 23 e ea ea er ep i ep and considering the EMF observer described by Eq 3 22 and the machine model from Eq 3 18 the dynamics of the estimation error will be described by the following state equations Tiel Gy O A Oielo o M alY 0 G 0 1 p Jo 0 Tp bd dt G 0 0 0 6 1 Of vm g es 0 G 0 0 0 fj A B Eq 3 24 Tia 0 0 1 O Yu Pape 0 0 l e C 6 with A being the state matrix and B and C the input and output matrixes respectively It must be noted that the same estimation error dynamics the same state matrix A is obtained if the EMF observer described by Eq 3 20 is used under the assumption V v In order to determin
90. dix 5 3 Supply Control Board 3uo 9beyoA Hullo Ove 6N139 gyi Relay Outputs 24V6 1 7 ULN 15V6 2002AN n Ems bud 50 1 0 H aj V d 0V6 0V6 s 1 6 4069N Ovi 9N139 ove CUM lt H_ X ACT A e Oo N 24V6 1 7 ULN m 2002AN j AI peoe peo f gt N V6 Te ila 6N139 gl 1 6 4069N CUT lt H_ EI JCT A e ol N 24V6 1 7 ULN ine 2002AN A 8 Be ale cAI lt H X ACT 1 7 ULN 2002AN 15V6 B lt ACT 0V6 TW 1 6 4069N Ovi 9N139 gyg 30k 24V6 15V6 Z15V1 0V6 DC Link Voltage Measurement Connection to PCI DASOS Figure 5 3 Diagram of the Suppy Control Board 130 lt ZOOPNT e lt Oo lt LOOPNT e lt o ZOOPNT Q lt o lt LOOVNT S lt o sng SadiAaq WO Appendix 5 4 Real Time Module Structure The real time module is compiled from the eleven source files listed below Their ertmodule c emonitoring h ecapture h econtrol h detailed structure including the defined functions and the main data structures is given in Figure 5 4 Figure 5 5 and Figure 5 6 Main file initialisation cleanup and the ISR Implements the Monitoring State Machine Implements the data capture functionality Here is implemented the machine control algorithm Headers for Linux kernel functions Headers for RTAI
91. double double double double x_b xm x e seg parameters 18 ic0 Kp Rel GaP GIE theta diff static double kpx static double kpv tiv static struct double control references static struct double v i u control limits void abc2alphabeta double w a double w b double w c double w alpha double w beta Jj void alphabeta2abc double w alpha double w beta double w a double w b double w c void alphabeta2dq double w alpha double w beta double theta double w d double w q 1 void dq2alphabeta double w d double w q double theta double w alpha double w beta Figure 5 5 Real time module detailed structure 2 3 132 Data structures containing the machine parameters Various constants used in control Segments parameters definition Position and speed controllers parameters Position speed and current references Speed current and voltage limits Transformation from three phased a b c to two phased a p coordinates system fixed statoric Transformation from two phased a p to three phased a b c coordinates system fixed statoric Transformation from fixed statoric a to vehicle attached d q coordinates system angle 0 Transformation from vehicle attached d q to fixed statoric a B coordinates system angle 0 e position h e s2d h e sls emf h Appendix Synch
92. drives system this start position can be determined when the vehicle passes through a sensor equipped section of the track A topology of the power electronics and the control system is proposed in 16 for a similar linear drive It consists on using two inverters connected through static switches to the different segments of the track This configuration however allows driving only one vehicle on the carriageway Topologies that allow for multiple vehicles on the same track are discussed in 5 Methods to deal with the vehicle transition from one segment to other are addressed in 17 18 for the case of no gaps between segments If there are gaps between segments proposals can be found in 19 20 In the previous proposals the position is assumed to be a known variable The usual approach for measuring the position is using a linear encoder with the read head in the vehicle and a scale on the track The position signal is then transmitted to the stationary controller by a wireless modem e g 18 2 This can be source of high delay in the control loop and low reliability In addition this arrangement needs an auxiliary energy supply on board of the vehicle Using a magnetostrictive position sensor as it is proposed in 4 the vehicle can be totally passive However the position information has a large delay for longer distances which will not allow high dynamics in the control of movement see also Section 2 1 In this dis
93. e Axy represents the length of the vehicle 10 tp 240 mm and xe and xe repre sent the beginning and the end positions of the segment defined as the first respectively the last position where the vehicle is completely inside the considered segment see also Figure 3 36 When the vehicle is completely outside the segment the current i produces no force Ke x 0 when the vehicle enters leaves the segment a linear increase decrease of Kr x is assumed For two consecutive stator segments m and n the electrical force F can be ex pressed as Fe Kem x Ke x lg Eq 3 46 In Eq 3 46 it is assumed that the force producing current i is controlled to the same value for both segments An illustration of the principle of determination of the current force coefficient at the transition between two consecutive stator segments m and n which have different EMF constants Ke m and Ken is given in Figure 3 36 Based on the mechanical model of the machine Eq 3 43 the mechanical observer will be implemented using the state equations ee dt d E B Se UG dt M M M V x Eq 3 47 Dee 110 Linear Drive for Material Handling Figure 3 36 Current force coefficients where Fi v and X are the estimated load force velocity and position respectively The reference value of the electrical force F is calculated based on the reference current i M is the mass the vehicle and B is the friction coeff
94. e corresponding sine signal and then for each period of each read head the amplitudes and offsets of the sine and of the cosine signals are calculated The resulting correction tables can be seen in Figure 3 31 This data is saved in files to be used online by the control program Figure 3 32 shows an online measurement of the reconstructed position signal compare with Figure 2 17 in section 2 2 6 Position Reconstruction in C Code The articulated vehicle was moved again by hand over the entire measuring length in both positive and negative moving directions with the control program running but with no inverter being addressed Entering into the sensors region and exiting from the sensors region in both directions are also covered by the measurement In the lower part of the figure the corresponding speed signal is displayed The speed was calculated online as the numerical derivative of the position over the 100 us control interval and without any filtering A zoom of the speed signal at the transition between read heads 1 and 2 can be seen in Figure 3 33 The high frequency variations in the speed are derivation noise which is generated by variations of the time intervals between successive position readings as explained below As described in section 2 2 3 the Sensor to Digital Board acquires new position Forward moving Reverse moving Position m oO oO Qo P Position m N 0 8 06 E See zoom i
95. e direction During this acceleration as the speed increases the estimation errors decrease again In Figure 3 40 c the force producing currents of the two segments igi and ig2 are shown When the vehicle is inside a stator segment completely or partially the current of that segment is controlled to the reference value obtained at the output of the speed controller When the vehicle is at the transition between the two segments the currents of both segments are controlled 115 Linear Drive for Material Handling Sensor l T 05r Observed E KO 9 x Oo Ae is O A Sensor Observed b Speed m s Speed Error m s q C Current i_ A l C1 d Obs EMF V e Obs EMF 6 V 0 5 0 6 Time s Figure 3 40 Field oriented control using the estimated position and speed a Sensed and estimated position position error b Sensed and estimated speed speed error c Force producing current in the two stator segments i4 ig2 d Obs EMF a in segment 1 e Obs EMF a in segment 2 116 Linear Drive for Material Handling In Figure 3 40 d and e the observed a components of the EMF vectors corresponding to the two segments are shown The thin grey curve in these two figures represents the sum of the a components of the two EMF vectors 3 3 4 Implementation on the Entire Length of the Machine Artic
96. e for Material Handling 3 1 2 Power Electronics and Control The power electronics and the Control PC are all hosted in a cabinet located near the linear machine centralised architecture From this central cabinet the outputs of the 18 power stacks are fed through cables 4 11 m to the 18 segments of the machine Two cabinets as described in section 3 1 2 1 were built at our department one for each university lab involved in the project The Control PC including the software architecture and the real time control software is presented in section 3 1 2 2 3 1 2 1 Cabinet Figure 3 10 shows the interior view of one cabinet The 18 Power Stacks from the Simodrive 611 series 44 Siemens company are located in the upper side of the cabinet were chosen according to the nominal currents listed in Figure 3 5 There are devices of two different power ratings ePS1 PS6 are of type 6SN1123 1AA00 0CA1 containing 50A IGBT modules Their nominal current at the 5 kHz switching frequency used in our system amounts to 19 A rms whilst with a S6 type load cycle they can provide a 22 A rms current These Power Stacks feed sections 1 and 5 of the linear machine which have the highest power demand e PS7 PS18 of type 6SN1123 1AA00 0BA1 contain 25 A IGBT modules and have a nominal current of 7 5 A rms at 5 kHz They feed the remaining three sections of the linear machine Power Stacks Inverter Bus SeewSSe Sea eer SR en m Um
97. e ideal sensor will produce the same circle The radius of this circle is actually not important as long as the arctangent function is used to calculate the incremental position the phase inside one given period see also Eq 2 2 That is the amplitudes of sine and cosine can change as long as their ratio remains unity Amplitude errors will change this ideal circle into an ellipse and offset errors will move its centre away from the origin of the xy plane The correction tables aim to bring the locus of the output signals from each read head as close as possible to a circle thus minimizing the systematic errors The influence of the corrections on the locus described by the output signals of the sensors will be further described in section 2 2 8 Experimental Results Two types of parametric corrections were tested mean corrections and period corrections In the case of mean corrections only three values are necessary for each read head the mean values of the offsets of the sine cosine signals and the mean value of their amplitudes ratio The sensor signals can be acquired for the entire length covered by a read head with a storage oscilloscope and then the data is transferred in a PC where the mean values are calculated This method is simple to implement and requires very few correction values three per read head but it cannot compensate for variations of the systematic errors along the sensor s length To further improve the quality
98. e the gains Gy and G of the observer the characteristic polynomial det sL A 0 Will be used i e s 2G s G4 2G Js 2G G s G3 0 Eq 3 25 Eq 3 26 By matching the coefficients from Eq 3 26 with the ones of a polynomial with two negative double roots 107 Linear Drive for Material Handling 2 2 S p S P 0 p PzEeR p lt 0 p lt 0 Eq 3 27 the gains of the observer can be written as ee Eq 3 28 G P paso q The transfer function associated to the state representation from Eq 3 22 having the EMF vector as input and the EMF estimation error as output BR z is given by F s C sL A B Eq 3 30 which is represented by the following expression 0 E E ee F s tp S G s G vn s G Eq 3 31 5 0 Tp S Gys G In the previous equations s represents the Laplace operator From Eq 3 31 it can be demonstrated 58 59 that the error in the estimated electrical angle Oer introduced by the assumption from Eq 3 21 is bounded by X Gy Derr S atan Sr Eq 3 32 Tp G For a given maximal allowable electrical angle error O the EMF observer s gain ratio I can be calculated c E Tp tan O err E a r gt 0 Eq 3 33 Combining this expression of the gain ratio with the expression obtained for the EMF observers gains Eq 2 28 the following relation between the observers poles results 1 Dep Eq 3 34 In order to have a stabile estimator the follow
99. e transition between the processing station and transport section the synchronisation between the sensor based and the sensorless position feedback was also implemented in the control algorithm and was experimentally tested 4 1 Future work Future work regarding the control of the linear drive for material handling described in this thesis will be oriented towards expanding the functionality of the drive Concerning the position acquisition and motion control the project will be continued in two main directions 1 The implementation at the existing linear drive of a sensorless method based on the evaluation of the machine s anisotropies by means of injecting of a test signal This will allow for positioning of the vehicles in the transport region too The two sensor less methods the anisotropies based method and the one described in Section 3 3 EMF based complement each other the first method yields good results at standstill and low speeds while the second one is more advantageous at higher speeds Both methods should be implemented in the control algorithm and a synchronisation switching procedure between them should also be investigated 2 The operation of the linear drive with more than one vehicle must be tested For this at least a second vehicle must be added to the existing drive Each vehicle will have its own Vehicle Controller and in order to coordinate the movement of the vehicles a Motion Coordination PC will
100. e vehicle was moved by hand in positive direction throughout the entire measuring length S A C A and R S B C B and R at the beginning of Hia at the end of HB ac AAI JA S A C A and S B C B S B C B and S A C A S4 A C A and S4 B de at the transition H A Hog at the transition Hog 5 H B at the transition Ha A gt H 8 A A Figure 3 30 Correction tables data articulated vehicle 99 Linear Drive for Material Handling Amplitudes and Offsets of H 1 A 150 2 c 100r gt Os gt S 50 Oc O 0 1800 A m S1 gt 1600 O C1 O 4400 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Period Number Na Amplitudes and Offsets of Ho M Osp gt 5 Oca O L C 1400 2 Aso gt 5 1200 C2 O 1000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Period Number Na Amplitudes and Offsets of Fata and Hat L T D gt c O O L f gt gt c O O i i i i i i i i i i 0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 Period Number N Period Number Ng Figure 3 31 Correction tables articulated vehicle 100 Linear Drive for Material Handling Only the sine cosine and reference signals were received in burst mode from the Sensor to Digital Board The period numbers for each logical read head are determined offline based on the positive zero crossings of th
101. ections was tested at the experimental setup shown in Figure 2 18 The optical read heads 1 5 were mounted along the track 9 of a linear permanent magnet synchronous machine with active mover The scale 6 along with the reference selection magnets 7 was mounted on the vehicle 8 The test machine was also already equipped with another optical sensor not visible in Figure 2 18 mounted in the configuration used for linear machines with active vehicle scale on the track one read head on the vehicle see also Figure 2 1 The scale of this sensor has a 4 um grating period one tenth of the period of the tested sensors system therefore it was possible to use it as reference sensor Figure 2 19 a shows the sine and cosine signals of read head 1 of the tested system for all the AN 5250 periods covered by the sensor Ideally all the points should be located on the circle also depicted in the figure in reality due to the amplitude and offset errors they form an elliptical annulus which is not centred in the origin of the xy sine cosine plane The thickness of the annulus is an indication of the variation of the systematic errors along the sensor Using mean corrections as described in section 2 2 7 the representation of the corrected sine cosine signals from Figure 2 19 b was obtained The thickness of the annulus remains unchanged mean corrections have no influence on the variation of the systematic errors
102. edges the communication party driving the bus lines writes data on the bus whilst on its rising edges the other party reads the data The low and high pulses of T gt to t te t t tio ty t Us ts tis tiz tis PCI Interface Sensor to Digital Table 2 3 Sensor Bus data frames 35 Position Sensing Systems for Passive Vehicles the control line are 200 ns Tp long each After the time ts the PCI Interface stops driving all the bus lines setting the bus drivers in high impedance state Z The Sensor to Digital FPGA detects the falling edge of the control signal at time t after the delay time Tp Measurements have determined that on the experimental setup this delay does not exceed 130 ns The bus delay has no influence on the pulse period Tp but must be known for the Sensor Bus protocol to function correctly The data will be saved in the Sensor to Digital FPGA registers at time to on the rising edge of the control signal Simultaneously a data request is signalled by asserting the SB REQ signal Between t4 and ts the bus lines are not driven by any of the two parties The high impedance state lasts for Tz 200 ns Between t and ts the data request must be acknowledged by the Synchronisation block in the Sensor to Digital FPGA and the position information saved in the bus registers The contents of the bus registers will be sent on the Sensor Bus in six data frames D4
