6. Further Thoughts¶

6.1. Tracking¶

Although the TimePix family of chips are used in many applications and to great satisfaction, the experiences with the two particular chips in this setup have not been great. We suffered from a broken chip, a new batch that was of questionable quality, and the ongoing development of RelaxD and RelaxDAQ. Hopefully these will all be resolved so that operating of the GridPix chip will be routine, and that the number of chips can be expanded to provide a greater tracking volume and a second tracking plane. Track fitting and 3D reconstruction will require a considerable effort.

For applications in medicine higher event acquisition rates are required, much higher than the 120Hz of the TimePix chip. A future version, TimePix 3, is currently being designed to attain a 40MHz pixel event acquisition rate, translating roughly to a 40kHz+ event if it were to be used in this setup. This gets the performance in the realm of medical feasibility and such an improvement is therefore eventually essential.

Dr. T. Tanimori gave a presentation at KVI on his research on large cubical drift chambers. While not using TimePix chips, his 3D tracker has a design similar to our GridPix: a 2D detection surface to one side of a cube, with precision time registration of each signal. Dubbed μPix, this is then used to recreate hits inside the drift volume. He reported successful performance of the device when applied in Dark Matter searches. While this research causes much lower pixel event rates than medical imaging, the array can register pixels at a rate of 10MHz. The experience with large drift chambers could serve as an example for when our drift volumes are scaled up (Takahashi et al., 2011).

A potential replacement for TimePix related chips could be SiPM. These chips integrate a small solid state photomultiplier above each pixel, allowing the detection of single photons per pixel with excellent timing. If each pixel is connected to a light conducting fiber, which are then fanned out and configured in a typical silicon strip detector plane, tracking hadrons could be a possible application. These fibers are still advantageous compared to silicon strips, because they can be of low-Z materials so that scattering and energy loss is minimized, as is the case with gas detection in GridPix (Philips Press Release, 2012, Clements, 2012).

6.2. Calorimetry¶

First the calorimeter itself should be replaced. The current device is old and therefore most likely discolored and no longer perfectly light tight. Once a new device is available, with known dimensions of the crystal, lab tests with energy reconstruction will be easier to conduct and performance will improve and will be better defined.

As for the calorimeter data acquisition system:

• The strange kink seen in the last curve of figure 5.7 must be investigated and resolved.
• An energy reconstruction must be performed at the earliest opportunity, to test the real-world performance.
• Pre-shaping of the calorimeter output could shorten the required event window, so that the effective event acquisition rate can be increased. It has been established that the device can acquire events up to 30kHz if the event window is small.
• A small firmware change in the HiSparc-box could disable the second channel, doubling the effective buffer space.
• The communication between the USB interface chip and the memory chip is currently capped at 1250kB/s. It seems that increasing this bandwidth and thereby acquisition rate would be fairly straightforward. The USB interface of the chip, the FT2232H, supports a data rate of 10MB/s so is not limiting us yet. Within Nikhef there already is a spin-off of the HiSparc-box, Muonlab, so there is some experience with application specific implementations of the same basic device.

6.3. Trigger¶

A small improvement that can be implemented with the recently developed Python bindings for RelaxDAQ is a deeper coupling to the CaloDAQ. It should be fairly straightforward to implement a unified starter script that prepares both tracker and calorimeter for data acquisition. A next step could be a post-acquisition script that verifies the number of acquisitions and timestamps for both tracking data and calorimeter data. The HiSparc message parser as of yet is too slow for such a task (as an example: processing the data in a 30 second run at a 31kHz event acquisition rate takes about an hour).

In a future application the accelerator itself may provide the trigger. This would remove the need for scintillators in the beam line, improving energy resolution and decreasing scattering, and provides a known signal for when particles should be seen, allowing for a better performance analysis of the setup. At this point an effort can be made into simulating the setup with software such as Geant.

6.4. Acknowledgments¶

I would like to thank Nikhef, Jan Visser and Sytze Brandenburg for the opportunity to work on this project: a crossover between disciplines and one of the few master projects in particle physics that are close to a real and lofty application. It was also a project that entailed almost any type of scientific activity:

• literature study: not only a bit on particle physics and the competing alternatives in hadron imaging, but also a tiny bit of medical knowledge,
• physics analysis: how we can combine our knowledge of physics with experimental data to fully understand what is going on,
• software development: apart from processing gathered data, the CaloDAQ driver had to be written which was my first experience with performance critical coding,
• clean room construction of the replacement GridPix,
• the data taking run at the KVI accelerator and all the excitement that goes with a live experiment,
• attendance of a 4-day conference in Darmstadt on the topic of hadron therapy, allowing me to speak with one of the persons cited quite often in this thesis,
• bugging many people in the Detector R&D group and elsewhere in Nikhef to help a novice out,
• setbacks: the broken GridPix or the missing documentation on the HiSparc-box, to test how difficult keeping a schedule can be,
• lab work: testing my abilities to solder, plug a cable or built a frame on the fly.

The breadth of tasks was astonishing, and I am very glad to have had my time as a master student on a project like this. In no particular order, a few names deserve to be mentioned. Vincent van Beveren was an essential help in both combating RelaxDAQ oddities and in getting the HiSparc-drivers to work. Bas van der Heijden was the go-to guy for all RelaxD problems, and didn’t mind reviewing some cabling every now and then. Arne de Laat, David Fokkema and Hans Verkooijen, the Hisparc guys, for helping me understand the hardware design of the HiSparc-box and developing most parts of the current driver. The two Martins, van Beuzekom and Fransen, taught me basically all I now know of electronics. I had not had any experience with cable latency or trigger logic before, and without their hands-on work there would be no trigger. Wilco Koppert, for his detailed knowledge on GridPix, gas and drift properties, track reconstruction and clean room construction of GridPixes. Panagiotis Tsopelas for giving me a running start, and Jan Visser for helping me keep running. Francesco, Mary, Enrico, Marten, Rolf, Matteo for all their support and the R&D group as a whole for showing me what the life of a researcher is like. Lastly, I want to thank my supervisors Jan Visser and Sytze Brandenburg for their thorough support with this thesis and for sharing their experience and knowledge in performing this project.

I hope this setup will soon see a successful test at the KVI accelerator and that the project will someday result in a competitive apparatus which can help cure cancer victims. I want to thank the reader for reading!