Index of /data/IRTS/docs/IRTS_isas/local/ipac/1st-Release

 Name                    Last modified      Size  Description
 README_pre_fix.html     2006-06-21 16:20   31K  
 README.html             2006-06-21 16:22   31K  
 README.html_ORIG        2006-05-19 15:28   28K  

IPAC

First Release of IRTS Attitude Data

CONTENTS

• ABSTRACT
• I. INTRODUCTION
• II. SOURCE EXTRACTOR
• III. POSITION RECONSTRUCTION
• IV. RESULTS
• V. COMPARISONS WITH THE SKY
• VI. UNCERTAINTIES and CAVEATS
• VII. SCHEDULE & OTHER WORK
• VIII. FORMATS
  • Header
  • Format
  • Definition of Variables
• FIGURE CAPTIONS

ABSTRACT

The purpose of this release is to provide the first in a series of staged releases of the IRTS position reconstruction. This is intended to provide the users a significant improvement in pointing compared to the nominal pointing information in att_lan files.

The IRTS position reconstruction in this preliminary release corresponds to packets 04210105-0423351; this covers the time period between 95/4/21 1:05:26.636 and 95/4/24 11:58:12. These data correspond to the repeated observations (with small position offsets) of the same regions of sky.

The "boresight" history files are called ipac.att.04210105, ipac.att.04220036, ipac.att.04230012, and ipac.att.04232351. These correspond to the concatenated files irts_04210105cc.lan, irts_04220036cc.lan, and irts_04230012cc.lan. The 4th file, ipac.att.04232351cc.lan, is not complete; it ends at 95/4/24 11:58:12, some time before the g-angle change from g = 18 deg to g = 19 deg.

Since this release is the first in a series, it does not yet meet the requirements of the final product. In this release, the star sensor position reconstruction has an assigned uncertainty (90% confidence) of 1' and 2' in the in- and cross-scan directions respectively. The position reconstruction has an rms uncertainty of 30" - 60"; when known systematic errors are taken out in future releases, the reported uncertainty envelopes will be much better than the current release.

The analysis of overlap data has indicated that the STS has a rotation of approximately 2.7deg with respect to its nominal position in the focal plane.

In the position reconstruction for boresight, the relative position of the STS has been adopted from the laboratory theodolite measurements done @ 300 K.

A very preliminary comparison with MIRS detections has been used to verify the relative coordinates used for the STS.

In this release several caveats exist with respect to transferring the star sensor position reconstruction to the BORESIGHT: i) the relative positions of the STS and boresight are based on nominal ground measurements @ 300K; ii) the scan angle with respect to the focal plane x-y coordinate system has been assumed to be the nominal value of 0deg since a sufficient amount of data from other instruments is not available; iii) the systematic variations in cross-scan in the STS reconstruction have NOT been removed, but instead the overall quoted error envelope has been expanded to cover such systematics.

I. INTRODUCTION

The processing of the data for each 24 hour packet was accomplished by splitting each into manageable pieces called "orbits". Each orbit is measured from Point-D to Point-D; the orbit numbering scheme was based on the list of Point-D times provided by Onaka-san; the first Point-D, being at 95-03-18 08:23:57, defines the start of orbit 1. Since packets could span an orbit, the orbit fragments were also assembled. This process yielded orbits 516 through 568, which were then processed through the pipeline. For the pointing reconstruction purposes, each orbit was further broken into segments, with segment A going from point-D to eclipse start, segment B going from eclipse start to eclipse end, and segment C going from eclipse end to point-D.

The star sensor data were time-tagged according to the ASCII times in the concatenated irts_lan files; the ASCII times and spare bytes sp3 and sp4 were used to determine contiguity in time for the STS data.

The gyro data (synchronized by the latest time correction equation provided by Nakagawa-san 9/22/95) were used to compute the initial scan rates for the source extractor; the STS position reconstruction module used this initial scan rate as a starting point and then modified it to correspond to actual star observations. The gyro data were also used to determine the time periods (put into the TGOOD file) during which the spacecraft performed smooth scans; 2.7 sigma "events" at the beginning of each segment were marked and used as settling time periods.

No STS data were rejected because of gyro data; the gyro data were used to identify time periods where the scans were not nominal. The STS position reconstruction has been extrapolated into these settling periods, but the reconstruction uncertainty has been increased to 30' as a warning and an aid in identifying these periods.

The stream of STS data was fed into the source extractor which detected point sources crossing the STS; the extractor computed the J magnitude and position of the source in a 2-dimensional coordinate system aligned with the scan direction.

The positions and magnitudes of detected sources were fed into the position reconstruction module; this module found counterparts to STS detections in the JCAT, and using the match information derived the best estimate for the position of the star sensor as a function of time.

