[About EXOSAT] [Archive] [Software] [Gallery] [Publications]

The ME Background

Although the general background rejection performance of the ME experiment is far better than expected, variability in the residual background finally limits the overall sensitivity of the instrument. A more detailed account of the analysis of the background will be provided elsewhere, however guidelines are presented here to provide observers with sufficient information for their analysis.

1. The background appears to be variable on many timescales but the following are the most important:

a. Flaring (1-10s or minutes).

Usually a rapid rise and slower decay. The strength may vary considerably between detectors. Observed most strongly in detectors B,C,F and H. Very heavily cut-off in Argon, (often not seen at all below channel 20).
Strongly correlated between detectors in the following way: A-E; B-F; C-G; D-H; thus Ql-Q3; Q2-Q4; Hl-H2 i.e. correlated across the H1/H2 line. For the best sensitivity to the detection of these flares light-curves should be produced for channels 25-60 in Argon, for each detector if possible. Do not use data from periods of time during which flaring has occurred.
For observations in offset configuration, true source flaring can be detected by the absence of the above correlation and by the softer spectra. For co-aligned observations, true source flares can be detected by their uniform brightness in detectors and by their spectra.

b. Slow changes (hours-days).
These changes appear to have a spectrum which peaks around channels 50-60 in Argon. and give an obvious error in background subtraction above channel 20. As with a. above the changes are largest with detectors B,C,F and G.
There is a weak correlation between detectors, however, the correlation coefficients provided on the CCF are something of an average and will probably not be significantly improved.

2. The Use of Slew data.
Ideally the slew data provides a background estimate just before and just after the observation. It is normally. a better estimate of the background than the correlated offset data. However slew data is often unsatisfactory for the following reasons:

Not enough time in the slews.
Other sources contaminating the slews.
Unaccountable background changes during the slew.
Only one slew.

Slew data should therefore be examined with caution and the Observatory cannot guarantee a priori quality.

3. The Argon energy range
The Ar detectors are operated at a gain such that there is no useful response above channel 70. For a Crab-like spectrum the most sensitive energy range of the Ar detectors 15 channels 6-24.
In this range the ME array in offset configuration would detect a 1 mCrab change In a faint source in 100 seconds at the 6 sigma level.

4. The use of Correlation coefficients.
Correlation coefficients have not proved particularly useful and furthermore introduce a considerable extra error in a background subtracted spectrum. For this reason they should only be used as a last resort.

5. Other background subtraction methods.
A number of other background subtraction methods have been considered; These may be classified as either 'Correlated' or 'Direct'.

Correlated:
Correlations between Argon-Xenon and between the high-low parts of either Argon or Xenon spectra have been used with mixed success. It seems that some variability in the shape of the background exists and none of the methods has been adopted universally. Note: this is not the case for the GSPC.

Direct:
A similar concept to the use of slew data is to obtain a background spectrum just before/after the source spectrum. This may be done by separate pointing to a blank field, -HOVER-, or by exchanging the offset halves, -NODDING-. The former has the great disadvantage that it uses extra manoeuvre gas and time. "Nodding' of the array halves was discontinued last autumn because of a component failure in one of the drive electronics, however careful use of the OBC time tag command facility does now allow operation of both drives and will be implemented for specific observations.

6. Presently adopted background subtraction method on specific targets
The following procedure leads to a good background subtracted spectrum:

  1. Slews to the source are executed in configuration, say H1 aligned and H2 offset.
  2. The first period of the observation, T1, has the same configuration.
  3. The second period of the observation, T2 = T1, has the array with H2 aligned and H1 offset.
  4. The slew away from the source is performed with the configuration unchanged.
  5. In the example below the offsets are such that, if H1 was positively offset, then H2 is negatively offset and vice-versa.
  6. For an observation of less than 3 units then T1 = T2 = half the observation.
  7. For an observation of greater than 3 units then an odd number of 'nods' are made such that T1 is never more than 1 unit.

    At present 'nodding' the array takes about 15 minutes.

  8. Let us consider an example:
  9. T1 H1 aligned H2 +ve offset.
    T2 H2 aligned H1 -ve offset.
    Let spectra in H1 be H1(1) and H1(2) etc.
    Then source spectrum for H1 = H1(1) - H1(2) - D(H1). Where D(H1) is the spectrum difference between aligned and offset positions for a blank field.
    The source spectrum for H2 = H2(2) - H2(1) - D(H2) and so the total spectrum

    W = H1(1)+H2(2)-H1(2)-H2(1)-D(H1)-D(H2)

Difference spectra have been determined from a number of observations of blank fields. It appears that the shape of the difference spectra is independent of array offset but that the magnitude is simply proportional to the offset. It differs for each detector, is greater for detectors B,C,F and G and for all detectors in the positive offset direction. For an array half it is typically 0.5 cts/second over the whole Argon range and less than 0.2 cts/s in the range 4-35 channels.
Should there be any variability in the difference spectra one might expect this to be correlated between halves and the method above would tend to cancel out such an effect.
Similarly, since the long-term variability is also correlated the extra background removed in one half will be compensated by the inadequate background subtraction in the other.

7. Sensitivities
The final sensitivity of the instrument depends upon the type and quality of background data available.
For the method outlined in 7. above one can expect to obtain a useful spectrum of a 1 mCrab source in 1 unit, (eg. photon index of -2.1 +/- 0.3). This should not be considered as any sort of lower limit.

8. Miscellaneous
Further information will be provided with the next AO, and in future issues of the Express.
Please let me know of your own experience and problems encountered in obtaining background subtracted ME spectra.

