We summarize IPAC processing of data from the Infra-Red Telescope in Space to obtain spacecraft pointing information using STS, the J-band Star Sensor. The data solved include all but a few orbits from those (180-283) included in the north-scan ("pre-flip") period, and all orbits (346-586) from the south-scan, or "post-flip" period. Pointing uncertainties based on STS for the south-scan data were ~40" in-scan and < 1' cross-scan. For the north-scan data, taken at low |b| where source confusion and background variation were complicating factors, uncertainties are typically significantly less than 1' in-scan and 2' cross-scan, respectively.
| I. | Introduction |
| II. | Processing Overview |
| III. | Source Extraction |
| IV. | Position Reconstruction |
| V. | Quality Checks |
| VI. | Special Situations and Caveats |
| Appendix 1: | Glossary & Notation |
| Appendix 2: | Summary File Format |
| Appendix 3: | Tgood Orbit Event File |
| Appendix 4: | IPAC Boresight File Format |
| Table 3: | Summary File Segment Fit Listing |
| Table 4: | Tgood.ver5 Orbit File Event Listing |
| Figures: | Figure Captions |
Reconstructions have been completed for all orbits in the pre-flip packets (179-283, 105 orbits total), except for the first orbit (179) and orbit 272. For the post-flip data, which were generally less problematic because of reduced source confusion, essentially all orbits (of 241 total) have been reconstructed, and all but a very few are quite satisfactory. Over 90% of the 105 pre-flip orbits passed a set of strict quality standards, detailed in Section IV; significant parts of about nine orbits could not be reconstructed to pass these criteria. An even higher fraction of post-flip orbits passed. Orbits failing the criteria in Section IV have been flagged with pointing uncertainties of 99' in the results.
(Dr. Minoru Freund when he was at IPAC used NIRS data to successfully solve the 42 orbits, from 305-346, when the STS was turned off; and solved as well a key portion of orbit 180. This NIRS analysis has been described in a separate document, and is not treated here.)
In addition to describing the meaning and limitations of the information directly present in the boresight, or output files, this document includes some background material on IRTS and on IPAC internal files and programs, in order to provide a self-contained reference against future need. As a consequence, it contains a mixture of information important for external users, as well as a compilation of reference information mostly of interest within IPAC.
Pre-Flip Packet Orbits 03292312: 179 - 196 03310335: 197 - 209 03312358: 210 - 226 04020237: 227 - 240 04030028: 241 - 255 04031330: 256 - 271 04050109: 272 - 283 Post-Flip (STS on) 04092249: 346 - 363 04110242: 364 - 378 04120140: 378 - 393 04130040: 393 - 409 04140046: 409 - 425 04150222: 425 - 438 04152348: 438 - 454 04170049: 454 - 470 04180057: 470 - 485 04190035: 485 - 500 04200009: 500 - 516 04210105: 516 - 531 04220036: 531 - 546 04230012: 546 - 561 04232351: 561 - 576 04250016: 576 - 586
The ISAS InfraRed Telescope in Space ("IRTS", Nakagawa 1995), was a 15 cm, f/4 Ritchey-Chretien design, the focal plane of which was cooled to 1.9 K. The spacecraft, Space Flyer Unit, (SFU, Murakami et al. 1996) was launched into a 486 km, 28.5 degree orbit on 1995-03-18 (Figure 1). Scientific data were collected from 1995-03-29 at 23:48:40 UTC through 1995-04-25 11:34:19 UTC, after the 90 liters of liquid He cryogen were exhausted. The telescope looked out radially, nearly perpendicular to the spin axis of the SFU. The spin axis was approximately aligned towards the Sun. In order to avoid looking at the Earth, IRTS maintained a spinrate of approximately one revolution per orbit. A description of the four scientific instruments (cf Appendix 1) on IRTS and their scientific objectives has been given by Murakami et al. 1996. Figure 2 shows the geometry of the IRTS focal plane, and Table 2 gives its nominal dimensions, based on laboratory theodolite measurements made at 300 K. Note especially that the co-ordinate axes in Table 2 have been chosen to agree with those in Figure 2; but that in every other place in this document the co-ordinate definitions differ so that x and y are interchanged, making x the cross-scan co-ordinate, and y in-scan. For the post-flip data this interchange of x and y is all that is needed. However, since the in-scan co-ordinate always increases with time, its sense is reversed from that shown in Figure 2, for pre-flip only.
In addition to the four main scientific instruments, the IRTS payload included a J-band (1.25 micron) Ge photodiode Star Sensor (Murakami et al. 1994), the STS, capable of sensing stars as faint as 6 mag in unconfused regions of the sky, and yielding positions with an accuracy of the order of 1 min of arc, depending on the brightness of the star and the spatial structure of the background. STS data, with augmentation by NIRS in special situations, were the basis for the position reconstructions obtained at IPAC. The STS had an approximately square field-of-view (FOV), nominally 18 X 18 arcmin in size, centered 1 degree below the center ("boresight") of the IRTS focal plane (Figure 2). A diagonally obscuring stripe resulted in a 2-lobed response, with the position of the dip between lobes giving the cross-scan (x) co-ordinate with RMS accuracy typically about 1.5'. In-scan (y) positions were typically measured with an accuracy of ~0.4'. Figure 3 ( a, b) shows maps of the STS response.
Pt. STS FILM MIRS NIRS C1 (-10.84,-63.0) ( 1.50, 51.0) (-55.16,-2.0) (60.50,-1.0) C2 ( 6.50,-63.0) ( 1.84, 72.0) (-55.16, 5.0) (60.50, 6.0) C3 ( 6.50,-46.0) (-6.17, 72.0) (-63.17, 5.0) (52.16, 6.0) C4 (-10.84,-45.0) (-6.50, 51.0) (-63.17,-2.0) (52.16,-1.0)
Nominal focal plane geometry adopted by IPAC in arc min. Cf. Figure 2. (Note the co-ordinate interchange between x and y, as discussed in the text.)
