Introduction to the RESIK detectors

Matthew Whyndham, MSSL-UCL

8 March 2004

Main features | Mechanical construction | Windows and internal structures
Event position encoding | Computation of Fixed Pattern Structure

The Coronas-F RESIK instrument uses two detectors (position sensitive proportional counters) to detect soft X-rays reflected from the four spectrometer crystals. The detectors are identical to those used in the Yohkoh BCS instrument. The detectors were designed and manufactured by MSSL.

Main features

The BCS/RESIK detector is a proportional counter with a one dimensional position sensitive readout. It is optimised for detection of X-rays in the 2-7 keV energy range, and for use in a space environment. It features low mass, small volume, low power consumption, and long life. Preamplifiers and test circuitry are mounted in a package on the rear of the detector. No gas supply system is required, as the detector is sealed permanently.

The gas filling is an equal Xenon-Argon mixture with 5% CO2 as a quench agent. The pressure of the gas filling is 1.2 atmospheres at 20 °C. This mixture of gases was chosen to provide the required quantum efficiency, signal characteristics and ageing properties.

The detector requires an external high voltage supply to establish the anode wire fields. When a photon is detected it emits shaped analog pulses which require further processing to yield the event positions. The detector has two anode channels so it can be used to observe two sources simultaneously.

The detector is essentially a hollow stainless steel rectangular box, with a thin (transparent to X-rays) metal foil forming one of the larger faces. Thin metal wires are suspended along the centre of the box, and high voltage is applied to these wires (the anodes). Other electrodes (cathodes) in the detector, including the position readout are held at ground potential. The configuration of the electrostatic field defined in this way is quite complex, but in the vicinity of the anodes it is very similar to that of a single-anode, coaxial cylindrical proportional counter. Near the window, the field lines are more parallel than coaxial.

The detector’s basic operation is like any other proportional counter. X-rays penetrate the window and interact with atoms in the gas. Soft X-rays (1-10 keV) can travel a few mm in the gas on average before being absorbed. Each absorbed photon ionises a few hundred gas atoms on average, the number being approximately proportional to the energy of the photon. These “primary” electrons drift toward the anodes under the influence of their electric field. The field strength increases as the electrons get closer to the wire, and so the energy of the electrons increases. Eventually the energy of the electrons is sufficient to cause further ionisation, and the resulting electrons quickly acquire enough energy to cause ionisations themselves. This multiplication process (the avalanche) produces enough charge for a signal to be detected on the anode wire.

This type of detector is called a proportional counter because the total quantity of charge in each avalanche is, on average, proportional to the photon energy, although the size of the avalanches is subject to random fluctuations. A charge sensitive preamplifier is connected to the anode and the number and approximate energy of incident photons can be determined.

The position sensing capability of the detector is accomplished by the Backgammon technique (so called because configuration of the readout cathode, interleaved wedges, resembles a Backgammon game board, or Jeu de Jacquet). When an avalanche occurs at the anode, charges are also induced on the readout – the amount of charge on each of the two electrodes depends on the position of the event – and these induced signals are used to determine the position, in one dimension, of the event.

See below for further details on the encoding system.

Mechanical construction

An overall view of the detector package is shown above.

The detector is mounted in the spectrometer by means of attachment points on the upper (window) side. In its flight configuration, an electronics package is integrated with the detector, and this package has an alloy enclosure and cable harnesses. A multi-wire harness terminated in a 25 way D type connector carries all power and signals except for the anode high voltage, which is fed via a separate cable terminated in a Reynolds screw connector.

It is permanently sealed by means of a weld around the seam where the base joins the window frame. Of these two subassemblies, the base is the more complex. It comprises: the electrical feedthroughs; mounting arrangements for the anode and cathode wires; and slots to hold the cathode readout plate in position. The base forms one face of the box that makes up the detector.

The window frame forms the other five sides of the box, with the window foil itself making up most of the face opposite the base. The window frame has a recess into which the window is bonded, and an attachment point for the gas filling tube. The four detector mounting points are also part of the window frame.

The detector is fixed to the spectrometer by means of four “feet”, which are part of the window assembly. A vibration-damping mounting system is used, consisting of small o-rings and continuity straps. This isolates the detector from vibrations and the structure (at launch and at other times during assembly and integration) whilst preserving alignment and maintaining electrical continuity.

