GHRS Instrument Handbook
Reference Information on Instrument Performance
Figure 8.8: Normalized Point Spread Function for the GHRS Small Science Aperture. The curve is normalized to 1.0 at the origin.
The Line Spread Function LSF is measured by observing the spectrum of a narrow-lined star with both the SSA and LSA and then determining what function must be convolved with the SSA observation to yield that seen in the LSA. These differential LSFs can be fitted well with Gaussians, and the Table below lists the FWHMs we found. The intrinsic LSF for the SSA is described by a Gaussian with a FWHM of 0.92 diodes (see Gilliland et al. 1992). Also listed in the Table are the expected FWHMs for the intrinsic LSF in the LSA determined by combining the FWHM of the SSA and the differential LSFs in quadrature.
Table 8.7: Differential Line Spread Functions
Each Digicon diode has its own discriminator, and if they are properly set the dark count rate measured is very low. On-orbit measurement has shown that both GHRS detectors have a dark rate less than the design goal of 0.01 counts diode-1 sec-1 at low geomagnetic latitudes and that the dominant noise source is cosmic rays. (For Side 2, the average dark away from the SAA is 0.012 xb1 0.008 and for Side 1 it is 0.005 xb1 0.002.) During SAA passage, the noise increases to a maximum of about 1 count diode-1 sec-1. The scheduling software for HST uses known contours of the SAA and does not accumulate counts when the spacecraft is within those contours.
Even outside the SAA, observations show that "dark" counts tend to come in bursts. Approximately 15% of dark counts are produced by events that occur within a time of 8 ms or less. The CENSOR feature in Accumulation Mode allows you to ignore integrations with such high dark rates. CENSOR works by summing all 512 channels of the Digicon every 8 ms, and if that sum exceeds a threshold a coincident event is recorded. The flight software can reject and then repeat a single 8 ms time slice for which the coincidence sum exceeds a given level. Electronic noise in the analog summing circuit effectively limits the ability to discriminate events that produce less than about 8 simultaneous counts, so that only the high-amplitude tail of background events can be rejected. It is estimated that use of CENSOR can reduce the dark count rate by about 20%. Using CENSOR=YES is unlikely to degrade an observation (unless the object is bright enough to cause a consistently high count rate), but the benefit is also low: only about a 10% gain in S/N in favorable cases.
The dark count rate is highly uniform over the diodes, so that a mean dark count rate is an excellent representation of what happens at the detector.
- The pre-flight noise specification was 0.01 counts diode-1 sec-1. The average measured value is 0.005 for D1 and 0.008 for D2.
- The background is sensibly constant between -20 and +20o geomagnetic latitude.
- At xb1 40o geomagnetic latitude, the extrema of HST's orbit, the rate is twice that at the geomagnetic equator.
- The background rate is correlated with the cosmic ray and trapped particle flux. Calculations show that the dark noise can be accounted for by cosmic-ray-induced Cerenkov radiation in the faceplates of the Digicons.
- The background due to the direct penetration of cosmic rays into diodes is very low (0.0004 counts diode-1 sec-1).
When using CENSOR:
- Use the default STEP-TIME of 0.2 sec.
- Do not use CENSOR if the expected count rate exceeds about 100 counts per second per diode because real photon events will be rejected. In a severe case, all the data could be lost.
- Using CENSOR can drop the noise level by about 20%. The nominal Side 2 dark count rate is 0.012 counts per second per diode, so CENSOR=YES can lower it to 0.010.
- The following table shows the expected effects of using CENSOR. Rate is the target count rate per diode per second; the second column shows the standard S/N and the third column lists the S/N achieved with CENSOR=YES if the dark is 80% of its value without CENSOR. Signal-to-noise was calculated on a per diode basis for a nominal exposure time of 10,000 seconds (which is reduced to 9,800 if CENSOR is used because of the loss of 2% of the exposures).
Table 8.8: Effects of CENSOR
FLYLIM is a special commanding option for rejecting noise in cases where the source signal is at or below the noise level. The idea is that as HST circles the Earth it finds itself in different noise environments due to the changing magnetic field, and that influences the detected noise background. Also, the background noise occurs as discrete events from radiation in the space environment. If the source count level is well below the background noise, then spectra integrated over a sufficiently short interval that contain multiple counts are probably just noise, and should be rejected, whereas spectra with single counts are more likely to contain real information. The rejected spectra are discarded, which wastes observing time, but there is a net gain in signal-to-noise. A test-in-principle run in 1993 showed that the net background rate could be reduced to as low as 0.002 counts sec-1 diode-1 for the Side 2 detector, which was a factor of four improvement over the mean dark rate. This gain was achieved at the cost of the loss of about 25% of the individual 0.2 second STEP-TIME integrations.
FLYLIM may now be used as an Optional Parameter with an ACCUM. If users believe that FLYLIM may be of help in their science program they should note the following:
- The use of the FLYLIM parameter should be arranged in advance, i.e., before the Phase I proposal deadline, through consultation with a GHRS Instrument Scientist.
- At the time this is written FLYLIM has not been fully tested and its use is at the observer's risk.
One alternative to FLYLIM is to use RAPID mode. The advantage of doing so is that all the observations are retained so that one may go back after the fact and try out algorithms for optimum signal extraction. If the FLYLIM parameter is set wrong due to imperfect knowledge of the source or because of variability, it would be possible to lose all the observations, but that would not happen with RAPID. The disadvantage of RAPID mode is that the spectrum is not fully sampled, so that the resolving power achieved is less.
If you decide you wish to use FLYLIM, we would suggest that you consider devoting one orbit of time as a first visit early in the cycle so that the true source count rate can be accurately and reliably determined.
For more information on using FLYLIM, we recommend that you read "Calibration of GHRS Burst Noise Rejection Techniques," by Beaver et al.in Calibrating Hubble Space Telescope, edited by J.C. Blades and S.J. Osmer (1993, STScI), p. 304.
Deviations from linearity in the way in which the Digicons count photons at high rates were illustrated above in Figure 7.2 on page 89. The effective deadtime for the GHRS detectors has been measured to be 10.2 ms for detector D1. The same value has been assumed to hold for D2. Deviations from linearity are imperceptible below 103 and can be corrected to an accuracy of 1% up to a measured count rate of 20,000 (in units of counts diode-1 s-1).
The images formed by the Digicons are vulnerable to the effects of the Earth's magnetic field. Over the course of a full orbit, the amplitude of the motion is about 50 microns per Gauss for D2 (about 15 microns peak-to-peak) and 10 microns per Gauss for D1. The 50 micron motion seen in D2 corresponds to the size of a diode. This geomagnetically-induced image motion ("GIMP"), together with thermal effects, is the underlying reason for breaking up long exposures into segments of no more than about 5 minutes each.
The Point Spread Function (PSF)
Figure 8.8: - Normalized Point Spread Function for the GHRS Small Science Aperture. The curve is normalized to 1.0 at the origin.
The Differential Line Spread Function (LSF)
Table 8.7: - Differential Line Spread Functions
Detector Dark Count and the CENSOR Option
Table 8.8: - Effects of CENSOR
Noise Rejection with FLYLIM
Count Rate Linearity