A real time electronics emulator with realistic data generation for
Contents
1. LHC s system 11 and to the TCC24 prototype board Its FPGA s receive dedicated firmware to perform the real time analysis for both the Data and the Trigger paths B Common Design Solutions To insure that all received data could be read and compared at the same time the extracted clocks for each channel are used to write into a FIFO buffer All buffers will be read by the same clock source not phased with any others but locally generated inside the TCC24 board To perform it a dedicated FIFO design has been developed In the Virtex2 architecture a block RAM can be used and concurrently accessed from both sides if and only if the read address and the write address are not the same To solve this problem both data and address information are stored in a single memory location Moreover a feedback mechanism is used With this method the FIFO is self addressing so no external counters are required A number of n address bits are fed back from the RAM output data port to its address port The same n bits are also fed to a simple incremental circuit The output of the incremental circuit is combined with the incoming data word This address plus the data word are then written synchronously into the next memory location The clock load for the incoming part is one thus it is not necessary to use the available global clock tree and their dedicated buffer limited in the Virtex2 family Up to 12 mcoming clock domains plus a local2. a FEC 9 board Due to the time delay between two FE boards this clock is provided to each board with a time lag which has to be compensated This is achieved for each FE with a dedicated hardware called PLL 10 All boards respecting the setup data time can not be necessarily synchronized within the same clock phase Consequently a different clock time domain is also created for each of them The data buffer made out of headers control words and real data is sent to the associated deserialisers of the TCC24 On the reception side each of them will be retrieved from the serial data stream by a deserialiser component The outputs of these components are connected to FPGA programmable devices On their own path the named functions TRIGPATH and DATAPATH embedded in these programmable devices will check that all received data for all channels are the same 25data bus ammm 12 Optical encapsulated Fibber p gt Token Ring bus LHC clock transport Optical fibber link Optical Links to TCC24 XFEST Test Platform 1 EF for stimuli 4 EF in test Trigger path C Test Bench Control FEC card inside ChipScope Console Figure 5 XFEST test bench in 4 boards configuration Figure 5 shows the XFEST test bench and all connections An AGILENT 81134A pulse pattern generator is used to emulate the LHC clock It is connected to a couple of boards TTCvi 11 and TTCex 11 part of the TTC
3. been designed and built to test up to 12 FE boards at the same time The same digital input patterns at a rate of 40 MHz are sent onto all Front End FE boards Each FE receives an equivalent of 25 channels 14 bits digital data streams at a 40 MHz clock frequency The data are processed by dedicated Asics called FENIX 4 In order to emulate the 25 digital input vectors a prototype version of the FE board equipped with commercial programmable circuits in place of the FENIX Asics has been used This board called EF is used in a reverse mode to its original usage as FE prototype One EF is installed on a motherboard designed and developed at LLR The main role of this motherboard is to insure the replication of the 25 digital signals in order to feed the FE boards In an initial phase of the project it has been foreseen to have three motherboards each one equipped with one EF in order to test at the same time 12 FE boards In the current running the system limits itself to one motherboard with four FE Figure 1 l Present address LPNHE Paris VI VII http www lpnhep in2p3 fr oly in2p3 fr Slow Control FEC interface LHC clock generation 0 XFEST Test Platform 1 EF for stimuli 4 EF in test 25 databus gt Token Ring bus SI Test Bench Control Parameter setting Command Figure 1 Display of the XFEST test bench Optical Links to TCC Data path The outputs of the FE boards under test are shipped
4. one have been taken into account The final goal is to collect data in all clock domains and access them in a single one This solution is well adapted to the needs and works fine To avoid working on the data buffer when entries are received a solution based on the comparison of the received data in real time has been adopted A channel must be defined as the reference for the comparison The user is able to define it through an internal register accessible through the VME interface The comparison errors the transmission errors or the loss of synchronization of the serial links are counted and written into registers accessible through the VME interface Figure 6 ChipScope Pro windows with TRIGPATH on the front displaying an error on channel 3 and DATAPATH waveform behind The channel synchronization loss debugging has been a real issue A specific solution based on the use of ChipScope Pro software 13 from XILINX has been adopted This software application can be seen as an internal logic analyzer that exhibits all stored signals values when a predefined event value occurs The signals to display are selected by the user Internal RAM blocks are used as acquisition memories and clocked inside the TCC24 with the local LHC clock Figure 6 shows a snap shot of this application A synchronisation loss can be seen on channel 3 This application has been really useful to investigate the data r