This photo shows a detector mounted in the RESIK spectrometer.

Windows and internal structures

X-rays enter the detector through its window, which is a single piece of 125 mm Beryllium foil brazed to the window frame. Although the window is securely attached to the window frame, the strength of the Beryllium foil is insufficient to support the outward pressure, and therefore the window frame has a number of apertures cut into it, leaving bars arranged along the length of the detector. These are stiff enough, and closely enough spaced, to fully support the window. These bars reduce the active area of the detector. Because they are parallel to the position sensing axis of the system, the presence of these bars has no other effects in the spectra.

There are seven electrical feedthroughs at the rear of the detector. Each pair of anodes has a feedthrough, shown on the right hand side of the sectional view in drawing A1-5219-308. The two parts of the readout cathode are connected to feedthroughs located at opposite corners of the base. There are three “stim” connections, approximately equally-spaced along the centre line but offset to allow for the strengthening members positioned along the centres of both axes.

The anodes are connected together in pairs, by conductors (inside the detector, but outside of the internal end walls). The anodes are 15 mm diameter (0.6 thou’) alloy wires (92 % Pt; 8 % W).

Reproduce this figure: c:\user\matt\thesis\xsect.cdr

Figure 2. Cross-section through the detector showing the relative positions of the anodes and cathodes.

There are four anode wires, nine cathode wires, and other cathode surfaces.

The detector volume is divided into two cells by a set of cathode wires. These wires are equally spaced along a line which divides the cross-section of the detector into two equal parts. The figure shows the arrangement of these nine wires, which are 25 mm in diameter.

The effect of these cathodes on the electric field is the same as if a solid conductor were to divide the detector. The potential is zero at the cathodes, and so the field configuration is symmetrical. The counter thus functions as two separate detectors.

The position sensing function of the detector is provided by the wedge and wedge cathode plate. This is a fused silica (quartz) plate, 2 mm thick, with a thin gold layer deposited onto it. A continuous line is etched into the conducting layer, dividing it into two electrodes (the wedges). When the detector is operated, a charge distribution is induced onto this cathode whenever there is an avalanche on the nearby anode wires. The wedges act to divide the quantity of induced charge according to the position of the avalanche. The signals from the wedges are processed to determine the position of each event.

Three circular metal pads have been deposited on the rear of the read-out plate. The function of the pads is to provide a means of testing the wedge electronics without having to apply high voltage to the unit. It also allows for rudimentary calibration of the position-encoding system. A stim pulse (tail pulse) is switched between each of the stim pads in turn (this is handled by the FEE package), with either anode preamplifier also stimmed with an inverted pulse. An externally generated stim clock controls the switching.

The voltage pulse on the pads induces a signal on the wedges (there is a capacitive coupling with the wedge electrodes). Because the pads are large enough to cover a few pitches of the readout, the size of the induced signals on the two electrodes depends on the position in much the same way does the signal from an avalanche (photon event) on the anode wire. When processed by the position decoding system, three peaks are seen in the position data, corresponding to the position of the pads.

Event position encoding

The positioning encoding system establishes a bin allocation for detector photon event based on its physical location. An avalanche arriving at the anode wire at a particular location causes induced charges to appear on the wedge electrodes. This is picked up as event signals in channels A and B which are digitised by the (8-bit) ADC's. The ADC data is then routed through the look-up table (LUT) to produce the encoded position datum.

In the Yohkoh BCS, position processing occurs in the unit “BCS-E”. In RESIK, the spectrometer electronics unit carried out this function, using identical circuit designs (although the LUT contents differed).

Implementation of the function in a fixed look at table, where the result is determined solely by memory content at an address determined by the values of the operands, is advantageous in space-engineering terms. At the time of the design of this encoding circuit for Yohkoh BCS, it was the only practical route given the prevailing performance, engineering and cost constraints.

However, this method of encoding positions does have some drawbacks. Mainly these are related to the degree of uniformity seen in the encoded data. At certain positions, it appears as if there are spikes and notches in what should otherwise be a flat or smoothly varying pattern of illumination. This is most pronounced at the centre of the readout, and is clearly visible at approximately one quarter and three quarters of the distance along it, and also results in less pronounced nonuniformities elsewhere.

These are seen both in the RESIK data and formerly in the BCS data (the Yohkoh Analysis Guide makes reference to these effects).