103. ements as the one shown in Figure 3 3 a which stator elements are formed by pressing from a Soft Magnetic Compound SMC The shape of these SMC elements allows for small angular and rotational displacements in every space direction between a Stator element with winding b Stator elements in up downhill configuration Figure 3 3 Stator elements Source 43 67 Linear Drive for Material Handling Vehicle Vehicle Translator oO EN ES BE ES EG ES EG ES E ES 1 I MEE v Eu Figure 3 4 Elementary machine Characteristic Value Number of poles 10 Pole pitch 24 mm Magnet height 8mm Vehicle translator length 240 mm Vehicle yoke material Solid iron Air gap 1 10 mm Number of stator slots 12 Slot pitch 20 mm Stator material Somaloy 500 Table 3 1 Linear machine characteristics Source 43 adjoining stator teeth In this way curved sections of the track can be constructed see e g in Figure 3 3 b a row of SMC elements arranged in up downhill configuration An elementary machine the smallest possible functional unit is sketched in Figure 3 4 It consists of 12 SMC elements teeth on which 6 concentrated coils are wound one coil every second tooth Corresponding to them there are 10 magnetic poles on the vehicle 10 12 configuration The main characteristics of one elementary machine are listed in Table 3 1 The linear machine used in
104. enance due to the lack of mechanical transmission elements 3 4 5 Figure 1 1 shows a simple example of combined transportation and processing of materials with linear drives In such applications the following properties are necessary for the linear drive system e The carriageway track must allow for horizontal and vertical curves as well as for closed paths and possibly switches S e A number of processing stations P1 P3 are spread along the track e On the carriageway several vehicles work piece carriers V1 V4 must be able to travel simultaneously with a high degree of independency eEach vehicle has to be controlled very precisely with an accuracy of a few micrometers when it operates within a processing station This requires the use of position sensors eWhile the vehicles are moving outside of the processing stations i e in the transport sections of the track usually a lower precision in positioning is sufficient S Switch V1 V4 Vehicles work piece carriers P1 P3 Processing stations Figure 1 1 Simple example of combined transportation and processing of materials using a linear drives system 17 Introduction so that a sensorless motion control should be used avoiding the complexity and costs introduced by position sensors The aim is to create a modular system which allows for different properties at different sections of the track e Straight sections
105. ers control signals For example in state H44 1 the signals from head 1 coming through MUX are used to determine the position In this state MCg 5 When the middle of the scale passes the head Na gt AN 2 a transition to state H44 occurs where MCg 2 CNTRs continues to remain blocked but LDg changes to 001 so the value 1 will be now loaded into the counter Thus the signals from head 2 are now coming through MUXs to the analog to digital converters and are ready for evaluation The multiplexers codes corresponding to each state of CHSM are depicted on the diagrams on the right of the state machine in Figure 2 11 The values of USE USEs LD and LDs are also listed for each state When in state Hiag and Na gt AN there is no direct transition to state Hog Because the control program must determine the offset of head 2 the position information of both heads from the same 2 us sample must be sent through the Sensor Bus So when NA gt AN the transition to state H442g occurs In this state LDg is set to 000 so CNTRa is also enabled and the position information from both heads is now valid indicated by USE and USE being both set This state is maintained until after a read request from the Sensor Bus is acknowledged i e until the falling edge of SB ACK comes When this happens the CHSM goes into state Hog where only the B position information will be used The transition between Hg and H 4 2 vehicle travell
106. estimation error is less than 0 05 m s in steady state At time t4 the range of the sensors is exceeded and the sensor based signals are set to zero The vehicle has exited the processing station and moves sensorless through the transport section towards the next processing station 3 4 2 Re entering the Processing Station In Figure 3 45 the position from the sensors xs the observed position x and the position used in control xc are shown when the vehicle enters the processing station At the beginning the vehicle moves through the transport region and the observed position is used in control xc X The vehicle enters the processing station at time to starting from this time the optical sensors provide valid position information xs with the observed position x amounting to ca 12 477 m which represents the total length of the track abstraction being made of the estimation error The synchronization process starts at tsync 2 one control interrupt after to so that the sensed speed vs is also available and it will be completed in two steps In the first step at Position usedin control Crossing of zero position of the track at time t Position m 0 06 0 04 0 02 Sensed position Estimation not available disabled SIS 10 5 0 5 10 15 20 25 30 35 Time ms Figure 3 45 Measurement of the position given by sensors of the observed position and of the position used in control
107. etic forces which would have lead to the breaking of the SMC stator elements The larger actual values of the air gap 0 require higher currents in order to achieve the nominal forces which currents are in some parts of the track higher than the ones that can by provided by the inverters It must also be noted that the air gaps are not only larger than their nominal values but within a given section they also vary leading to variations of the inductances and of the EMF constants of the segments The machine was delivered to our department in more stages First only section 1 together with a rigid vehicle was available as shown in Figure 3 6 This temporary setup was used for the initial testing of the control algorithms see also section 3 3 3 When the other sections of the linear machine together with an articulated vehicle capable of travelling through the curved section of the track were available it was possible to test the control algorithms on the entire linear machine see section 3 3 4 section 3 4 Figure 3 7 shows a transversal cut through the double sided section of the machine together with the vehicle 43 The two sides of the stator are mounted at a 70 angle in retaining brackets which brackets are spaced along the track at distances equal to the length of the vehicle 240 mm On the brackets there are also mounted two parallel tubes on which the vehicle travells on the carrying rollers and which also ser
108. f the reference signal respects the tolerances defined in Figure 2 6 the falling edge of Rs will come at the latest in the middle of quadrant 4 of period ANss so the state transition in CHSM will be triggered by Ng gt ANsp At time ts CNTRs counts up and Ns becomes ANss 1 One FPGA clock later time t4 CHSM switches to state Ho setting LDg 100 and another clock later time ts CNTRa is blocked outputting Ng ANsp d Entering sensor s region moving in negative direction The vehicle is outside of the sensors region moving in negative direction and approaching read head 5 from the right i e logical head 5B CHSM is in state Ho At time t the rising edge of Rs comes and all four sine cosine signals are sampled In this case Sp S and Cg Cs are relevant After first conversion CHSM changes to state Hsp 2 at time to CNTRg is enabled before it was blocked to ANsg and position information B is now valid e Transition between heads H and H4 moving in negative direction The scale covers heads 2 and 1 with the vehicle moving in negative direction CHSM is in state Hog CNTRa is counting down and CNTR is blocked at Na AN When Ng lt 1 time t the transition to state Hog44 occurs and CNTR is enabled LD 000 At t2 quadrant A changes from 1 to 4 and CNTRA counts down Na AN 1 The two counters run in parallel and the position information from both heads is ready for evaluation until t when on the rising
109. for all the stator segments of the linear machine are given in Table 3 3 The error vector is defined as the difference between the reference and the measured values of the controlled quantity at the sampling instant k e k x IK x k Eq 3 1 The integral component of the PI control is then calculated Jen OLE LE oes Eq 3 2 where Ts is the sampling interval 100 us and Kp and T are the proportional gain and the integral time constant of the controller respectively Then the output of the controller actuating variable is determined first without limitation w k u k K e k Eq 3 3 89 Linear Drive for Material Handling Ts Kp 2l Figure 3 21 Discrete Pl Controller with anti windup Sec Stator Phase Phase Current control parameters Resistance Inductance Ron Q Lon mH Ko V A Tii ms tion Segments Table 3 3 Current controllers parameters Afterwards the actuating vector is limited to Ymax and the output of the controller results w k i wik Y max yIk wik MN if wik gt Veron jwi wIk gt y Eq 3 4 If a limitation of the output occurred the integral term is also adjusted anti windup en Im if wIK lt Ymax y k Kp efk if wIk gt Ymax After the reference voltage vectors of the controlled stator segments were calculated as outputs of the current controllers they are transformed from the vehicle attached reference system dq back to the s
110. functions Vendor IDs and Device IDs of the three PCI Interface Boards used for their identification The two FIFOs used command and data Definition of the commands of the Command FIFO DC Link state definition Machine state definition static enum Communication state definition Module initialisation Command FIFO read handler static void isr void Interrupt Service Routine Module clean up rtmodule c include lt linux kernel h gt include lt linux module h gt include lt linux pci h gt Hl include lt rtai h gt include rtai sched h gt include rtai fifos h gt include lt rtai_math h gt define VCI BRD VID 0x1172 define VCI_BRD_DID 0x0004 define POS BRD VID 0x1172 define POS BRD DID 0x0005 define DAS 8 BRD VID 0x1307 define DAS08 BRD DID 0x0029 define FIFO CMD define FIFO DATA 1 define CMD SET PARAMS define CMD SET CORRS define CMD DC LINK POW define CMD MACHINE ON OFF define CMD CAPTURE static enum DC LINK DISABLE DC LINK ENABLE dc link state l static enum MACHINE OFF MACHINE ON machine state COMM OK COMM ERR UI COMM ERR DASO8 COMM ERR S2D comm state include monitoring h include control h include capture h int init_module void void get ui command void monitoring OF machine_control data_capture void module cleanup void
111. g The conversions are finished after a maximum time of 750 ns 27 which is indicated by LD 3 Signification 000 001 010 100 Other Table 2 2 Possible values of LD CNTR is enabled and counts based on MSB S4 and MSB Ca Value 1 is loaded in register CNTRA Value AN is loaded in register CNTRA Value AN 2 is loaded in register CNTRA Illegal Changes in MSB Sa and MSB CA have no influence on the counter s value 29 Position Sensing Systems for Passive Vehicles the de asserting of the four busy signals BS S BS Cg At this point the four read signals RD Sa RD Cg are used to sequentially bring the conversions results into four temporary registers R TMP S4 12 R TMP Cp 12 This sequential reading through the ADC BUS 12 takes 200 ns After that the ADC Control State Machine asserts the ADC FINISHED signal determining the Synchronisation State Machine to assert the UPDATE SC REG signal thus simultaneously updating the content of registers R S 12 R C4 12 R Sg 12 and R Cg 12 Based on the new values of S4 and CA their sign bits the Quadrant A State Machine will be set in the correct state The first conversion was made soon after the rising edge of R4 This means that the quadrant determined by Sa and Ca can be 2 3 or 4 depending on the width of R4 see Figure 2 6 As previously stated CNTR is blocked at the value 1 by CHSM Because the scale does not cover head 5B the va
112. g between consecutive stator segments which must be taken into consideration by the control algorithm As previously stated position sensors are required within the processing stations Currently there exists a multitude of linear position sensors based on various physical principles optical inductive capacitive Hall effect magnetoresistive magnetostrictive etc 6 7 However not all of them are suitable for our application The following re quirements must be fulfilled by the position sensing system 1 The vehicles are fully passive i e there is no auxiliary energy source available on board of the vehicles Measuring accuracy of a few micrometers Measuring length of several meters Measuring interval of 100 us or less The delay introduced in the position and speed control loops by the measuring system has to be as low as possible in order to allow for a high dynamic load stiffness 8 A detailed study 9 concerning different position sensing principles and commer cially available position sensors was conducted at our department Based on this study two types of position sensors suitable for linear drives with passive vehicles were selected for evaluation optical sensors and capacitive sensors In the transport sections of the linear drive for material handling described in this work there are no requirements for high dynamics or high positioning precision In these sections the complexity and costs intr
113. g com specifications conventional Link retrieved in June 2010 Dr Johannes Heidenhain GmbH Exposed Linear Encoders Specifications July 2003 Linear Technology LTC1410 12 Bit 1 25Msps Sampling A D Converter with Shutdown Datasheet 1995 http cds linear com docs Datasheet 1410fa pdf Link retrieved in June 2010 B hr A and Mutschler P Systematic Error Correction Methods for Sinusoidal Encoders and their Application in Servo Control in Proc of 10th European Confe rence on Power Electronics and Applications EPE 2003 Sept 2003 H scheler B Erh hung der Genauigkeit bei Wegme systemen durch selbst lernende Kompensation systematischer Fehler in Proc of SPS IPC DRIVES 1999 pp 617 626 Nov 1999 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 Bibliography B nte A and Beineke S High Performance Speed Measurement by Suppression of Systematic Resolver and Encoder Errors IEEE Trans on Ind Electron vol 51 no 1 pp 49 53 Feb 2004 Netzer Precision Motion Sensors LE C Linear Encoder With Cable free Read Head Specifications March 2004 Netzer Y Linear Absolute position Output Cable free Electric Encoder in Proc of Power Conversion Intelligent Motion Conf PCIM 2003 pp 135 139 May 2003 Mihalachi M Characterization of a Capacitive Incremental Encoder and Design of an Improve
114. g time of the control algorithm for one vehicle does not exceed 22 us 21 7 us in Figure 3 14 thus it is possible to implement the control for two or three vehicles in the same PC It must be noted that as previously stated the Linux kernel code executes during the time when the Service Routine of the control interrupt is not running so if the execution time of the control algorithm is close to the 100 us control cycle it can have a negative impact on the responsiveness of the non real time part of the operating system Figure 3 15 shows a more detailed overview of the software architecture The machine parameters as well as the correction tables for the optical sensors are stored in files on the hard disk of the Control PC When the User Interface is started it will read the contents of these files and transfer it using the real time FIFOs to the Real Time Module which at this point is already loaded into kernel 79 Linear Drive for Material Handling pJeog 2e U1 IA 24 Aq payesaueb jsenbai ydnsequy mme K ugnoy 321S 1dn449 uU O414 8 amp 3eq 94 ul MN UONIIM SI Sen eA Jo pojq YJOMJON Mou e en p jje Sr DICE CICERO ccc areata cd jeuJou 3 1940 JojsueJ 2q SONO Ba AE nn u a RENI 4l aea ee Be peinqydes i OJIJ e3eq Ae ds q snyeIs sisAjeue uoneinje So ge L sega c uonoauo i az JOUL syepdn S OJI4 pueululo
115. g was used for the speed signals in any of the cases A major reduction of the variation due to derivation was obtained by using the periodic correction tables 2 2 9 Implementation of the Optical System The optical system described in this section was implemented on the linear drive for material handling described in section 3 of this work The details of the implementation of the optical sensors system are given in sub section 3 2 45 Position Sensing Systems for Passive Vehicles 2 3 Capacitive Sensor The optical sensors system described in the previous section achieves a high accuracy especially after the correction of the systematic errors but due to the large number of read heads is expensive It also requires complex electronics and complex processing algorithms for the synchronisation between read heads Therefore simpler and cost effective alternatives should be investigated One of the alternatives can be the position sensor described in this section The sensor shown in Figure 2 22 works based on the capacitive principle an electric field is generated by stationary electrodes on the fixed part of the sensor the mobile part attached to the mover of the linear machine slides between the stationary electrodes causing a position dependent modulation of the electrical field The exact arrangement of electrodes and an electrical model of the sensor will be presented in subsection 2 3 2 The fully passive slider makes the