A. Smith
ESTEC/SSD



The GSPC Calibrations


1. Introduction

The Gas Scintillation Proportional Counters (GSPC) on EXOSAT and Tenma are the first such instruments to be flown on X-ray astronomy satellites. On EXOSAT the GSPC is performing well, although there have been a number of unexpected differences between the in-orbit performance and pre-launch predictions. This report contains an account of how the GSPC has been calibrated and the optimum method for analysing data. A description of the instrument can be found in Peacock et al. (ref. p.16).

2. Background Subtraction

In the 2-17 keV energy range, the background count rate is dominated by particles. It is constant around one EXOSAT orbit except for the few hours before and after entry into the van Allen belts. Background reduction is achieved by filtering out events with a very long or short UV photon burst rise time using the instrument Single Channel Analyser burst rise time (length) discriminator. Three different sets of lower and upper SCA thresholds have been used to define an acceptable burst length window, viz:

  1. All channels.
  2. Channel 89-104.
  3. Channel 89-107.

All AO-1 and A0-2 observations have been carried out in configuration (c),while many of the early PV/Cal. phase observations were in configuration (b), which excludes an unacceptably high fraction of X-ray photon events relative to the number of subtracted background counts. Some valid photon events are also of course rejected with configuration (c). In Figure 1 typical back ground spectra are shown for the three different burst length windows used. The ratios indicate the fraction of the particle background remaining i.e. ~20% for 89-104 and ~25% for 89-107. The spectra in Figure 1 show two line features at 10.74 keV and 12.09 keV which are believed to originate from particle activation of the lead collimator and detector shielding. These features are very stable, have no orbital or long term evolution and provide a useful in-flight calibration (Section 3). Count rates In the lines are 0.019 and 0.0275 ct s-1 for the lower and higher energy features respectively. The shape of the background spectrum remains constant although its overall intensity can vary on a timescale of weeks. In addition, the gain of the detector varies by a few percent from orbit to orbit and this must be corrected before a standard background is subtracted. The current procedure for background subtraction depends on the source strength.

  1. Weak Sources - When the source is sufficiently weak for the lead lines to be detected the data can be cross correlated against a standard background re-binned to correct for any gain change. Because there may still be some excess or deficit in the higher channels ( >~ 12keV), the background should be re-normalised to give the same count rate as the data in these upper channels.
  2. Bright Sources - If the lead lines are swamped by the source or the source is detected up to 12 keV, then the slew file data or surrounding observations can be used to obtain the correct gain. Since the detector gain changes slightly each time the high voltage is switched on, only data prior to the following switch off should be used to derive the gain corrections. On most bright sources the uncertainty in the absolute background level is small but again corrections can be applied using the slew file data.

Short timescale "solar events" similar to those seen in the ME can affect the background and must be excluded from the analysis. There is no foolproof way using the GSPC data alone to distinguish these from real photon events and the ME (e.g. EMAX) must be used. When the instrument is operated at beta symbol angles less than 90°, solar X-ray flares are seen through the calibration source aperture in the lead shielding. Spectra during these events are typified by strong S, Ar, Ca and Fe emission. It is not easy to ensure the inter- flare data is free from contamination and at the present time beta symbol < 90° GSPC data should not be used.

3. The Gain

After the high voltage is turned on, the gain of the photomultiplier stabilises over a period of time which is a function of the duration of the 'off' time. Figure 2 shows, for a long 'off' period such as perigee, the channel number of the LED stimulation of the photomultiplier tube as a function of time after high voltage switch-on, characterised by an initial jump with an exponential 6 hour scale factor stabilisation. This may give rise to a maximum three channel (gain 2) drift during a 6 hr observation. LED stimulation data can be used to estimate the drift throughout the observation.

Thermal control of the LEl CMA normally requires that the Al 28V line to all experiments must be turned off briefly prior to a slew. Since Jan 84, therefore, the GSPC HT (derived from the Al line) has been turned off and on prior to each slew, and gain variations must be expected for most observations. However, the recovery period is usually of the order of an hour for such short 'off' times. For future GSPC prime observations, the observer should request that the power lines are left on for the preceed ing slew (ref. p.4)

During the performance verification phase the in-flight calibration source was used to monitor the instrument performance. This procedure was discontinued following two occasions of 'malfunction' of the telecommand to move the source from a 'closed' to 'open' position, although in each case re-transmission of the command was successful and no explanation exists for the 'failure'. It is now clear that the absolute energies given by the calibration source are offset from those given by a cosmic X-ray source because the calibration source illuminates the side of the detector where the electric field is non-uniform.

Under the assumption that the S, Ar and Ca lines are at the same energies as measured by the Einstein SSS, energies of 10.74 and 12.99 keV are deduced for the lead lines with a possible systematic offset of +/- 0.04 keV. These energies are not consistent with either lead La and b or Bi La and b. The reason for this is not clear but the result may indicate that the lines are a blend of the two or that a region of the detector with a slightly different gain is illuminated.

There is a feature around 4.75 keV (determined from the above lead line values) in the spectra of all cosmic X-ray sources measured with the GSPC caused by a small change in the gain of the detector at the Xenon filling gas L edge, resulting from a jump in the mean ionisation energy. We have chosen to model this in the form of a line feature at 4.75 +/- 0.02 keV as a means of calibrating the detector gain. It has a strength of 0.08 times the continuum with a FWHM of 0.2 keV, independent of the spectrum. On bright sources it is a very sensitive measure of the gain. Line energies measured below 4.75 keV should be increased by 40 eV to account for the effect.

Recently an LED has been used to illuminate the photomultiplier to calibrate gain variations. LED stimulations are well correlated with the gain variations in the energy of the lead lines on timescales of an observation. On longer timescales there is apparently an additional dependence (possibly temperature) that has yet to be calibrated.

In order to establish the gain of the detector either the lead lines or the feature at 4.75 keV should be used. The LED's are then a measure of the gain stability throughout an observation. The nominal channel to energy conversions are given by:

E = GPnxN + GCn

where n is the gain and N the channel no.