A glossary of special terms appears in the Appendix.
In this document we have adopted the convention that computer
programs and filenames (eg, manfit) appear in
computer type style, whereas normal mathematical symbols and
multi-character program variables names (eg, dsy) appear
in italics.
Figure 1 shows the IRTS orbit geometry with key events and segment boundaries indicated. Times in the boresight ("IPAC att_lan") files appear as "IRTS-time" (or "launch time"), seconds since launch on 1995-03-18 at 08:01:00 UTC, to be distinguished from "IPAC-time", seconds since 1995-03-15 at 00:00:00 UTC. Thus for the times given in seconds, IPAC-time = IRTS-time + 288,060 s. In general, times in IPAC files and programs other than the boresight files are in IPAC time. The beginning of the orbit is denoted "Point D", and occurs 90 degrees in mean anomaly after the mid-point of orbit day. The "A" segment of the orbit extends from Point D until orbit sunset. The "B" segment, which is identical with orbit night, follows. The final, or "C" segment extends from sunrise until the succeeding Point D. Other portions of orbits have been labeled with a 2-letter code, for the beginning and ending segment. The designations "AB" and "CC" are particularly common and important, the former including A+B, the latter all of the C segment. These have normally been used as reconstruction fit intervals.
Appendix 2 and Table
3 summarize the solutions obtained for each segment, as
described in the summary files. Here and elsewhere,
all celestial positions are given for epoch and equinox B1950.
Appendix 3 and Table
4 give details of the modified orbit event, or
"Tgood" files, which contain comprehensive
information about a variety of on-orbit events and conditions.
It became apparent early in the analysis that the pre-flip data would be considerably more difficult to interpret than the post-flip data due to the greatly increased source confusion problems associated with the low inclination to the galactic plane of the pre-flip scans. Thus the post-flip data were in general processed earlier, and the pre-flip data later, so that the pre-flip could benefit from the improved techniques and lessons learned on the post-flip. Thus the discussion must often distinguish between pre-flip and post-flip, as the analysis methods, fitting, and quality criteria were sometimes different.
The total number of match stars for the 103 reconstructed pre-flip orbits is ~6000, for an average of ~58 stars per orbit. The most detailed information about the accuracy of the solutions obtained can be found in the time-series plots of residuals and histograms of residuals. In this section we present plots of pre-flip in-scan and cross-scan residuals against angular distance from Point D, and histograms of the same, for the seven pre-flip packets altogether and for each packet individually. Since the distributions are for single STS star observations, the least-square aspect solution should normally be considerably better than the RMS width of the histograms, depending on the number of stars fit in the segment. On the other hand there are situations, as for example one can see by the cross-scan ripple (cfSection VI.D. below) where the residuals are clearly not normally distributed, and the error may be larger. Figure 4a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in the seven pre-flip packets. Figure 4b shows the corresponding residuals' histograms.
Figure 5a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 03292312, including orbits 180-196. Figure 5b shows the corresponding residuals' histograms.
Figure 6a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 03310335, including orbits 197-209. Figure 6b shows the corresponding residuals' histograms.
Figure 7a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 03312358, including orbits 210-226. Figure 7b shows the corresponding residuals' histograms.
Figure 8a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 04020237, including orbits 227-240. Figure 8b shows the corresponding residuals' histograms.
Figure 9a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 04030028, including orbits 241-255. Figure 9b shows the corresponding residuals' histograms.
Figure 10a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 04031330, including orbits 256-271. Figure 10b shows the corresponding residuals' histograms.
Figure 11a shows in-scan and cross-scan residuals versus scan angle from Point D, for all the astrometric match stars in packet 04050109, including orbits 273-283. Figure 11b shows the corresponding residuals' histograms.
The first was the creation of a J-band astrometric star catalog, (JCAT), by B. Smith of IPAC. The IRAS Faint Source Survey (Moshir et al. 1992), the associated FSS Optical Identification database (Conrow et al. 1994) and the PPM Positions and Proper Motions catalog (Bastian & Röser 1991) were used to generate a catalog with astrometric positions and predicted J-band magnitudes, with approximately half a million entries, going down to J~10 mag. The sources included stars known either optically or from IRAS, or both. For stars with no IRAS counterpart, the J magnitudes were extrapolated from V magnitudes using their spectral types along with an extinction model (Jarrett 1992). The V magnitudes were not homogeneous, but depended on the optical catalog from which they were taken. JCAT contains ~73,000 stars with J < 6.5 and ~517,000 stars for J < 10. The predicted J magnitudes have been found to agree to within ~1.4 mag for sources observed by 2MASS and CIO.
Next was the extraction of candidate astrometric stars from
the STS data. This step is described in Section
III below. The reconstruction itself, described in Section IV below, was an iterative process. Once a good
solution was obtained for each orbit, the program
makebph created the corresponding boresight file.
After concatenation into packets, various quality checks have
been performed, as described in Section V.
att_lan" files with IRTS times for each frame,
orbital information and coarse aspect, estimated a priori.
All files received were converted into ASCII, generally standard
table files. For standard table files, header rows at the
beginning of the file define the contents, width, and type of the
information in each column, so that they are largely
self-documenting.
| Input STS binary TLM data files: | sout_ppppppppcc-xxxo11 |
Input att_lan: |
eatt_ppppppppcc-xxxo11 |
| Input orbit event data: | Tgood.ver4 |
Here pppppppp is packet, and xxx is
orbit number. Input STS binary TLM data files were converted to
give an ASCII sequence of (time, voltage) pairs for the STS;
input att_lan files were converted to table files
with times converted from IRTS time to IPAC time, and the
orbit-event information was converted to
Tgood.ver4.
