Next: 4.4 Effects of Ionizing Up: 4 Instrumental Characteristics Previous: 4.2 Detector Properties
Next: 4.4 Effects of Ionizing
Up: 4 Instrumental Characteristics
Previous: 4.2 Detector Properties
The CRE output voltage of the first few readouts on a ramp does not follow a linear increase in time but can remain constant or even show a decrease in voltage. The effect becomes more pronounced in case only a small fraction of the CRE dynamic range is used. This can happen in observations where the source + background flux is severely overestimated by the observer or in the low brightness regions of maps with high dynamic range.
To exclude this non-linear part of a ramp, the first fraction of the total number of non-destructive readouts is discarded in the SPD processing. The value of the fraction is stored in a dedicated calibration file (see section 13.4)
Integration ramps are not exactly linear but exhibit higher order variations or in some cases ``knees'' which are caused by the CRE. The deviations from linearity only depend on the absolute value of the CRE output voltage and not on e.g. the slope of the ramp (see [28]). Non-linearities can cause systematic errors up to 50% when comparing signals from the same source using ramps with different dynamic range in CRE voltage. This effect introduces an error in the flux calibration in case the FCS readouts span a different dynamic range than the source readouts. Improper correction for ramp non-linearities can also cause serious systematic brightness errors in maps with high dynamic range between background level and target. In addition, ramp non-linearities introduce a larger formal uncertainty in the signal when deriving a signal value by fitting a straight line through the ramps.
The deviations from linearity for all detectors except for PHT-SS and SL have been determined with respect to a standard linear ramp which was constructed from a stack of several thousand individual ramps. An example of the resulting deviations is presented in Fig. 4.5.
 |
In the SPD processing the CRE output voltage for each readout is corrected before deriving the signals. Calibration tables for the P, C100, and C200 detectors have been created containing CRE output voltages and the corresponding voltage corrections. These calibration tables are stored in Cal G files (see section 13.5).
Non-linearities in the CRE output of PHT-SS/SL data are presently not corrected for. It is expected that calibration tables for PHT-S will become available in a future OLP version.
Measurements of a given target obtained with the same detector/filter/aperture combination but with different readout timing (ROT) parameters have indicated that the signal can depend on the commanded reset interval. For a constant detector illumination the signal level is found to vary systematically from one reset interval to the other.
For C100 and C200 these variations are of the order of 10-20% between measurements which differ in reset interval by a factor of 2. For reset intervals a factor of 8 apart the signal difference can be as large as 40%. P1, P2 and P3 show much less variations. For the PHT-S detectors the situation is still not completely clarified.
As long as the FCS calibration measurement is obtained with the same reset interval as the target measurement no corrections are necessary. However, in case of multi-filter photometry, the reset interval used in each filter measurement is generally different from the one in the FCS measurement. Also for multi-aperture observations where the signal range is large, different reset intervals are used for different aperture measurements.
The effect can be removed by correcting all signals with respect to a single reference reset interval. Correction tables are included in OLP Version 7 (cf 7.3.1).
Detectors PHT-SS, PHT-SL, and PHT-P1 are high bias, PHT-P2 medium and the PHT-P3, -C100, -C200 low bias detectors (see Table 4.3). For the low bias detectors de-biasing can occur when the full integrator voltage is used. Infrared photons induce charge carriers in the detector which are accumulated on the feedback capacitor of the readout electronics (Cint, cf. Fig. 3.2) during the integration. An electrical field is built up which reduces the actual bias. The stronger the illumination, the stronger the field and herewith the bias reduction. For the C200 detector the effect is strongest: the bias can drop by factor about 2 from 80 down to 40 mV for very high dynamic range. This effect causes the ramps to become non-linear with a downward curvature.
| Detector ID | Default Bias (V) |
|---|---|
| SS | -37 |
| SL | -37 |
| P1 | -90 |
| P2 | -10 |
| P3 | -0.25 |
| C100 | -0.2 |
| C200 | -0.08 |
In contrast to the CRE ramp non-linearity (section 4.3.2), the non-linearities caused by de-biasing depend on i the CRE reset level (cf. section 3.2) which can float and ii the strength of the photo-current. This complicates possible solutions for correction. Presently no distinction is made between de-biasing and CRE non-linearity when correcting the ramps. For a given detector, there is one correction table taking care of both effects. Section 7.2.7 describes the correction that has been applied by the SPD processing software.
