Bragg Crystal Spectrometer (BCS)
    2.1  Instrument Characteristics
        2.1.1  Wavelength Range
        2.1.2  Data Compression
    2.2  Instrument Calibration
        2.2.1  Sensitivity Functions
        2.2.2  Wavelength Resolution
    2.3  Detector Performance
        2.3.1  Detector Gain and Energy Resolution
        2.3.2  Position Resolution
        2.3.3  Digitization Spikes
    2.4  Detector Background
        2.4.1  Energy Discrimination
        2.4.2  Particle Background
        2.4.3  Germanium Fluorescence
        2.4.4  Effect of Gain Depression
    2.5  Dynamic Range and Count Rate Effects
        2.5.1  Deadtime Correction
        2.5.2  Saturation
        2.5.3  Gain Depression
        2.5.4  Rate-dependent Distortion
    2.6  Data Features
        2.6.1  Queued Data
        2.6.2  Timing of Spectra
        2.6.3  Last Spectral Bin
        2.6.4  Medium-rate Spectral Data in Autumn 1991
        2.6.5  Single Event Upsets
    2.7  Contact Persons
    2.8  References


2  Bragg Crystal Spectrometer (BCS)

The Bragg Crystal Spectrometer (BCS) is designed to study plasma heating and dynamics during the impulsive phase of solar flares. It consists of two bent crystal spectrometers, BCS-A and BCS-B, that observe the H-like line complex of Fe XXVI, and He-like complexes of Fe XXV, Ca XIX and S XV. Each spectrometer consists of a double detector placed behind a pair of germanium crystals that diffract the incoming X-rays into the detectors. The two structures are mounted at the front of the Yohkoh spacecraft, on either side of the central panel, with a thermal filter mounted in front of each spectrometer. The crystal dispersion axes are oriented north-south.

There is much more about the performance of the BCS instrument than is reported here. If additional information on the design and performance of the BCS is needed, the user is encouraged to consult the documents listed in Section 2.8.

2.1  Instrument Characteristics

2.1.1  Wavelength Range

The designed and laboratory measured wavelength ranges are given in the table below:

Chan    Designed Wavelength     Bin Range      Lab. Measured Wavelength
 1       1.7636 -- 1.8044       212 -- 28        1.7597 -- 1.8121
 2       1.8298 -- 1.8942       224 -- 36        1.8284 -- 1.8957
 3       3.1631 -- 3.1912       27 -- 229        3.1633 -- 3.1933
 4       5.0160 -- 5.1143       40 -- 234        5.0163 -- 5.1143

It should be stressed that these wavelengths are only valid at the designed, nominal BCS boresight. Because of the nature of the way a bent crystal spectrometer works, if a source is observed at other than the boresight then the wavelength range observed by the detector is shifted. In order to allow the SXT to observe more of the northern polar corona, the Yohkoh spacecraft is normally pointed several arcmins north of its designed pointing; because the dispersion axes of the crystals are oriented in the north-south direction, with this offset pointing channels 1 and 2 are normally seeing wavelengths that are slightly shorter than designed, and channels 3 and 4 slightly longer than designed. Consequently, during times of the year when the north pole of the Sun is tilted towards the Earth (i.e. Oct-Jan), if the active region is in the southern solar hemisphere, then this offset pointing severely compromises the ability of the BCS to observe the blue-wing in Ca XIX (channel 3).

The bin shifts caused by an offset in the pointing were measured in the laboratory to be (6.634±0.039) 10-2 bins/arcsec, (4.730±0.025) 10-2 bins/arcsec, (8.218±0.062) 10-2 bins/arcsec and (4.305±0.028) 10-2 bins/arcsec for channels 1 to 4 respectively (note: these are per single bin - channels 3 and 4 are often double binned). These values are being confirmed by in-orbit measurements which compare the bin of the resonance line with the source position observed by SXT. The above values correspond to 15.074, 21.142, 12.168 and 23.229 arcsec/bin.

2.1.2  Data Compression

fig_bcs_decomp.gif

Figure 2.1: The BCS data compression algorithm

All the BCS spectral data are output to telemetry as compressed counts. The compression is done by a hardware lookup table held in ROM (read-only memory) using a compression scheme that minimizes the errors. It is a complex function and is shown in Fig 2.1.