GP2.0 =  0.06904, GP1.0 = 0.13661, GP0.5 = 0.2740
GC2.0 = -0.035,   GC1.0 = -0.104,  GC0.5 = -0.104

The values of GPn are arbitrarily chosen and should be scaled to position the background lines or 4.75 keV feature in the correct channel in the spectrum. To combine gain 2.0 data with gain 1.0 there is an offset of one gain 2.0 channel such that:

 
CH1.0(l) = CH2.0(2) + CH2.0(3)
CH1.0(2) = CH2.0(4) + CH2.0(5)

etc...

Gain 1.0 to 0.5 is not offset.

4. The Effective Areas

Figure 1 and Table 1 summarise the effective area of the GSPC as a function of energy for the three burst length windows used and also gives the pre-launch expected area. Even for the wide open setting there is a 30% shortfall in effective area. These areas have been deduced by forcing a fit to the Crab Nebula spectrum, giving a normalisation of 9.5, a photon index of 2.1 and a column density of 3.5 x 1021 H cm-2. In addition to the 30% overall shortfall there is an extra reduction in efficiency below 4 keV. Figure 1 also shows the effect of additional attenuation by the burst length windows. The origin of the pre and post-launch discrepancies in effective area is not clear. Since the spectral dependence of the corrections applied to obtain the correct fit to the Crab are unknown, spectral measurements below 4 keV, and in particular column density measurements ~< 1022 H cm-2 should be treated with caution. Below 2 keV (i.e. below the window cut off) there is an excess of counts significantly above that expected from the L and K escape fractions. This is caused by partial collection of charge from photon interactions occurring close to the detector structure (eg. the window). Because of this, data below channel 30 (in gain 2.0) should not be included in spectral fits. Since the effective area is ~10 cm2 at this energy, no real loss of data will occur.

5. The Resolution

The in-flight calibrations give the nominal pre-flight resolution for this instrument as described In the FOT Handbook.

6. A Warning

On a FOT the Initial record in an observation immediately following an LED stimulation may also contain some LED data (see J. Sternberg's note in this Issue). To avoid contamination of spectra, users should discard two formats (16s) of data as shown at the time of the stimulation.

N.E. White


Figure 1: GSPC background spectra for various burst window settings. The data is in gain 2.0 and has been compressed from 256 to 64 channels.
GSPC background spectra
Figure 2: The evolution of the gain of the GSPC following a turn on using an LED stimulation of the photo-tube.
Gain evolution
Figure 3: The GSPC effective area deduced from forcing the response matrix to give the correct spectrum from the Crab. The pre-launch prediction is also shown.
Effective area
effective area table


The ME Gain Calibration

1. Introduction

There have been a number of changes to the ME calibration data since the last CCF was issued. The aim of this article is to inform the community of these changes, to catalogue the history of the Argon gain settings and to indicate how the effects might best be incorporated into users' software. An updated ME CCF is being prepared and will be released shortly. Updated CCF's will have a time of last update (bytes 12-15 of record 1) of 145981505 or 1984 day 229 14:25:05.

Note that the observed gain is dependent on two factors, the amplifier gain setting (Ag) and on any leaks/drifts in the detectors.

Work is continuing on the Xenon calibrations.

2. Gain History

The first ME observation took place on 1983 day 168 (D1) All the Argon detectors had an Ag setting of 1.220 (the maximum) and the Xenon detectors 0.972. After an initial examination of this data the Argon Ag's were changed to achieve approximate alignment of the overall gains on 1983 day 173 13.23 (D2). The Ag's were then:

Table 1

Det. Ar Ag
A 1.184
B 1.176
C 1.184
D 1.212
E 1.220
F 1.220
G 1.184
H 1.204

The Ag's remained at these settings until 1983 day 198 20.13 (D3) when they were set to:

Table 2

Det. Ar Ag Xe Ag
A 1.220 0.972
B 1.220 0.988
C 1.188 0.968
D 1.208 0.972
E 1.216 0.972
F 1.200 0.972
G 1.180 1.004
H 1.192 0.988

Two observations of the Crab Nebula have been carried out on 1983 day 276 (D4) and 1984 day 086 (D6) enabling checks to be made of the overall gains. Examination of the data from the second observation has shown that the majority of the Ar gains were slowly increasing viz:

Table 3

Det. Gain change/day = Delta g
A 0.000027
B 0.000031
C 0.000030
D 0.000437
E 0.000003
F -0.000004
G 0.000160
H 0.000027

On 1984 day 211 13.47 (D7) the Argon Ag settings were changed to align the gains to a new standard, at which the channel number corresponding to an energy of 6.7 keV is the same for all detectors.

The relative gains between the D6 and D7 standards are:

Det. D6/D7
A 1.008
B 1.014
C 1.004
D 1.026
E 0.995
F 1.011
G 1.040
H 0.987

Regular trims of the detector D and G Ag's are scheduled so that they remain within 0.5% of their nominal values, established on day D7 as a fiducial point. It is expected that by February 1985 the gains of the other detectors will have drifted by <0.5%. The Ag settings will be reviewed after subsequent Crab and Cas A observations.

The Ag settings on day D7 were:

Det. Ar Ag Xe Ag
A 1.208 Unchanged
B 1.200 "
C 1.180 "
D 1.116 "
E 1.220 "
F 1.188 "
G 1.122 "
H 1.212 "

An examination of data from the Fe55 cal. source showed that the gain of detector D did not start to drift appreciably before 1983 day 353 (D5)

The gain data (i.e. the channel boundary coefficients) on the present ME CCF is derived from an analysis of data obtained during the second Crab observation, i.e. they are only good on day D6. For observations at other times a set of relative gain coefficients must be calculated and applied to the channel boundary coefficients provided in the CCF. Future CCF's will have the D7 standard.