5. A real time electronics emulator with realistic data generation for reception tests of the CMS ECAL front end boards T Romanteau Ph Busson C Collard A Debraine D Decotigny L Dobrzynski A Karar N Regnault Laboratoire Leprince Ringuet 91128 Palaiseau France romanteau Abstract The CMS 1 electromagnetic calorimeter ECAL 2 uses 3 132 Front End boards FE performing both trigger and data readout functions Prior to their integration at CERN the FE boards have to be validated by dedicated test bench systems The final one called XFEST eXtended Front End System Test and for which the present developments have been performed is located at Laboratoire Leprince Ringuet In this contribution a solution is described to efficiently test a large set of complex electronics boards characterized by a large number of input ports and a high throughput data rate To perform it an algorithm to simulate the Very Front End signals has been emulated The project firmwares use VHDL embedded into XILINX Field Programmable Gate Array circuits FPGA This contribution describes the solutions developed in order to create a realistic digital input patterns realtime emulator working at 40 MHz The implementation of a real time comparison of the FE output streams as well as the test bench will be also presented J INTRODUCTION The present XFEST project is part of the electronics boards validation system described in 3 It has
6. d when all trigger and data outputs are identical during the testing time at the bits level I PATTERN GENERATION SOLUTION A Overview Five Virtex2 FPGAs from XILINX present on the EF board are reprogrammed in a reverse mode to emulate the signals send by the VFE electronics These FPGA devices called PATGEN work with an embedded firmware providing pattern generator data A sixth FPGA device on the EF board is used to control and synchronise all the others by a start stop software command An IC interface is used for the internal parameters setting To test more efficiently the discrimination capabilities of the FE board it has been decided to favour a signal amplitude distribution with low level amplitudes rather than the high one Die Ll Piglogrem TES OR wal ues madd 152 EL iS i METI errr rere 14ea ere ere nt ay LELT Shee ete pb pbb IEEE a EL PF PPE eee EEE EEE ae ee O75 PPP ESS Seer E EE eS Pees eee Hid 1G EO EE 14 a Ba eRe i A A Be a By fil TOIT OCCT PP ORAS ORGS PEPPER Cert TEETER dil ToC Pere Pee te ett Serer Pre aa tee Per ree eee ery MEUN ee 2 a ee ee ee Se a ee 4157 ESH zdim METS AN E mA ES PAT TERM Wiles Hi ajo iz BE HH Valus ic Trea ESSE scgui red maria auras Cl Phologree eD wal used ER 1718 VSP PEF ER iHi iiiidiibhiiikihi iii 1225 foterrrer er tite rie tee ere rey EL PITtett cP eeer ree ttt rte rere ee tinny 745 ARR a ATAT da PPE EEE ee SPP PPE Pee ee
7. ee eee fut ee eee en l wm meg ae ie att oan Hatt WOT ADDER Yalu WAH elo ja POT WIH falusa iz TS frea DES soy irad mania birea Ll Piliin EDI wal il EFF i PPP ES ee PEPE ee Pe ee Wi i a a a a iii PTT OP PPP Een STS ee ere ett a OPP eer eta eee err h 135 LITTE a hk ANITINA EEP EPEETAN ET OTTEET OPE Te TPT CCP eee tte te ete Peer rs fete ter EErEE T hel EE Serer rrr rer eee rere Pee rer eer ae t ag PERPE ee ek a Hg PPR SSeS EPP EEE eae EE ee PP E n ai pd PPP repr EEO ee ee Pe eee AF EY iirinn jonr CY SC j biik bdid dii pi i i B h Bia CPP Pe oa Ba B Ba da ia Biba i palia i ai b Ba ha da E B creer a o E ii ADO rintd iad OLA SOAD LER gP ADDED EPEE EE PERITE ts ee m fe e m oe cme m fe a m m mamaa a aise ie EN mun ain isisi Hid P LFE ale Bit sali ok HIE Value 5s from DIEI wple ea Figure 2 Shape of the signal amplitude distribution top provided by the EF pattern generator It is obtained by mixing middle several non correlated LFSR low distribution A single side triangle distribution as shown of Figure 2 top histogram is used for the signal model The algorithm also allows event pileup at a level of 10 To obtain the signal distribution a uniform pseudo random number generator has been synthesized in the FPGAs providing a flat distribution as shown on Figure 2 bottom histogram By mixing several non correlated LSFR one obtains a triangle distribution as shown on Figure 2 middle histogra
8. eturned through the VME interface when an error happens To perform the comparison all channels have to be synchronised on the same data It is then necessary to correctly adjust the PLL values on all FE boards in test as depicted before This is achievable if the cumulated delay time between several FE boards does not exceed one clock cycle For more than 4 boards this request is no more satisfied consequently a specific hardware has to be implemented in the DATAPATH and the TRIGPATH FPGA s firmwares Then it has been foreseen for the comparison alignment and for each channel the ability to automatically introduce a programmable delay to adjust the phase of the different signals with the reference one up to 3 clock cycles An effective comparison starts when all channels have adjusted their necessary latency and return a Phase OK information flag A scalable design has been created by using VHDL generic parameters With them it is easy to adapt the hardware designs to the number of FE board on the test bench C Specific solution for the DATAPATH The main specificity of the DATAPATH concerns the data storage sent in a burst mode when a LIA event occurs A well defined data block format 4 is then transferred to the TCC24 Each FE board handles 25 crystals data named Trigger Tower TT It concerns 5 Strips of 5 Channels each The format includes a TT header a Strip header and a Channel header Moreover eac