116. gital Board Sensor Bus Soft Magnetic Compound Stator Segment Transistor Transistor Logic User Interface Universal Serial Bus Vehicle Controller Vehicle Controller Interface Symbols Optical Sensors System Section 2 2 BS ICS RD AN C4 C5 Ca Cp Do D7 H1 Hs Ha Hs LDa LDg LDI LDO MCA MCs Na Ng P R Ri R5 RB R TMP 94 95 Sa SB SB ACK SB IN SB OUT SB RD SB REQ SB WR Tp Tp Tr Tz USE USEs Analog to Digital Converter Busy Signal active low Analog to Digital Converter Chip Select Signal active low Analog to Digital Converter Read Signal active low Number of periods covered by an optical read head Cosine signals from optical read heads 1 5 Cosine signals routed through Multiplexers A B Data Frames 0 7 Sensor Bus Communication Optical read heads 1 5 Optical read heads whose signals are routed through Multiplexers A B Load signals of Period Counters A B Local Data Input PCI Interface Firmware Local Data Output PCI Interface Firmware Control Codes of Multiplexers A B Values of the period counters A B Period pitch of the optical sensors Register Sensor to Digital Firmware Reference signals from read heads 1 leftmost and 5 rightmost Sensor Bus Register Sensor to Digital Firmware Temporary Register Sensor to Digital Firmware Sine signals from optical read heads 1 5 Sine signals routed through
117. hal ko the hardware abstraction layer rtai_sched ko the real time scheduler rtai_fifo ko real time FIFOs related functions rtai_math ko this module redefines the mathematical libraries functions for use in the real time environment e After the RTAI modules are successfully loaded the real time module used for the machine control rt_control ko referred to as the Real Time Module throughout this section is also loaded When first loaded the Real Time Module is not addressing any Power Stack Interface the feeding of the DC Link is disabled and no control algorithm is running 77 Linear Drive for Material Handling LS ste ri res Control Sensar 2 Digital DC Link is enabled Position mm Press to DISABLE 268 158 m Machine is stopped Press ta START o e 9 2B rData Capture Monitoring Fl E2 EM K2 K3 Points 50000 a 2222 9 Decimation 5 Active Int Mr Ude V 966563 1566 02 Synchronize with Machine Start Start Save Discard Q Q Q LU I DAS 52D Capt Spt t Communication Status Figure 3 13 Screenshot of the User Interface window e After the Real Time Module is loaded the script runs the User Interface program e Through the User Interface commands can now be sent to the Real Time Module power up the DC Link start the control algorithm capture data
118. he Inverter Bus is a 16 bit parallel bus implementing the RS 485 differential signalling standard identical with the Sensor Bus described in Section 2 2 4 actually the Inverter Bus was developed at our department prior to the Sensor Bus and the latest was designed in such a manner as to take advantage of the already existing hardware The Inverter Bus Master requires no actual hardware being integrated in the FPGA firmware of the first Vehicle Controller Interface Figure 3 2 presents a simplified diagram of the Inverter Bus communication protocol 42 As aforementioned there are three types of communication partners the Inverter Bus Master the Vehicle Controllers through their respective VCls and the Power Stacks also through their PSIs Within every 100 us control cycle the Inverter Bus Master allocates a communication time slot of 10 us for each Vehicle Controller E g at time t O us in Bus Vehicle Controller VC Power Stack PS Master u ERR ee 4 3 3 53 n2et m 1 2 3 4 spere c n 2 n 1 Vel Modulation Information PS n 5s IRQ No other PS is being controlled Call 10us H VC2 Actual Currents PS 2 1 154s T S T Sun Poss m being controlled 20us Call VC3 gt de 25us lt x 30us Call Ms S Mod Modulation Inf PS 3 a _Actual Currents PS 3 35u S 8 i lt a re e JE othenies 1s t 4 2
119. he Real Time Module see Figure 3 15 and updates the Status display section or if it is the case of a user initiated capture gathers all the received data in an internal buffer which can be subsequently saved on the hard disk The Real Time Module will write in the Data FIFO a number of data blocks as required by the number of points argument of the capture command received the decimation argument defines how many interrupts occur between successive data savings One block of data contains up to 30 status and control variables of the Real Time Module all acquired during the same control interrupt Data included here e The number of the interrupt when the data block was acquired eThe state of the monitoring for details see Figure 3 18 and the accompanying explanations e The value of the DC Link voltage if the DC Link was powered up e The position and speed given by the optical sensors system if the vehicle is inside sensors region including the current read head number s e Observed position and speed when the vehicle is outside sensors region e The currents reference voltages and observed EMFs for each of the two controlled stator segments In the following the structure of the Real Time Module is briefly described A more detailed structure including all the defined functions and the main data definitions is given in Appendix 5 4 The Real Time Module consists of four principal functions e Mod
120. he board is shown in Figure 2 14 1 PCI Interface FPGA 2 EPROM storing the FPGA firmware 3 80 MHz oscillator 4 PCI Bus buffers 5 PCI connector 6 JTAG programming connector 7 Active Serial programming connector 8 Auxiliary inputs outputs used for debugging 9 5V and 3 3 V power supplies connected to those coming through PCI connector 10 DC DC converter for the 1 5 V supply required by the FPGA 11 Optocouplers separating the potentials of PC and Sensor to Digital Board 12 Sensor Bus drivers and receivers 13 Sensor Bus connector 14 Separate 5 V supply for the Sensor Bus drivers and receivers 36 Position Sensing Systems for Passive Vehicles E LETTIENTTIT Figure 2 14 PCI Interface Board Figure 2 15 depicts a simplified block diagram of the PCI Interface FPGA firmware The PCI Interface Function block provided by the FPGA s manufacturer handles the PCI bus protocol Data coming from the PCI bus is available on the 32 bit Local Data Input bus LDI and data can be sent on the PCI bus using the 32 bit Local Data Output bus LDO The Data Transfer Control State Machine is responsible for decoding the PCI bus addresses received from the PCI Interface Function block through ADDR 32 A number of five addresses are recognized e ADDR D The control program writes at this address the data frame Do as defined in Table 2 3 e ADDR D4D5 ADDR D D and ADDR D D The
121. hine s winding at the transition between segments there will be a reduction in the modulus of the sum of the two EMF vectors even at constant speed and constant air gap e There are variations of the air gap especially in the curved sections of the track In order to overcome these problems a mechanical observer will be used which observer only uses the phase information contained in the estimated EMF vector s for determining the position and speed The following 3 order model will be used to describe our linear machine s mechanical subsystem PUT Ba Eq 3 43 109 Linear Drive for Material Handling In Eq 3 43 F and F represent the electrical force and the load force respectively v is the vehicle s speed and x represents its absolute position M is the mass of the vehicle and B is the friction coefficient The load force F is assumed to be constant When the vehicle is completely inside one stator segment the electrical force F can be expressed as Fc 3 2 Ke ig Eq 3 44 where Kg is the mean EMF constant of the considered segment and i is its quadrature force producing current If the vehicle s entering and leaving of the segment is also considered then the current force coefficient Ke x of that segment will be defined as follows 0 X Xg AX X X Ax 3 2 Ke ea er Ne Axy Ke x 3 2 Ke Xg lt XSX Eq 3 45 X AX J X jog UE Xe lt X Xc AXy V 0 X Xg AX wher
122. his state has no exit in order to restart the control algorithm after disabling the feeding of the DC Link the user must stop and restart the Real Time Module and the User Interface Interface Electronics Power Supply Failure When the Monitoring State Machine is in one of the states MO M4 and a failure occurs in the electronics power supply EM 0 K2 and K3 will be opened thus deactivating the feeding of the DC Link and the operation of the state machine will block in this state DC Link Power Supply Failure The DC Link supply was stopped because E2 became inactive Cause the Off button T1 on the cabinet s door was pressed before deactivating the DC Link power supply from the User Interface Infeed Monitoring Module Failure The DC Link supply was stopped because E1 became inactive while E2 was still active transition from M4 or 3 seconds after the closing of K3 Ato Ats E1 was still not active transition from M3 Table 3 2 Description of the states of Monitoring State Machine States M5 M8 are all final states they have no outbound transitions with the same outputs K2 0 K3 0 They were introduced in order to communicate to the User Interface and further to the user the reason why the normal execution was stopped A block diagram of the machine control function is given in Figure 3 19 and Figure 3 20 At the start of the control function a position update request is sent to the Sensor to Digi
123. icient Gr Gy and G are the gains of the mechanical observer The correction term ez is discussed below For only one stator segment the correction e is calculated x m mx e E sgn v COS SIn a Eq 3 48 Tp Tp 85 In order to analyse the action of the correction term Eq 3 14 will be introduced in Eq 3 48 yielding _ TX S TUX TX Tp e sgn v v Ke cos sin Eq 3 49 Tp Tp TX cos Tp By assuming that sgn v sgn v e can be reduced to T a TC ex K usin 2 2 x Eq 3 50 Tp Tp If x and x are close to each other then s can be liniarised ES e M amp x P Eq 3 51 For analysing the dynamics of the mechanical observer the same method of the error dynamics will be used as in the case of the EMF observers Defining the estimation errors F F F V v V X x X Eq 3 52 111 Linear Drive for Material Handling the equation describing the dynamics of the estimation errors can be derived from Eq 3 43 and Eq 3 47 0 0 G Keln tp l d 1 B dt _ M M Eq 3 53 X X 0 1 G Kelvin p AT Ye A Matching the coefficients of the characteristic polynomial det sI A 0 Eq 3 54 with the ones of a 3 order polynomial with the poles p p2 and ps the following expressions are obtained for the mechanical observer s gains G Mv Pi P2 Ps Kg v 2 te 2 G _ B M B M P4 P2 3 P4 P2 D2 P3 5 P Kent 000 E
124. ignals were measured at the inputs of the four analog to digital converters whilst the reference signals were measured at the outputs of the two comparators see Figure 2 7 Ideally the sine cosine signals generated by the optical sensors would have an amplitude of 1 Vpp over the entire measuring length the analog to digital converters have an input range of 5 Vpp so a gain of 5 would be necessary at the Sensor to Digital board s signal conditioning stage However the gain was set to 4 3 86 so that in the case of small local increases of the amplitudes over the nominal value the analog to digital converters outputs would not saturate With this gain 4 3 Vpp signals are expected at the inputs of the analog to digital converters i e 2 15 V which corresponds to approx 1760 converter units The per unit values in the following figures Figure 3 28 Figure 3 30 take into consideration this reduced gain In Figure 3 27 a and g the relationship between the reference signals and the a H Beginning of Read head 1 Sa S4 Ca Cy 02 0 0 2 0 4 Time ms b H S V C M Middle of Read head 1 Sa S4 Ca Cy S IV C IV i H H on Transition Read head 1 Read head2 Sa 7 84 Ca Ci m Ss S2 Cg C2 Time ms Figure 3 27 Measured sine cosine signals 95 Linear Drive for Material Handling d H S V CIV Middle of Read head 2 Sp S5 Cg Co S V C M
125. igure 3 24 the straight double sided section of the linear machine stator segments 1 and 2 is depicted with the optical read heads mounted Because this section is just four elementary machines i e 0 96 m long a travelling distance of only 0 72 m is available with the 0 24 m long vehicle remaining fully inside the section and only three read heads can be completely covered Therefore the leftmost and the rightmost read heads were not used The FPGA code on the Sensor to Digital Board and the position calculation function in the control program were modified accordingly by reducing the number of states in the Current Head State Machine and in the Reconstruction State Machine respectively see section 2 2 3 and section 2 2 6 It must be noted that the distance between the read heads was also slightly reduced to 0 2m 5000 sine cosine periods so the total measuring length now amounts to 0 6 m This 3 read heads configuration was also kept when the entire linear machine all the 18 segments together with the articulated vehicle was available Even though now the travelling length is extended in both directions by stator segments 3 and 18 when covering the leftmost and the rightmost read heads in the original 5 read heads configuration the articulated vehicle is already partly into the curved section of the track and the scale is rotated with respect to these read heads making the sine cosine signals unusable for position dete
126. in a nutshell ISBN 978 0 596 10079 7 O Reilly Media Inc 2007 pp 127 Racciu G and Mantegazza P RTAI 3 4 User Manual rev 0 3 Oct 2006 Dozio L and Mantegazza P Real Time Distributed Control Systems using RTAI in Proceedings of the 6 IEEE International Symposium on Object Oriented Real Time Distributed Computing ISORC 03 pp 11 18 May 2003 Schwebel R Echtzeit unter Linux mit RTAI Elektronik No 7 2002 pp 72 77 Glade A User Interface Designer http glade gnome org Link retrieved in June 2010 The GTK Project http www gtk org index php Link retrieved in June 2010 Intel Corporation Intel 64 and IA 32 Architectures Software Developer s Manual Volume 3A System Programming Guide Part 1 pp 221 241 March 2010 http www intel com Assets PDF manual 253668 pdf Link retrieved in June 2010 Cupertino F Giangrande P Salvatore L and Pellegrino G Sensorless position control of permanent magnet motors with pulsating current injection considering end effect Energy Conversion Congress and Exposition 2009 ECCE 2009 pp 1954 1961 Sept 2009 Leidhold R and Mutschler P Specific features of Position Sensorless Methods in Synchronous Linear Motors in Proc of the 2010 International Exhibition amp Conf for Power Electronics Intelligent Motion Power Quality PCIM 2010 pp 124 129 May 2010 Tomita M Senjyu T Doki S and Okuma S New sensorless co
127. in this situation Xc t xs t for ts tons Eq 3 62 The time when the sensor based position xs approaches its maximal or minimal value is called tsync1 After this time the control algorithm will start using the observed position and velocity Taking into consideration that the observed position x is not absolute as opposed to the one given by the implemented sensors system the difference Axc between the two positions will be calculated at tsync 1 AXc Xg toync1 X teynca for t teyncs Eq 3 63 The difference Axc will be used to correct the estimated position After tsync 1 the position used in control will be calculated as Xc t x t Axe for t gt tamar Eq 3 64 This method of synchronization also ensures a smooth transition of the electrical angle used for field orientation For more details see the discussion of the experimental results in subsection 3 4 1 When the vehicle enters the processing station it must be taken into consideration that the position given by sensors is more accurate than the estimated one so in this case using a position correction similar to the one defined by Eq 3 63 will only introduce an error in the sensed position xs The straightforward way to do the synchronisation in this case would be simply to use Xs as soon as it is available as soon as the vehicle enters the processing station But if the estimation error in the observed position is large a step will be introduced in the
128. ing in the negative direction happens similarly passing in this case through state H g 44 Some examples of transitions of the CHSM states will be illustrated in the last part of this section based on the CHSM representation in Figure 2 11 and on the simplified signal diagrams in Figure 2 12 a Entering sensor s region moving in positive direction Initially CHSM is in state Ho and CNTRA is blocked to 1 When the rising edge of R4 comes time t the sampling of Sa S4 and Ca C4 begins At time t when the first conversion is finished CHSM changes state to H14 14 and CNTR is enabled At time ts after passing through the zero position of the sensors region CNTRA counts up yielding NA 7 0 and continues counting based on the sign bits of SA and CA 32 Position Sensing Systems for Passive Vehicles tit t cHsM Hy GE o y f Leaving sensors region moving in negative direction c Leaving sensors region moving in positive direction d Entering sensors region moving in negative direction Figure 2 12 Examples of state transitions in Current Head State Machine 33 Position Sensing Systems for Passive Vehicles Any reading of the position information before to will result in USEA 0 and USEs 0 vehicle outside sensor s region If the position information is read between tz and ts the control program will calcul