The IPAC output, or boresight, files are ASCII table files, with one row for each ~1 second of data, and contain the sky co-ordinates of the boresight, STS (C2 point) and other information, all supplied in fixed-width columns. Appendix 4 describes the format in further detail.
| Program: | Input Data: | Output Data: | Function: |
|---|---|---|---|
pointless |
sout_* |
ssrc_* |
Extract sources |
manfit(or: autmch,) |
ssrc_* manmchTgood.ver4 |
summary |
Fit one segment |
makebph |
summary eatt_*Tgood |
output_xxxSS* |
Write boresight file |
cmpoutext |
manmch or extcat |
cmpoutext.xxx |
Check internal consistency |
cmpoutnirs |
NIRS.cat |
cmpoutnirs.xxx |
Check NIRS consistency |
cmpoutmirs |
MIRS.cat |
cmpoutmirs.xxx |
Check MIRS consistency |
The detection step involves the use of a matched filter. In this step, the time sequence of data samples for each scan of the star sensor was passed through a linear filter whose output was in the form of a 2-dimensional image. This filter had a better performance than a zero sum filter since the background was independently removed from the data. The matched filter output was detected with a threshold set at a specified level (typically 4-sigmas, corresponding to m(J) < 6.5 approximately) and a list of candidate sources produced. For each candidate detection, the most probable 2-d position and magnitude were estimated, together with the associated reduced chi squared. Finally, source verification was performed. In order to determine whether an extracted source was genuine, two tests were performed: (a) chi-squared test, and (b) a check on qualitative morphological structure of the extracted profile (specifically, that it should possess precisely two broad peaks).
Figure 3a shows the original STS response map (PSF), as measured prior to launch. To improve the ability of the algorithm to discriminate against spurious detections, the original STS response map was refined using the observed profiles of bright sources. Approximately 500 sources of magnitude 4.5 or brighter, detected during the post-flip scans of April 21-24, 1995, were employed for this purpose. Comparing the expected profile from the original PSF with the actual data, we adjusted the original PSF. Typical adjustments were of the order of 10% of the peak response.
In addition, the distribution of in-scan position offsets with respect to known stars showed a trend with respect to the cross-scan position of the reference star. This indicated that the star sensor scanned the sky at angle of ~2.7 deg around the +Z_I axis of the instrument. Comparison with a very limited amount of MIRS ( cf Section V.D) data indicated that at that time the satellite itself scanned along the y-axis of the focal plane as planned. The 2.7 deg angle was therefore treated as if it were due to a rotation of the STS relative to the focal plane, and a rotation was applied to the STS response function. The validity of the refined response map was verified by performing a further iteration, whereby the residuals were found to be gaussian, and of the order of 1% of the peak. The refined, rotated map appears in Figure 3b.
In the post-flip scans, approximately 50,000 STS point source
detections were obtained at S/N > 5. These point source
detections resulted in sightings of over 28,000 JCAT sources.
During the shorter duration pre-flip period, over 20,000 STS
detections were obtained at a S/N > 5; these resulted in
sightings of over 3,600 JCAT sources. The list of extracted stars
is placed in the ssrc_* file for the orbit.
In general, an initial solution was used to give approximate positions for all the extracted stars. Extracted stars could then be looked up in JCAT, and if good candidates were found, they were added to the list of identified sources, and used to determine an improved solution. However, starting the process was sometimes difficult, especially for the pre-flip orbits.
For the post-flip scans, where source confusion was not an impediment, the following method was developed:
eatt_* file for the orbit), and increasing the
size of the swath to allow for the uncertainties, select from
JCAT a list of astrometric sources which could potentially be
identified with sources in the extracted source list.
The least-squares fit is described further in Section IV.B. below.
The pre-flip scans were done in an analogous manner except that due to increased confusion and extinction close to the galactic plane, the extraction / reference matching step described above was more difficult for the pre-flip scans.
A very powerful technique, once a few orbits have been solved, is to look for stars repeating in adjacent orbits, at least one of which has been solved. Figure 12 shows a portion of a series of well-behaved pre-flip orbits with several obvious "repeaters". This method can be misleading when applied to fainter stars, because of the large number of possible candidates, but it has been found to be the most generally useful method for starting the solution for most orbits when the method described for the post-flip analysis fails.
For about half of the orbits in the pre-flip period, the following series of semi-automatic steps, based on repeating sources, was successful:
manfit, a minimum of three
good stars were needed to start the solution. Once a good
preliminary solution was found, identifications could be made in
JCAT for many fainter extract stars.
To fit astrometric candidate stars to a model for the scan, we
first read from an input file manmch, a list of
candidate assignments between extracted stars (each specified by
the integer part of the IPAC time), and positions for each
(normally found by searching JCAT for tentative identifications).
The program manfit then finds the best least-squares
solution to fit the given candidates.
manfit model for a segment of IRTS
data is uniform rotation about the nominal spin axis fixed on the
sky, (and fixed in the body axes of the SFU), with the boresight
fixed (90 degrees + epsilon ) from the spin pole, where
epsilon is a small angle. Parameters determined from the
data by manfit include the celestial coordinates of
the spin axis, the spin rate omega (defined in terms of
dsy, the fractional deviation from the nominal value
omega0 given in the Tgood file, as described
in Appendix 2), the phase angle of the
boresight at the beginning t0 of the orbit, and the
parameter epsilon. RMS in-scan and cross-scan residuals,
sigy and sigx, as minimized by the fit, also appear
in the fit output or summary file, along with RMS
difference sigm between the STS-observed J-magnitudes of
the extract stars and the J-magnitudes expected from their
tentative JCAT identifications.
manmch stars was close
positional agreement. For almost all of these orbits the
rule-of-thumb was 2' in-scan and 5' cross-scan. An important
secondary criterion was magnitude agreement, although (due to the
uncertainties involved in generating JCAT magnitudes) we found
that the presence of a candidate with a measured IRAS 12 micron
flux was more important that magnitude agreement itself.
