III.A. Time-Ordered Detector Data Improvements

ISSA Explanatory Supplement
III. PROCESSING
A. Time-Ordered Detector Data Improvements


Chapter Contents

Table of Contents | Index | Previous Section | Next Section


  1. Positional Improvements
  2. Calibration Improvements
    1. Detector Response Function
    2. Zero Point Calibration
    3. Other Calibration Enhancements
  3. Deglitching

A. Time-Ordered Detector Data Improvements

The ISSA images, like the SkyFlux images (Main Supplement §V.G), were made from in-scan, time-ordered detector data that were calibrated, positionally phased, compressed, position-tagged, filtered and resampled. The compressed, time-ordered database used by ISSA incorporates improved boresight and calibration information. In addition, radiation hits were removed from the time-ordered data used in making the ISSA. These improvements are discussed below.

Time-ordered detector data were smoothed using the same algorithm as described in the Main Supplement, §V.G for SkyFlux. This algorithm smoothed and resampled the IRAS detector data from 16, 16, 8 and 4 samples per second at 12, 25, 60 and 100 µm, respectively, to two samples per second at each wavelength. Additional information is found in Appendix B of this Supplement.


A.1 Positional Improvements

Positional calculations were improved since the SkyFlux processing by the following corrections and modifications. Most important was the correction of an error that advanced the in-scan position by 115" for half the mission data. This error was found in the SkyFlux images and the ZOHF Version 2.0. No other data products were affected. A second improvement was in the data phasing. Phasing is the process by which the individual detector data streams are realigned with respect to each other to bring together samples taken at the same in-scan sky position. The satellite scan rate used for this process was changed from the initial scan rate of an observation to its average rate. A third improvement involved implementation of a new algorithm for the position computation. The cumulative effect of the position corrections and improved interpolation scheme is quantified for the ZOHF in Appendix H, Table H.4.

Although pointing reconstruction errors were a relatively small contributor to the original position errors, improvements in the satellite pointing reconstruction made to support the IRAS Faint Source Survey were also incorporated in the time-ordered detector data (Explanatory Supplement to the IRAS Faint Source Survey, §II.B). In general, pointing reconstruction improvements reduced the in-scan 1-sigma boresight uncertainties from 3.0" to 1.5" and the cross-scan 1-sigma from 4.5" to 3.0". In addition, many scans that had anomalously bad pointing were improved to bring them to the same accuracy as the other scans.


A.2 Calibration Improvements

Several important changes were made in the IRAS calibration software. An improved model of the detector response function that corrects the first-order effects of the radiation-induced and photon-induced responsivity enhancement was implemented. Improvements were made to the model of the Total Flux Photometric Reference (TFPR), which was used in maintaining the zero point of the IRAS detectors. Improved estimates of the solid angles of the detector fields of view were used and the measurement of the internal reference source was derived using a more robust algorithm. In addition, an empirical method for reducing scan-to-scan variations was implemented at 25 µm (§III.C.1).

A.2.a Detector Response Function

The response of each detector was known to be enhanced due to radiation and photon exposure (Main Supplement §VI.B.4). This responsivity enhancement is referred to as the hysteresis effect. A response function for each detector was implemented that models hysteresis at the point source frequency. The model corrects all detectors for radiation-induced responsivity enhancement due to the South Atlantic Anomaly (SAA) and the 60 and 100 µm detectors for photon-induced responsivity enhancement. At 12 and 25 µm, the photon-induced responsivity enhancement that created the point source tail artifacts was not removed by this model. Point source tails remain in the data.