129. ing relation must also be fulfilled Po p lt T Eq 3 35 The gains of the EMF observer can be now expressed as T p D p 1 Eq 3 36 a emer Gy CEN 108 Linear Drive for Material Handling The pole p can be chosen in order to obtain the desired dynamics for the EMF observer while the gain ratio I will ensure that the orientation error will not exceed the maximal allowable value O For example if a maximal orientation error 05 25 is required this corresponds to a reduction of less than 10 in the electrical force at a maximal speed of 10 m s then the resulting gain ratio yields V 24 10 m tan 25 i OR m 356 233 10 s rad Eq 3 37 Choosing e g p 5000 rad s Eq 3 38 satisfies the condition p lt I 2807 153 rad s Eq 3 39 and the second pole of the EMF observer results p 6400 rad s Eq 3 40 The resulting observer s gains are Gy 11 4 10 rad s G 32 10 rad s Eq 3 41 3 3 2 Mechanical Observer Several contributions can be found in literature e g 60 61 in which the components of the estimated EMF vector are used directly for the calculation of the estimated position and speed X atan E xe V mM Eq 3 42 e Kc Kc This approach is relatively simple and can yield good results e g for rotating PM synchronous machines but it cannot be used at our linear drive because e Due to the gaps in the mac
130. ing technique was adopted because it allows for a frame length that is not dependent on the delays on the bus delays caused by the RS 485 drivers and receivers the propagation delay on the bus cable and especially delays caused by the optocouplers on the PCI Interface Board Figure 2 13 shows the transfer protocol between the Sensor to Digital and the PCI Interface Board In the upper part of the figure the control and data signals are sketched as seen from the PCI Interface FPGA while in the lower part as seen from Sensor to Digital FPGA Initially the PCI Interface drives the bus holding all the lines at logical 1 When required from the control program time to it starts sending a data request to the Sensor to Digital Board Data frame Do sent from control program accompanies this data request The content of Do is listed in Table 2 3 bit 14 ENA will be saved in the R ENA register of the Sensor to Digital FPGA and will enable or disable the position acquisition algorithm Bits 13 O contain the real number of periods covered by head 5B The sign bit of ANsg is not sent on the bus it is always 0 but instead the SETsg bit will be sent This bit is used to save ANsg in R ANss only once at the beginning of the control program Frame Do is held on the bus until t4 400 ns In this time the control line is held low for 200 ns and then back high for the remaining 200 ns This line is acting as bus clock signal on its falling
131. initial state H o It is then assumed in Figure 2 17 that the vehicle re enters the sensors region but now from the right time tg Now the offset of the firstly covered read head 5B is no longer zero the offset of head 5B calculated during a run in positive direction can be used For example xo to xo t The reconstruction state machine will change state to H sg and the information from head 5B Sg Ss Cg Cs and Na will be used together with Xo to to calculate the global position The procedure is identical with the one discussed for positive direction entering from left 2 2 7 Correction of Sine Cosine Signals Measurements on the experimental setup have shown that the sine and cosine signals produced by the five read heads contain amplitude and offset errors that cannot be neglected These systematic errors of the sine cosine signals will result in an error of the calculated position and must therefore be minimized Parametric correction tables 28 29 were used for the correction of the sine cosine signals Even though a reference position sensor is available at the experimental setup its use for the generation of the correction tables was avoided The corrections are generated based solely on the shape of the sensors sine cosine signals in a xy representation the sine and cosine signals coming from an ideal sensor will describe a circle with the centre in the origin of the said xy plane and all the periods of th
132. ion of the communication protocol on Sensor Bus is given in section 2 2 4 The vehicle can enter the sensors region from left or from right so the sine and cosine signals from read heads 1 and 5 must be simultaneously available This is achieved when the signals from head 1 come through MUX MC 1 and the ones from head 5 through MUXg MCs 5 On the other hand the signals from read head 4 are 26 Position Sensing Systems for Passive Vehicles routed through MUXs so when the vehicle is travelling in positive direction at the transition from head 4 to head 5 the sine and cosine signals of the last must be available through MUX otherwise it would not be possible to sample the four signals simultaneously This is why S5 and Cs are routed through both MUX and MUXz The distance covered by read head 5 will be split in two in the first half the signals from MUX will be used and in the second half the signals from MUXg So to the physical read head 5 correspond two logical read heads 5A and 5B each covering half the distance covered by head 5 At the transition between the two logical heads the position phase difference is zero In the following heads 5A and 5B will be treated in the signal processing as two different read heads The realisation of the pre processing unit Sensor to Digital Board is shown in Figure 2 8 The five optical sensors are connected to the board through connectors 1 5 In the sign
133. is based on sinusoidal excita tion and position determination by phase measurement Table 4 1 gives a comparative overview of the evaluated position sensing systems nn Capacitive Encoder zen Optical Sensors System Square Wave Sinusoidal Evaluation Excitation Excitation criterion Pitch of the sensor Position error without any corrections peak to peak Integral Local Position error with parametric corrections peak to peak Integral 120 um Local 30 um see note Very High Moderate Low four ADCs plus syn only one fast only digital chronisation firmware ADC is required processing Complexity of the signal processing hardware High High arctangent arctangent divisions etc divisions etc Computational resources High Large number of read heads required Cost ofthe system incl signal processing Note For evaluation of the capacitive encoder with sinusoidal excitation only a mean correction was used Using individual period corrections as in the other two cases would certainly improve this result Table 4 1 Comparative overview of the evaluated position sensing systems 125 Conclusions For both optical and capacitive sensors correction tables were used in order to improve the accuracy these corrections tables were generated based solely on the information provided by the evaluated sensors i e without the use of the reference sensor which is typically not availa
134. is shown by the dark line The local deviations are reduced to 15 um This remaining local deviation may be due to harmonics in the signals of Figure 2 34 which are not compensated by the parametric correction table The integral deviation remains unchanged it does not depend on the quality of the sampled signals This deviation along the whole measuring length is likely to be due to imperfections in producing the pattern of transmitting electrodes and temperature expansion of FR4 One way to improve this would be a very precise manufacturing of the transmission electrodes on a substrate of temperature stable ceramics Another way might be the usage of a reference sensor when the correction table is generated This can compensate global imperfections during production but not temperature expansion of FR4 Figure 2 36 shows in red line the results of applying a correction table which was generated using the reference sensor The integral error is compensated now Such a correction table may be generated at the manufacturers test bed during the final production test before the sensor leaves the manufacturer 58 Position Sensing Systems for Passive Vehicles 2 3 4 Sinusoidal excitation From Figure 2 33 it is obvious that the transients generated by switching of the exciting voltages are a major source of problems It is possible that transients may not have decayed sufficiently at the sampling instant when analog to digital conversion starts
135. is shown in green line in Figure 2 43 Figure 2 43 and Figure 2 36 show very similar results of the position deviation for the two different methods sinusoidal and square wave excitation respectively but we have to note that only an average correction was used in Figure 2 43 Using a correction individually for each period of the sensor should improve the result of Figure 2 43 The main difference between the two methods is the expense of the implementation Using the square wave excitation method requires a fast 12 bit analog to digital converter and computational resources to perform divisions and calculate the atan2 function For production of sensor systems in large quantities it is profitable to do the necessary signal processing in an ASIC It is much easier to implement the signal processing used for the sinusoidal excitation method in an ASIC no analog to digital converter is necessary and no resource demanding calculations like division or atan2 function are required This will result in a rather simple and cheap ASIC for the sinusoidal excitation signal processing For lower production quantities it is no problem to implement the processing required by the sinusoidal excitation method in an FPGA 63 Position Sensing Systems for Passive Vehicles Position deviation over the entire measuring length lj M I m Ir M IN ilii TI I n T M ill il L IL vii Position deviation um vA
136. is started and after it is finished new values of S4 Ca Sg and Cg will be available leading to the update of the quadrant state machines and of the period counters and based on the updated counters values to the update of the state of CHSM During this update time the position information can be inconsistent The redundant temporary sine cosine registers in the ADC Control block where introduced to minimize this inconsistency time Instead of 200 ns as it takes to sequentially update the four conversion results through the ADC Bus the time for updating of the sine cosine registers R Sa R Cg is reduced to one FPGA clock 12 5 ns Adding to this time the time necessary to update the quadrant state machines the period counters and the CHSM an inconsistency time of 50 ns results Any data request coming from Sensor Bus during this time will be delayed until the position information registers are consistent again and the delay must be short enough not to impair the timing requirements of the Sensor Bus protocol When the Sensor Bus Control State Machine receives a position information request it asserts the SB REQ signal The Synchronisation State Machine responds by asserting the acknowledge signal SB ACK with a delay no longer than 50 ns if necessary as discussed above 30 Vehicle outside sensor region initial state Scale covers head 1 first half Scale covers head 1 second half Synchronisation between
137. ished through real time FIFOs Linux character devices used for inter process communication From the Real Time Module the RTAl provided FIFO manipulation functions create read write destroy etc are used whilst the User Interface located in the user space can access these FIFOs like normal sequential files 16 Hardware Other System Hardware Kernel Space Linux Device Drivers Linear Drive for Material Handling User Other User Interface Space Linux Processes Linux Process System Call Interface Linux Kernel memory management process management Real time virtual file system etc Mo dule Peg RTAI Hardware EN Abstraction Layer HAL Direct Hardware Real time Scheduling Access RTAI Real time Kernel m n Ku umm umm umm um uum Rum mum mum um um mum mum mum mum oum e e c c c a a a a C a a a a ru ru DE m mm e e e e e e m mm Other Hardware Control Interrupts Interrupt hard disk network card peripherals etc PCI DAS08 Vehicle Control Position CO CO fo O O ea ea D eb O O d9 S fa oO 4 4 I Figure 3 12 Linux RTAI scheduling In order to coordinate the starting and stopping of the Real Time Module and User Interface a small Linux shell script was written e The shell script first prepares the real time environment loading into the kernel the necessary RTAI modules gt gt gt gt rtai_
138. itted from the PCI bus to the interrupt pin of the CPU and by the time necessary for the context switch when the CPU receives the interrupt request it finishes the current instruction saves the values of its internal registers in the stack and then loads the appropriate interrupt vector the start address of the Interrupt Service Routine into the Program Counter 55 For the considered system the interrupt latency has a mean value of 6 2 us 78 Linear Drive for Material Handling HMO3524 HAMEL SRT O229 Instruments TB 5us CH1200Y DC 100MSa Parallel port output set at the beginning of the ISR and reset at the end of the ISR a Mak processing time CH1 2V2Py CH2 2 2Fy Figure 3 14 Measured interrupt latency interrupt jitter and processing time The variation of the interrupt latency is the interrupt jitter One cause for this variation is e g the Direct Memory Access DMA When a device is granted Direct Memory Access it will take control of the Memory Bus so if an interrupt request comes during a DMA transfer the CPU must wait until the DMA transfer is complete in order to regain control of the Memory Bus For reducing the interrupt jitter and also the data transfer and interrupt overhead on the PCI bus some un essential devices audio hardware USB serial port etc were deactivated from the BIOS For our Control PC an interrupt jitter of 2 us was measured see Figure 3 14 The maximum processin
139. ks the self holding contactor K1 must also be closed This is realised in two steps first the control software must close contact K2 located on the Supply Control Board which will be described later in this section It is not possible to supply the power electronics if the control program is not running In the second step the T2 On push button located on the front door of the cabinet must be pressed by the operator Pressing the T1 Off push button opens the contactor K1 and disconnects the supply of the Siemens devices Closing the K1 contactor supplies the electronics of the Siemens devices but not the DC link An extra step is needed for this the control software must also close the contact K3 located like K2 on the Supply Control Board This determines the closing of an internal contactor of the Infeed Module thus supplying the power sections of the Siemens devices 73 Linear Drive for Material Handling Pre Processing Sensor to Digital SAS GAO Id 04 U0D N N 6 0 PX GIN PX y CLN sallddns 9MOd4 CASH oq aa Ba We 6 0 2 A N ENS lt e N S2IUOJ329 3 PAO ZA IM by 6 vC 6AvCT o EIT EAO ETT EAG Past bO lt o N Mi A oj Ju ORNOUCUMUCUCU VAST lt al lt lt See 9 PIN EAST GGUUGBBOWONN 3 Jd ZA0 GA
140. l traversing speed 480 m min Vibration 55 to 2000 Hz lt 200 m s IEC 60 068 2 6 Shock 11 ms lt 500 m s IEC 60 068 2 27 Operating temperature 0 to 50 C Power supply 5 V 5 lt 150 mA Incremental signals 1 Vpp 40 um Table 2 1 Manufacturer specifications for LDIA 181 Source 26 generates two reference pulses one close to each end of the scale The zero position of the entire system is defined as the zero position of the leftmost read head assuming left to right positive direction The zero position of the leftmost read head is defined as the position where a zero crossing of the sine signal occurs whilst the cosine signal is positive and a reference pulse is generated by the rightmost sic part of the scale 2 2 3 Pre processing Unit Sensor to Digital The block diagram of the pre processing unit is shown in Figure 2 7 The differential signals from the five read heads are first converted to single ended ones Analogue sine and cosine signals S4 Ci Ss Cs are routed through two multiplexers MUX and MUXs to the inputs of four 12 bit analog to digital converters ADC S4 ADC C4 ADC Ss ADC G The outputs of the analog to digital converters are connected through the ADC Bus to the board s FPGA which also generates the control signals for the analog to digital 180 90 180 90 Ref Cosine Sine N Quadrant A142 A3 A4 A1J2A3A4 Perio