In a few cases, good candidates which would ordinarily have been acceptable could not be distinguished from others which were equally probable. If both candidates were very close together the choice would not affect the final solution very much. However, in some cases the ambiguous stars were not close to each other, and in such situations the decision was generally taken that it was better to reject both, rather than risk an incorrect assignment's degrading the results.
In the iterative process of fitting to pre-flip data, after
each manfit iteration, residuals were checked for
satisfactory agreement. For flexibility, manfit has
provision for assigning each of the astrometric candidate stars
in the manmch file a weight. For the pre-flip
processing described here, this facility was used only to permit
manmch files to be set up to completely ignore
groups of stars by setting their weights to zero. This allowed
the same list of stars to be used for both AB and CC segments of
an orbit, by setting the weights to 0 or 1, which was a
convenience. Unacceptable match stars were removed by setting
their manmch weights to zero (but usually left in
the file, to show that they had been considered). Then a search
of JCAT was performed to see if an alternate acceptable match
candidate could be located.
manmch astrometric stars than some processed
earlier. Most of the orbits processed later have 50-90 match
stars, with at least 20 in each fit segment. An attempt was made
to find matches for all the brighter extract stars in each
segment; 4th magnitude was the rule-of-thumb threshold, though
not uniformly applied in the orbits analyzed earlier. If no
brighter extract stars could be matched, fainter candidates were
tried, until any large (eg, > 500 s, ~35 degrees) gaps in
coverage were filled. Failure to find any reasonable candidate
for several bright extracted stars was taken as a sign that the
solution is not reliable, and is reflected in the value for Flag1
in the Tgood.ver5 file and the corresponding
uncertainties assigned in the final boresight files.
manfit model described previously.
summary file (Table 3), but for a few the solution parameters
are identical for AB and CC. This indicates that those orbits
were actually fit as one segment, rather than two.
Figure 13 and Figure
14 compare the solutions before and after the B-C boundary
for one of the first pre-flip orbits processed manually which was
found to require segmentation. For each figure all the
manmch stars listed for the orbit are plotted, but
those excluded from the fit (ie, with 0 weight) have been boxed.
The amplitude of the characteristic sinusoidal cross-scan
residual indicates the approximate value of the spin pole motion
is 12', and may be compared with the value of ~0.2 degrees from
the summary file.
Input to manfit includes the manmch
file, the list of extract stars for the orbit in the
ssrc_* file, and the Tgood.ver4
file.
The output of manfit includes a list of output
parameters for the new best-fit solution, saved as one line in
the summary file, plus fit residuals
(manres file) for each manmch star, and
an estimate, according to the solution, of the celestial position
of each extract star in the orbit (mancat). Finally,
the manres file was used to generate time histories
and histograms of the residuals in x and y, such as
those in Figure
24.
For the final, most troublesome group of pre-flip orbits,
manfit was used exclusively, but more automated
versions (autmch and multch) were
generally used to implement the iterative processes described in
Sections IV.A.1. and IV.A.2., especially for the post-flip
data.
Given the final summary files, the program
makebph (cf Section II.D.) then generated the
boresight file for the orbit. Because of the occasional need to
interpolate across gaps, this step was generally performed only
after all the orbits for a complete packet had been solved.
Besides the summary file with the fit input from
manfit, makebph uses the
Tgood files (both ver 4 and ver 5), and the
eatt_* file with the time stamps. It writes a piece
of the boresight file (named output_xxxSS.ver2,
where xxx is orbit number and SS fit segment, either AB or CC)
for each fit segment that it finds in the summary
file for the orbit. The segments were finally concatenated
together into a single combined_att_lan.* for the
packet.
The overall fit quality obtained is evident from the RMS residual plots in Figures 4-11, and the information concerning acceptable orbits and segments in Section III. Because of the policy of normally accepting astrometric stars for which a reasonable candidate could be found within 2' in-scan and 5' cross-scan, there remains a slight danger that a spurious solution could be obtained. However, the typical RMS agreement was normally so much less than the rejection criteria (the latter being 3-4 sigma) that this danger is believed to be small. A very few orbits or segments have in-scan and cross-scan residuals so much more than the normal (< 0.5' and < 1.5', respectively) that there is doubt. Orbit 273 is an example, one of the worst. These have been flagged with large uncertainties.
As a general overall check on each packet, we have run
so-called "postage-stamp" plots, which are of the difference
between the original ISAS a priori estimate of the aspect
and the final IPAC solution from the
combined_att_lan.* file, both in-scan and
cross-scan. Since the final and original solutions are sometimes
very different, these plots are not reliable indicators of
problems, but may call attention to errors and provide some
reassurance that the processing has been well-behaved for most
orbits.
The program cmpoutext checks for any
inconsistencies between the boresight file and the orbital fit,
comparing the STS position from the boresight file with the
manmch positions for the input match stars. It
generates in-scan and cross-scan difference histograms for each
orbit. These are generally similar to the manres
histograms of residuals, and to the RMS fit errors sigx
and sigy reported for each segment in the
summary file.
We have verified the external consistency of our
reconstructions using MIRS and NIRS data, kindly provided by Drs.
Freund and Yamamura on behalf of the MIRS and NIRS teams,
respectively. We have converted the MIRS and NIRS detection times
into positions using our reconstructions, and then found the most
likely JCAT matches for those sightings. Using a galactic
latitude |b| < 5 degress criterion, we produced a list
limited to the most unambiguous sightings. For NIRS, a test for
reasonalbe color was used as well. Then cmpoutnirs
and cmpoutmirs compared the JCAT match positions to
the STS-reconstructed positions a the times of the NIRS / MIRS
detections. This procedure measured the focal plane offsets of
these detections with respect to point C2 of the STS, to provide
a check on the boresight solution and the nominal focal plane
geometry.