Figure III.A.1 Point Source Hysteresis Comparison - 100 µm. This figure shows the 100 µm point source flux discrepancy due to the hysteresis effect across the Galactic plane. A set of 100 µm point sources were selected along ecliptic longitude 270°. Fluxes were measured and compared from ascending and descending scans and the percent difference between the ascending and descending scans was computed. The percent differences were averaged within a 5°× 10° bin and plotted. The Galactic plane crossing is around -15° ecliptic latitude. The solid line represents the values from uncorrected scans and the broken line represents the values from hysteresis-corrected scans.
larger largest
The parameters for the hysteresis model were derived based on the history of the point source responsivity as measured by the internal reference source after SAA crossings and Galactic plane crossings. The internal reference source was used at the beginning and end of each survey scan to monitor the point source responsivity of the system. A history was kept of the point source responsivity for each detector throughout the mission. From this history, the responsivity measurements were sorted and organized based on time from SAA crossing and again based on time from Galactic plane crossing. These two different datasets were used in deriving time constants for each detector to represent the exponential decay of the responsivity after particle radiation or photon exposure. The hysteresis formula is given in Table III.A.1. The detector response due to photon exposure as shown in the equation is R(t) and is only applicable to the 60 and 100 µm detectors. For the 12 and 25 µm detectors the value of R(t) is zero. The derived detector time constants are listed in Tables III.A.2(a)-(d). The improvement at 100 µm in tracking the point source responsivity is seen in Figure III.A.1. This figure shows the average percent difference within a 5°× 10° bin in the point source fluxes as measured from ascending and descending scans along ecliptic longitude 270°. At this longitude the Galactic plane is crossed around -15° ecliptic latitude. For this set of point sources, the original uncorrected response function resulted in the ascending scans overestimating the point source flux values after the Galactic plane crossing by 10%-15% compared to the descending scans (solid line). The hysteresis-corrected point source fluxes (broken line) show a reduced effect across the Galactic plane. Discrepancies on the order of 6% RMS remain.

Table III.A.1 Hysteresis Equation
R(t) = [A + B e- t/B]/[1 - R(t)]    for T1 < t < T2
A = R(T1) - [ R(T1) * R(t)] - B*e-T1/B
B*e- t/B = e-(t - T1)/B/ [1 - e- (T2 - T1)/B] * [(R(T1) - (R(T1) *R(T1)) - (R(T2) - (R(T2) *R(T2)))]
R(t) = min(R(t - )*e/p + K*Fint(t - ), Rmax)
K*Fint(t - ) = K* Fint(t - ) if Fint(t - ) >= Threshold
           = 0 if Fint(t - ) < Threshold

R = total detector response
R = detector response due to photon exposure
B = bias boost time constant
p = photon exposure time constant
K = max %/saturation (Joules)
Fint = integrated flux over time interval measured in Joules
= Delta time
t = time from last bias boost
T1 = time of first internal stimulator
T2 = time of second internal stimulator

Table III.A.2(a) Time Constants, 12 µm
Detector #Tau for Bias
Boost (sec)
231200
241200
251200
261200
271200
281200
291200
301200
471200
481200
491200
501200
511200
521200
531200
541200
Mean Time Constant
Standard Deviation
1200
0

Table III.A.2(b) Time Constants, 25 µm
Detector #Tau for Bias
Boost (sec)
391200
401200
411300
421300
431700
441500
451500
461000
161000
17 --
181200
191000
20 --
211000
221200
Mean Time Constant
Standard Deviation
1238
222

Table III.A.2(c) Time Constants, 60 µm
Threshold = .6E-11 Joules, 1.27E-6 Wm-2sr-1
Saturation = 3.2E-10 Joules, 6.76E-5 Wm-2sr-1
Detector #Tau for Bias
Boost (sec)
Tau for Photon
Exp. (sec)
Max.
Effect (%)
86333836
9 782 400 3
10 914 407 6
11 10000 476 6
12 10000 420 12
13 785 568 5
14 828 351 8
15 10000 250 8
31 10000 476 7
32 10000 439 10
33 10000 340 10
34 910 350 10
35 10000 626 5
36 -- -- --
37 10000 430 13
Mean Time Constant
Standard Deviation
--
--
419
93
8
3

Table III.A.2(d) Time Constants, 100 µm
Threshold = .6E-11 Joules, 0.57E-6 Wm-2sr-1
Saturation = 3.2E-10 Joules, 3.04E-5 Wm-2sr-1
Detector #Tau for Bias
Boost (sec)
Tau for Photon
Exp. (sec)
Max.
Effect (%)
55 1200 1590 22
56 980 756 23
57 2200 1554 16
58 1400 1540 20
59 1200 1565 16
60 1200 1667 20
61 1600 1616 20
62 1450 1560 18
1 1320 1460 24
2 1490 1415 17
3 1600 1867 8
4 1100 1547 23
5 1415 1420 16
6 1000 704 13
7 1000 401 12
Mean Time Constant
Standard Deviation
1344
316
1377
413
18
5

Detector responsivity is a function of spatial frequency. Although the hysteresis model was derived from data taken with the internal stimulators, which measure the point source or AC frequency response, it was assumed that this model would represent the hysteresis effect at all spatial scales. Only the factors discussed in §II.B.2 were used to scale the point source responsivity to an extended emission responsivity prior to producing the ZOHF and ISSA products. To verify that the hysteresis model was effective for extended spatial scales, ascending and descending scans (before and after hysteresis correction) in the 0.5° ZOHF were compared. The result of this comparison showed the same hysteresis effect existed for extended spatial scales at 60 and 100 µm as for point sources. After hysteresis corrections were applied at 60 and 100 µm, a 5%-6% discrepancy remains between 6° and 15° of the Galactic plane. Larger uncertainties still occur within 6° of the plane.