3. Calculation of Relative Gain Coefficients

The calculation is similar irrespective of whether the D6 or D7 standard is being used and is as follows:

  1. Determine the analogue amplifier gain settings (Ag's) at the time of the observation. Corresponding digital values (Dg's) can be obtained from the Memory Load Commands (FOTH Section 3.2.2) or from the Housekeeping data (FOTH Section 3.4.2.1). The relation between the two is:
  2. Ag = (Dg - 200) * 0.004 + 1.0

    Dg's corresponding to the time of the standard will be part of an updated CCF and the above relation will apply.

  3. Determine the change associated with the Ag values, ie
  4. R = Agt/Ags
    
    

    where Agt is the relevant value for the time of the observation.
    Ags corresponds to the value at the time of the standard.

  5. Determine the change in gain within the counter:
  6. C = 1 - (Delta)g * (S - t)

    where S = time of standard, (Delta)g = as per section 2 above.

    Note, for detector D, Ag = 0.0 for t < 1983 day 353.

    The relative gain coefficient is then just R*C

    For t > D7, R*C = 1.000 +/- .005 (D7 standard)

4. Correction of the Argon Channel Boundary Coefficients

The Argon channel boundary coefficients have the form:

mV = a1 * (1 - e-a4 * E) + a2 * E x e-a3 * E)

where mV is the output of the front end electronics in mV and E the Energy in keV. The relation between mV and K, the channel number, is given by:

mV = (k - 0.5) * 1500/128

see the FOTH Section 3.7.2 5/23. To correct the coefficients (ai) supplied in the CCF to another time simply multiply the values of a1 and a2 only by R*C for each detector.

5. Comment

The ME detectors, as most proportional counters, will show small erratic variations In gain above the drifts mentioned above. It is estimated that the gain of any ME detector will always have an uncertainty of ~0.5% (= 30 eV at 6.7 keV). This will cause an increase in the value of chi-squared from spectral fitting to data from a bright source, when the statistics are very good.

6. Summary

The above results show how to correct the given ME gains at an arbitrary time. We Intend to monitor carefully the future trends in gain change using data from the on-board Fe55 source, Cas A and Crab observations. The results will be published in the Express and used to update the CCF.

A.N. Parmar
A. Smith


EXOSAT Observatory Computer Environment

This article describes the computers, software functions and data flow for EXOSAT telemetry and, although not all aspects of the computers involved in the EXOSAT ground system are discussed, the aim is to give some background information relevant for visiting observers.

For historical reasons the computer configuration may be divided into two parts which have evolved independently:

  1. Operational Support Computers (shown on the diagram as MSSS/EX1/EX2). MSSS = Multi-Satellite Support System. These systems were specified and developed by ESOC, following consultations with all the institutes involved in the development of the payload. Much of their software has been used for supporting other satellites in the past besides EXOSAT. The emphasis is on fast reaction, spacecraft safety and security of telemetry. An observer in the EXOSAT Control room communicates only with these specific computers via the keyboards and displays. Their functions are described more fully in an article in EXOSAT Express, No.4, p.30.
  2. Scientific Support Computers (shown as HP1 to HP4 on the diagram). HP computers were originally procured some 9 years ago at the start of the EXOSAT hardware development programme and various systems were used to support the assembly, integration and test phases of the industrial contract and instrument calibration. Because of the accumulated software development experience and expertise, an extension of the use of these systems to the operational phase of EXOSAT was a logical development. Initial specification of the hardware and software requirements for analysis of data at ESOC was undertaken by the pioneer members of the Observatory Team, again after very detailed consultations with institutes, and improvements and expansion have continued up to the present. The emphasis is on a very flexible data analysis facility which can respond to changing requirements. An observer can communicate with these systems from the HP terminal area adjacent to the EXOSAT control room. Most of the remainder of this article deals with this second group of computers and the background concerning their software.

The production of FOT's (final observation tapes) is performed using data from system 1, i.e. the HP systems and the Observatory Team are not directly involved.

EXOSAT Data Structures

Currently there are 38 available on-board computer (OBC) programs of which about 10 are commonly used, generating roughly 30 different data structures ranging in size from 2 to 8192 bytes. There is in general no synchronisation between different programs or with the fixed pattern of frames of the housekeeping telemetry. The layout and frequency of occurrence of each "floating" data structure depends on the OBC programs running at the time and their workspace parameters.

This presents two problems for the user. Firstly, there are structure clashes between formats/frames/OBC-generated records/ message packets. Secondly, no software can assume that information will appear regularly and with constant layout in the floating telemetry, unless it knows in advance that the 0BC status remains constant during the period in question.

The FOT was defined with a view to solving both these difficulties. Different types of data are separated by sorting, so that each can be accessed as a fixed-record-length direct access file, which has advantages in simplicity and efficiency. The advance knowledge of OBC status is given in the "observation directory", which contains the history of the 0BC status and details of the records in the telemetry at given times. However, the idea of the observation directory was extended in order to record all status relevant to one instrument, not only OBC status but also telecommand and attitude information, which provides an analysis program with all the essential status history on the FOT without the necessity of reading the actual telemetry.

All the sorting/directory construction/data extraction, as well as image linearisation and sorting into small maps, is performed at ESOC during the FOT production and therefore need not be done by observers. It is also performed by the HP1 computer, which acts essentially as an interface between EX2 (message packet data) and the other HP's (FOT data). Consequently, an observer normally never comes into contact with the floating telemetry in its message packet form, except on the EX2 graphics system.

Despite the apparent benefits of the FOT structure, there is a drawback: the telemetry is no longer a single chronological sequence but instead a two-level file structure (directory plus record types). This structure reflects the requirement specification of the 0BC software and should be seen as a consequence of the demand for flexibility in data encoding, processing and telemetry. The cost is in added complexity of the software required to use the files from a FOT.