For example, Figure 15 shows focal plane positions of sources detected by MIRS, for the pre-flip period. (We thank Prof. Onaka for providing these pre-flip MIRS detections and associations.) Using the STS boresight solution, an approximate celestial position was determined for each MIRS detection. Each source was then identified using JCAT, and the resulting true celestial positions compared with the reconstructed STS positions. Sources with |b| < 5 degrees were removed from the detection list, to reduce the effect of false identifications; 353 sources survived this editing process. If the boresight solution were perfect, the identifications were all good, and the focal plane geometry exactly as determined from the pre-launch 300K calibrations, the result would be a cluster of stars filling the nominal 8' X 8' MIRS field-of-view, at the nominal position (Figure 2).
Figure 16 shows in-scan and cross-scan histograms of the data in Figure 15. Figure 17 and Figure 18 show similar scatter plots and histograms for 1090 NIRS-detected sources in the pre-flip period. For NIRS, in addition to the |b| < 5 filtering, sources were removed based on a J-12 micron color test to further reduce the effects of false identifications. For the post-flip data, Figure 19 and Figure 20 show corresponding plots for 596 NIRS sources, and Figure 21 and Figure 22 for MIRS, respectively,
These figures show that the offsets differ by ~2' from the nominal measurements.
For the pre-flip data, both MIRS and NIRS detections offsets have been examined extensively. The offsets have shown consistent small discrepancies from the nominal focal plane measurements provided by the instrument teams. Currently, these discrepancies in the in- and cross-scan directions are 2.6' and 1.65' for MIRS, and 2.28' and ~0' for NIRS. Thus, our suggested measurements of the center of the MIRS and NIRS detectors (in units of minutes of arc) from the pre-flip data are as follows:
SUGGESTED center of MIRS NOMINAL center of MIRS IN-SCAN CROSS-SCAN vs. IN-SCAN CROSS-SCAN -60.94 0. -58.34 1.65 SUGGESTED center of NIRS NOMINAL center of NIRS IN-SCAN CROSS-SCAN vs. IN-SCAN CROSS-SCAN 54.05 2.5 56.33 2.5
The exact causes of these small deviations from nominal measurements in the STS, MIRS and NIRS statistics may be related to several possibilities, such as: 1) a slanted scan pattern, 2) a physical rotation of the STS, 3) electronic timing delays, 4) possible beam pattern asymmetries in NIRS and MIRS.
For the post-flip data, a small set of MIRS detections were used to verify external consistency. This data set, however, is too small to independently establish the center measurements for the detector or to distinguish between the possible causes for the deviations listed above. We are have analyzed a somewhat larger set of NIRS data, and the results strongly suggest that a slanted scan pattern is present, as indicated in Figure 23. A larger number of NIRS and MIRS detections will be needed to complete this analysis at ISAS.
During the course of the work a number of special situations and conditions arose, of which the user should be aware. We describe these here.
Frequently no or very few good match stars could be found in the AA segment, from Point D to the end of orbit day. This situation is apparently related to the change of aspect at Point D, and a corresponding wobble or nutation associated with it. Also, in many orbits, the scan rate changes during this 500 sec interval. As a result, the uncertainties in the boresight files have been systematically increased during the first 500 seconds of the AA segment to account for this circumstance.
In a number of orbits significant spin-axis changes occurred
during the middle of a a fit segment. Examples were orbits 194,
209, 216, and 224, among others. Some of these were known a
priori from the Tgood file, and some were
discovered during analysis. In either case no credible match
stars can be found beyond the change, and the fit must be broken
off. Data in the region with no good fit has been interpolated,
and flagged with Flag1 = 5 in the Tgood file, and
uncertainty 99' in the boresight file.
A change in IRU MODE from "High" to "Low" occurred 1995-03-30 at 02:26:13, during the CC segment of orbit 180. A change in IRU MODE from "Low" to "High" occurred 1995-04-04 at 23:30:13, during the CC segment of orbit 270. Partly as a result, substantial portions of both of these orbits could not be solved. A few other orbits (eg, 201) were affected by gyro problems for short periods, which have been flagged.
Almost all orbits displayed a characteristic cross-scan ripple, of amplitude up to 2', at certain points in the day / night cycle. In particular, the reconstruction residuals from eclipse exit back to Point-D show a time variation akin to a limit cycle (Cf Figure 24a for post-flip, and Figure 4a for pre-flip). For post-flip, this oscillation had a peak to peak range of about 4' in the cross-scan direction, and a much smaller range in the in-scan direction. For the post-flip data only, we have removed this periodic trend. Figure 24b shows post-flip residuals after removal of the ripple. Figure 25a and Figure 25b show the corresponding residual histograms. As a result, our internal reconstruction uncertainties have been reduced to 40" and 1' in the in-scan and cross-scan directions, respectively. The effect is smaller (< 1') and less consistent in the pre-flip data, and entangled with the other problems which afflict the pre-flip period. For this reason we have not attempted to apply a similar correction to the pre-flip solutions.
The cause of this variation (with a period of approximately 1500 seconds) is not known in detail. As in the AA segment, most of the effect can be attributed to small adjustments by the aspect control system, as eg, at orbit dawn, when the sun-sensor would often make a small correction.
For the pre-flip data, no good solution has been obtained for the following orbits: 182, 185-187, and 271-272. Orbits 180-181, 270, and 273 have significant intervals when the solution is absent or degraded, and have been flagged. All post-flip orbits have been solved.
Finally, we stress again that it is essential for the user to check the uncertainty information in the boresight file when using these data. Otherwise inaccurate aspect information is liable to be used.
We thank Dr. Minoru Freund, of NASA Ames and ISAS, for many helpful comments and other contributions to the STS work during his NIRS analysis at IPAC. This work was carried out at the Infrared Processing and Analysis Center, with funding from NASA under contract to the California Institute of Technology and the Jet Propulsion Laboratory.
Bastian & Röser 1991 Sterne und Weltraum, 30: 592.
Conrow et al. 1993 BAAS 183:03.03.
Jarrett, T. 1992 Ph.D. Thesis, University of Massachusetts.
