Subsections

• 7.2.1 Discarding destructive readouts
  • Ancillary data required:
• 7.2.2 Discarding disturbed non-destructive readouts
  • Ancillary data required:
• 7.2.3 Discarding data with OTF or OPF `off'
  • Ancillary data required:
• 7.2.4 Data identification and separation
  • Ancillary data required:
• 7.2.5 Determine chopper position and step number
  • Ancillary data required:
• 7.2.6 Conversion of digitised numbers to CRE output voltages
  • Ancillary data required:
• 7.2.7 Discarding saturated readouts
  • Ancillary data required:
• 7.2.8 Correction for non-linearities of the integration ramps
  • Ancillary data required:
• 7.2.9 Ramp deglitching
  • Ancillary data required:
• 7.2.10 Extraction of the signals and their uncertainties
  • Ancillary data required:

7.2 Ramp Processing

7.2.1 Discarding destructive readouts

Detailed description: Section 2.4.2

The destructive readout is the last readout of a ramp in parallel to the reset of the readout electronics and is in general disturbed by the pulse pattern of the electronics. Destructive readouts usually do not follow the highly linear relationship between time and readout voltage as is the case for the non-destructive readouts.

In Derive_SPD the destructive readout is assumed to be unreliable and is generally discarded. The only exception is for PHT32 processing, see Section 7.7

Ancillary data required:

None

7.2.2 Discarding disturbed non-destructive readouts

Detailed description: Section 4.3.1

At the beginning of an integration ramp immediately after a reset, the first non-destructive readout is unreliable. In case of a large number of non-destructive readouts per ramp, more than one readout can be disturbed.

In Derive_SPD a consecutive number of readouts at the beginning of an integration ramp is discarded. The value depends on the detector and the length of a ramp in terms of non-destructive readouts.

Ancillary data required:

The number of readouts to be discarded as a function of (1) detector and (2) the length of a ramp are stored in Cal-G file PSELNDR, see Section 14.4.

7.2.3 Discarding data with OTF or OPF `off'

Detailed description: none

During a sky measurement the source can drift out of its nominal pointing due to telescope pointing instabilities. When this happens the On-Target Flag (OTF) is automatically set to `off'.

Similarly, if the chopper is not correctly positioned, a chopper On-Position Flag (OPF) is also set to `off'.

Initially, until revolution 524, the OTF is set to `off' when the actual pointing drifts out of a cone with radius of $10''$ around the intended pointing. Due to the good pointing performance of ISO the OTF cone has been reduced to $2''$ for ISOPHOT as of revolution 524.

The OTF and OPF flags are contained in the readout records of the ERD products. If either of these flags is set to `off' then the current integration ramp is abandoned and excluded from further processing. A record is kept per pixel per measurement of the number of integration ramps that are rejected for this reason.

For raster scans, it is assumed that the spacecraft moves off-target to the next raster point, when one of the two raster point identifiers changes. To avoid the inclusion of off-target data while the spacecraft is slewing and repositioning all data are rejected for a fixed (1 s) period after the change in the raster point identifier. After this period the OTF is checked: if the OTF is `off' then data are rejected until the OTF is `on' again.

Ancillary data required:

None

7.2.4 Data identification and separation

Detailed description: none

The different measurements that were performed as part of an AOT (like sky, FCS, and dark measurements) need to be identified and separated before they can be processed. For a given TDT all output data records of the same detector and measurement type will be stored inside the same SPD product.

Different measurements are sorted out by performing a cross-correlation between the ERD record and the corresponding CSH record with the same time key. The following flags are then inserted at the beginning of each data record in Derive_SPD to identify the type of measurement:

• Raster point ID
• Aperture ID
• Chopper position
• Filter ID
• Polariser ID

Note that the filter ID also determines the detector.

Ancillary data required:

None

7.2.5 Determine chopper position and step number

Detailed description: Sections 2.4.4, 2.8 and Appendix B.3

The telemetry contains the actual chopper position which is recorded every 2 s. The data are provided as voltages and must be converted to a deflection angle. The conversion algorithm can be found in Appendix B.3.

The chopper dwell time is directly related to the instrument readout timing parameters (see Section 2.4.4) and is therefore a reliable parameter in the instrument commanding. Using the chopper dwell time corrected for the chopper movement (Section 2.8), the moment (in ITK units) is determined when the chopper should have changed its position.

From the inferred chopper transitions and the ITK of the readouts the relative position of the chopper is obtained. In principle the same can be done from direct analysis of the chopper position in the telemetry data. The reasons why time is used rather than the chopper position are:

• the chopper position associated with the first integration ramp after a chopper transition is unreliable (on-board software feature), and
• the algorithm allows recovery of the correct position in the chopper cycle, if there is a gap in the data due to telemetry dropouts.