141. lues of Sg and Ca are at this point irrelevant Two clocks after the UPDATE SC REG signal was asserted leaving time for the sine cosine registers and for quadrant state machines to stabilize the synchronization state machine asserts R leading to the transition of the CHSM from state Ho into state Hai The complete Current Head State Machine is shown in Figure 2 11 In state Hi LD 000 thus CNTRA is enabled and the value 1 is stored in R USEA indicating valid position information from head A in this case 1 The multiplexers control signals remain unchanged At this point the content of all the position information registers sine cosine registers period counters multiplexer code registers and use registers is consistent indicating the new state of the system the vehicle is at the beginning of the region covered by of head 1 and its position can be correctly calculated based on the A values The B position information will be ignored until the scale covers a read head whose signals are routed through the B multiplexer The reference signal R4 or Rs if entering from right is used only to trigger the first conversion cycle and when this first conversion is finished the first transition in CHSM All the subsequent reference pulses are ignored until the vehicle is again outside the sensors region When the conversion timer started at the same time with the first conversion hits 2 us a new conversion
142. machine The arguments for this command are gt The segment number for which the parameters are to be set gt The parameters of the respective segment current PI controllers parameters EMF observers parameters etc eCMD S2D CORRS The correction tables for the optical sensors are sent to the Real Time Module as arguments of this command The command is also sent during initialisation once for each logical read head see sections 2 2 6 and 2 2 7 for the definition of the logical read heads and for the calculation of the correction tables The command s arguments are gt The number of the logical read head for which the corrections are sent gt The number of periods of the given read head for which the corrections were determined gt The amplitude and offset corrections of the sine and cosine signals for each period eCMD DC LINK POW Enable or disable the feeding of the DC Link e There is only one argument for this command namely the required state of the DC Link eCMD MACHINE ON OFF Starts or stops the control algorithm of the linear machine This command also has only one argument the required state of the control algorithm The control algorithm can be started from the User Interface only after the DC Link was powered up while the command for DC Link powered down can only be given after the stopping of the control algorithm eCMD CAPTURE Used to acquire data from the Real Time Module As a result of this command da
143. mentation of the Optical Sensors System 2u0224202000002nn0 nennen nnnn nenne 91 3 3 EMF Based Sensorless Speed Control eese 103 Sul EMF OVS en een ae 104 Contents 3 3 2 Mechanical Observer Zee 109 3 3 3 Implementation at the Straight Section of the Machine Rigid Vehicle 114 3 3 4 Implementation on the Entire Length of the Machine Articulated Vehicle 117 3 4 Sensor Sensorless Transitlorni cuc eer oni en a 119 3 4 1 Leaving the Processing SESIIOD nennen 120 3 4 2 Re entering the Processing Station u0222s00 20000000080 nennen nennen nennen 122 COMES OWNS ziesanaoaes een ee ee ee nee 129 21 F RS WOK nee 126 S ADDEN dT ee 127 5 1 Imaging Scanning Principle mans een 127 5 2 Dimensions of the Capacitive Sensor uunuueenssuenenuenennenennnnnnnnnennenennennnnn ernennen 128 5 9 Supply Control Board nennen ea 130 5 4 Real Time Module Structure sen nD aque aeta ptodedidcteeaniiebonfon einen 131 5 5 Mounting of the Optical Sensors at the Linear Drive for Material Handling 134 5 6 Measurement of the EMF constants uice unsre 135 See siS 19 q AET E E AE ee ee ee 137 Academic PHOTIC cacrscacossa saat ccecedacsenanacinsisidaaloatseisasantalonscignsesicladusnice aentmebadensandsioderaptdcuseeigagadesacn 143 3D A D ADC ADDR APM ASIC atan2 BIOS CHSM CNTR COMP CPLD DC DMA DRV DSP EMF EMI EMK ENA EPROM FEM FIFO FPG
144. munication FIFOs Again if an error occurs it will be recorded and the function terminates with a failure code At this step the read handler for the Command FIFO is also installed The initialisation then continues by starting the first analog to digital conversion of the DC Link voltage on the PCI DASO8 Interface Board and as a last step the interrupt generation on the Vehicle Control Interface Board is enabled The module is now loaded and prepared to respond to the hardware control interrupt and to commands from the User Interface Before the Real Time Module is unloaded the module clean up function will be called see Figure 3 17 for a block diagram of this function The first thing to do here is to disable the control interrupt generation then the Interrupt Service Routine can be unregistered After that the two FIFOs are destroyed and the relay outputs K2 and K3 are deactivated just as a precaution measure The module can now be removed from the kernel The Interrupt Service Routine calls in its turn three sub functions in this order e monitoring Implements the Monitoring State Machine discussed below emachine_control Here is the control algorithm for the linear machine implemented edata_capture If data is requested from the User Interface see CMD_CAPTURE then this function is responsible for writing it in the Data FIFO Module clean up entry point Disable the interrupt generation VCI Board
145. n Adjustable distance between lt the collision protection plate protection L and the load carrying plate on the vehicle Figure 5 8 Collision protection 134 Appendix 5 6 Measurement ofthe EMF constants In order to determine the EMF constants the following measurement was conducted for each segment of the linear machine with the segment disconnected from its power stack the vehicle was moved by hand inside the segment while the voltages induced in the three phases were acquired using a storage oscilloscope Figure 5 9 a shows the result of such a measurement the induced phase voltages a ep and e for stator segment 3 e V e V e IV e M ey M lel IV XI v m s k Vs m 0 50 100 150 200 250 Time ms Figure 5 9 Determination of the EMF constant a Measured EMF in the three phases of Segment 3 b The a components of the EMF and the EMF module c Position and speed d EMF constant 135 Appendix After filtering the high frequency noise the ea and eg components of the EMF were calculated using the Clarke transform 1 e 2e amp 6 1 ED Ni Bi The module of the EMF vector see Figure 5 9 b was then determined le fee e For calculating the EMF constant the speed information is also required First the position x was determined using the phase information contained in the measured EMF x E atana Sa T ep Then the speed v was
146. n Q next Figure Q amp 0 4 a 0 0 0 2 0 02 04 06 08 1 2 29 24 26 28 3 Time s Time s Figure 3 32 Measurement of the reconstructed position and of the corresponding speed 101 Linear Drive for Material Handling 820 sol a a ae Ga Ga es Ga bud arse Red 800 LH Upc ch Reb demie dede deg menle nenne mn nenn m am en ne 780b ZEE MM mop Mp 2 I 760 ER AHLEN nee eee TIME eee NERONI TL AESELB cones ores teens eee Speed mm s mob Allan MOTH Mv eee ee poesie VL E E S dne P20 edece eek en a e E Br EEE En 700 524 526 528 530 532 534 536 538 540 542 544 546 Time ms Figure 3 33 Zoom of the measured speed at the transition between heads 1 and 2 information every 2 us These 2 us time intervals are generated with the help of the board s 80 MHz oscillator On the other hand the position requests are generated by the control program at the beginning of the interrupt service routine which is triggered by a hardware interrupt request coming from the Vehicle Control Interface Board The hardware interrupt timing is based on the Control PC s PCI clock Because there is no synchronisation between the two clock sources a clock drift can occur so the time elapsed between the sampling of two successively read positions can have a variation of 2 us with respect to the 100 us control cycle This can be verified by analysing Figure 3 33
147. nd Electron vol 51 pp 821 826 Aug 2004 Seki K Watada M Torii S and Ebihara D Discontinuous arrangement of long stator linear synchronous motor for transportation system in Proc 1997 Internatio nal Conf on Power Electronics and Drive Systems vol 2 pp 697 702 May 1997 Suzuki K Kim Y J and Dohmeki H Proposal of the section change method of the stator block of the discontinuous stator permanent magnet type linear synchronous motor aimed at long distance transportation in Proc of 18 Interna tional Conference on Electrical Machines 2008 ICEM 2008 pp 1 6 Sept 2008 MTS Sensors Temposonics absolute ber hrungslose Positionssensoren R Serie Katalog http www mtssensor de fileadmin medien downloads datasheets rs erie katalog d pdf Link retrieved in June 2010 MTS Sensors Magnetostriktion physikalische Grundlagen http www mtssensor de fileadmin medien downloads mts messprinzip pdf Link retrieved in June 2010 Rettenmaier T Positionserfassung und Kommunikation zwischen zwei DSPs in modularen Servoantriebssystemen Diploma Thesis Nr 1349 Institut f r Strom richtertechnik und Antriebsregelung TU Darmstadt 2009 Texas Instruments The RS 485 Design Guide Application Report Feb 2008 http focus ti com lit an slla272b slla272b pdf Link retrieved in June 2010 PCI Special Interest Group PCI SIG PCI Local Bus Specification Revision 2 3 March 2002 http www pcisi
148. nsmitting Plate with transmitting electrodes in the shape of rectangular copper strips Four adjacent copper strips named a b c and d cover one pitch P of the sensor 2mm Each fourth strip is fed by one of the voltages Ua Ug All the electrodes j j a b c d from all the periods of the sensor are connected in parallel The transmitting electrodes and their connection are shown in Figure 2 29 2 A mobile Modulating Plate slider covered on one side with a sinusoidal copper area which is the modulating electrode On the other side is a copper plane the coupling electrode which is electrically connected with the modulating electrode The modulating plate contains 30 periods 60mm covering slightly more than 10 of the total length of the tested sensor This plate encodes the position dependent information in the electric field generated by the excitation voltages Ua Ua 3 A second stationary plate the Receiving Plate on the other side of the slider with one receiving electrode also an unstructured copper plane the charge received on this electrode contains the position information The real electric field distribution in the space between transmitting and receiving electrode may be calculated using 3D Finite Element Method FEM With a large number of 3D FEM calculations where the modulating electrode slider is being moved by a small distance between successive FEM calculations the output signal
149. nterface between the Sensor Bus and the Peripheral Component Interconnect PCI Bus 25 of a PC where the control code implemented in C language is running The second role of the PCI Interface Board is to galvanically isolate the PC from the rest of the position acquisition system through optocouplers The position calculation routine is integrated in the control program and will be called in every control cycle 100 us When a call occurs it will send a request via the PCI Interface Board to the pre processing unit which will send back the updated position information Based on this information the routine calculates a new position value and passes it to the control algorithm The detailed function of the component parts of the position acquisition system will be presented in the following sections 2 2 2 Sensors Used For the proposed system the linear optical encoder LIDA 181 26 from Heidenhain company depicted in Figure 2 5 will be used The encoder is based on imaging scanning principle for details see Appendix 5 1 and has a grating period pitch of 40 um The analogue sine and cosine signals with a cycle length equal with the pitch delivered by the sensor are differential in order to reduce the influence of noise and have a magnitude of 1 V peak to peak It must be noted that in order to function as specified the sensors require a rather small gap between scale and read head 0 75 mm with very small tolerance 150 um
150. ntrol for brush less DC motors using disturbance observers and adaptive velocity estimations IEEE Trans Ind Electron vol 45 no 2 pp 274 282 Apr 1998 Leidhold R and Mutschler P Speed sensorless control of a long stator linear synchronous motor arranged in multiple segments IEEE Trans Ind Electron vol 54 no 6 pp 3246 3254 Dec 2007 Kim J S and Sul S K New approach for the low speed operation of PMSM drives without rotational position sensors IEEE Trans Power Electron vol 11 no 3 pp 512 519 May 1996 De Angelo C Bossio G Solsona J Garcia G O and Valla M I A rotor position and speed observer for permanent magnet motors with nonsinusoidal EMF waveform IEEE Trans Ind Electron vol 52 pp 807 813 June 2005 Leidhold R and Mutschler P Speed sensorless control of a long stator linear synchronous motor arranged by multiple sections Proc of 31 Annual Conference of IEEE Industrial Electronics Society 2005 IECON 2005 Nov 2005 Bibliography Papers published in peer reviewed conference proceedings 63 64 65 66 67 68 69 70 71 Mihalachi M Leidhold R and Mutschler P Long Primary Linear Drive for Material Handling accepted for publication in the IEEE Industry Applications Magazine Mihalachi M Leidhold R and Mutschler P Motion Control for Long Primary Linear Drives used in Material Handling in Proc of the 14th
151. nvestigated For high dynamic drives the delay introduced into the control loop by position acquisition must be as short as possible Therefore the PLL like tracking filter 36 used in the commercial system will be avoided and a new instantaneous demodulation method will be proposed Figure 2 31 shows the four rectangular excitation voltages U Ug and the two line to line voltages They have a period of 15 us and 50 duty cycle The phase shift between two successive voltages is 90 In order to extract the position from Uo Eq 2 25 instantaneous demodulation will be used in each phase Ph1 Ph4 the output voltage is sampled by an analog to digital converter The four resulting samples are IER EUI x sen c Usas X K Upe osf Usut a X K Upc cos Phi Ph2 Ph3 EN ENS P m g Excitation U gt NN pem j Voltages 0 E 0 UN Line to line Uc n Voltages am Une Siem T Uout 1 Usut 2 Figure 2 31 Ideal square wave excitation voltages 53 Position Sensing Systems for Passive Vehicles The sine and cosine sums in the above equations can be rearranged yielding p a Usu2 X KUpe 42 cos x zi K Upe V2 sin x4 z AKUgc V2 cos Ex 5 Constant K Upc V2 sin Eq 2 27 From the sampled output voltages of two consecutive phases position values can be calculated using the four quadrant inverse tangent function atan2 P X12 5 atan2 Ue Ua
152. o phase digital filtering MATLAB online function refere nce http www mathworks com access helpdesk help toolbox signal filtfilt html Link retrieved in June 2010 Benavides R and Mutschler P Controlling a System of Linear Drives in IEEE 36 Power Electronics Specialists Conference 2005 IEEE PESC 2005 pp 1587 1593 June 2005 Canders W R Hoffmann J Maurus Q and Mosebach H Modular linear drive system with soft magnetic composite elements for 3D tracks in 7 International Symposium on Linear Drives for Industry Applications LDIA 09 2009 Siemens AG Simodrive Umrichter Projektierungsanleitung Manufacturer Service Documentation Aug 2002 Benavides R Entwurf und Realisierung eines RS 485 basierten Multimasterbus systems Master Thesis Nr 1307 Institut f r Stromrichtertechnik und Antriebs regelung TU Darmstadt 2003 139 Bibliography 46 47 48 49 50 51 52 53 54 95 56 57 58 59 60 61 62 140 Measurement Computing Corporation PCI DASO8 Analog input and Digital I O User s Guide http www mccdaq com PDFs manuals pci das08 pdf Link retrieved in June 2010 Vector Linux 5 8 documentation http vectorlinux osuosl org docs vl58 manuals index html Link retrieved in June 2010 Linux Kernel Documentation Index http kernel org doc Link retrieved in June 2010 Kroah Hartman G Linux Kernel
153. oduced by position sensors will be avoided by using sensorless motion control There are several sensorless methods for PMSM that are found in literature They can be divided in two main classes O AON Introduction e Methods based on the evaluation of the Electromotive Force EMF or of the mover s flux 10 11 12 these methods are relatively robust but have the disadvantage that they lose performance at low speed and do not work at standstill i e they can be used for sensorless travelling at speeds higher than a minimal value but they do not allow sensorless positioning e Methods based on the evaluation of the machine s anisotropies 13 14 15 usually by means of injecting a test signal These methods are suited for standstill and low speed range When compared to the rotating PMSM the segmented long primary linear ones present additional challenges with regard to the sensorless acquisition of the mechanical quantities e The translator vehicle of a long primary linear machine covers only partially the stator segments which makes the extraction of the mechanical quantities from the measurable electrical ones more difficult e The mechanical quantities must also be acquired at the transitions between stator segments All sensorless methods work incrementally i e the absolute position cannot be determined without the knowledge of an absolute start position In the particular case of the discussed linear