To propagate the STS reconstruction to the BORESIGHT, the focal plane geometry from Table 1 is used.

In the above table the values are in degrees and the boresight is defined to be at point (0,0).In this scheme, during the post-flip observations, an object is seen first by MIRS and then by NIRS. The above focal plane geometry and coordinate system is seen in Figure 1 (./figure.01.ps).

A small sample of data from MIRS were consulted to check the correctness of the focal plane geometry.

II. SOURCE EXTRACTOR

The source extractor utilizes a two-step procedure for the detection and position-estimation of point sources. The detection step involves the use of a matched filter derived from the 2-dimensional response of the detector, the STS response map; the matched filter output is then thresholded at a specified S/N level to produce a list of candidate detections. Next, the position and magnitude of each candidate is estimated by using a maximum-likelihood procedure. In addition, the quality of the fit is assessed in order to determine whether the candidate detection is, in fact, a genuine point source.

In order to improve the ability of the algorithm to discriminate against spurious detections, the STS response map was refined using the observed profiles of strong sources. Approximately 500 sources of magnitude 4.5 or stronger, detected in the overlap region, were used to modify the original response map (provided by Onaka-san). The maximum correction was of the order of 10% of peak response. The validity of the refined response map was verified by performing a further iteration, whereby the residuals were found to be noiselike, and of the order of 1% of the peak. The response map generated in the first refinement step was therefore retained for all further analysis. The original response map and the refined map are shown in Figures 2 (./figure.02.ps) and 3 (./figure.03.ps).

The source extractor was set to detect point sources brighter than 6.5 mag. Over 12000 detections were obtained over the overlap region.

III. POSITION RECONSTRUCTION

The STS position reconstruction was accomplished by first matching, then fitting, the pattern of extractions in xy-scan coordinates to a set of reference stars taken from the JCAT catalog which was developed at IPAC for this purpose.

The reconstruction proceeded on an "orbit" by "orbit" basis. Each orbit was further broken into segments A, B, and C as defined in the INTRODUCTION. For the overlap region it was possible to combine segments A and B, leaving us with only two fits per orbit. The data were broken so as to allow linear fits within each segment with discontinuities between segments. It is expected that outside the overlap region the orbit may have to be broken down further to handle events such as g-angle changes.

The first step in the reconstruction is to identify matches between the STS extractions and reference stars with accurately known positions. For each segment this is accomplished by a two-peg pattern matcher, in an iterative fashion. Two bright, well separated, extractions are selected and forced to match candidate pairs of reference stars from the JCAT catalog. Each trial is evaluated based on how many of the remaining STS extractions have close reference star matches. For the correct candidate pair, most of the remaining extractions will have close matches. The algorithm works best on segments which are neither too long nor too short. Experience has shown that, for the IRTS data, a match-segment of 600 to 700 seconds works best.

Once the initial placement of the scan on the sky is achieved, the goal of the STS position reconstruction is to determine the direction of the spin axis, an adjustment to the scan rate determined from the gyros, as well as adjustments to the in-scan and cross-scan positions at the start of the orbit. For each segment this fit is linear. In the future delivery the residuals remaining after the linear fit may be fitted in turn, in order to further reduce the reconstruction errors.

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_sub_I axis of the instrument.

Comparison with a limited amount of MIRS data showed that the satellite itself scanned along the y-axis of the focal plane as planned, and that this 2.7 deg angle represents a rotation of the STS relative to the focal plane.

IV. RESULTS

The residuals leading to the detection of rotation angle are seen in Figure 4 (./figure.04.ps). After the rotation of the PRF, the same distribution is nearly flat as in Figure 5 (./figure.05.ps).

There is a periodic trend in the cross-scan residuals whose cause is still undetermined, see Figure 6 (./figure.06.ps). No attempt has been made to correct for the periodic residuals in this release.

After the rotation of the STS is taken into account, the distribution of residuals of the in-scan and cross-scan fits have sigmas of 0.56' and 1.07' respectively. The histograms can be seen in Figures 7 (./figure.07.ps) and 8 (./figure.08.ps)

The in- and cross-scan residuals as a function of time for one orbit typically have a behavior like that seen in Figures 9 (./figure.09.ps) and 10 (./figure.10.ps).

The residuals in the in- and cross-scan directions as a function of time (measured from Point-D) show a systematic periodic trend in the C segment of the orbit as seen in Figures 11 (./figure.11.ps) and 12 (./figure.12.ps). The error envelope of 1' and 2' selected for this release covers these systematic effects.

Back to CONTENTS

V. COMPARISONS WITH THE SKY

A comparison of positions and magnitudes of sources repeatedly seen more than three times shows that the STS source extractor provides positional repeatabilities in the range of 10" - 40" and magnitudes are repeatable to 0.2 mag. A plot of the range of repeatability of the positions and magnitudes is seen in Figure 13 (./figure.13.ps).