To remove possible contamination by unreliable chopper positions in the data products, a median filter is applied to the positions associated with each integration for a fixed chopper position.

Before the chopper position can be calculated it is necessary to determine the step flag $I_{step}$ within a cycle. The total number of chopper plateaux in a chopper cycle depends on the chopper mode as follows:

Staring	1
Rectangular	2
Sawtooth	$2 \times steps + 1$
Triangular	$4 \times steps$

Where $steps$ indicate the number of chopper offset positions at one side of the central field of view (CFOV). The ITK is used to calculate the position in the cycle. In general, for rectangular, sawtooth and triangular chopping, the first chopper position is an `off-source' position, the second position is an `on-source' position (Section 2.8). This is then digitised to the step flag $I_{step}$ as follows:

Staring:

$I_{step}$ = 1, unless pointing at the FCS, in which case it is defined to be $-$1.

Rectangular:

$I_{step}$ = $-$1 on-source, 1 for background.

Sawtooth:

For all chopped AOTs exept PHT32: $I_{step}$ = $-$1 on-source, 1 for background. For PHT32: $I_{step}$ varies from $-steps$ at maximum negative angle to $+steps$ at maximum positive angle. Step flag 0 represents the CFOV.

Triangular:

$I_{step}$ = $-$1 on-source, 1 for background.

Ancillary data required:

None

7.2.6 Conversion of digitised numbers to CRE output voltages

Detailed description: Section 2.4.3

The Detector Interface Electronics (DIE) in the PHT external electronics unit subtracts a commandable offset from the CRE voltage and amplifies the difference with a commandable gain factor before the analogue to digital conversion. The CRE output voltage must be reconstructed from the Digitised Numbers (DN).

There are two DIE chains. Each detector subsystem has a default connection to one of them (see Section 2.4.3). A change in default connection is indicated by the `cross status' flag in the CSH (actually never used during the mission). The CSH also contains the value of the selected gain of the differential amplifier in the DIE electronics. The CRE output voltage $U_{CRE}$ and DNs are related by the following formula for a given DIE electronics chain and selected gain:

$$U_{CRE} = (D_{o}-DN-G_{signal}\times (DI-2048))\times \frac{20.0}{4096\times G_{off}} + U_{off}~~~~~~{\rm [V]}$$ (7.1)

where,

$D _{o}$ = fixed offset

$G_{off}$ = offset gain

$DI$ = selected OFFSET data word (0...4095)

$G_{signal}$ = signal gain

$U_{off}$ = $G_{signal}$ dependent offset

Ancillary data required:

1. The selected OFFSET from the measurement record in the CSH.
2. The DIE transfer function tables are stored in Cal-G files PDIE1TRANS and PDIE2TRANS, see Section 14.3.

7.2.7 Discarding saturated readouts

Detailed description: Section 4.3.5

If the source is significantly brighter than anticipated, integration ramps can saturate, i.e. the CRE output voltage has reached its maximum value prior to reset. The readouts taken during times of detector saturation are useless and must be discarded. Saturation also occurs in case the responsivity of the detector exceeds the nominal value by a large factor. In practice this frequently happened towards the end of the science window where both FCS and sky measurement got saturated due to high responsivity. For detectors P3, C100 and C200 many FCS measurements taken during revolutions 94-191 were saturated due to over-illumination by changed FCS behaviour.

Threshold voltages for each detector are listed in Section 4.3.5. As soon as the threshold voltage has been crossed, all following readouts up to the end of the ramp are discarded.

Ancillary data required:

It is assumed for all detectors that saturation is reached for CRE output voltages greater than 1.0 V, this is with some safety margin below the actual thresholds given in Table 4.4.

7.2.8 Correction for non-linearities of the integration ramps

Detailed description: Sections 4.3.2 and 4.3.4

Integration ramps are not perfectly straight but show deviations from linearity. The non-linearity is caused by two independent effects:

• non-linearities due to the CRE (Section 4.3.2)
• downward curving due to de-biasing. Only low-bias germanium detectors P3, C100 and in particular C200 show this effect (Section 4.3.4).

The integration ramps are corrected for non-linearities before signals are derived.

In Derive_SPD it is assumed that the corrections are a function of only the absolute value of the CRE output voltage. It is assumed that non-linearities due to both CRE and de-biasing (as is the case for the germanium detectors P3, C100 and C200) can be corrected using one function which only depends on the CRE output voltage.