2.2  Instrument Calibration

The BCS instrument was calibrated at the Rutherford Appleton Laboratory (RAL). This work is described in Lang et al., 1993.

2.2.1  Sensitivity Functions

The response of a spectrometer channel depends on the transmission of the detector window (as a function of its length), the detector linearity, the crystal curvature and the transmission of the thermal filter. All three of these factors were independently measured during construction of the BCS and an all-up check was then performed at RAL. The response files for each of the channels are in the directory defined by ``$DIR_BCS_CAL'' under the names BCSA1.nnn, BCSA2.nnn, BCSB3.nnn and BCSB4.nnn, where nnn is the file version number. The sensitivity functions contained in these files are listed in tables in an appendix of Lang et al., 1993.

These functions represent the departure from the nominal efficiency, or effective area of the spectrometers. The measured effective areas, with uncertainties, for the flight wavelength ranges are 0.104±0.009 cm2, 0.114±0.012 cm2, 0.303±0.032 cm2 and 0.071±0.010 cm2 for channels 1 to 4 respectively when the relative response within a channel is taken to be uniform.

2.2.2  Wavelength Resolution

The wavelength resolution for a particular BCS channel depends on the rocking curve of the crystal for that channel, on the position resolution of the detector, and on the angle that the photons enter the detector in that channel. The latter point is more relevant to the higher energy photons of channels 1 and 2 - these photons penetrate further into the detector volume and if their path is not perpendicular (to the window) they will produce an electron cloud that spans a longer length of the wedge-and-wedge cathode, and hence a broader positional distribution.

2.3  Detector Performance

2.3.1  Detector Gain and Energy Resolution

The gain and energy resolution of the detectors are dependent on the gas mixture and the high-voltage (HV) setting. The detector performance is regularly checked using a Fe55 radioactive source. There are 8 setting of the high voltage at  30 volt intervals that can be used to compensate for changes in the detector gain due to contamination (gain decreases) or leakage (gain increases) - each step represents a change of a factor of ~1.28 in gain. The trims on both HV units were set to the nominal value of 4 (1476 V) at launch. Although both detectors are showing a small, steady decrease in gain (1.95% p.a. in BCS-A and 0.92% p.a. in BCS-B), it has not yet been necessary (Jan/94) to change the HV trim setting. The history of the HV trim settings are contained in file ``$DIR_BCS_LOGS/hv_log.dat'' .

During the radioactive source calibration, in order to make it possible to detect the 5.9 KeV photons emitted by the Fe55 source, the gain of the detector for BCS-B is changed by setting an HV trim level of 2 (1418 V).

The results of the PHA (pulse-height analyzed) pre- and post-launch detector calibrations performed with a radioactive source are contained in the file ``$DIR_BCS_LOGS/gain_log.dat.'' A history of the the gain and energy resolution of the two BCS detectors can be obtained by typing the IDL command:
 
IDL >  .run gain_plot4
The variation of the gain with temperature (determined from measurements made during the pre-launch Thermal Test) can be obtained by typing the IDL command:
 
IDL >  .run gain_temp_plot

2.3.2  Position Resolution

The position resolution of channels 1 and 2 are approximately 350 mm, and channels 3 and 4 somewhat larger.

The wedge-and-wedge pattern of the detector can be stimulated through a set of three pads located on the rear of the quartz plate. When a signal is applied to a pad, it induces a charge on the pattern in the same way that an incoming photon would, but at a known location. During a stim test, each pad is stimulated in turn. The position resolution and linearity of the detectors are regularly measured using these electronic stims.

2.3.3  Digitization Spikes

The position of a photon in the detector is calculated from the signals measured on the two wedges (Wa and Wb) of the wedge-and-wedge cathode. The position is given by the formula: x = [(Wa)/( Wa + Wb)] In the BCS this calculation is done using a look-up table held in ROM. Because of the quantized nature of this digital technique, solutions sometimes fall either side of the value that would have been determined with an analogue technique and spikes and dips are seen in what should be a flat field. The position of these can be calculated, and corrected - there is some dependence on the energy of the detected photon. No photons are lost, they are merely displaced into adjacent bins so integrations under curves are not affected, but when fitting spectra it is best to avoid those bins.

2.4  Detector Background

2.4.1  Energy Discrimination

To allow the rejection of photons whose energies lie outside the desired range, a pair of adjustable energy discriminators (single channel analyzers, or SCA's) are provided for each channel of the BCS.

