Expansion of the detector bank of the Necsa neutron powder diffractometer

1.1.PITSI at SAFARI-1

The South African Fundamental Atomic Research Installation, SAFARI-1, is a 20 MW tank-in-pool type materials testing/research reactor operated by the South African Nuclear Energy Corporation (Necsa) SOC Limited. In addition to the production of radio isotopes, SAFARI-1 offers various irradiation services, neutron activation analysis, as well as hosts a number of neutron beam-line instruments [5]. The Powder Instrument for Transition in Structure Investigations (PITSI) [6] was commissioned in mid-2014 and has since provided the South African research community with a unique capability to perform temperature dependent angular dispersive neutron powder diffraction studies. It supports both low and high temperature sample environments to facilitate in-situ parametric studies of crystallographic and magnetic phenomena covering the temperature range 1.5 to 1800 K. Its flexibility in the choice of the monochromator reflections, focus conditions and take-off angles allow selection of instrumental characteristics that suit investigation over a wide range of scientific problems.

1.2.Detector bank

The PITSI detector assembly employs 15 vertically stacked Reuter-Stokes model RS-P4-0824-208 linear position sensitive ³He-filled tubes to cover an effective active area of ∼610×375 mm2 and subtends a 2θ angle of ∼20∘ at a sample-to-detector distance of 1.6 m. The detector tube properties are given in Table 1. The linear resolution of the bank is 2 mm (H)×25 mm (V) which corresponds to diffraction resolution of Δd/d=∼3×10−3. A full diffraction pattern that covers the diffraction angle range 10∘⩽2θ⩽115∘ is acquired in 6 independent partially overlapped datasets by stepped-scanning an angle slightly smaller than the angle subtended. The data is rebinned into equal steps and the overlapped regions of adjacent frames stitched together as part of the post-processing protocol. As an example of the data acquisition protocol, should a data set be taken by employing a data acquisition time of 3 hours, the time required to acquire a complete diffraction pattern amounts to 18 hours.

1.3.Data acquisition and reduction

The analogue signals from each detector tube are processed with an Instrumentation Associate’s (IA) Position Encoding Module (PEM) through a pre-amplifier [1]. Each PEM fits in a standard NIM crate slot and communicates with the host computer via a USB interface. The PITSI detector bank therefore requires 15 PEMS and 15 USB cables. The PEMs operating parameters are configured via the IA software PSDControl which calibrates each detector using data of isotropically scattered neutrons transmitted through a cadmium mask with regular vertical slits that is positioned in front of the detector array. Calibrated data is requested from the Ansto Histogram Memory Server (HM) via a TCP/IP protocol which provides it to the SINQ Instrument Control Software (SICS) [3]. The spatially scattered data is stored together with a complete list of instrument parameters in a NeXus data file. The in-house developed ScanManipulator [4] data processing software is used to apply various data correction algorithms as shown in Fig. 1 and amongst others outputs an intensity vs. 2θ data file which can be imported into Rietveld refinement programs for further analysis.

Fig. 1.

Illustration of the main neutron optical components of PITSI and data processing steps required to obtain a complete diffraction pattern.

1.4.Beam time optimisation

Improvement of the overall instrument efficiency has become a priority to accommodate a steadily growing user base, as well as a requirement for the investigation of dynamic effects that occur in crystallographic and magnetic transitions. The latter requires analysis of full diffraction patterns, or a wide range of diffraction peaks, simultaneously in an ideal situation. Notwithstanding the instrument rendering medium resolution diffraction data quality, it has become a priority to reduce the long measurement times currently required with investigations. Through the implementation of a large double focused monochromator the flux at the sample position is already maximised. An outstanding contributor that can be readily improved is to enlarge the detector subtention angle, whilst retaining the highest possible capture efficiency at each diffraction angle.

2.1.Multi bank detector array

Various expansion concepts have been explored with the aim at minimising cost and implementation effort whilst maximising the detector active area and versatility. The configuration shown in Fig. 2 was selected as it incorporates the existing detector elements and signal processing, thereby minimising costs. The complete detector array will consist of three banks each containing 16 vertically stacked tubes (48 in total) which in conjunction with an adjustable sample-to-detector distance, between 1.2 m and 1.6 m, will provide an additional capability to either select highest peak intensity, or highest angular resolution. The incorporation of an additional detector in the bank height is a logical expansion since each (new) analogue electronic module can process 8 detectors. These expansions will thus increase the active area of each bank to ∼610×400 mm2 and the subtended angle at 1.6 m to be ∼3×20∘. A dead space of 18.5° (at 1.6 m) between banks is covered by one stepped-scan of the array. An equivalent diffraction pattern would therefore be acquired with only 2 datasets. The increased active detector area will result in an acquisition time improvement factor of 3.2, which translates to a total saving of 12.4 hours per diffraction pattern with reference to discussed example. The extent of the dead space can be minimised by reducing the sample-to-detector distance.