Consequently, signals are more underestimated the stronger the signals are. Signals derived from partially saturated ramps (section 4.3.5) suffer most noticably from this effect. Observers who collected maps with high dynamic range in bright regions using P3, C100, or especially C200 should be aware of this effect before interpreting their results.
In addition, for high signals on low bias detectors charges on the feed-back capacitor can cause an increased glitch rate (section 4.4) of the detector.
If the astronomical source is brighter than anticipated it may happen that parts of the integration ramps are saturated, i.e. the maximum CRE voltage level is reached before the next destructive readout. For most detectors the subsequent readouts after saturation remain at the maximum CRE voltage level until the next destructive readout. However, for some detectors the voltage of the saturated readouts can drop back below the threshold voltage.
The same can occur in case the actual detector responsivity exceeds the nominal responsivity by a large factor. A nominal responsivity was used to a-priori set the ROT parameters before execution of the observations.
The readouts taken during CRE saturation cannot be used and should be discarded. During SPD processing, all readouts in a ramp that are subsequent to a saturated readout are removed (section 7.2.6). The threshold CRE voltages for saturation are presented in Table 4.4.
| Detector | Threshold (V) |
|---|---|
| P1 | 1.091 |
| P2 | 1.091 |
| P3 | 1.097 |
| C100(1,2,3) | 1.063, 1.088, 1.095 |
| C100(4,5,6) | 1.097, 1.098, 1.099 |
| C100(7,8,9) | 1.098, 1.098, 1.097 |
| C200(1,2) | 1.091, 1.090 |
| C200(3,4) | 1.091, 1.088 |
| PHT-SS(1..64) | 1.040, 1.093, 62 1.095 |
| PHT-SL(1..64) | 1.007, 1.099, 621.102 |
The signal can still be determined from the remaining readouts that are below the CRE threshold voltage.
The first ramp of each chopper plateau is always disturbed due to the electronics commanding the chopper movement. Even in case of a staring measurement the chopper is still commanded to perform a zero deflection after integrating for 128 sec. This virtual chopper step can be noticed in the datastream of a long staring measurement where one disturbed ramp occurs every 128 sec.
For the disturbed ramps the CRE reset level (cf. section 3.2.2) is high and also the ramp signal is higher than the average signal derived for the corresponding chopper plateau.
As part of the SPD processing, the first ramp on each chopper plateau is discarded and therefore not taken into account when computing the average signal per chopper plateau.
The CRE for the ISOPHOT detectors were fabricated in CMOS technology. A known phenomenon of CMOS devices is the occurence of high supply currents after switch-on, which is called latch-up. The effect is due to the presence of a parasitic transistor in the CRE. A small amount of charges left in the basis region of this transistor can cause it to conduct thereby creating the high supply current.
During the instrument activation sequence at the beginning of each revolution a latch-up recovery is performed which resets the CREs into their nominal status. The recovery procedure consists of a pseudo measurement which selects and subsequently deselects a given CRE.
Besides the switch-on effect some CREs tend to show latch-up effects during the science window. As a result, the integration ramps become very noisy. The latch-up is accompanied by an increase of the CRE supply voltage above a certain critical limit. The CRE supply voltages for each detector-CRE combination are monitored by the instrument controllers.
When a latch-up event occurs the measurement is aborted in real time by the instrument controllers. Subsequently they must initiate a latch-up recovery procedure. Due to the duration of this procedure also the following AOT could be skipped. All affected AOTs are flagged as failed and the observations enter the queue for rescheduling.
During the first 270 revolutions it was found that the P2 detector suffered from latch-up events when switching the device on. This problem was solved by introducing a proper command timing and introducing for safety a pseudo measurement in the AOT. The inclusion of this measurement required a change in AOT logic which adds an extra overhead of 10 sec to the AOT on-target time. The pseudo measurement was inserted when one or more P2 measurements were selected in PHT03, PHT04, PHT05, or PHT17. From revolution 690 onwards the AOT logic was changed such that no pseudo measurements were taken anymore. The timing overhead remained the same.
The pseudo measurement is present in the telemetry data (of revolutions 270-690) but is removed in Derive_ERD and should be transparent to the observer who starts processing from ERD.
ISOPHOT Data Users Manual, Version 4.1, SAI/95-220/Dc