The history of the SCA settings are contained in file ``$DIR_BCS_LOGS/sca_log.dat''.

2.4.2  Particle Background

The BCS detectors are effected by a background caused by particles trapped in the Earth's magnetic field. Many of these are rejected by the setting of the lower level SCA, but occasionally some are seen in channels 1 and 2 - they appear as a flare-like spike, but are not seen at the corresponding time in channels 3 and 4.

Over the South Atlantic Anomaly, the background caused by trapped particles is particularly intense and the High Voltage supplies to the BCS detectors are switched off (using time-tagged commands).

2.4.3  Germanium Fluorescence

When germanium is illuminated by photons whose energy exceeds 11.2 keV, the crystal fluoresces, emitting a photon whose energy is 9.9 keV. Since the energy resolution of the BCS detectors is around 20%, the tail of the distribution of these fluorescent photons falls within the energy range that the BCS is sensitive to and indeed part of the tail lies under the distribution of the photons detected by channels 1 and 2. In order to try to reject the fluorescence, the upper level discriminator is set midway between the two energy distributions. germanium fluorescence is less of a problem in channels 3 and 4 since there is a greater difference in the energy between the channel photons and the fluorescent photons.

2.4.4  Effect of Gain Depression

When the observed count rate increases the gain of the detector becomes depressed and the position that the photons from a given channel are recorded at (in SCA space) slips to lower values. As a consequence, the perceived energy distribution of the photons can move relative to the energy window defined by the lower and upper level discriminators and some photons may be rejected as being out of limits. This effect has made it difficult to reject the photons caused by Ge fluorescence when the count rates are high.

2.5  Dynamic Range and Count Rate Effects

The dynamic range of a proportional counter of the type used on the BCS is limited and it is not possible to observe both very weak and very intense flares. The BCS on Yohkoh was designed to have high sensitivity so that it can observe as early as possible at the onset of a flare and the dynamic range is therefore biased towards the smaller flares. As a consequence the BCS saturates during larger flares, i.e at high count rates. During saturation all positional information is lost but the total event counter may still be valid.

2.5.1  Deadtime Correction

All analogue signals produced by the detectors take a time to process that is significant when count rates are high. Although the anode signals only take 1 ms to process, the processing of the wedge signals takes 3.5 ms and additional time needs to be allowed for the signal levels to settle back to zero so that in total 35 ms must elapse between successive photons. Consequently, during this ``deadtime'' the processing of any new position signals is inhibited. At low count rates this effect is small, but it becomes more important as the count rate increases. If the photon arrival times were evenly spaced, then the pair of channels in one detector could see a maximum of ~28×104 counts/sec. However, in practice the random nature of the events, and the onset of saturation, results in a maximum of less than half this.

In the BCS there are three sets of counters for each channel: the total event counter, the limited (or in-window) event counter and the encoded event counter. The gate for the limited event counter only accepts events that fall within the allowed energy window, and in addition inhibits any event that arrives less than 35 ms after the previous one. From a knowledge of the event counters, it is possible to calculate the effects of the deadtime and hence reconstruct what the count rate should be. This is done within the MK_BSC and MKBSD routines.

It should be noted that while the event counters used for deadtime corrections are in the DP-synchronous (DP_SYNC) part the telemetry, the spectral data are in the PH stream and after passing through the queue memory are asynchronous. As a consequence, if there are holes in the data coverage, it is possible for the DP_SYNC data to be absent at the time of the spectral observations. At such times it is not possible to make a deadtime correction.

2.5.2  Saturation

The problem of saturation is exacerbated by the double detectors used on the BCS. During construction, constraints on mass and volume required that the two channels on a spectrometer shared the same detector gas volume. Although the anode signals from the detectors (which allow energy discrimination) are processed separately for each channel, the positional information is provided by a wedge-and-wedge pattern that is shared by the channels. Thus, the processing deadtime for position encoding for a detector is affected by the sum of the total count rates for both channels of that detector.

That the onset of saturation is determined by the counts in two channels can produce ``interesting'' effects. For instance, if the flare is hot, the detector of spectrometer BCS-B will saturate at a higher count rate in Ca XIX (channel 3) than if it is cool. Normally the count rate in S XV (channel 4) will dominate in BCS-B, but for a hot flare the saturation will depend more on the Ca XIX than on the S XV count rate. In spectrometer BCS-A, Fe XXVI (channel 1) has only been observed in a few flares because often the count rate in Fe XXV (channel 2) is already very high by the time that Fe XXVI is likely to be observed.

