Detector Group

During the past 12 months the Detector Group at Diamond has been working on two major development projects. The first project is the completion of the Xpress4 digital pulse processor, and the second the development of the large area detector for time resolved experiments, now called Tristan. The latter is still ongoing but has made substantial progress towards the delivery of the first prototype.

Figure 1: The Xpress4 digital pulse processor installed in I20.

In January 2018 the Detector Group successfully completed the final installation and commissioning of the Xspress4 digital pulse processor (DPP) on a branchline of the Versatile X-ray Spectroscopy beamline, I20-Scanning (Fig. 1). Xspress4 is Diamond’s new spectroscopy DPP and has been designed to significantly improve the spectral and rate performance that can be measured using the existing 64-element monolithic hyper-pure germanium (HPGe) detector – the workhorse for fluorescence mode X-ray Absorption Spectroscopy on I20-Scanning.

Analysis of the HPGe detector signals had shown that a major barrier to high rate, high resolution performance was the presence of element to element crosstalk within the HPGe detector. A digital domain based signal processing algorithm was developed (and patented) to effectively remove crosstalk, allowing further downstream processing to more accurately measure fluorescence X-rays. Xspresss4 is the physical and firmware realisation of this algorithm. It is a channel count scalable, custom designed, rack mounted hardware architecture whose salient features include a new high performance front-end digitiser, the latest high capacity high performance digital processing FPGA hardware and high speed serial links to cross-connect all channels in the digital domain. New firmware implements the crosstalk correction algorithm, enhances X-ray energy measurement precision, and reliably manages the inherently high data flows within the system in real time.

The Xpress4 DPP manages to considerably enhance the energy resolution of spectra at high counting rates (Fig. 2). For a molybdenum sample (K-alpha fluorescence at 17.47 keV) the maximum input count rate (ICRmax) increases six-fold from 210 kcps per pixel to 1.25 Mcps per pixel for the same usable FWHM resolution (e.g. < 600 eV on all pixels). At lower energies where crosstalk has less impact, ICRmax increase four-fold e.g. manganese K-alpha (5.9 keV) ICRmax increases from 200 kcps to 800 kcps for a usable FWHM resolution < 300 eV.

Figure 2: Molybdenum fluorescence spectrum acquired from channel 10 at 760 kilocounts per second by using algorithms with no cross talk correction (blue line) and with cross talk correction (red line). The energy resolution is greatly enhanced as it goes from 2318 eV FWHM down to 500 eV FWHM.

With challenging user samples it is sometimes not be possible to operate at very high count rates simply due to the diluteness of the sample under study. In such cases, using Xspress4 still improves measurement accuracy by enhancing the resolution of the small fluorescence peak of interest while supressing the impact of the inevitably large close-by elastic peak in an XAS scan. Comparative tests with ultra-dilute nickel nitrate solutions shows a two-fold improvement in dilution for a given performance using Xspress4 compared with previous results. Finally the characteristics of the spectra delivered by Xpress4 do not depend as heavily as before on the counting rate, which helps in improving the data quality of absorption spectroscopy beam lines when the beam intensity changes during a scan.

In parallel to the development of the Xpress4 DPP some work has been carried out to characterise a demonstrator of a 19-channel monolithic segmented HPGe detector with a pad size smaller than the pad size currently in use at Diamond. The pads of the demonstrator have hexagonal shape with 1 mm apothem as opposed to the current ones that have square shape with 5 mm side. The demonstrator has been equipped with Cube preamplifiers1 and it was tested twice at Diamond’s Test beamline (B16). The results of the first beam test has been published at the IEEE 2017 Nuclear Science Symposium2. During the second beam test the demonstrator was read-out with Xpress4, a very preliminary analysis shows that the detector can operate at about 1 million counts per second per channel with a count loss less than 13% that is still in the usable range of X-ray Absorption Fine Structure (XAFS) experiments. The energy resolution is still sub-optimal because in order to keep the cost within the budget the optimisation of all the parts was not pursued; however the tests done give a clear indication that this is a possible path to enhance the throughput of the fluorescence systems for XAFS by increasing the segmentation of the detector and by enhancing the counting rate of the individual channels.

Figure 3: Chassis of the Tristan 1M with all the electronics installed. The system currently drives a single Timepix3 chip connected to the rest of the electronics through an adapter board and a cable as shown in the picture.

The Tristan project aims at building a detector for time resolved experiments based on the Timepix3 chip³. Sixteen Timepix3 chips are bonded to a monolithic pixelated silicon sensor with more than a million pixels. The Tristan project is being carried out in two phases: the first phase is the development of a prototype with a single sensor (Tristan 1M), which serves as pilot detector and the second phase delivers a detector with ten sensors (Tristan 10M), which will then be installed on the Small-Molecule Single-Crystal Diffraction beamline (I19).

Most of the parts of Tristan 1M have been developed and assembled. The detector is undergoing tests with a single chip connected to the rest of the electronics through an adapter board (Fig. 3). The front end printed circuit board to which the Timepix3 chips are wire bonded and routes the signals to the back end electronics is currently under manufacturing. When this printed circuit board is delivered the Tristan 1M will be able to be tested in full. In parallel the mechanics of the Tristan 10M has been designed and the parts are currently being manufactured. The timescale for the completion of the project is to deliver a working Tristan 1M during the summer 2018 and then to assemble the Tristan 10M in autumn 2018 with the target to install it on the I19 during the winter of 2018/19.

1. XG LAB, www.xglab.it/low-noise-preamplifier.shtml, (2018).
2. Tartoni, N. et al. High channel density germanium detector demonstrator for high throughput X-ray spectroscopy. IEEE-NSS Conference Proceedings (2017).
3. Poikela, T. et al. Timepix3: a 65K channel hybrid pixel readout chip withsimultaneous ToA/ToT and sparse readout. Journal of Instrumentation 9, 5, (2014).