It is important to realise that the FOT contains sorted raw telemetry. It does not contain analysis results, such as images or spectra, because this would imply throwing away a great deal of essential information from the original telemetry, e.g. time variability, differences in characteristics of ME quadrants possible selection of events etc.

Summary of Main Functions of Individual Computers

EX1 (Siemens 330). Adding derived information to telemetry. Archiving to magnetic tapes.

EX2 (Siemens 330). Maintenance of "short history file" (SHF), a circular file of the most recent 40 hours (at 8 kpbs) data in message packet form. Support of DCR graphical displays, with data received in real time or read from the SHF.

MSSS (Multi-satellite support system) General table-driven software for display and telecommanding, plus EXOSAT-specific software for operating the 0BC.

HP1 (Hewlett Packard 1000-M). Conversion of SHF data into the FOT structure described above, including sorting linearisation etc., for near-real-time use by HP2/HP4. Filing FOT's to disc (function shared by all HP's). Production of real-time observation log.

HP2 (HP 1000-F). Automatic analysis on a routine basis, calibration analysis (to determine or refine calibration data for future FOT's), and interactive analysis, all using data in FOT-format either from an actual FOT or from the SHF. Note that the distribution of disc drives among the HP's is flexible and may be modified in the light of experience.

HP3 (NP 1000-F). Mission planning. Maintenance of FOT request file, tape catalogues and off-line observation log covering the whole mission. All HP2 software may run on HP3 also, except that there is no access to the SHF. Although HP3 is included on the configuration diagram, it is physically in a separate location from the other HP computers.

HP4 (HP 1000-F). All HP2 software can also run on HP4, but it is primarily intended for interactive analysis, and has additional CPU power via a link to the Siemens/Fujitsu computer.

Siemens/Fujitsu. This is the current ESOC general-purpose main- frame. It's EXOSAT functions are:

  1. FOT production,
  2. OBC software maintenance
  3. providing additional processing power to HP4, via a remote job entry interface.

Software Packages in use on HP Computers

During the past 5 years the Observatory computers have evolved through many stages. The original concept was to provide an extension of the EX2 display system, including all software needed for in-flight calibration. Later the routine "automatic analysis" of FOT's was included in their functions. After the build-up of the Observatory Team prior to launch, the HP computers also became a general purpose computer for the team's use and an important tool for mission planning. The most recent stage in the expansion has been the, implementation of facilities for visitors to analyse data (FOT's or real-time), the so-called interactive analysis system; these new facilities will be described in a subsequent EXOSAT Express.

One factor which has helped this evolution was the definition of a number of common software packages at an early stage. For example, the rather complicated structure of FOT's has been partly concealed from the scientific software by a so-called "co-ordination package", which is linked into any program reading telemetry data, whether automatic analysis, calibration, or interactive analysis, and whether from actual FOT's or from the SHF. Currently there are about 30 HP programs using this package. Some of the functions it supports are:

  • access to payload & OBC status history (observation directory).
  • definition of chains of observations (which are subsequently accessed as if a single observation)
  • open/read/close for a chain and for a particular record type, with implicit use of file pointers in observation directory.
  • access to auxiliary, calibration and orbit data.
  • scheduling other programs with data transfer.
  • access to "parameter characteristics file" (used in decoding and calibrating housekeeping data).
  • checking for errors in FOT structure.
  • direct access via time key, or (for images) direct access to a particular small map.
  • efficient and correct use of HP file management system.

In listings of observatory software all entry points to the package can be recognised by names beginning with CS.

Several different graphics packages are available on the Observatory computers. The initial emphasis was on automatic analysis, the output from which is written to Gould and Versatec electrostatic plotters. If a conventional (device-independent) graphics package is used with a raster plotter, it is necessary to create an environment whereby the user is able to plot information in any order on the paper, which conflicts with the basic capabilities of the plotter - drawing lines of dots as its paper moves. For data handling on a large scale the cost of defining such an environment is excessive in terms of disc access and therefore a special-purpose raster plotting package which does not use the disc has been developed.

Conversion of plotted information into raster lines is performed in memory and concealed from the calling program; a special grey- scaling facility is available. The improvement in efficiency is very large, with the accompanying drawback of the unconventional calling interface (Names beginning with GR in listings of observatory software). The package is partly in FORTRAN and partly in HP assembler.

An additional "layer" of graphics software (names beginning with GS) provides the usual axis drawing and scaling facilities, as in conventional packages.

Two different Ramtek colour VDU's are attached to HP2 and HP4, and a compact package has been written in HP assembler to support their capabilities.

With the gradually increasing emphasis on interactive analysis other packages are still being developed within the Observatory Team and are concerned with manipulating processed results such as images and energy spectra (details in a future issue).

J.R. Sternberg


EXOSAT ARRIVAL TIME ANALYSIS

For a detailed time study of an X-ray source it is often necessary to correct event times recorded by EXOSAT to a common reference frame such as the Solar System barycenter.

This note illustrates how the EXOSAT Observatory software makes this correction with particular emphasis on those aspects that are specific to EXOSAT. Subroutines to correct ground station time to MJD, to calculate the Earth's position relative to the Solar System barycenter and EXOSAT's position relative to the centre of the Earth are available from ESOC, on request (Contact: J. Sternberg).

The calculation of accurate barycentric times is as follows:

  1. Determine the packet reference time to Ground Station time (which is UT) relation during the observation of interest. This can be derived each format (every 8 secs) since both times are contained within each HK packet. Normally, it is only necessary to determine the relation at the start and end of the observation unless very high precision is required (see FOTH, Section 4)
  2. Find the reference time half-way through the spectrum or intensity sample of interest. This can be obtained by examining the workspace parameters and packet structure of the data.
  3. Calculate the ground station time corresponding to this reference time.
  4. Calculate the UT at which the data was detected by EXOSAT. To do this the position of EXOSAT, relative to the centre of the Earth, can be obtained using subroutine ORBIT. Note that for precise work an additional correction, since Madrid is not located at the centre of the Earth, must be applied!
  5. The equivalent time at which the event would be detected at the centre of the Earth can now be calculated using ORBIT and the source RA and DEC obtained from auxiliary data.
  6. The event time can now be corrected to the Solar System Barycenter using the source position and subroutines SOLUN and SUNBAR.