154. on Sensing Systems for Passive Vehicles U U generation rP 1 Dif Input Active Filter IC Bm a mpliiTier mpliiTier n U j d Gain Errors 2 4 order low pass Correction m EPE E E m u u EE E E E a EE m nr u E GP TIMER 3 a a a a Tee U U generation identical with the U U generation GP TIMER 1 a a um a ee ee ee ee Se ee ee a ae eee ee eee ae ae a ea r 4 TMS320F2812 DSP JOSUSS oAnmede5 GP TIMER 2 Analog Diff Receiver Diff Driver Amplifier Figure 2 42 Proof of concept implementation of the sinusoidal voltages generation Usu In reality such a phase shift occurs biasing the absolute position of the capacitive sensor Eq 2 33 can be rewritten Usu X Ani cos t Xp c Eq 2 34 The unknown phase qc introduced by the filter and the amplifiers can be determined during initialisation when the point of reference is crossed Figure 2 43 shows in blue line the difference between the reference sensor s position and the capacitive sensor s position measured position without correction A correction table was generated without the use of a reference sensor by an off line zero phase filtering process using Matlab function filtfilt 41 Only a correction for the average values was used as there is not enough storage for a period individual correction table in the evaluation board The result with this simple correction
155. on reconstruction state machine and ki of the calculation of the position x Initially the vehicle is outside the sensors region This is signalled from Sensor to Digital by USE 0 and USEs 0 The reconstruction state machine is in state H o and the position x cannot be calculated At time to Figure 2 17 the Sensor to Digital board sends for the first time USE 1 This means that valid position information from head 1 in this case A is available Firstly based on Sa Ca and Na the local position given by head A is calculated Xa ta P N en t Eq 2 1 where P is the pitch of the sensor and Xinc a to is the incremental position of head 1 as calculated based on Sa to and Ca to using the arctangent function p Xinc A to T 2n atan2 S t5 6 to Eq 2 2 The atan2 function returns a value between r and x In 3 and 4 quadrants Sa 0 its result must be corrected by adding 2x or equivalently by adding P at the incremental position if S ty lt 0 Xinc A to Xinc A to P The variable Xo will be used to store the offset of the current read head with respect to the zero position of the system Because zero position of the system was defined as the zero position of the head 1 its offset will be zero Eq 2 3 38 Position Sensing Systems for Passive Vehicles Initialisation gt H Vehicle outside sensors region USE 1 __ USE 0 Hj Scale covers head 1 USE 1 USE 1 amp
156. onary a reference frame Components of the estimated current dependent flux vector of a stator segment in the stationary a reference frame Components of the estimation error of the current dependent flux vector of a stator segment in the stationary a reference frame Ypma Ypmg Components of the flux linkage generated by the permanent magnets in the stationary a reference frame Electrical angular speed 1 Introduction Linear electrical motors are presently gaining an increasingly widespread use in application fields like transportation 1 2 and industry machining actuators etc Linear motors have the capability to produce thrust directly thus avoiding the use of rotative to linear transmission elements like belts and pulleys racks and pinions or screw systems which are associated with the rotative motors whenever they must achieve linear motion At present linear drives are used in industry preferably for precise motion with no backlash high accuracy of positioning and high dynamics acceleration on a straight track e g in advanced machine tools In industrial production plants materials must be worked upon in processing stations e g for milling drilling grinding etc and transported between these stations Using just one system of linear drives for processing as well as for transportation will result in benefits like high precision high dynamics high productivity and reduced wear and maint
157. osition Sensing Systems for Passive Vehicles digital to analog converters For simplicity the components not used in the evaluation are not shown in Figure 2 32 The four excitation voltages for the capacitive sensor are generated by two dual MOSFET drivers one driver generates two complementary voltages The two control signals for the drivers are generated by the Reference Position Acquisition Board s CPLD and they are synchronised with the acquisition of the reference position from the optical sensor The charge amplifier is located on a small PCB which is mounted directly on the receiving plate of the capacitive sensor in order to keep parasitics and susceptibility to noise as small as possible For the charge amplifier a low noise low input current Bi Mosfet operational amplifier 37 was chosen 38 The output of the charge amplifier is transferred differentially to the Capacitive Sensor Acquisition Board also to reduce the influence of the external disturbances Here the signal is sampled by the board s analog to digital converter four times in each cycle of the excitation voltages once in each phase Ph1 Ph4 see Figure 2 31 Synchronisation between the generation of the excitation voltages and the conversion start signal of the analog to digital converter occurs via the IRQ11 signal on the ISA Bus This signal is asserted by the Reference Position Acquisition Board s CPLD every 100 us When this signal is active
158. ossible number of controllers An overview of the power electronics and control topology is given in Figure 3 1 There an Inverter Bus provides communication between each Vehicle Controller VC and each Power Stack PS within every 100 us control cycle All power stacks are equipped with a CPLD based Power Stack Interface PSI and each vehicle controller is attached via an FPGA based Vehicle Controller Interface VCI to the Inverter Bus Access to the Inverter Bus is controlled in a stringent time slot regime by an Inverter Bus Master All actions within the controllers and all power stacks including switching of the IGBTs and sampling of the fundamental component of the stator currents are strictly synchronized to the timing generated by the Bus Master Vehicles V Stator Segments SS V2 X N Ya v3 V4 l N v1 i A PS PS d PSL1 J PSI2 PSI3 P PSI n 3 PSI n 2 PSIn 1 PSIn Power Inverter Bus Stacks PS rN N Power Stack Interface VCI Vehicle Controller Interface Inverter Bus a Modulation information from controller to power stack b Actual current values from power stack to controller Located e g in one cubicle Industrial PC IPC Industrial PC IPC Phy according to IEEE 802 3 Industrial PC IPC Figure 3 1 Overview of the system architecture 65 Linear Drive for Material Handling Physically t
159. overed by each read head In Figure 2 20 a it looks as there is a jump in the position deviation calculated without corrections the blue waveform at the transition between read heads 3 and 4 but this is not the case as Figure 2 20 b shows Figures b and c are zooms of Figure 2 20 a at the transitions between heads 2 3 and 3 4 respectively In these zooms it can be seen that the position deviation is continuous meaning that no jump occurs at the transition between read heads in the position calculated based on the Sensor to Digital data Two components of the position deviation can be identified in Figure 2 20 the first Perc Position deviation over the entire measuring length og corrections Nemo corrections Period corrections Position deviation um 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 Reference position m b Transition 3 4 C Transition 4 5 E E c c e o9 E s 2 2 9 D o o c c g 9 O O A a Reference position m Reference position m Figure 2 20 Position deviation 44 Position Sensing Systems for Passive Vehicles Without corrections Mean corrections Period corrections From reference sensor 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 Reference position m Figure 2 21 Measured speed without and with position correction one has the same wavelength as the pitch of the sensors 40 um it is due to systematic errors and can be reduced by
160. p 1 8 5 9 Oct 2008 Mihalachi M and Mutschler P Evaluation of Two Position Acquisition Systems for Permanent Magnet Linear Motors with Passive Movers in Proc of the 11th International Conference on Optimization of Electrical and Electronic Equipment OPTIM 2008 pp 37 44 22 24 May 2008 141 Academic Profile Marius Alexandru Mihalachi Born in C mpina Romania on August 29 1981 2005 2010 1999 2005 1995 1999 Working as assistant at the Department of Power Electronics and Control of Drives Technische Universitat Darmstadt Germany Study of Electrical Engineering and Computer Science at Transilvania University of Brasov Romania with speciali sation in Automatics and Industrial Informatics Technical High School Energetic C mpina Romania with specialisation in Informatics 143
161. q 3 55 E TU Up B M p p ps G Kg v 2 tp x From Eq 3 55 it can be seen that the eigenvalues of the observer will move as the speed changes if fixed values are considered for the gains The gains of the observer will be calculated in such a manner as to achieve a desired dynamic for a certain minimal speed vo and then the locus of the observer s poles will be analysed for the entire speed range Let us assume that the dynamic of a 3 order Butterworth filter with cut off frequency at 20 Hz is desired for a minimal speed vo 0 5 m s Following values result for the mechanical observer s poles p 125 66 P gt 62 83 j 108 83 Equi Introducing these values in Eq 3 55 together with the vehicle s mass M 13 2 Kg and the pole pitch tp 24 mm and considering the EMF constant of stator segment 1 Ke 17 72 Vs m and a friction coefficient B 50 Kg s the following gains result at Vo 0 5 m s G 23 542 10 G 27 543 Eq 3 57 G 0 2225 Figure 3 37 shows the root locus of the obtained mechanical observer over the entire speed range v 0 10 m s It can be seen that for speeds higher than vo the dynamic of the dominant poles p and ps changes relatively little With increasing speed the pole p4 moves further away from the origin of the s plane and for speeds close to the maximal value of 10 m s it approaches the poles of the EMF 112 Linear Drive for Material Handling Iv
162. raction Layer HAL was applied to the Linux kernel source code and the kernel was re compiled In this process certain features which might have interfered with the real time control were also disabled e g APM Advanced Power Management or CPU frequency scaling Then RTAI 3 5 was installed After installation the RTAI components can be loaded and unloaded using the normal Linux mechanism insmod for inserting a module into kernel rmmod for removing a module from kernel etc so it is possible to turn on and off all or some of the real time capabilities of the operating system as required The control of the linear machine is implemented in the Real Time Module Figure 3 12 which is running in the kernel space Because in this module very fast response times are necessary the use of some Linux system functions e g keyboard input display output hard disk access must be avoided due to the delays they can introduce A second application program the User Interface will be used to manage these functions user input data saving on the hard disk etc This application has no real time requirements so it can run in the user space like any normal Linux application The User Interface program a screenshot of which is presented in Figure 3 13 was generated with the help of Glade 2 12 1 53 an interface builder based on the GTK 54 toolkit The communication between the Real Time Module and the User Interface is accompl
163. ratings move in relation to each other the incident light is modulated if the gaps are aligned light passes througn If the lines of one grating coincide with the gaps of the other no light passes through Photocells convert these variations in light intensity into electrical signals The specially structured grating of the scanning reticle filters the light current to generate nearly sinusoidal output signals The smaller the period of the grating structure is the closer and more tightly toleranced the gap must be between the scanning reticle and scale Practical mounting tolerances for encoders with the imaging scanning principle are achieved with grating periods of 10 um and larger Source 26 Structured Window sensor Scale aie Index grating Condenser lens Scanning reticle Light source LED Figure 5 1 Imaging scanning principle Source 26 127 Appendix 5 2 Dimensions of the Capacitive Sensor Geometrical dimensions e Pitch of the sensor P 2mm e Width of the transmitting electrodes w 0 35mm 70 of P 4 e Height of the upper lower sinusoidal copper pattern A 10mm e Height of middle rectangular copper pattern h 2mm e Total length of the sensor L 560mm e Total number of periods pitches N L P 280 Figure 5 2 e Length of the slider mobile plate Dimensions of the 60mm position dependent overlapping areas e Number of periods on the slider Nmp Lm P
164. rmination The articulated vehicle is less stable in the z axis than the rigid one mainly due to the spring mounting of the outer guiding rollers In order to increase the guiding stiffness along the measuring axis the collision protection system shown in Figure 3 22 and in Figure 5 8 was converted into a guiding system for the load carrying plate as sketched in Figure 3 25 The load carrying plate now travels on the guiding rollers the height of which was set about one millimetre higher than the normal level of the load carrying plate assuring that the contact is always maintained The guiding rollers are spaced so that at every point inside the sensor s region the load carrying plate is guided by at least two of them thus maintaining the correct orientation of the scale with respect to the optical read heads Figure 3 26 shows the underside of the load carrying plate with the guideway for the rollers The ends of this guideway are angled upwards to allow smooth entering into and exiting from the guided section 93 Linear Drive for Material Handling Figure 3 26 Underside of the load carrying plate 94 Linear Drive for Material Handling Figure 3 27 shows measurements of the sine cosine and reference signals at different positions in the sensor s region after the adjustment of the three read heads The scale was mounted on the rigid vehicle without the guiding rollers mounting as in Figure 3 22 The sine cosine s
165. roller sampling time Components of the voltage vector of a stator segment in the stationary a reference frame Components of the reference voltage vector of a stator segment in the stationary a p reference frame Speed of the vehicle Reference value of the vehicle s speed Estimated speed Speed estimation error Speed value used as feedback in the control algorithm Speed provided by the optical sensors system Position of the vehicle Reference value of the vehicle s position Estimated position Position estimation error Begin position of a stator segment Position value used as feedback in the control algorithm End position of a stator segment Position provided by the optical sensors system EMF observer gains ratio AXc AXy x Eya Gps 0 6 Up Meroen LIT Tig Tia Pop Was Y Symbols Difference between the position given by the optical sensors and the estimated position Length of the vehicle Correction term of the mechanical observer Difference between the measured and the estimated current dependent flux vectors components in the stationary a reference frame correction terms of the EMF observer Electrical angle of the vehicle Estimated electrical angle of the vehicle Pole pitch of the linear machine Components of the total flux linkage vector of a stator segment in the stationary a p reference frame Components of the current dependent flux vector of a stator segment in the stati
166. ronisation function includes the two headers below Position and speed given by the optical sensors system position h include parameters h include s2d h include slsemf h Position and speed used in control Synchronisation function between optical sensors and observed position and speed EMF observer definition general EMF observer update function Definition of the two EMF observers for the two controlled stator segments m and n Mechanical observer definition general Mechanical observer update function Mechanical observer definition static double x c static double v c void synchronisation double x s2d double v s2d double x emf double v emf double x c double v c include parameters h typedef struct double psi alpha psi beta double emf alpha emf beta void update void t emfobs void emfobs update t emfobs h static t_emfobs emfobs m void emfobs update static t_emfobs emfobs n void emfobs update o typedef struct double M B double GF Gv G x double eps x FL v x void update void t mecobs void mecobs update t mecobs h Ir static t_mecobs mecobs void mecobs update Fi DE DE POS BRD RESET 0x00 POS BRD DO 0x04 POS BRD STAT 0x20 POS BRD D1D2 0x24 POS BRD D3D4 0x28 POS BRD D5D6 0x2C S2D PITCH 40e 6 S2D OFFS 3B 0 500135356
167. rovided by the optical read heads will further increase the position resolution Active Vehicle Moving Windings Passive Track Stationary Magnets Power and Information Transfer by Drag Chain Stationary Scale Moving Read Head Attached to Track Attached to Vehicle Figure 2 1 Position sensing with optical encoder and active vehicle Moving Scale Passive Vehicle Stationary Read Heads Attached to Vehicle Moving Magnets Attached to Track Active Track Stationary Windings Figure 2 2 Position sensing with optical encoder and passive vehicle Passive Vehicle Moving Magnets d Active Track Receiving Electrode y Stationary Windings Attached to Track y gs Transmitting Electrode Modulating Electrode Attached to Track Attached to Vehicle Figure 2 3 Position sensing with capacitive encoder and passive vehicle 21 Position Sensing Systems for Passive Vehicles nanometers range The capacitive sensor schematically depicted in Figure 2 3 is a low cost low complexity alternative to the optical system from Figure 2 2 The passive slider attached to the vehicle modulates position dependent an electrical field produced between the stationary transmitting and receiving electrode the position can then be extracted by demodulation Besides optical and capacitive sensors there exist some more physical sensing principles which may be interesting for the discussed application e g 2