9. gs for the communication with the Control and Communication Unit CCU modules see also 12 10 QPLL User Manual v1 1 January 2005 CERN Microelectronics Group More information at WEB address http proj qpll web cern ch proj qpll 11 Timing Trigger and Control TTC Systems for the LHC initially developed by RD12 for CERN More information at WEB address http ttc web cern ch TTC intro html 12 F Drouhin amp al The CERN CMS Silicon Strip Tracker Control System CMS CR 2004 032 August 2004 CERN 13 ChipScope Pro real time debug and verification tool from XILINX http www xilinx com ise optional_prod cspro htm
10. h channel header includes the data frame length This length can be dynamically set to zero by the FE board to prevent the data overflow when too many LIA appear in a short period of time In this case the header the bunch crossing identifier information BCID the event identifier EVID and the computed checksum are only present into the DAQ data block format The performed comparison electronics must have the capability to detect and extract from the internal headers and identifiers the presence of data A variable frame length has then to be supported corresponding to various block sizes produced by the FE boards To manage this a real time finite state machine has been implemented for each channel The percentage of data with empty blocks is automatically computed during the comparison process and written to a register for each channel under test These results can be accessed by the VME interface IV CONCLUSIONS During the development of this project it has been checked that original hardware solutions could be developed for very challenging tasks Real time emulation of existing electronics boards has been a key start point for the test bench They are required to validate electronics boards which have to be integrated in the detector without any possibility of maintenance like for the space projects The validation of several thousands of electronics boards for the CMS electromagnetic calorimeter requires a robust test bench s
11. i 12 bis MGPA Gain Satiching Emulator 2 bita r Charure pS Bx Mulieres j x addera 1240 bite LFSR 11 x 16 bis Rages 2a RAM 32 x 18 bes Figure 3 Block diagram and resources of a single channel Pattern Generator device PATGEN B Simulation phase To check by simulation the signal amplitude distribution two methodologies have been developed The first one uses Matlab Simulink software and is a very fast methodology but the connection with the HDL model did not exist when our tests have been carried out The second one uses a library of specific procedures developed in VHDL It has been implemented in the test bench at the top level The standard HDL simulator console then exhibits a pseudo graphic distribution curve Several data files including the values produced with the HDL simulation have been produced and successfully imported within the ROOT 8 framework for analysis Another important point has been to check the access to the internal registers They are used to implement the stop start commands and study the effect of the parameters change Thanks to the re use of an FC IP and the CERN simulation library it has been possible to carry out these tests easily Performing these tests in a HDL simulation Multiple Grain Pre Amplifier Electromagnetic Calorimeter readout chip for CMS For each of the three available gains the user has a set of internal registers to define the value of gain pede
12. m The noise is also emulated with another LFSR whose repeatability sequence Based on a Logical Shift Feedback Register LFSR Initialisation with a seed equal for which all bits are at 1 or 0 is forebidden 7 6 hours is low by using a high bit length number The possibility to work within a pre defined range of energy can also be set by defining minimum and maximum amplitude values accessible through internal registers To achieve it a new IEEE VHDL package 7 has been experimented The digitally coded waveform of the signal is stored in a memory and can be modified by the user Two memory outputs can be mixed to produce a realistic pileup shape The automatic gain switching as well as the pedestal and the noise of the ECAL VFE electronics MGPA Asics chip can also be emulated This specific implementation is now in operation All parameters of the PATGEN devices can be set by the user via the embedded PC interface The user can also access and control the registers providing the test status The interface is controlled by a software application described in 3 Figure 3 shows the block diagram of a single channel pattern generator A pipelined architecture has been chosen to rename no it fie run at the LHC clock frequency CAN mere here Sharad Aiesoiunces 1 Multiplier 3x adiens 1x40 bits LFSR Sx 16 bite Poegisters pP 40 bits LFSR rear Noise G12 Gain i Noise Coe Pedestal