The main “Notches” are easily seen in the data set below, which is a night-time exposure with no filtering or post-processing. The energy deposited in the detector is due to orbital background counts, whose spatial distribution is uniform over the detector.

These nonuniformities are due to the way in which ADC output data is processed by the instrument position-encoding lookup table. This is because the lookup table implements a digital division operation at a fixed arithmetic precision. The attributes that are seen are not present if the same operation is carried out in floating point arithmetic (or indeed at suitably higher integer precision).

The BCS/RESIK lookup table stores, essentially, the result of the following expression in the lookup table address (A, B):

256 . A / (A+B)

If the result is not exact, then the nearest integer value is used.

The nature of the resulting transfer function, from real position to encoded position, is not smoothly varying. The most important determining parameter, apart from the position of the event, is the total pulse height (A+B). At small values of pulse height the adjacent values in the look at table may differ by values greater than one. Here the encoded precision is less than the optimum. At larger values of pulse height, adjacent entries in the lookup table may be identical. Along a particular locus, where the pulse height is exactly 256, the encoded position changes by exactly one at each available address location. This locus is the main diagonal from (0, 255) to (255, 0) on the map of the lookup table represented in the figure below.

When we consider this map, representing the grid of available ADC coordinate pairs, we can appreciate the origin of the “notches”. It resembles the optical effect seen when driving near a field of regularly spaced plants, military gravestones or other objects (e.g. this installation). Along certain directions of view, clear corridors seem to open up. In other directions we perceive an unbroken mass of objects. The angles along which we see a lower apparent density of objects are those where there is close alignment of our view angle and the symmetry of the grid. These are analogous to the locations of spikes and notches in the spectrum data.

The effect has been called The Spike or The Notch, but more properly it should be called Fixed Pattern Structure or Flat Field Structure since the effect can influence any bin in the data and not only the main “directions” along the LUT map. It should be noted that the structure seen will depend on the locations being sampled by the ADC-data, and is hence data-dependant. Technically it is not a true Fixed Pattern, and cannot truly be removed by scaling by a constant flat field. However, for practical comparison of spectra, a convenient approximation can be made to such a scaling. This is possible since the appearance of the structure is relatively stable over a range of similar observations. In order to apply this scaling in an analytical context (i.e. quantitative analysis of spectrum line properties), the uncertainties inherent in the detector data and any correction methods should be considered.

Computation of Fixed Pattern Structure

A simulation technique has been followed successfully to reproduce the fixed pattern details. This work was done by Phillips (MSSL, 199?) and Sylwester (SRC, 2004), respectively, for the Yohkoh BCS and RESIK instrumentation setups. This section outlines the general approach.

The centre point of the main diagonal (and all other diagonals parallel to it) corresponds to detector events occurring at the midpoint of the position readout. The distance of the event from the origin (bottom left) in this map, corresponds to the event of pulse height (A+B).

Since the detector's pulse height distribution (seen as the instrument PHA data) is broad (about 20% for the typical RESIK photon energy), then at any given position on the detector, a certain length of particular radial spoke in the lookup table map will be involved in encoding the position of the events. The frequency of sampling a LUT coordinate is determined by a weight given by the pulse height distribution.

Note that in both the BCS and in RESIK a single channel discriminator is applied to the event of pulse height before acceptance of the event data into the encoding stream. Therefore a limited swathe (or stripe) of the table will be used in any observational setting, determined by the upper-level and lower-level discriminator settings.

A theoretical “flat field” for a given set of experimental conditions can be generated from the following inputs:

  1. discriminator settings
  2. detector HV
  3. pulse height distribution, including any fluorescence
  4. relationship between ADC values, discriminator setting and PHA scale
  5. LUT scaling parameter (compression of occulted bins)

The resulting function can be used to invert the observed data to mask the effect, by scaling, of the division non-uniformities. It should be noted that there are uncertainties involved in the determination of all the above parameters, and in the encoding process itself, and that any such inversion will introduce a finite additional uncertainty into the processed data.

It should also be noted that the mechanism of the effect is not a scaling law (like a sensitivity change) but a redistribution of counts from one location, inaccessible due to the finite table, to another. In general no counts are lost due to the encoding process, and any re-scaling function should maintain this property.

Matthew Whyndham : Other RESIK documents