168. rrupt Line Number of VCI Board Create Command FIFO Free the Interrupt Line uninstall the previous interrupt handler abessaw 1011 OUIOY e pJooas Install isr void as the new interrupt handler Install the function get ui command void as read handler for the Command FIFO Look for the VID and DID of the Position Interface Board Create Data FIFO Position Interf Record a kernel error message Read Base Address Re gister 0 of the Position Interface Board abessaw 1011 JguJa e pJooa Start the first A D conv on PCI DAS08 board Look for the VID and exit Failure DID of the PCI DASO8 Interface Board Enable the interrupt generation VCI Board RT Module succesfully initialised Figure 3 16 Real Time Module initialisation 83 Linear Drive for Material Handling The Base Addresses of the PCI devices are not hardwired as it is the case e g with the Industry Standard Architecture ISA Bus but allocated by BIOS at start up They also have to be read from the PCI structures associated with the three PCI interface boards Additionally the interrupt number corresponding to the PCI slot where the Vehicle Control Interface Board is located must also be determined in order to be able to install the Interrupt Service Routine When all the PCI boards were located and initialised the module initialisation function will proceed with the creation of the two com
169. s Spacing between modulating and receiving plates Thickness of each of the three plates of the capacitive sensor Position dependent overlapping area between the transmitting electrodes a and the modulating electrode Frequency of the sinusoidal excitation voltages Position dependent overlapping area between the transmitting 1 LJ electrodes a and the coupling electrode Height of middle rectangular copper pattern on the slider Number of sensors periods covered by the slider modulating plate Total number of periods of the capacitive sensor Period pitch of the capacitive sensor Quality factor Charge amplifier feedback resistance Laplace operator Excitation voltages of the capacitive sensor Line to line excitation voltages DC Link Voltage Output of the charge amplifier Unc Uas Uac Uasr Uo Ww X Xp 0 3 Qc Ch Symbols Outputs of the two MOSFET H Bridges used to generate the sinusoidal excitation voltages Filtered outputs of the two MOSFET H Bridges Sum of the four excitation voltages Width of the transmitting electrodes Position given by the capacitive sensor Normalised position of the capacitive sensor Absolute permittivity of vacuum Relative permittivity of epoxy FR4 Phase shift of the normalised position Pulsation of the sinusoidal excitation voltages Linear Drive for Material Handling Section 3 gt o O9 WwW W lt State Matrix State Space Representation Input M
170. s checked whether the vehicle is in a region where sensorless control is necessary If yes updated values of the observed position and speed are calculated The EMF based sensorless control is described in detail in section 3 3 Then the synchronisation function between the sensor s position and speed and the observed ones is called The transition between sensors equipped and sensorless regions is described in section 3 4 As result of this function the values of the position and speed to be used in control are obtained and they will be given as feedback values to the position and speed controllers from which the reference value of the current results In the following depending on which stator segment s are covered by the vehicle the corresponding current controllers are calculated First the electrical angle of the vehicle associated to each of the controlled stator segments is calculated based on vehicle s position Because of the gaps in the machine s winding see sections 3 1 1 and 3 3 different values of the electrical angle can result for consecutive stator segments The electrical angles are then used for field orienting the d and q components of the currents in the vehicle attached reference system are calculated and given as feedback values to the current controllers Figure 3 21 presents a vectorial block diagram of a discrete PI controller with anti windup general case The parameters of the PI current controllers
171. s from the read heads are brought to a pre processing unit Sensor to Digital lt Scde Mu Position Calculation Routine C Code A IM Signal Conditioning PCI Bus Multiplexers Comparators FPGA i i A D Converters gt FPGA t Optocouplers t RS 485 Interface RS 485 Interface Sensor to Digital Board Sensor Bus Figure 2 4 Block diagram of the optical position acquisition system I Interface Board PC 23 Position Sensing Systems for Passive Vehicles Board The analogue sine and cosine signals from two neighbouring heads are fed via multiplexers into four analog to digital converters where they are digitised and sent further to an Field Programmable Gate Array FPGA The FPGA calculates the coarse position based on the sign bits of sine and cosine It also keeps track of the current read head and decides based on the coarse position of the current read head when to start evaluating the signals from the next read head The reference signals from the first and last read heads are also necessary to detect when the vehicle enters the sensors covered region They are fed through comparators to the Sensor to Digital Board s FPGA The pre processing unit is connected through the Sensor Bus to a PCI Interface Board The Sensor Bus is a 16 bit parallel bus implementing the RS 485 differential signalling standard 24 As its name suggests the main purpose of PCI Interface Board is to provide an i
172. s position estimation in the transport sections of the track and the synchronisation between the position given by sensors and the estimated one Finally in Chapter 4 the conclusions are drawn and future work regarding the expansion of the functionality of the linear drive for material handling presented in this work is outlined 20 2 Position Sensing Systems for Passive Vehicles 2 1 Overview The linear drive for material handling presented in this work consists of processing stations connected by transport sections Within the processing stations highly precise positioning is necessary thus the position of the vehicles must be acquired using position sensors The measured position is used as feedback value of a high precision position control loop while its approximate numerical derivative is used by the subordinated speed control loop Optical sensors are the industry standard for applications where high precision is required In typical applications e g machining the passive scale of the sensors is mounted on the stationary side track while the active read head is on the vehicle see Figure 2 1 For passive vehicles the passive scale must be attached to the vehicle and several active read heads will be mounted along the track as in Figure 2 2 Incremental optical scales are commonly used with a narrow grading in the micrometers up to tens of micrometers range Fine interpolation of sine cosine signals usually p
173. sation between the estimated and the sensed speed is completed a speed close to zero would lead to instability of the observers structure At the time tsync2 AT the position used in control xc equals the sensed position xs and the synchronisation is completed The observers can now be disabled the sensed position and speed will be used in control 123 4 Conclusions In this work aspects concerning the position acquisition and control for linear drives with passive vehicles were investigated For industrial material handling a combined transportation and processing system based on a segmented permanent magnet linear drive is developed In order to attain high accelerations passive lightweight vehicles are used The track of the drive contains processing stations which must be equipped with position sensors for high accuracy of positioning and high dynamics Passive vehicles present a restriction for the position acquisition system through the fact that neither energy nor information must be transmitted to the moving parts In Chapter 2 the realisation and testing of two position sensing systems which comply with this requirement were examined the first one is based on high resolution optical sensors while the second one uses a comparatively lower resolution capacitive sensor For the ca pacitive sensor two evaluation methods were tested the first one uses square wave excitation and instantaneous demodulation and the second
174. sertation a system and control method is proposed for a multi segment linear drive with totally passive vehicles which allows for gaps between consecutive segments sensor based operation in some sections and sensorless based operation in others The main contributions are eA novel method for measuring the vehicle position with high accuracy and high dynamics using optical encoders which method is suitable for use with passive vehicles e Two new and improved evaluation methods for a commercially available capacitive sensor that has a relatively lower resolution This capacitive sensor could be used in certain applications as a less complex and more cost effective alternative to the optical sensors e A method for handling the gaps and distinctive parameters among stator segments 19 Introduction eA procedure that allows the smooth handover transition of the vehicles between the sensor equipped and the sensorless sections of the track This thesis is composed of two main sections In Chapter 2 the two position sen sing systems for passive vehicles based on optical and capacitive sensors respectively are presented and their performance is evaluated In Chapter 3 the position acquisition and the control of the linear drive for material handling currently under development at our department is discussed including the overall system architecture the sensor based position acquisition inside the processing stations the sensorles
175. sing stations In this sections sensorless control based on the evaluation of the electromotive force EMF is implemented The distinctive parameters of the different stator segments are taken into consideration Due to mechanical constraints there are gaps in the winding arrangement between consecutive segments of the machine which means that the EMF vectors of two consecu tive segments can have an arbitrary phase difference providing additional challenges especially for the sensorless control At the transition between processing stations and transport sections a synchroni sation procedure between the measured position and the estimated one is described and experimentally evaluated Kurzfassung Ein Langstator Linearsynchronantrieb mit leichten passiven Fahrzeugen welcher sowohl f r die Bearbeitung als auch f r den Transport von Materialien in industriellen Produktionsanlagen dienen kann wird entworfen und experimentell gepr ft Diese Arbeit konzentriert sich auf die Positionserfassung und Bewegungssteuerung des vorgeschlage nen Systems Um eine hohe Unabh ngigkeit der Bewegung mehrerer Fahrzeuge zu erm glichen wird der St nder Prim rteil der Linearmaschine in zahlreiche Speiseabschnitte Segmen te unterteilt Jedes Segment wird von einem zugeordneten Wechselrichter gespeist Die Kontrollinformation wird zwischen allen Wechselrichter und allen Fahrzeugkontroller durch einen Wechselrichterbus ausgetauscht Mehrere Bear
176. sisting of stator segments 1 and 2 The rigid vehicle was used in this implementation With this vehicle it was possible to set the air gap at ca 2 5 mm smaller than the 3 mm air gap set for the articulated vehicle so a slightly higher EMF constant was attained in this implementation about 20 Vs m Figure 3 39 shows an offline measurement of the phase a component of the EMF at the transition between stator segments 1 and 2 The induced voltages in the two segments are in phase so the correction term of the mechanical observer will be Figure 3 38 Block diagram of the mechanical observer 114 Linear Drive for Material Handling EMF V 0 2 0 1 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 Time s Figure 3 39 Offline measurement of the EMF at the transition between stator segments 1 and 2 Rigid Vehicle calculated as described in Eq 3 58 m 1 n 2 In Figure 3 39 it can also be seen that there is no noticeable drop in the sum of the two EMF components which was to be expected as there is no gap between the two segments Figure 3 40 shows a measurement in which the observed position and speed were used for field orienting as well as for feedback to the position and speed controllers The sensed position and speed were used for starting from standstill at the beginning of the control algorithm not shown in Figure 3 40 see section 3 4 for a description of the synchronisation procedure between the sensed and the e
177. stimated position and speed after that the sensed position and speed were only used for evaluating the estimation errors In this measurement position control was used with the reference position step changing between 5 and 595 mm close to the limits of the sensors region the speed was limited to 2 m s and the force producing currents igi and ig were limited at 10 A about the half of the maximal current In order to reverse the direction of motion taking into account that positioning is not possible under EMF based sensorless control the following procedure was used the proportional gain of the position controller was set high enough so that the controlled position would overshoot the reference position This means that when the vehicle reaches the reference position its speed is decreasing but it still has a high enough value for the mechanical observer to remain stable When the reference position is first reached by the controlled position a step change occurs in the reference position determining the vehicle to accelerate in the opposite direction It can be seen in Figure 3 40 a and b that when the estimated speed is close to zero the estimation errors increase The maximal position error is about 4mm corresponding to an error in the electrical angle of ca 30 and to a reduction of ca 15 in the electrical force The reduced electrical force is however still high enough for the vehicle to accelerate in the opposit
178. ta will be sent from the Real Time Module to the User Interface using the Data FIFO discussed below There are two arguments for this command the number of points to be acquired and the decimation of the data points The capture command is used for two purposes gt To periodically acquire data from the Real Time Module using a 200 ms soft ware Update Timer implemented in the User Interface in order to update the 81 Linear Drive for Material Handling status and variables displays of the interface In this case the command will be sent with the arguments 1 for the number of points and 1 for decimation gt To make a longer user initiated data capture using the Data Capture section of the User Interface see also Figure 3 13 In this case the user can choose the number of data points up to 50 000 and the decimation between 1 and 10 The length of the capture buffer was chosen by experience and it proved sufficient in most cases but if necessary the amount of captured data can be increased the practical upper limit is only the amount of available memory of the system During a user initiated data capture the periodical data request command is suppressed The second FIFO used in the communication is the Data FIFO The Real Time Module writes in this FIFO as a result of a capture command from the User Interface see above The User Interface is receiving the data on a read handler function similar to the one in t
179. tal Board via the Position Interface Board Because the communication protocol on the Sensor Bus takes about 3 5 us see section 2 2 4 some other tasks can be completed before the position information becomes available Next the state of the analog to digital conversion on the PCI DASO8 board is 86 Linear Drive for Material Handling checked the conversion was started in the previous interrupt or in the case of the first interrupt by the module s initialisation function If the conversion is finished then the value of the DC Link voltage is read If not a communication error flag is set based on which the addressing of the Power Stack Interfaces will be disabled This flag is also sent to the User Interface which will notify the user about the encountered error After reading the DC Link voltage the values of the measured currents for the two controlled stator segments are read from the Vehicle Control Interface Board If the control Control function entry point Is the S2D position information ready S2D positon request Write at ADDR D Read the PCI DAS08 A D conversion state Is the PCI DAS08 A D conversion ready Monitoring state M4 active No communication error and machine ON from UI Read the DC link voltage Set the DASO8 com munication error flag Y Set m 0 and n 0 so that no inverter is being addressed Update the reference