Note that effects such as electronic processing times within the experiments have been ignored (ref. Express No.5, pp.31-47).

A.N. Parmar


Observation Boundaries on FOT's

It is important to understand clearly the definition of an "observation" on the FOT (see section 3.1 of the FOT Handbook). The division of data into observations is based only on information in the housekeeping record, and the real time (on-board) of any status change is therefore uncertain by up to 8 seconds (the format period). This is illustrated in the following diagram. F1, F2 and F3 mark the start times of 3 consecutive telemetry formats. If a parameter appears in the telemetry at time A1, and is observed to change its status in the following readout at time A2, then the real time of the status change may lie anywhere between A1 and A2. Nevertheless, on the FOT the observation boundary will be exactly at time F2.

F1      A1      F2      A2      F3
|_______________|________________|
|               |                |

An additional blurring of observation boundaries arises from the on-board buffering of floating telemetry - normally the OBC- generated data lags behind the housekeeping by one record.

The few seconds at the beginning or end of an observation may therefore include some data which has spilled over from an adjacent observation. Normally this effect is not significant in an analysis of an observation lasting for a few hours. In one particular case however caution is needed, namely when accumulating a GS spectrum during a time period adjacent to a measurement with the LED stimulation. If the stimulation precedes the observation it will be found that typically the first two histograms include contamination from the LED, one histogram from each of the two effects described above. Observers are advised that any OBC-generated information which is within 8 seconds of an observation boundary should be disregarded. For this purpose the reference time in an individual record may be compared with the spacecraft clock limits given in the observation directory.

J.R. Sternberg


Specification of the Commonly Used OBC Application Programs

1.GSPC Programs

  • Direct mode: GDIR (program no. 1)
  • All channels may be sampled i.e. energy, burst length and time tag, with flexible sampling rate. The following data combinations are possible -

    1. Valid energy values (energy word non-zero) plus a count of invalid energy samples since the last valid energy event are transferred to the output buffer. This allows the time of each event to be determined according to the sampling frequency of the channel.
    2. Valid energy (non-zero) and burst length values are transferred directly to the output buffer.
    3. Valid energy (non-zero) are transferred to the output buffer for those events with a burst length within defined limits.
    4. Valid energy (as before), burst length and time tag values are transferred directly to the output buffer.

    A minimum time between transmission of output buffers may be selected in order not to overload the telemetry in the case of high count rates.

  • Energy/Burst length histogram mode: GHEBL4 (program no.32)
  • The energy and the burst length channels may be sampled with a flexible sampling rate.

    256 channel energy histograms for all non-zero energy events are formed and transmitted to ground at a flexible pre-selectable rate.

    The following options are available:

    1. Either energy or burst length histograms may be formed.
    2. Compression to 128, 64, 32, or 16 channels may be selected such that for the selected transmission rate.
    3.         2 x     128 channels
      or      4 x     64     "
      or      8 x     32     "
      or      16x     16     "     will be sent to ground.
      
    4. Only those valid energy events with a burst length within defined limits will be used to form the histogram.

2. ME Programs

  • Direct Mode: MDIR2 (program no. 71) ref. p. 30.
  • The energy channel is sampled at a flexible rate. The number of invalid (zero) samples between valid (non-zero) energy events within a defined energy range is transferred to the output buffer. At 8K telemetry rate, 442 words per second are allocated to the OBC floating telemetry. Continuous transmission of data is possible for events rates ~< 600 c/s with 100% telemetry allocated to the ME experiment. A minimum time between transmission of output buffers may be specified in order not to overload the telemetry.

  • Energy/Identifier histogram modes:
    1. MHER4 (program 33)
    2. The energy and detector identifier channels are sampled at a flexible rate. Full resolution (256 channels Ar + Xe) energy histograms are formed for each (Ar+Xe)detector (10 sec integration time), each pair of detectors (5 sec Integration time) or for each half-experiment (2.5 sec integration time). Intensity profiles for each detector, pair of detectors or half experiment are also formed. If half experiment data is selected, the following options are also available.

      1. Compression to 64, 32, or 8 channel energy histograms may be selected such that
      2.         4 x 64 channel
            or  8 x 32  "
            or 32 x  8  "
        

        histograms will be transmitted in the integration time.

      3. Selection of an energy range for inclusion in energy histograms and intensity profiles may be given (first channel, and number of channels).
    3. MHER5 (program 9)
    4. The energy and detector Identifier channels are sampled at a flexible rate. For each half experiment, separate Ar and Xe energy histograms are formed, each having separate flexible integration times. 64 energy channels only are used to form the histograms, with the first channel for both Ar and Xe being selectable. Intensity profiles for the Ar data only are formed, with flexible minimum resolution (in units of 31.25 ms). An option to compress the 64 channel histograms to 32 or 8 channel histograms is available.

    5. MHER6 (program 70)
    6. The energy and detector identifier channels are sampled at a flexible rate. For each half experiment, intensity profiles are formed using valid energy events from either one or two selectable energy ranges. The length of the intensity profile time bin is flexible and is defined as the number of samples of the energy channels to be used to form the bin e.g. at a 4K s-1 sampling rate of the energy channel, 64 or 32 energy samples per time bin correspond to 16 and 8 ms time resolution respectively.