A.2.b Zero Point Calibration

The detector calibrated zero points were maintained by daily reference to a patch of sky of measured brightness near the north ecliptic pole called the Total Flux Photometric Reference (TFPR) (§II.B.4). The brightness of the TFPR varies with time largely due to the Earth's annual motion through the cloud of interplanetary dust surrounding the Sun. A model of this variation was developed for use with the daily calibration observations. The method used to measure the brightness of the TFPR and the assumptions made to develop the TFPR model are the same as used for SkyFlux processing. This is described in the Main Supplement §VI.B.3. A brief description is repeated below for completeness. Improvements to the TFPR model used in the ISSA processing are explained below.

A sinusoidal variation added to a constant term was found to be a reasonable model for the TFPR brightness. The largest annual variation is due to the tilt of the symmetry plane of the zodiacal dust distribution with respect to the orbital plane of the Earth causing a variation in the line-of-sight path length through the dust cloud toward the north ecliptic pole. A secondary contribution is due to the eccentricity of the Earth's orbit that causes changes in the temperature and density of the interplanetary dust as the Earth's distance to the Sun changes. Some of the constant term in the TFPR model is due to the Galactic emission toward the north ecliptic pole.

To determine the constant term of the TFPR model, the brightness of the TFPR was measured between eight and ten times, depending on wavelength, during the IRAS mission using a special observation called the Total Flux CALibration, TFCAL. The TFCALs were based on the fact that two observations of the TFPR at different responsivities would yield both the absolute brightness of the TFPR and the zero point of the electronics. The change in the responsivities for the 12 µm detectors was achieved by use of the alternate bias level available to those detectors. For detectors at 25, 60 and 100 µm, the TFCAL observations made use of the responsivity enhancement caused by the heavy exposure of the detectors to the protons trapped in the South Atlantic Anomaly (SAA). Normally, a bias boost was applied during and immediately after SAA passages to anneal the detectors and return the responsivity to normal. For execution of the TFCALs, the bias boost annealing was delayed for a fraction of an orbit until the satellite could point to the TFPR. Two observations of the TFPR were made separated by the bias boost annealing cycle. Flashes of the internal stimulators during both TFPR observations calibrated the responsivity before and after the bias boost. Under the assumption that the electronic zero point remained unchanged by the bias boost, the brightness of the TFPR was extracted using this method. Responsivity variations of 300 to 400% were obtained at 60 and 100 µm, while variations of 30% were typical for 12 and 25 µm.

An important detail of the implementation of the TFCAL observations is the assumption that the bias boost did not alter the electronic zero point of the detectors. This was in fact an erroneous assumption. The bias boost did indeed change the electronic zero point of the detectors in most boosted modules due to heating of the cold electronics by the boosted bias current. This however was successfully modeled for removal in the TFCAL reduction process.

Independent information was obtained concerning the initial zero point for each detector from a single `chop' experiment performed during the first week of the IRAS mission. The cryogenically cooled cover which was still in place allowed zero background conditions for detectors at 12 and 25 µm. Measurements agreed with results from the TFCALs to within 6% and 10% at 12 and 25 µm. No measurements were obtainable at 60 and 100 µm because of uncertainties in the 60 and 100 µm background levels with the cover on (Main Supplement §VI.B.3.a).

In principle, the sinusoidal parameters of the TFPR model could be determined from the TFCAL measurements alone. However, the limited number of TFCALs were insufficient to yield an accurate phase and amplitude of the sinusoidal component. Instead, a measure of the annual variation was available in the form of differences between the north and south ecliptic pole brightnesses derived from about 200 IRAS survey scans. Each scan observed both poles within 50 minutes. The difference between the polar brightnesses removed drifts on time scales greater than 50 minutes. The annual variation in the brightness at the TFPR was then derived by fitting a sinusoid to the polar differences. The amplitude of the annual variation at the TFPR is then half the variation derived from the differences. This observationally determined the effect of the Earth's motion with respect to the symmetry surface of the zodiacal dust cloud. The polar difference had the undesirable effect of canceling out the TFPR brightness variations due to the eccentricity of the Earth's orbit. To account for the eccentricity of the Earth's orbit in the TFPR model, it was necessary to add back a model which represented this variation. When results of the TFCALs observations were combined with data extracted from survey scans connecting the north and south ecliptic poles, a smooth, sinusoidal variation in the TFPR brightness was apparent. Two significant changes were made in the TFPR model used to produce the ISSA and ZOHF Versions 3.0 and 3.1. Unlike the previous TFPR model, the current model includes the effect of the eccentricity in the Earth's orbit about the Sun as calculated from the zodiacal emission model of J. Good (Appendix G). The special calibration observations, the TFCAL observations, which determine the constant term of the TFPR model (also described in §VI.B.3 of the Main Supplement), were re-analyzed with noticeably improved results. The improved values for the TFPR model are found in Table III.A.3.