13. stal and noise environment allowed a fast development and a precise analysis on the generated data C Implementation All functionalities required for a five channels pattern generator have to fit the one million logical gates device available on the EF board The system needs up to 41 multipliers and 53 adders working in signed or unsigned arithmetic As the targeted device does not embedded all requested dedicated multipliers a XLINX IP generator software library has been used to place some pipelined multipliers on the programmable matrix To ease the placement and the routing and respect the temporal requested performances it has been necessary to manually define physical sectors on the matrix for each large subset of the design This process called floor planning can also clearly reduce the necessary time to place and route the cells Figure 4 shows the affected physical areas for each of the 5 pattern generators sub blocs and for an I C interface functions 5 x Pattern generators 1x IC Interface Figure 4 General and detailed implementation views of the PATGEN FPGA All kind of available resources are used Since the whole design can be viewed as a data path system a pipelined architecture has been used The insertion of some pipeline registers along the path allowed reaching the working frequency The total latency does not have a real importance here but the system has to work at the nominal frequency Ho
14. to a prototype of the Trigger and Concentrator Card 5 TCC24 by optical fibres The TCC24 is equipped with commercial programmable circuits In order to perform the real time comparison of the FE outputs the TCC24 has to run with a dedicated firmware To achieve the validation goals two different firmwares have been developed for the project as well as the interface to drive the setup e For the 25 signal emulators the following signal characteristics have been implemented the analogue signal shape its potential pileup the electronics noises and pedestals the trigger occurrence distribution and the signal amplitude which have been randomly generated Amplitude gains signal range and level of noise are all adjustable via an integrated IC interface 6 This dedicated firmware is loaded into the FPGA of the EF board e For the comparison system the main issue is to ageregate and compare in a same clock domain all data from different clock time buckets The real time computing of all input data is a key point to solve the buffer control problems The dedicated comparison firmwares are loaded to the TCC24 FPGA s e Parameters setup like the reference card selection is possible through an embedded VME interface Diagnostic and access to error counting are included in the design The interface is described in 3 The validation test for a set of FE performed for about one hour equivalent to 10 different patterns is obtaine
15. wever this type of architecture increases the number of used registers in a significant way The developed architecture gives access to a large set of internal resources available in the Virtex2 device Block RAM multiplier tri state buffer It represents 860 000 equivalent gate count and adie occupancy ratio of 70 The developed design is a good candidate to perform as an efficient testing platform of the FE boards This project proves that complex signals provided by the Front End electronics of HEP experiments can be emulated within a dedicated and versatile digital electronics system gt Intellectual Proprietary electronic block based on an assembly of logical primitives and provided by an outside source II TRIGGER AND DATA PATH ANALYSIS A Overview The main function of the XFEST test system is to provide the same signals to the FE boards plugged on the motherboard Each of them sends via two optical links their output information to a TCC24 5 prototype board One link Trigger Path continuously transmits computed trigger vectors the second one Data Path sends the internal buffered data when a LIA event accept is sent to the FE via the Token Ring Daisy chain bus used for inter connexion of the FE board The emulated signals are built out of 25 data buses arriving through 5 connectors soldered on each FE board Each FE board synchronizes the data with the LHC clock distributed via the Token Ring port and delivered by
16. ystem A limited design time has forced us to reduce risk by reusing existing prototype electronics board made with re programmable devices The use of VHDL to produce FPGA firmwares allows flexible hardware solutions and realistic nditions of test A deep interaction between the software and the hardware has been successfully achieved V REFERENCES 1 CMS Collaboration The Compact Muon Solenoid Technical Proposal CERN LHCC 94 38 2 CMS Collaboration ECAL Technical Design Report CERN LHCC 97 33 CMS TDR 4 15 December 1997 3 C Collard et al CMS ECAL Front End boards the XFEST project contribution paper to the current conference 4 FENIX ASIC version 1 0 HDL description version 8f Specification Mars 2003 CERN by Magnus Hansen 5 Pascal Paganini et al TCC24 reference to Tests of the boards generating the CMS ECAL Trigger Primitives from the On Detector electronics to the Off Detector electronics system Proceedings of LECC2004 Boston from 13 17 September 2004 6 Inter IC bus from PHILIPS It is a serial control bus that provides in a system the communications link between integrated circuits 7 Fixed point VHDL packages by David Bishop see at address http www eda org vhdl 200x vhdl 200x ft packages files html 8 ROOT an object oriented data analysis framework for physic experiment http root cern ch 9 Front End Controller an interface board that hosts token rin
Download Pdf Manuals
Related Search