Moshir, M. et al. 1992 Explanatory Supplement to the IRSA Faint Source Survey, Version 2, JPL D-10015 8/92 (Pasadena: JPL).
Murakami, Hiroshi et al. 1994 Ap.J 428: 354.
Murakami, Hiroshi et al. 1996 Proc. Astron. Soc. Jap. 48: L41-46.
Nakagawa, 1995
Term: Definition:
AB Segment of orbit from Point D to orbit sunrise.
boresight Nominal center of IRTS FOV, co-ordinate (0,0)in
focal plane
C2 Physical corner of STS, adopted as reference point;
cf Figure 2 & Table 2
CC Segment of orbit from orbit sunrise to following
Point D.
FILM Far-IR Line Mapper, instrument on IRTS
FIRP Far-IR Photometer, instrument on IRTS
g-angle Sun-to-spin axis angle.
IPAC Infrared Processing and Analysis Center, Caltech
IPAC-time Time in seconds since 1995-03-15 at 00:00:00 UTC.
IRAS InfraRed Astronomical Satellite.
IRTS InfraRed Telescope in Space
IRTS-time Time in seconds since launch, at 1995-03-18,
08:01:00 UTC.
ISAS Institute of Space and Astronautical Science
JCAT J-band IR star catalog constructed at IPAC for IRTS
STS data analysis.
MIRS Mid-IR Spectrometer, instrument on IRTS
NIRS Near-IR Spectrometer, instrument on IRTS
Point D Initial point of orbit, defined to be 90 degrees
past orbit noon.
Post-flip "South-Scan Data", second part of IRTS mission.
PPM Positions and Proper Motions Catalog of Optical Sources.
Pre-flip "North-Scan Data", first part of IRTS mission.
PSF Point Spread Function, = STS response function.
repeater Stars seen in STS on adjacent orbits.
RMS Root-mean-square.
SFU Space Flyer Unit
STS STar Sensor, J-band diagonal mask type on IRTS.
|packet |orb|sg|t0 |tf |alfp |delp |alf0 |del0 |eps |dy0 |dsy |mchf|mchp|nfit|sigx|sigy|sigm|datime0 |dt |comment| |int |int|ch|double |double |real |real |real |real |real |real |real |int |int |int |real|real|real|int |i |char | |mmddhhmm|-- |--|sec |sec |deg |deg |deg |deg |amin |amin |-- |-- |-- |-- |amin|amin|mag |yydoyhhmm|min|-- |
The header lines for this ASCII table file show the variable
names, variable types, and units for each column. The first
header line of an ASCII table file gives a list of variable names
for each column, separated by "|"'s, indicating the field width.
The second header line gives the type of the data, such as char
or ch for character, int or i for integer, and real for floating
point. Each row is the result of a segment fit by
manfit (cf Section IV in the
text). Columns 1-5 give packet number, orbit, and fit segment
designation, and start and stop IPAC times. Columns 6-9 give the
RA and Dec of the spin pole, and the RA and Dec of the normal to
the spin axis, in the plane of the boresight, at the beginning
t0 of the orbit. Here and elsewhere, all celestial
positions are given for epoch and equinox B1950.
Columns 10-11 give epsilon (defined in the text, shown
as eps in the header), and dy0, the phase
offset at t0. For the earlier analysis only, especially
the post-flip, the phase was defined so that dy0 was the
fitted parameter, and typically non-zero. For the later analysis,
especially the pre-flip, dy0 was defined to be zero, and
the phase variation was taken up in the Ra and Dec of the
spin-axis normal vector; thus dy0 is always 0 for these
later data.
Column 12 gives dsy, the fitted fractional deviation
from the nominal spinrate omega0, from the
Tgood file, described in Appendix
3. The nominal omega0 was determined from gyro data.
The final fitted spinrate omega for the segment is then
given by:
The final columns in the summary file give the RMS width of
the x (cross-scan) and y (in-scan) fit residuals,
of the magnitude agreement, and the year, day-of-year, and time
of manfit processing.
Table 3 summarizes the solutions obtained for each segment.
The summary files, pre-flip.summary
and post-flip.summary are critical to the analysis,
and have been placed under Sun Source Code Control System (SCCS)
configuration control, in the /irtsdev/sccs/
directory.
The modified orbit event, or "Tgood.ver5" file,
has, in addition to the nominal orbit event and segment
information, has the sub-segmentation information for the A, B,
and C segments which was found necessary for some orbits due to
situations discovered during the analysis, and quality
information resulting from the fits, encoded in Flag1 (see
below). Another version of Tgood, Version 4, is used
by manfit and makebph, and is generally
similar except that it lacks the sub-segmentation of the A, B,
and C segments, the quality information in Flag1, and also lacks
the Moon flag, Flag9. (Thus it has one less flag, and flags
beyond 9 are shifted accordingly.)
Table 4 gives a listing of the
"Tgood.ver5" file,
Column 1 gives the packet number.
Column 2 gives the orbit number.
Column 3 gives the segment designation letter (A, B, or C).
Column 4 gives the subsegmentation as a 2-digit code "nm", where n is subsegment number out of a total of m subsegments in the segment. For example, "12" means first subsegment out of two. The most common case, "11", means no subsegmentation was required. Subsegmentation typically either occurred because of an a priori known unusual orbit event (eg, thruster use or aspect control mode change), or the occurrence of a region where no solution could be obtained, as determined after processing.
Column 5 gives segment start time.
Column 6 gives the time interval for gyro rates to settle (sec) after the segment start time.
Column 7 gives segment stop time.
Column 8 gives the time interval of bad gyro behavior at the end of a segment (sec). (This occurs when a gyro event starts slightly before an orbital event.)
Column 9 gives omega0, the nominal spinrate, in arcmin per second..
Column 10 gives the uncertainty on omega0.