  • Pulsar modes:
    1. MPULS (program 8)
    2. The energy and time-tag channels are sampled at either 4K s-1 or at 1K s-1 This mode is used for observations of periodic sources where the period is defined by t x 128 + r, in units of 2-17 seconds (7.63 µ s). If the exact period of the source cannot be specified thus, r can be updated by +/- 1 every i periods; t,r and i are selectable.

      129 bin intensity profiles of non-zero energy events from the Argon data only are formed. The first 128 time bins are of equal length t, and the 129th is equal to r. The profile can be summed over a number of periods, T.

      Energy histograms of valid events are also forned, the integration time being 8 x the intensity profile integration time. The following options exist.

      no. of EN histograms no. of channels per histogram no. of time bins per energy histograms
      129 8 1
      33 32 4
      9 128 16
      9 256 16
      17 128 8

    3. MPULS2 (program 14)
    4. The energy, detector identifier and time-tag channels are sampled at either 2K s-1 or 1K s-1 This mode is also used for observations of periodic sources, where the period is defined as for MPULS, but no updating of r every i periods is available.

      This program performs basically the same functions as MPULS, but has the additional feature of checking the detector identifier to determine whether the event originated from the half experiment pointing at the source, or the offset half experiment. 'Source' data is used to form intensity profiles and energy histograms as described in MPULS. 'Background' data is used to form a simple background histogram, with the same number of channels as the source energy histogram. The energy histogram options are as follows:

      no. of energy histograms no. of channels per histogram no. of time bins per energy histogram
      129 8 1
      33 32 4
      17 64 8
      9 128 16
      17 128 8

  • Intensity Mode
  • MHTR3

    The qualified events counter is sampled at a rate selectable between 32 s-1 and 4 K s-1. Each sample is transferred to an output buffer, which is transmitted when full. Output buffer size can be selected from one of four values: 256, 512, 1024 and 4096 words. Continuous data is possible according to telemetry and sample rates, otherwise discrete time intervals can be covered depending on the size of the output buffer and sample rate.

    Note that this counter integrates the qualified event rates from all detectors and can be electronically conditioned to accept Ar only, Xe only or Ar and Xe events - it is a physically different counter from the housekeeping QE counter.

3. LE Programs

  • Direct Mode: LDIR2 (program no. 35)
  • All channels may be sampled, i.e Energy, Rise Time, Position and Time Tag. The sampling rate is fixed at 512 s-1.

    Valid data is selected by checking that the position is non zero, specifically that MSBXY (most significant bits XY) and LSBX (least significant bits x) not equal to 0. The following data combinations per event are possible.

    1. Valid position data (3 bytes) transferred directly to the output buffer.
    2. Valid position data and energy values transferred directly to the output buffer plus the count of invalid samples between events.
    3. All data mode - all channels for a valid event are transferred directly to the output buffer, plus the count of invalid samples between events.

    A square shaped or diamond shaped filter to perform background rejection may be applied to any of the above data combinations. A minimum time between transmission of output buffers may be specified in order not to overload the telemetry in the case of high source count rates.

C. Durham


A Guide to the Use of the OBC Programs

The aim of this article is to provide information necessary to define the optimum OBC configuration for a particular observation and to define a number of standard OBC configurations in which observers may choose to conduct their particular observations. Note that the raison d'être of the OBC is to obtain the maximum amount of scientific information with a telemetry stream limited to 8 kbits/second.

Currently, observations are planned with a standard OBC configuration for LEl, a simple choice of GSPC modes (95% of the time = GHEBL4) and a more complicated trade-off of spectral/temporal resolution and telemetry usage for the ME.

For faint ( ~< 10 cts/sec/half) sources a default configuration shown below is nearly always used:

Instruments 0BC Configuration Telemetry usage
LEl LDIR2, tmin=22 ~25%
GSPC GHEBL4 256 channel spectra every 8 sec.(Gain 2.0 nominal) 4%
ME MHER4, det. ID, 8 256 channel spectra every 10 secs + intensity profiles every 0.25 secs. 27%
MHER6, channels 6-24 (2-6 kev, background ~10 cts/s/half) 12 ms time resolution 20%


76%

MHER4 in det. ID mode is the most sensitive ME mode and should always be used for faint sources unless high time resolution is needed. The intensity profiles (8 every 0.25 secs i.e. 1 from each detector) contain data from both Argon and Xenon and are therefore not too sensitive since the background in the Argon chambers is ~40 cts/sec/half while in Xenon it is ~250 cts/sec/ half.

Generally observations of a pulsar with a period P require a time resolution of about t = p/10. Since the ME is normally the prime instrument, the above OBC configuration is therefore only useful for pulsars with periods >~100 sec Similarly if a source is thought to be variable or a timescale of q then time resolution with t = q/10 is required.

An alternative OBC configuration with two advantages and two disadvantages is:

Instruments 0BC Configuration Telemetry usage
LEl LDIR2, tmin=22 ~25%
GSPC GHEBL4, t = 8 sec 256 channels 4%
ME MHER5, 64 + 64 channel spectra every 1 sec. Intensity samples every 31 msec 39%


68%

The advantages are the higher time resolution of the ME spectra (1 sec) and of the intensity samples (31 msec) and the much im proved background in the intensity samples (30 cts/sec/half rather than 300 cts/sec/half) since HER5 puts Argon data only in to the intensity samples. The disadvantages are that only 2 spectra are transmitted per sample (i.e. one per half rather than 1 per detector) and this can complicate the background subtraction. In addition only 64 channels of Argon and 64 channels of Xenon are transmitted. Normally all the science data is contained within these channels (unless the source is visible in Xenon at >~ 30 keV), however, the data above channel 64 can be useful in checking that the background is stable.