Fig. 2.

Illustration of the new expanded PITSI detector array respectively positioned at 1.2 m and 1.6 m from the sample position.

2.2.Data acquisition electronics

Adding an additional 33 detector elements to the current electronic system would require an additional 33 PEMs and 33 USB cables to be installed, i.e. 48 respectively in total. In order to reduce the cabling and physical size of the electronics, it was decided to replace the IA system with the Mesytec psd+ implementation [2]. Eight detector elements can be connected directly to a single Mesytec MPSD-8+ readout module using high voltage coaxial cables. The readout module contains a preamplifier and window discriminator with the gain and threshold values adjustable via computer control. Up to eight readout modules can in turn be connected to the data bus of the MCPD-8 data collector module using Lemo connections and communicates with the host computer through a single Ethernet cable. The complete PITSI detector implementation using the Mesytec system therefore requires only 7 NIM slots.

2.3.Software system

The HM receives raw event data directly from the Mesytec system via TCP/IP as 3×16 vertical channels each divided into 1024 horizontal bins (Fig. 3(a)). The data is then tiled and rotated to produce 16 horizontal and 3×1024 vertical channels (Fig. 3(b)). Pre-determined tube width and offset corrections are finally applied to produce a uniform dataset (Fig. 3(c)). Tube offsets are obtained by determining the individual centre positions of an isotropically flooded dataset and calculating the offsets from the theoretical detector centre. Tube width corrections are applied by re-binning the measured active width thereby stretching the data to fill the entire 1024 channels.

Fig. 3.

Simulated dataset of the expanded PITSI detector showing (a) raw, (b) tiled & rotated and (c) calibrated diffraction data.

As the dataset contains information from three individual detector banks, the set must be split into three frames, each representing a different 2θ range. The complete 2θ range then contains six frames obtained from two measurement steps (Fig. 4(a)). Once separated, the frames can be processed (Fig. 4(b)) as usual with ScanManipulator to produce the final diffraction pattern (Fig. 4(c)). It is envisaged to automate the dataset sectioning in ScanManipulator.

Fig. 4.

Illustrations of the sectioned datasets: (a) before and (b) after flat field & geometric corrections have been applied; (c) Final 1D diffraction pattern of an Al₂O₃ powder after corrected frames has been integrated, normalised and stitched.

3.1.Electronic noise

The electronic noise was defined as the total number of events registered by the detection system when the neutron beam was closed. It was assumed that background radiation levels are sufficiently low so as not to influence the measurement. Ideally, the threshold and gain values should be such that the electronic noise in the system are minimised. The optimised domain shown in Fig. 7(a) confirms that increased electronic noise is present with high gain and low threshold values.

3.2.Signal-to-noise ratio

A common measure in electronic optimisation is the signal-to-noise ratio. This was optimised (maximised) by equating the diffraction peak intensity to the required signal, and the background to the noise. The optimised domain is presented in Fig. 7(b).

3.3.Intensity

Even though the signal-to-noise ratio is a very important factor, the resulting optimisation may present a solution where both the intensity and background levels are very low, which will result in unrealistic long measurement times. To avoid this situation the domain of maximum peak intensity was identified as shown in Fig. 7(c).

3.4.Intensity vs. standard deviation of background

Overlaying the optimised domains up to this point would result in a relatively large viable domain. It was however noticed that the spread of measured background values varies drastically throughout this domain and the intensity vs. standard deviation of the background was therefore maximised as shown in Fig. 7(d).

3.5.Summary

The optimal setup parameters of the analogue electronics were identified from the convergence of the respective saddle points for each criteria under consideration in the overlaid representation shown in Fig. 7(e) as: Pre-amplifier gain=0.1; Discriminator threshold=42%.

Fig. 7.

Illustration of various evaluation parameters with respect to pre-amplifier gain and discriminator threshold values to determine the optimal range of the detector electronic system.