It should be noted that adjusting the SCA values does not affect the onset of saturation. The same number of photons are still entering the detector and each is seen by the positional encoding circuitry no matter how many are gated as having valid energies.

2.5.3  Gain Depression

As mentioned under the ``Detector Background'' section, when the count rate is high, the gain of the detector becomes depressed and the perceived energy distribution slides down in the SCA space, and possibly partly or totally out of the energy window set by the SCAs. As a consequence, the number of photons accepted as having valid energy is reduced and since this signal is used to determine whether the decoded position should be used, the number of encoded photons is reduced.

2.5.4  Rate-dependent Distortion

When a large number of counts are put into a small length of the detector volume, a space charge effect develops which attracts the electron cloud towards that region of the detector. This results in lines that are taller and narrower than they should be and locally distorts the linearity of the detector. This effect is not confined to spectral lines and can occur in any region of the detector where the photon flux is large; the total number of counts is conserved, but there is a loss of positional information. It is not clear that this effect can be deconvolved, and the matter is still being studied. More information is given in Trow, Bento and Smith (1993).

2.6  Data Features

2.6.1  Queued Data

The BCS spectral data are stored in a queue memory prior to insertion in the telemetry frame. These data are therefore asynchronous to the data in the DP synchronous area of telemetry and occasionally the DP_SYNC data are not present at the same times as the spectral data. During these periods, deadtime correction to the spectral data are not possible. Key items relating to the mode being executed by the BCS are included in header information included in the queued data and it is always possible to know what mode the BCS is in.

At times a mode called ``Fast Queue'' is used. In this the data accumulators are sampled at intervals without clearing and then the data are stored in two different queue - normal and fast. The final integration before clearing of the accumulator is stored in the normal queue and all others are stored in the fast queue. In this mode, if the flare-flag is raised the fast queue is dumped to telemetry, but normally only data from the normal queue goes into telemetry. The mode is used at times when the spacecraft is in a mode whose bit-rate does not allow as short an accumulation time as is desired. The fast queue allows high cadence data to be stored and only output if a flare switches the spacecraft to a higher bit-rate.

Fast Queue data can be recognized by a sawtooth-type ramping in the light curve. These type of data are normalized by the IDL function BCS_NORM.

2.6.2  Timing of Spectra

The time contained in a mode record for a BCS spectrum represents the start of the integration time of that spectrum. Thus a correction (equal to half the integration period) needs to be added to the points in a light curves for all plots that are NOT in histogram mode (psym=10). The routine BCS_TVEC does this.

2.6.3  Last Spectral Bin

The last bin in each channel is used to accumulate positionally encoded counts when the sum of the signals on the two wedge, or the signal on any individual wedge is very small or very large. Although useful as a diagnostic, these bins should ignored for analysis purposes.

2.6.4  Medium-rate Spectral Data in Autumn 1991

Just after the launch of Yohkoh, it was found that there was an error in the way the DP inserted BCS PH-data into the telemetry frame during medium bit-rate. This can be seen a ``hole'' that steadily marches through the data in successive spectral integrations; when viewing the light curve, a periodic, but structureless reduction in the count rate is seen in medium bit rate. When the hole is not in an interesting part of the spectrum, the data can be used, otherwise it should be rejected. The problem was corrected by a patch to the DP in late November 1991.

2.6.5  Single Event Upsets

At intervals, and particularly when the spacecraft has its apogee over the South Atlantic, the memory of the BCS microprocessor can be corrupted by a Single Event Upset (SEU). Often this affects a part of the memory that does not disturb the running of the BCS, but whenever an SEU is detected (most are flagged as a result of a checksum calculation), the BCS microprocessor is rebooted. Occasionally, several hours, or days of data have been lost before this is done.

2.7  Contact Persons

If you have any question about BCS software and/or instrument, or if you need any advice in BCS data analyses, contact:

                rdb@mssl.ucl.ac.uk               (Bentley)
                mariska@aspen.nrl.navy.mil
                zarro@smmdac.gsfc.nasa.gov

2.8   References


Converted at the YDAC on Apr 14, 1999
(from LaTEX using TTH, version 1.92, with postprocessing)