If 1 second spectral data and 31 msec intensity profiles are not sufficient, then the spectral data can be compressed by a factor of 2 to double the time resolution with only a small loss of spectral information. Also, the HER5 intensity profiles can be replaced by HER6 intensity profiles thus:

Instruments 0BC Configuration Telemetry usage
LEl LDIR2, tmin=22 25%
GSPC GHEBL4, t = 8 sec 256 channels 4%
ME MHER5, 32+32 channel spectral every 0.44 sec. No intensity profiles 36%
MHER6, 12 msec intensity profiles for channels 6-24 20%

85%

85% telemetry usage is an upper limit which allows a margin for solar flares etc. This configuration has a potential problem of CPU usage. For bright sources ( >~ 600 cts in the ME) the CPU usage can become too high causing the OBC to halt with a resulting loss of about 15 minutes of data. Caution should therefore be exercised if the source bursts or flares - the most interesting data could be lost! If there is any possibility of the source being this bright then it is probably better to use MHER5 or MHTR3 to provide the intensity profiles. The disadvantage of this is that the background will be ~40 cts/sec with MHER5 or ~80 cts/sec with MHTR3 as opposed to 10 cts/sec/half for channels 6-24 using MHER6.

Other options to consider for higher time resolution are:

  1. compress MHER5 spectra to 8 channels (0.125 sec time resolution available)
  2. Use MPULS or MPULS2 modes. These provide spectra folded over a given period. The problem with these modes is that if the period is not well known then information can be lost in the folding process.
  3. Use MDIR2 mode (ref. p. 30) 125 msec time resolution, over a selectable range of energy channel 5, for count rates of between 200-600 cts/sec.

GHEBL4 is the standard mode for the GSPC and is appropriate for 95% of all observations. It provides 255 channel energy spectra with 8s time resolution. Higher time resolution and compression of the spectrum are available.

For fine-time resolution spectral studies such as bright source observations, with significant counts in short periods, or observatlons of pulsars with periods ~< 100 s where the data is folded over several periods, use of GDIR is recommended. This provides spectral and timing data per photon to a resolution of the sample interval, typically 0.5 or 1 ms selected according to source strength.


Principal OBC Modes

LE1

  • LDIR2 Standard LE program.

GSPC

A choice of two, viz:

  • GHEBL4
    Standard GS Program. Normally 256 channel spectrum every 8 sec but t can be changed and spectra compressed to increase time resolution.
  • GDIR
    An alternative GS program for use with pulsing or rapidly varying bright sources.

ME

There are a number of programs available for the ME. They either provide spectra or intensity profiles or a combination the two.

  • MHER4
    Standard ME Program. Provides 256 channel spectra with (selectable) det ID, quad ID or det half ID, giving spectra every 10,5, 2.5 sec respectively and intensity profiles. MHER4 in det ID is normally the mode for faint ( 10 cts/sec/half) observations where high time resolution is not required. Compression possible but not normally used now that MHER5 is available.
  • MHER5
    Principle alternative ME program. Provides higher time resolution than MHER4. 64 channel Argon and 64 channel Xenon spectra only. Compression available to increase time resolution. Time resolution is selectable and limited by telemetry considerations. Typical values are ~1 sec with no compression and .25 sec with compression = 1 (32+32 channels).
  • MHER6
    Provides intensity profiles over a selectable range of channels (normally 6-24 - where the ME is most sensitive) or 2 ranges of channels (i.e. hardness ratios).
  • MHTR3
    High time resolution intensity profiles for all of Ar, Xe, or both (Ar only normally used for timing studies) Opt = 3 is the normal mode but in special circumstances use in opt = 1 gives up to .25 msec time resolution non-continuous data. Particularly useful instead of HER6 for observations of bright sources carried out in the coaligned configuration or as a dead time monitor (low sample rate 32 Hz - p. 30).
  • MDIR2
    Direct ME mode (ref. p. 30). Provides high time resolution (up to .125 msec) intensity profiles over a selectable range of channels. For continuous data at a sample rate of 4 K s-1 the count rate should be between 200 and 600 cts/sec in the channel range of interest.
  • MPULS
    Provides spectra and intensity profiles folded over a given period for bright pulsing X-ray sources. No det ID, so normally only used in coaligned configuration. Normally MPULS and MPULS2 are only necessary for sources with periods ~< 1.5 seconds.
  • MPULS2
    As above but used in offset configuration. Data from one half only folded over given period, data from other half is put into background spectra. For use with fainter (~< 100 cts/sec) sources.


Telemetry Usage

A few general formulae to calculate approximate telemetry usage (at 8 kbits s-1).

LDIR2
~25% depending on count rate, whether a diamond filter is used and the value of tmin.
~3% radius = 128; diamond filter.

GHEBL4
4% with t = 1 sec
35% with t = 8 sec

MHER4
27% always

MHER5
39% with tAR = 1 sec, tXe = 1 sec tin = 1/32 Sec. No compression
50% with tAR = 0.31 sec, tXe = 0.31 sec. tin = 8 sec, compression factor = 1.

The telemetry used is given by:

tel % = 500 (0.015 x tAR x tXe + tXe x tin + tAR x tin)
___________________________________________________________
tAR x tXe x tin

where the t's are in units of software cycles (31.25 ms)
and   tAR = Argon accumulation time
      tXe = Xenon accumulation time
      tin = Intensity samples accumulation time
MHER6
15% with Nsamp = 64 (t = 16 msec) and 1 set of channels
40% with Nsamp = 48 (t = 12 msec) and 2 sets of channels

The telemetry used is given by:

tel% = ( 959 / Nsamp) x Nchan

where Nchan = 1 or 2 depending on the number of channel ranges chosen.

MHTR3
15% For 8 msec, opt 3 data - the default mode.

A.N. Parmar


[About EXOSAT] [Archive] [Software] [Gallery] [Publications]


HEASARC Home | Observatories | Archive | Calibration | Software | Tools | Students/Teachers/Public

Last modified: Thursday, 26-Jun-2003 13:48:32 EDT