The ramps are linearized using tables which contain for a given CRE output voltage the correction voltage to be subtracted. In the tables the sampling of the CRE voltages is sufficiently fine to allow searching for the table value closest to the measured CRE voltage and using its corresponding correction.

Ancillary data required:

The CRE Transfer Function Table per detector pixel and clock frequency are stored in Cal-G files PC1CRELIN, PC2CRELIN,, and PPCRELIN, (Section 14.5). There is no ramp linearisation for PHT-S measurements.

7.2.9 Ramp deglitching

Detailed description: Section 4.4

A radiation hit or glitch shows up as a voltage increase between two subsequent readouts which is larger than the increase expected from the steady photo-current produced by the celestial or internal source illumination. In case the hit is very energetic, the voltage increase can be so high that it saturates a ramp or even causes a responsivity transient. In SPD two different deglitching algorithms are applied: the first algorithm is described in this section and is based on an analysis of the readouts per ramp; the second one checks for any outliers in the signals of a given chopper plateau and is described in Section 7.3.4.

For ramp deglitching, an iterative algorithm was implemented which identifies and removes excessive increases in CRE output voltage. The algorithm has the following settings:

$N_{it}$	=	4	number of iterations
${\kappa}_1$	=	4	minimum number of standard deviations for glitch detection (first threshold)
${\kappa}_2$	=	1	minimum number of standard deviations for second threshold
$N_{min}$	=	25	minimum number of readouts per ramp for application of algorithm
$N'_{min}$	=	32	minimum number of readouts per ramp for application of second threshold

 
For integration ramps with less than $N_{min}$ readouts no ramp deglitching is applied and the deglitching can only take place at signal level (Section 7.3.4). For an integration ramp consisting of $N > N_{min}$ readouts with voltages V(1), V(2),$\dots$V(N), taken at times t(1), t(2),$\dots$t(N), the slope between each consecutive readout is calculated:

$$s(k) = \frac {V(k+1) - V(k)} {t(k+1) - t(k)}~~~~~~~~~~~~~[V/s].$$ (7.2)

Outliers in $V(i)$ are always positive because of the extra photocurrent due to ionising radiation. The maximum $s_{max}(k)$ is removed from the $N-1$ differences prior to calculating the mean $S$ and standard deviation $\sigma$ of the voltage differences. The exclusion of the most extreme element makes the computation of the mean and sigma of the distribution more robust and efficient.

In case

$$s(k) > S + \kappa_1*\sigma~~~~~~~~~~~~~~~~~[{\rm V/s}],$$ (7.3)

then a glitch is detected in readout $k$, which will be flagged. If the ramp contains more or equal than $N'_{min}$ readouts, then this detection triggers the algorithm to change the threshold $\kappa_1$ into $\kappa_2$ for $l\,>\,k$ until

$$s(l) < S + \kappa_2*\sigma,~~~~~~l>k~~~~~~~~~~~[{\rm V/s}].$$ (7.4)

As soon as Equation 7.4 is satisfied, the threshold is reset to $\kappa_1$. All readouts which do not satisfy Equation 7.4 are flagged. This two-threshold deglitching is very efficient in removing the `tail' of a glitch. This tail is due to transient behaviour of the detector signal.

If readout $k$ is flagged then the voltages $k+1$ to $N$ are corrected by subtracting the excess voltage due to the glitch:

$$V_{corrected}(k+1) = (V(k+1)-V(k)) - (V(j+1)-V(j)) + {\Delta}V~~~~[{\rm V}], ~~~k=j,...,N-1$$ (7.5)

where ${\Delta}V=S*{\Delta}t$ is the mean voltage difference for all readouts of a given ramp.

The procedure is repeated $N_{it}$ times to remove successively smaller glitches.

A check is made on the parameter which defines the minimum number of readouts that can be processed. If this parameter is less than 7, then an error message is logged (PRDE) and the deglitching algorithm is not executed. This avoids the possibility of an execution error occurring if the minimum number of readouts parameter is set too low in the Cal-G file PCONTROL.

Ancillary data required:

None, the thresholds and other deglitching parameters are hardcoded.

7.2.10 Extraction of the signals and their uncertainties

Detailed description: Section 2.4.2

Of each ramp $j$, the slope $s(j)$ (or signal in V/s) is proportional to the photo-current which is a measure of the number of photons falling on the detector per unit time. In the SPD processing all valid readouts between two reset intervals are used to fit a first order polynomial.

The uncertainty ${\Delta}s(j)$ is the rms of the fit residuals.

Ancillary data required:

None