It was summer 2002. As a student I've just started 3-month Summer Student Programme at CERN under supervision of Dr Richard Jacobsson in PH division (LHCb experiment on Large Hadron Collider accelerator). I had several tasks - measurement and debugging of various boards and some feasibility studies. As a last task I was asked to prepare design of beam position and intensity measurement system. It was quite challenging. How to measure electrical pulses that come from the LHC electrostatic position pick-up which are 1ns long and occur every 25ns. One needs to know the energy of the pulse and its relative position to 40MHz machine clock. In 2002 ADCs were not fast enough, good-enough TDCs did not exist so I had to do the front end in the analog domain.
The idea was simple - build fast analog integrator followed by 40MHz ADC and FPGA that does the magic. That was challenging. Interesting issue was to rectify the very short pulse that was bipolar with very high speed Current Feedback Amplifiers
Here is presentation with more details https://slideplayer.com/slide/9430892/
The idea was verified with simulations.
Another problem was measurement of the beam phase. Since TDCs that were on the market did not have enough throughput, I designed my own version based on RS flip flop and ADC working as pulse-length measurement.
The final circuit board drawing is below.
Take into account that the VME communication interface was not used. Every VME board had its own CreditCard PC module which was x86 SBC with another module that served as PCI to parallel bus converter. All communication went over Ethernet. It was due to the problem with VME interoperability. Then the summer ended, I left CERN to continue studies in Warsaw and the project was continued by Zbig Guzik from NCBJ in Swierk. In the meantime, 40MS/s TDC chips from ACAM (now AMS) became available so Zbig used one instead of the analog TDC. The project update is here http://lhcb-doc.web.cern.ch/lhcb-doc/presentations/conferencetalks/postscript/2006presentations/JacobssonPcdgs.pdf
I consider this project a success. The initial design process took less than one month.
In the meantime, I applied for another CERN job and in March 2003 I've started the Technical Student programme at CERN under the supervision of David Belohrad in the Beam Instrumentation group (BE/BI/PI). The task was challenging - renovate the old measurement system for CERN Proton Synchrotron Booster accelerator. The aim of the system was a measurement of the beam intensity. In other words, I had to count protons in a vacuum tube that traverse with a near speed of light. And do it for every individual bunch of particles! Easy task because I did a similar thing for LHCb. Here I could use entirely digital design because signals had much lower bandwidth.
Moreover, the old system was purely analog. One poor guy had to periodically travel around the accelerator with a screwdriver and calibrator to calibrate the equipment. It could be fine in the 1970s when the original system was installed, but in the 21-st century, we got used to different methods. So my task was also to build an embedded calibration current source that has quite challenging parameters: 4Amps at 50Ohm load. And a few ns of the rise time. Which means 200V of signal amplitude. And the idea was to do the calibration after every measurement. Why the current source? The protons that travel along the tube are charged particles. So, in fact, they are current. The measurement sensor is a coil wound around the toroidal core that is mounted around the vacuum tube equipped with some slot in which the electromagnetic field can interact with the measurement transformer windings. The calibration input is a single turn of wire on the same core. That's why one needs to pass 4A of current - such is the maximum sum of proton charges in the bunch that flows within the bunch time.
I found some old photos of me in the lab. It was in building 37, under the floor, there is the PS Booster installed. In front of me the VME board prototype is visible.
Below is a photo of the VME board I designed at CERN. It was equipped with two 210MS/s ADCs, front-end with variable gain (right) and HV supply with calibrator (right side). On the bottom side, a standard VME interface is located. At the time I didn't know that to get real 12bit accuracy I need to drive ADC with low jitter clock. But it worked quite nicely. No serious issues were found. TRIC stands for TRansformer Integrator Card because its task was numerical integration of the signal.
Here is the second version of the card installed in the machine crate.
During 1-year Technical Student Programme I designed a few other interesting circuits. One of them was sub-ns ToF meter utilizing ACAM TDC and Constant Fraction Discriminator. Another one was a very fast gated integrator for LHC beam signals..
Here is the paper that describes the results
https://www.spiedigitallibrary.org/conference-proceedings-of-spie/6347/63470F/Beam-intensity-measurement-system-for-proton-synchrotron-booster/10.1117/12.714530.short
In the meantime, I defended the master thesis at the WUT, finished CERN contract, applied for Ph.D. studies and also got a new Ph.D. student contract at CERN.
During the summer and fall of 2005, I was involved in the development of Gigabit Ethernet version of the camera for Pi of the Sky experiment.
