This summer, I had the extraordinary opportunity of being a CERN Summer Student. I was ecstatic when my CERN Summer School application had been accepted; I could not wait to contribute to cutting-edge research, attend the world-class Summer Student lectures, visit various CERN sites and meet people from around the world!
For the placement I worked alongside CERN intern Agustina Quesada, supervised by Dave Barney, EP-CMX group leader.
The project was for the High Granularity Calorimeter (HGCAL), which will replace the endcaps of the CMS detector to survive the increased radiation levels and pile up of the High-Luminosity LHC. To find out more information about the HGCAL, please click here.
The HGCAL includes 600m2 of hexagonal silicon modules, comprising baseplates (for mechanical support, thermal stability etc.), gold-covered Kapton sheets for biasing of the silicon sensor, the silicon sensor itself and a printed circuit board known as a “hexaboard” due to its shape. A prototype module and its components were presented by Agustina and me before and after a summer student lecture, as shown in Figure 2.
The main components bonded to the hexaboard are an FPGA (Field Programmable Gate Array) and four Skiroc2-CMS ASICs. In the first stage of the project, we carried out reception tests of a batch of hexaboards, newly delivered from the manufacturer to CERN. The hexaboards had been previously tested at the manufacturer, but the latest set of tests were carried out after some non-functioning ASICs had been replaced and a protective “globtop” layer was added over each ASIC. Testing was to determine if ASIC replacement had solved the previous faults observed, or if there was a fault with the printed circuit board. For the testing, hexaboards were connected to a custom test system based on a Raspberry Pi microcomputer and another FPGA. The system carried-out a number of functional tests (communication to the hexaboard and configuration of the ASICs etc.) followed by a dummy “run” to acquire data. We found that a majority of boards that had had ASICs replaced exhibited the same errors that were found prior to the replacement – a surprise that indicated a problem with the PCB itself. We also found some boards that did not work, despite working perfectly at the manufacturer. We attempted to understand more about these failures by using an X-ray source coupled with the “Timepix” chip (with Michael Campbell and Jerome Alozy from EP-ESE) but the resolution was not sufficient to see the ASIC wire bonds, which are suspected to be the source of the problem. The CERN metrology lab will help with this later in September and we hope to identify the faults and feed-back information to the manufacturer.
We then proceeded to our main project: the replacement of old Nuclear Instrumentation Modules (NIMs) with a small programmable system known as NIM+. NIMs are used extensively at CERN, being used for example for triggering in data acquisition in beam tests. Modules include discriminators (which produce square wave output if the input voltage is above a set threshold), coincidence units (which give an output for simultaneous input signals) and level adapters (which format the output for use as a trigger in DAQ systems). NIMs are very easy to use and ubiquitous at CERN. However, NIMs are relatively large, mainly old and many are obsolete or only partially working. In addition, some parameters, such as the threshold, can only be adjusted manually, potentially losing precious beam test time.
Several developments in laboratories around the world are aimed at replacing NIMs with more modern programmable systems. An example is the compact and inexpensive NIM+, a system developed at Fermilab by Lorenzo Uplegger et al. The version of the NIM+ that we had consisted of a commercial “ZedBoard” – an FPGA-based general-purpose board with an FMC connector and a custom discriminator FMC-based mezzanine providing a number of inputs and outputs (using standard LEMO connectors). The FPGA can replace NIMs such as the coincidence unit and level adapter, while the custom discriminator can replace the discriminator NIM. For our project, work was carried out towards this goal by learning how to use the Vivado software suite to program the ZedBoard’s Zynq FPGA. Programming FPGAs is very different to the type of programming we were familiar with (C, Python etc.) due to the real-time nature of FPGAs, requiring a deep understanding of logic blocks and clock-based state changes. Using Vivado, a logic block diagram of what we wanted was created and we aimed at populating these blocks with custom software. Although we did not complete the project, we did succeed in talking to Zedboard components such as hardware switches and LEDs through the FPGA. For example, we made the LEDs represent binary sums of two 4-bit “nibbles” (half of an 8-bit byte) determined by the switch positions, as shown below.
Our research was presented in group presentations and a project report was published. We also wrote documentation to help others build-upon our work and head towards a final NIM+ system to be used for the HGCAL and, potentially, many other CERN projects. Agustina and I also presented at the CERN Summer Student Poster Session, where our poster had an Instagram-themed design, which proved to be very popular with the audience!
An unforgettable summer at CERN has now come to an end. A massive thank you to everyone who helped us, including Dave Barney for being an outstanding supervisor, taking the time to explain topics and giving us lots of helpful advice for pursuing a career in research. Also, thank you to Agustina Quesada for being an amazing person to work alongside, and Paul Rubinov for all of his help with explaining electronics concepts and solving problems with Vivado. Finally, thank you to the CERN Summer Student team for providing the opportunity for me to be a part of this once-in-a-lifetime experience!
This article was originally published in the CERN EP Newsletter.
The views expressed in CMS blogs are personal views of the author and do not necessarily represent official views of the CMS collaboration.