Column 11 gives an array of 14 flags, describing various aspects of the data. The first of these, Flag1, is particularly important, describing the overall reliability of the solution obtained.
The final columns translate the segment start time into normal calendar date and time.
The Tgood (both versions 4 and 5) files, named
Tgood.ver4 and Tgood.ver5, are critical
to the analysis, and have been placed under Sun Source Code
Control System (SCCS) configuration control, in the
/irtsdev/sccs/ directory.
|packet |orb|seg|cc|t0 |dt |tf | dt_end | phidot |sigpd |flags | UT | | char |int|c |i | real | real | real | real | real | real | int | char |
format(a8,1x,i3,3x,a1,1x,2(i1),1x,f10.2,1x,f8.2,
+ 1x,f10.2,1x,f8.2,1x,
+ f10.7,1x,f9.7,1x,14(i1),1x,14(i1))
Digit of "flags" parameter = Flag number:
format(i1,'/',i2,1x,i2,':',i2,':',i2).The STS and boresight history files described above, have the following format:
The file header:
\ Reconstructed Attitude File (IPAC- Phase I)
\
|date-time |LAUNCHtime |ra_sts |dec_sts |sigi|sigx|posang |sigpa |FPang |sFPang |ra_bs |dec_bs |s1bs|s2bs|pa_bs |spa_bs |packet |qflag |spare |
|double |double |double |double |d |d |double |double |double |double |double |double |d |d |double |double |char |int |char |
|mmddhhmmss.sss|seconds |deg |deg |amin|amin|deg |deg |deg |deg |deg |deg |amin|amin|deg |deg | | | |
Format for reading the data:
integer*4 month, day, hour, minute
real*8 seconds
real*8 time, ra_sts, dec_sts, sigin_sts, sigx_sts, pa_sts,
real*8 sig_pa_sts, fp_ang, sig_fp_ang, ra_bs, dec_bs,
real*8 sigin_bs, sigx_bs, pa_bs
integer*4 numpacket, qflag(15)
character*6 spare
read("ipac_att_lan", 100) month, day, hour, minute, seconds,
time, ra_sts, dec_sts, sigin_sts, sigx_sts, pa_sts,
& sig_pa_sts, fp_ang, sig_fp_ang,
& ra_bs, dec_bs, sigin_bs, sigx_bs, pa_bs, sig_pa_bs, numpacket,
& qflag, spare
100 format(1x, 4i2, f6.3,
& f12.3, 2f9.4, 2f5.1, f9.4,
& f8.4, f9.4, f8.4,
& 2f9.4, 2f5.1, f9.4, f8.4, i8, 1x, 15i1, 1x, a6)
Definition of Variables:
-----------------------
date-time = month (mm), day(dd), hour(hh), minute(mm), second(ss.sss)
LAUNCHtime = time in seconds since launch (95-03-18 08:01:00 UT)
ra_sts = reconstructed right ascension (B1950) of STS lower right
corner as seen in Figure 7 of Murakami et al.
(1994, ApJ, 428, 354) This point is labeled "C2"
in the memo on the theodolite measurements of the
relative positions of the four corners of each FPI
dec_sts = reconstructed declination (B1950) of STS lower right corner
sigi = twice the standard deviation of the in-scan reconstruction (for
the star sensor data)
(for Nov. 1995 release, this is set equal
to 1' for all times except in the gyro settling periods
after an orbital event. In those intervals, this is set to 30')
sigx = twice the standard deviation of the cross-scan reconstruction (for
the star sensor data)
(for Nov. 1995 release, this is set equal
to 2' for all times except in the gyro settling periods
after an orbital event. In those intervals, this is set to 30')
posang = angle defining in-scan direction (east of north B1950).
(this is 90 deg off from the positional uncertainty
ellipse position angle for the star sensor data
if sigx > sigi, and equal to the error ellipse position
angle if sigi > sigx.)
sigpa = standard deviation of position angle for the star sensor position
(in Nov. 1995 release, this is not calculated, but is set to
zero)
FPang = the angle between the assumed in-scan direction according
to the nominal focal plane orientation and the actual
in scan direction
(in Phase 1, not reconstructed, but assumed equal to 0 deg)
sFPang = standard deviation of the FPangle
(not calculated in Phase 1; set equal to zero)
ra_bs = ra of boresight
(in Phase 1, calculated using FPang = 0 deg)
This is calculated using the theodolite measurements
of the distance between the nominal boresight to the STS
lower left corner. These measurements are 0.1083 degrees in
in-scan direction and -1.05 degrees in cross-scan direction.
dec_bs = dec of boresight
(in Phase 1, calculated using FPang = 0 deg)
This is calculated using the theodolite measurements
of the distance between the boresight to the STS
lower left corner.
sbsi = in-scan uncertainty of the boresight position
(in Phase 1, this is calculated using sigi, assuming an
additional systematic uncertainty of 1' due to the uncertainty
in the C2-boresight distance.)
sbsx = cross-scan uncertainty of the boresight position
(in Phase 1, this is calculated using sigi, assuming an
additional systematic uncertainty of 1' due to the uncertainty
in the C2-boresight distance.)
pa_bs = position angle of the boresight error ellipse
(east of north B1950)
(in Phase 1, this is set equal to posang)
spa_bs = standard deviation of the position angle for boresight uncertainty
ellipse
(in phase 1, assumed equal to sigpa)
packet = packet number of concatenated input ATT_LAN and IRTS_LAN files
qflag = quality flags
--------------------------
digits of quality flag parameter (1 corresponds to the leftmost digit):
1 = Did Not Match Flag => Did Not Match STS Stars in
this Time Interval
2 = Thruster Flag => Thruster Mode
3 = Bad/Missing Data Flag => Bad/Missing Data
4 = Bad Gyro Flag => Bad Behavior of Gyros
5 = Aperture Cover Flag => Aperture Cover On
6 = STS off Flag => Star Sensor Off
7 = G Angle Flag => G Angle Change
8 = Spin Rate Flag => Commanded Spin Rate Change
9 = Moon Flag => Moon Dominates Star Sensor
Data
10 = Split Packet Flag => Segment Split Between 2
Packets
11 = Low Latitude Flag => Low Latitude Scan (before
flip)
12 = Pt D to Eclipse In Flag => During Pt. D to Eclipse In
13 = Eclipse In to Eclipse Out Flag => During Eclipse In to Eclipse
Out
14 = Eclipse Out to Point D Flag => During Eclipse Out to Pt. D
15 = Could Not Fit Flag => Problem Fitting STS Stars in
this Time Interval
Figure 2: IRTS Focal Plane Geometry The focal plane geometry adopted by IPAC (from Table 2) for position reconstruction. Measurements in degrees.