In 2006 I started the Ph.D. Student programme at CERN where I continued the beam position measurement related tasks. An improved version of the TRIC card was developed by me, where the second measurement and calibration channel was added a well as some fast RAM memory. FPGA was upgraded to a newer version (Cyclone II). Even though FPGA has 484 balls, I managed to route most of the balls using 4 layer PCB. This time ADC clocking was done properly using a low jitter clock generator IC.
In the meantime, other groups realized that my board is also attractive to them so they asked for a few modifications and the next version arrived. The main difference was the SO-DIMM SDRAM socket so the board could be used as a digitizer. Several boards were built and installed across the CERN site.
Upgraded version that utilises newer FPGA and improved calibrator was also designed that could be use either as a 250MS/s digitizer or for beam instrumentation other tasks but only 2 prototypes were built. Previous version 3.2 was well tested and simply good enough.
Intensity measurement was not real problem that could satisfy me in terms of PhD. CERN wanted to build new Beam Trajectory Measurement System (TMS) for CERN Proton Synchrotron. It was quite challenging task because the beam signal varies a lot depending on types of particles that are accelerated. Moreover, the relationship between the beam signal frequency and the RF cavities frequency is not fixed. Existing system was done in old-fashion way and required expensive maintenance. It occupies a few racks.
What could go wrong after 30 years of service ?
Some details:
The signal gets acquired by the electrostatic pick-up.
The electric field generated by moving charges interacts with electrodes and produces signals
One can notice that they vary a lot. We need to know intensity of every bunch, bunch by bunch.
Below are some slides presented by Jeroen Belleman during DIPAC05 conference.
Not only structure of the pulse changes, its frequency varies as well during the acceleration cycle. The new beam gets accelerated every second. The magnetic field is also ramped to keep the beam within the vacuum tube centre.
The new system must cope with not only such conditions but also beam splitting!
During the acceleration cycle, at certain moment the beam bunches get splitted to more bunches.
The image below shows several scope traces from consecutive revolutions placed on a singe chart:
That's why old Trajectory Measurement System occupied 7 racks! My task was to build something simpler that fits within one 19" crate, is built with recent industrial standards which will last for at least 20years. And of course FPGAs, the newer the better. My colleague, Jeroen Belleman proposed the idea of the digital Phase Lock Loop that automatically adapts to the signal frequency and its phase. He did some simulations with real recordings which proved the concept. My task was to build a working system with FPGAs. The same PLL approach was also adopted to the PS Booster Intensity measurement implemented on TRIC3.2 cards. The final TMS algorithm was implemented on COTS modules from AlphaData. The whole system occupies almost two 19" Compact PCI chassis.
The system was described with details in my thesis:
https://cds.cern.ch/record/1482181
http://inspirehep.net/record/886999/files/CERN-THESIS-2010-081.pdf
The results are quite remarkable.
https://slideplayer.com/slide/9548720/
In 2009 I decided to leave CERN and focus on projects in Poland. One of them was the Creotech company, another were activities at the WUT.
However, involvement in beam instrumentation did not finish!
In 2012 we started building an open source beam position measurement platform based on Micro TCA standard for Brazilian synchrotron called Sirius.
The results are described in this thesis
https://cds.cern.ch/record/1977919?ln=en
and here
https://indico.cern.ch/event/743699/contributions/3072694/attachments/1750591/2837151/sirius_bpm_fofb.pdf
http://accelconf.web.cern.ch/AccelConf/ICALEPCS2013/talks/wecocb07_talk.pdf
The system was built as open-source so commercialization was straight-forward. We tried at the WUT other paths of commercialization but were extremely difficult due to several reasons, not only legal.
The early version of the system, without front panels is presented below. All modules except for chassis, power supplies and MCH controllers were designed in-house.
Several derived products were developed since that time which shares a lot of initial ideas and design choices https://www.ohwr.org/project/afc/wikis/home https://www.ohwr.org/project/afck/wikis/home https://github.com/gkasprow/AFCZ/wiki https://github.com/sinara-hw/Sayma_AMC/wiki https://www.ohwr.org/project/utca-rtm-8-sfp/wikis/home
https://www.ohwr.org/project/fmc-adc-250m-16b-4cha/wikis/home
https://www.ohwr.org/project/fmc-adc-130m-16b-4cha/wikis/home
Some of the modules were produced by Creotech in hundreds of pieces and became standard components in many labs including LNLS, CERN, GSI, DESY to count a few.
Whole LNLS Sirius beam diagnostics based on the modules above were delivered by Creotech in 2016/2017
So,the Beam Instrumentation adventure continues...
Comments