Table III.A.3 TFPR Model Parameters1
Effective wavelength (µm)122560100
Parameter:2
B0 (MJy/sr)312.523.38.19.6
statistical uncertainty50.31.20.080.2
total uncertainty61.63.10.471.3
B1 (MJy/sr)31.732.660.670.19
uncertainty70.10.10.10.05
phi (day)4-38.3-32.8-34-31
uncertainty71.61.589
1The parameters have been converted to sky brightness (MJy sr-1) in order to illustrate the relative magnitudes of the parameters. The parameters were originally derived relative to the flashes of the internal reference source.
2At a time t in days the model assumes B[TFPR] to be:
B[TFPR] = B0 + B1 × sin[(2π/365.25)× (t-phi)]
3The usual convention of using a flat spectral distribution for the sources was followed in deriving the flux densities.
41983 January 1 (UT) is t = 1.0 days.
5The statistical uncertainty corresponds to the standard deviation in the fit to the observations.
6The total uncertainty incorporates uncertainties from stimulator flash stability, baseline drift corrections, frequency response, feedback resistor nonlinearities and solid angle uncertainties.
7This uncertainty is obtained from a combination of statistical uncertainties within model fits and the dispersion among fits to different subsets of the IRAS pole-to-pole scans.

The internal consistency of the TFCAL observations is now 2% or better of the TFPR brightness at 12, 60 and 100 µm and 5% at 25 µm. The zero point uncertainties in the TFPR model based upon internal inconsistencies are now 0.36, 1.2, 0.17, and 0.4 MJy sr-1, at 12, 25, 60 and 100 µm, respectively. The uncertainties in the basic responsivity calibration of the IRAS data traced back to standard stars and the asteroid model remains 2%, 5%, 5% and 10% at 12, 25, 60 and 100 µm, as discussed in §VI.C.2.c on page VI-24 of the Main Supplement. The actual zero point uncertainties of the survey observations are larger than those of the TFPR model due to baseline drifts on time scales shorter than one day, variation of the sky position observed as the TFPR (§II.B.4) and other systematic effects discussed in §IV.


A.2.c Other Calibration Enhancements

The accuracy of the calibration was enhanced by the use of more accurate solid angle measurements for the detectors (see §I.D.7 and the Explanatory Supplement to the IRAS Faint Source Survey Version 2, §II.D.2) and a more robust method of extracting internal calibration flashes in confused areas of the sky. The improved solid angles resulted in sky brightness shifts of 13% at 12 µm, 8% at 25 µm, 3% at 60 µm and 6% at 100 µm. These band average estimates are calculated for full-sized detectors and only affect the extended emission calibration. Calibration stability was improved by a few percent due to the improved accuracy in measuring the internal reference source.

A.3 Deglitching

The time-ordered detector data were improved by the removal of radiation hits. Even though IRAS used an onboard deglitcher, many artifacts with an amplitude of less than 100 times the sample noise remained in the data. These artifacts were typically due to charged particles impacting the detectors (Main Supplement §IV.A.6). The deglitcher used by the ISSA is the same as that used for the IRAS Faint Source Survey (Explanatory Supplement to the Faint Source Survey, §II.C.2.a). The algorithm operated on the time-ordered detector data prior to phasing and compressing. The processor monitored the output of a high-pass filter over the detector data streams for events that exceeded the local detector noise by a factor of five (SNR > 5). Events with power at a frequency greater than the point source frequency threshold were identified as glitches. The glitches were replaced using linear interpolation between the data points on either side of the offending glitch and a quality flag for the interpolated sample was set appropriately to signal downstream processors that deglitching had occurred. No interpolated data were used in creating the ISSA images. The deglitch filter removed more than 95% of the radiation hits with SNR > 5. Most of the removed glitches were at 12 and 25 µm.


Chapter Contents

Table of Contents | Index | Previous Section | Next Section