Figure 3: STS Response Map. Figure 3a: The original point spread function (PSF) for the STS, determined by pre-launch calibration. Figure 3b: The final STS response map, refined from the original based on sightings of approximately 500 stars brighter than 4.5 magnitude, and rotated by 2.7 degrees, as described in the text.
Figure 4: Pre-Flip Residuals, All Packets. Figure 4a: History of residuals versus angular distance (~time) from the initial point D of the orbit, for all pre-flip orbits, 180-283. Figure 4b: Histograms of the distribution of the residuals in Figure 4a.
Figure 5: Residuals for Packet 03292312, Orbits 180-196. Figure 5a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 180-196. Figure 5b: Histograms of the distribution of the residuals in Figure 5a.
Figure 6: Residuals for Packet 03310335, Orbits 197-209. Figure 6a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 197-209. Figure 6b: Histograms of the distribution of the residuals in Figure 6a.
Figure 7: Residuals for Packet 03312358, Orbits 210-226. Figure 7a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 210-226. Figure 7b: Histograms of the distribution of the residuals in Figure 7a.
Figure 8: Residuals for Packet 04020237, Orbits 227-240. Figure 8a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 227-240. Figure 8b: Histograms of the distribution of the residuals in Figure 8a.
Figure 9: Residuals for Packet 04030028, Orbits 241-255. Figure 9a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 241-255. Figure 9b: Histograms of the distribution of the residuals in Figure 9a.
Figure 10: Residuals for Packet 04031330, Orbits 256-271. Figure 10a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 256-271. Figure 10b: Histograms of the distribution of the residuals in Figure 10a.
Figure 11: Residuals for Packet 04050109, Orbits 273-283. Figure 11a: History of residuals versus angular distance from the initial point D of the orbit, for orbits 273-283. Figure 11b: Histograms of the distribution of the residuals in Figure 11a.
Figure 12: Typical Pre-flip Data Showing Repeaters. A ~500 sec segment of STS data from orbits 242 (bottom) - 250 (top), showing repeated appearances of some dozen strong to moderate sources. From such information it was usually possible to find an initial set of match stars for starting an unsolved orbit, based on comparison with a previously solved orbit adjacent to it.
Figure 13: Orbit 256 Fit Before Segment BC Boundary. The sources shown are all extracted stars, but only those between the sunset and sunrise (unboxed) have been used in the fit. The large discontinuity in the cross-scan residuals at sunrise is due to a motion of the spin pole.
Figure 14: Orbit 256 Fit After Segment BC Boundary. Same as Figure 13, except only extract sources after the B-C boundary have been used in the fit. As can be seen from the amplitude of the segment B cross-scan residuals, the best-fit poles between the AB and CC segment fits differ by about 12'. For this orbit, only about 1500 sec of data from the CC segment were available.
Figure 15: MIRS Source Focal Plane Positions for Pre-Flip. Focal plane positions of 353 sources detected by MIRS, for the pre-flip period. The distribution of points reflects the MIRS FOV, while the position checks the co-ordinates in the focal plane. See text.
Figure 16: Histogram of MIRS Source Focal Plane Positions for Pre-Flip. The data from Figure 15 are histogramed in x (cross-scan) and y (in-scan).
Figure 17: NIRS Source Focal Plane Positions for Pre-Flip. Analysis similar to that for Figure 15 above, as detailed in the text.
Figure 18: Histogram of NIRS Source Focal Plane Positions for Pre-Flip. The data from Figure 17 are histogramed versus y (in-scan) and x (cross-scan).
Figure 19: NIRS Source Focal Plane Positions for Post-Flip. Analysis similar to that for Figure 15 above, as detailed in the text. In-scan position is with respect to STS C2.
Figure 20: Histogram of NIRS Source Focal Plane Positions for Post-Flip. The data from Figure 19 are histogramed in x (cross-scan) and y (in-scan).
Figure 21: MIRS Source Focal Plane Positions for Post-Flip. Focal plane positions of sources detected by MIRS, for the post-flip period. Analysis similar to Figure 15 above, as detailed in the text.
Figure 22: Histogram of MIRS Source Focal Plane Positions for Post-Flip. The data from Figure 21 are histogramed in x (cross-scan) and y (in-scan).
Figure 23: NIRS Detections with respect to NIRS Center. The in-scan (y) positions for post-flip sources versus orbit number for the post-flip period. The NIRS source detections were processed similarly to those shown in Figure 19 above, but are based on an earlier version of the solution. The in-scan positions are with respect to the NIRS center. Significant variations with time, especially from about orbit 480 to 550, are evident.
Figure 24: Post-Flip Residuals Before (a) and After (b) Removal of Ripple. Figure 24a shows the cross-scan (upper panel) and in-scan (lower) residuals initially obtained for the post-flip data. Figure 24b shows the residuals after correction for ripple, as described in the text.
Figure 25: Histograms of Post-Flip Residuals Before (a) and After (b) Removal of Ripple. Figure 25a shows histograms of the cross-scan (left panel) and in-scan (right panel) residuals initially obtained for the post-flip data. Figure 25b shows histograms of the residuals after correction.