180. tator attached one a Then the modulation information switching states and times is determined 45 and written to Vehicle Control Interface Board from where it will be sent via the Inverter Bus to the corresponding Power Stack Interfaces Eq 3 5 As a last step a new analog to digital conversion of the DC Link voltage is started on the PCI DASOS8 board this concluding the control algorithm for a given sampling instant 90 Linear Drive for Material Handling 3 2 Implementation of the Optical Sensors System In this section the implementation of the optical sensors system presented in section 2 2 of this work at the linear drive for material handling see the description of the linear machine in section 3 1 1 will be discussed First the mechanical mounting of the optical sensors will be presented and then measurements concerning the quality of the sine cosine signals the correction of these signals and the reconstructed position will be shown Because on the linear drive for material handling there is no reference position sensor available only the quality of the sine cosine signals will be used as a measure of the quality of the resulting position signal Figure 3 22 shows the mechanical mounting of the optical read heads and scale at the double sided section of the linear machine with rigid vehicle The read heads are mounted on an L shaped profile attached to the machine s carrying guiding tube The measuring
181. tb t c double ita 1 b 1 6 double i alpha i beta double i d iq double u alpha u beta double u d u q double theta seg m seg_n void current_cond Vehicle Interface Board addresses definitions Definitions of the current control vari ables for the two simultaneously controlled stator segments m n Currents condi oy tioning function void modulator struct segment seg Modulator double u_dc function PI controller parameters and states definitions general PI controller update function definition general static t pic pi seg m id typedef struct double kp ti e u double update void double double Rupie double pic update t pic h double e double ymax Definition of the two ge Une current controllers of stator segment m static t pic pi seg m ig void pic update static t_pic pi seg n id Definition of the two voidt pic update current controllers of stator segment n static t pic pi seg n iq void pic update static t_pic pi velocity Definition of the speed PI controller ond pe caup date E Axis transformations functions PI controllers currents speed Calculation of the switching states and times SV PWM D parameters h N define PI 3 141592654 define SQRT3 1 732050808 define TS 1008 6 define TAUP 24e 3 static struct double k lem double i max double ia ibO
182. three observers structure is disabled until the vehicle gains some speed time t in Figure 3 44 After being enabled the observers need the time interval t t ca 100 ms to converge In the interval to t4 valid position and speed information is available from the optical sensors as well as from the observers so the estimation errors can be analysed Position m Pos Error mod To mm Electrical Angle Error deg Estimated Speed m s Speed Error m s 0 01 02 03 04 05 06 07 Time s Figure 3 44 Measurement of the position and speed when leaving the processing station From top Sensed and estimated position Position electrical angle estimation error Sensed and estimated speed Speed estimation error 121 Linear Drive for Material Handling Before synchronisation which occurs at time t3 tsync 1 according to Eq 3 63 the mean position estimation error amounts to 96 mm i e 4 tp due to the fact that the observers were not started close to the zero of the sensed position Starting with time t3 the estimated position and speed are used as feedback in control In Figure 3 44 the position estimation error is represented modulo 2tp so the synchronisation at time t3 does not affect the error curve After the estimated position converges the estimation error remains less than 1 mm corresponding to an orientation error of 7 5 electrical degrees The speed
183. tion Sensing Systems for Passive Vehicles ccccceecceceeeeeeeeeeeeeeeeeeeeeeeseaeeeees 21 uo E 21 2 2 Optical Sensors System usa ana 29 2 2 1 General Description of the System sees 23 2 22 SENSOR SO PREMO REC occ 24 2 2 3 Pre processing Unit Sensor to Digital sees 25 2 2 4 Sensor BUS Communication an ae 35 2 2 5 PCI Interface BOON nu ns TM 36 2 2 6 Position Reconstruction In C Code 38 2 2 7 Correction of Sine Cosine Signals u0220200000000000nen nennen nnnnnn nennen nennen 41 2 2 8 Experimental Results 2202220002000020000nnn eo nnnnnnnn nnne 43 2 2 9 Implementation of the Optical System seseeeeeeeeereenn 45 2 3 Capacitive Sensor 46 VANS NEM mil Tee NETTE 47 2 3 2 Electrical Model of the Capacitive Sensor s222002020000000nnnn none nennen 48 2 3 3 Square wave excitation ueenseessssenesnennnnnnnnnnnennennnnnnnnnn nene nnns 53 2 3 4 Sinusoidal excitation ee ee een 59 3 Linear Drive for Material Handling 002220000220000000800Rnnnnn nne 65 3T System UGC le een 65 3 1 1 Linear Machine ee ee pen een enter 67 3 1 2 Power Electronics and Control eee 72 RA is ABIDE ee ee een ne en ne ie GELBE EIN eee ee ee 72 3 1 2 2 Control SONWA E T 76 3 2 Imple
184. tion protocol egenerates the 6 control signals for the IGBT drivers based on the values of the switching times received through the Inverter Bus from the Vehicle Controller e digitises the three current signals from the external LEM sensors and sends them back to the Vehicle Controller emonitors the two internally measured currents of the Power Stack over current protection emonitors the temperature signal available from the Power Stack over temperature protection All 18 Power Stack Interfaces are connected through the Inverter Bus to the Vehicle Control Interface Board located in a PCI slot of the Control PC The Vehicle Control Interface Board and the Position Interface Board are two identical PCI cards as described in section 2 2 5 The only difference between the two is the FPGA firmware which implements the Inverter Bus communication protocol for the first one and the Sensor Bus protocol for the second one The Position Interface Board is connected through the Sensor Bus to the sensors pre processing unit Sensor 2 Digital which is located outside of the control cabinet The cabinet circuitry is connected to the 400 V grid by the hand operated main switch KO located on the front door of the cabinet When this switch is closed at power up only the Control PC and the interface electronics are supplied 1L3 N In order to supply the Siemens devices the Infeed Module the two Monitoring Modules and the 18 Power Stac
185. to calculate the correction term 62 replacing with and amp with amp amp respectively k mx nx e E sgn 0 cos sn ee Me Eq 3 58 Tp Tp Cam Ean n In case there is a phase difference between the two observed EMF vectors the correction term e will be calculated as a sum of two partial corrections one for each controlled segment x Eym TE xn Eq 3 59 113 Linear Drive for Material Handling The two partial corrections amp m and xn are calculated based on the estimated EMF vectors corresponding to the two controlled segments and on the two estimated electrical angles n and 6 ne sgn cos 6 sn n 5 s 7 sgn V cos 8 sin Me Eos The estimated electrical angles of the two controlled segments are determined from the estimated position x taking into account the different initial phases Oom and Oon of the segments a TX nX 0 0 0 tU m Tp 0 m n Tp 0 n Eq 3 61 the initial phase of each segment being considered with respect to the first segment of the track SS1 A block diagram representation of the mechanical observer described by Eq 3 47 Eq 3 59 Eq 3 60 and Eq 3 61 is given in Figure 3 38 3 3 3 Implementation at the Straight Section of the Machine Rigid Vehicle The three observers structure described in the previous subsections was first implemented at section 1 of the linear machine straight double sided section con
186. tween Uac and Uo e g by measuring the time interval between the positive zero crossings of the two signals the position of the capacitive sensor can be determined Figure 2 38 shows an overview of the setup used for the evaluation of the capacitive sensor with sinusoidal excitations The same Reference Position Acquisition Board as for the evaluation with square wave excitations will be used for the acquisition of the optical sensor s position and for the reference current s output For the acquisition of the capacitive sensor s position an eZDSPF2812 develop ment board equipped with a TMS320F2812 DSP 39 from Texas Instruments will be used The PWM outputs of the DSP are used to generate the control signals for the sinusoidal voltage generation as discussed below Comparators are now used to determine the zero crossings of the phase modulated signal Uou The outputs of the comparators are fed to the Capture Units of the DSP 59 Position Sensing Systems for Passive Vehicles Excitations o C I T gt I o zm Pos mod P mm nn Sensor Output 30 40 50 60 70 80 90 100 Time us Figure 2 37 Phase modulation There are two synchronisation signals between the DSP and the Reference Position Acquisition Board s CPLD The first one IRQ SYNC is used to synchronise the 100 us control interrupt IRQ11 and the acquisition of the reference position with the generation of the excitation signals of the
187. ulated Vehicle Due to the modular construction of the linear machine there are gaps spacings in the stator winding between consecutive stator segments see Figure 3 41 a These gaps are generally not equal between each other and not a multiple of 2tp tp being the pole pitch of the linear machine so they will generate phase displacements between the estimated EMF vectors of two consecutive stator segments which displacements must be taken into consideration in the implementation of the mechanical observer The phase differences between each two stator segments were determined by offline EMF measurements at the transitions between segments as illustrated in Figure 3 41 b The phase differences between all 18 stator segments are listed in Table 3 5 Figure 3 42 shows the a component of the observed EMF vectors as well as the observed speed during motion of the vehicle from segment 4 to segment 6 At the transitions between segments the phase differences of the observed EMFs can be seen It can also be noticed that the amplitude of 645 varies significantly while its frequency remains approx constant implying that the amplitude variation is caused by changes of the air gap Segment k Segment k 1 Induced voltage in phase a V 0 50 100 150 Time ms a Spacing in the machine s winding b Resulting phase difference in the EMF Figure 3 41 Phase difference in the EMF at the transition between stator
188. ule Initialisation function called when the module is loaded into the kernel e Module Clean up function called at the module s unload from the kernel e Command FIFO Read Handler called whenever a command is sent from the User Interface e The Interrupt Service Routine called every 100 us through the Vehicle Control Interface Board s interrupt and where the control algorithm is implemented In Figure 3 16 a block diagram of the module initialisation function is shown When the module is loaded into the kernel the required hardware the three PCI interface boards is first looked for The boards are detected using the Vendor ID VID and Device ID DID fields of the PCI structures associated with every PCI device 25 If at least one of these interface boards is not present then the control algorithm cannot function properly so the initialisation function will not proceed it will register an error message in the kernel message buffer and then exit with a failure code 82 Linear Drive for Material Handling RT Module entry point Look for the VID and DID of the VCI Board Was the PCI DAS08 found JJ 2 o 2A 3n cl Q3 o mo Read Base Address Re gister 2 of the PCI DASOS Interface Board the VCI Board error message Record a kernel Reset the register BAR2 0x02 of the PCI DASOS Board K2 K3 Read Base Address Re gister 0 of VCI Board Read Inte
189. ve as guides see 69 Linear Drive for Material Handling Rigid Vehicle without upper plate Figure 3 6 Straight double sided section of the machine with rigid vehicle the horizontal guiding rollers in Figure 3 7 On the bottom side of the vehicle there is the magnets carrying block which slides between the V shaped stators On this block the permanent magnets are mounted 10 on each side On the topside of the vehicle there is a load carrying plate which also serves as a support for the measuring scale of the optical sensors In Figure 3 8 a photo of the articulated vehicle 43 is shown The main difference between the rigid vehicle and the articulated one is that the magnets carrier block of the last is split in two halves which are connected through a spherical joint allowing them to Load carrying plate Tube for both carrying and guiding A Permanent a o magnets Stator SIR n g Figure 3 7 Transversal cut through the double sided section of the machine Source 43 70 Linear Drive for Material Handling rotate with a small angle horizontally as well as vertically with respect to each other The two magnet carriers and the spherical joint are held together by a spring tensioned steel cable which passes through the centre of the sphere The entire linear machine at our experimental setup is depicted in Figure 3 9 The curved sections of the track are super elevated in order to allow for high
190. wn in Figure 2 10 can be divided in seven major functional blocks 1 ADC Control controls the four analog to digital converters 2 Sin Cos Registers store the conversions results 3 Counter A Control controls a 14 bit counter CNTRA which counts the periods coarse position of the read head currently routed through MUXA based on the sign bits of SA and CA 4 Counter B Control identical with Counter A Control CNTRB counts based on the sign bits of SB and CB 5 Current Head Control controls the Current Head State Machine CHSM which keeps track of the read head s currently covered by the scale Additionally this block generates the control signals for the two multiplexers MCA and MCB 6 Sensor Bus Control implements the interface with the Sensor Bus 7 Synchronisation generates the timing and the synchronisation signals for all previous blocks It is assumed that initially the vehicle is outside the sensors region The Current Head State Machine is in state Ho signifying that no read head is covered by the scale The CHSM uses two 3 bit buses LDa 3 and LDg 3 to control the behaviour of the two period counters CNTRA 14 and CNTRga 14 The possible values for LD and their meaning are given in Table 2 2 Those of LDg are similar with the only difference that for LDg 100 the value ANsg instead of AN 2 is loaded in CNTRe In the initial state LD 001 and LDg 100 meaning that the
191. wurde zun chst analysiert und ein Model ermittelt Basierend auf diesem Model werden zwei Auswertungsmethoden dargestellt eine verwendet unverz gliche abtastungsbasier te Demodulation und die andere basiert auf Phasenmessung Innerhalb der Transportabschnitte des Linearantriebs sind die Anforderungen an Genauigkeit und Dynamik der Positionserfassung weniger anspruchsvoll als innerhalb der Bearbeitungsstationen In diesen Abschnitten wird sensorlose Regelung implementiert basierend auf der Auswertung der Elektromotorischen Kraft EMK Die unterschiedlichen Parameter aller Statorsegmente werden dabei ber cksichtigt Aufgrund mechanischer Beschr nkungen sind L cken in den St nderwindungen zwischen aufeinanderfolgenden Speiseabschnitten vorhanden Dies bedeutet dass die EMK Vektoren zweier aufeinanderfolgender Statorsegmente einen beliebigen Phasen unterschied haben k nnen wodurch sich insbesondere f r die sensorlose Regelung zus tzliche Herausforderungen ergeben Beim bergang zwischen Bearbeitungsstationen und Transportabschnitten wird ein Synchronisierungsverfahren zwischen der gemessenen und beobachteten Position dargestellt und experimentell ausgewertet Contents PDS 3 Kurzfassung n 9 Bio QI dg ee UR TER e OX T Xe cep ee E E 9 EIL eS E ee E E gue ee ee see E E 11 1 O CUO aeee cS 17 2 Posi
192. y exist between the two heads and secondly since the position data is updated in the Sensor to Digital firmware every 2 us but read through the Sensor Bus every 100 us by the control algorithm there can be a difference between the position where the transition starts and the position where the transition data is for the first time available to the position calculation function This has however no influence on the position calculation since the Sensor to Digital firmware guarantees that the signals from the previous and from the incoming head are sampled simultaneously at every transition until they are read through the Sensor Bus and made available to the position calculation function While the scale covers head 2 state H 28 a new reversal back to positive direction is assumed and the vehicle travels now the entire region covered by the sensors exiting at its right The state of the reconstruction state machine changes recalculating the offsets of the incoming heads as indicated by the data received from Sensor to Digital Board and in parallel the global position x is calculated This continues until time ts when the last 40 Position Sensing Systems for Passive Vehicles position x tg is calculated In the next control cycle tg 100 us USE 0 and USEs 0 is read from the pre processing unit meaning that the scale is beyond head 5B No position information is now available and the reconstruction state machine returns to its

Position Acquisition and Control for Linear Direct Drives

Contents

Download Pdf Manuals

Related Search

Related Contents