The in- and cross-scan residuals of the STS reconstruction with respect to the sky indicate that for the time period from Point-D to exit from eclipse, the reconstruction has very little time variation. However, the reconstruction residuals from eclipse exit back to Point-D show a time variation akin to a limit cycle. The cause of this variation (with a period of approximately 1500 seconds) has not been determined yet; in a future release, these systematic variations will be removed from the data when possible.

In the time period of 900 seconds to 2450 seconds following Point-D, when there are no systematic time variations, the in- and cross-scan position residuals have rms errors of 20" and 33" respectively. The residuals distributions for the in- and cross-scan directions for this time period are seen in Figures 14 (./figure.14.ps) and 15 (./figure.15.ps) respectively. This shows that for these time periods, the reconstructed positions have a much smaller uncertainty than indicated by the overall error envelope.

VI. UNCERTAINTIES and CAVEATS

The reconstruction of STS position relies on the knowledge of the detector geometry and extent. The nominal data from Table 1 were used for the detector size; the refined response map was used to estimate the relative position of point C2 within that map. The reconstruction of the STS is therefore limited by the accuracy of this determination. Furthermore, the residuals statistics point to a rotation of the STS with respect to the focal plane; however, this has been corrected to first order. There may still be a small rotation angle between the scan direction and the y-axis of the focal plane, which will be detectable when a larger volume of data from other instruments becomes available to us.

Therefore an extrapolation of the STS position to the boresight will amplify these uncertainties. In fact, since the location of the boresight is not known very accurately, this release gives the position reconstruction for the point (0,0) in the focal plane (per Table 1). Currently a 1' uncertainty has been convolved with the STS positional uncertainties to yield the positional uncertainty for point (0,0) (resulting in error envelopes of 1.4' and 2.2' for the "boresight")

When gyro events indicated a change from the nominal scan rate at segment boundaries (typically running from tens of seconds to approximately 100 seconds), the time period overlapping these events was assigned a large uncertainty of 30' to indicate that such gyro events occurred.

VII. SCHEDULE & OTHER WORK

Following the release of the overlap region data, the next milestone is the release of the post-flip data (packets 04092249 through 04250016) by Mid-December (the problem of g-angle change and the periodicity of cross-scan residuals will be dealt with); we hope that we can deliver the post-flip data earlier than this date.

The pre-flip data will be treated immediately afterwards. The major area of concern for the pre-flip data is the source confusion problem. At present, final tests are being performed to validate the STS source extractor for the low latitude sky. By January 21 (and hopefully earlier), we will be able to provide the preliminary reconstruction of these data.

We regret that this delivery has taken longer than expected. On the positive side, this delay resulted in a more accurate understanding of the focal plane geometry, which will be improved as more and more data will be processed.

VIII. FORMATS

The STS and "boresight" history files described above, have the following format:
(ATTENTION: This format is slightly changed in the latest version. see release note on March 6, 1996. )

Header: (5 lines)

\ 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         |dummy|
|double        |  double   |double  |double  | d  | d  |double  |double |double  |double |double  |double  |  d | d  |double  |double |char   |int           |char |
|mmddhhmmss.sss|  double   |deg     |deg     |amin|amin|deg     |deg    |deg     |deg    | deg    |deg     |amin|amin| deg    |char   |char   |int           |char |

Format:

Definition of Variables:

FIGURE CAPTIONS

• Figure 1. The focal plane geometry adopted by IPAC (from TABLE 1) for position reconstruction; the measurements are in units of degrees.
• Figure 2. The original STS response map provided by Onaka-san.
• Figure 3. The improved STS response map; improvements based on sightings of approximately 500 stars brighter than 4.5 magnitude.
• Figure 4. The mean in-scan residuals (dy) as a function of cross-scan (xe) position of reference star sightings before discovery of rotation of STS.
• Figure 5. The mean in-scan residuals (dy) as a function of cross-scan (xe) position of reference star sightings after a rotation of 2.7 deg was applied to the star sensor.
• Figure 6. The mean cross-scan residuals (dx) as a function of cross-scan (xe) position of reference star sightings; the periodic variation in cross-scan has not been corrected for in this release.
• Figure 7. The distribution of in-scan residuals from reference stars (for all packets of the overlap region) used for position reconstruction.
• Figure 8. The distribution of cross-scan residuals from reference stars (for all packets of the overlap region) used for position reconstruction.
• Figure 9. The distribution of in-scan residuals from reference stars as a function of time for orbit 532.
• Figure 10. The distribution of cross-scan residuals from reference stars as a function of time for orbit 532.
• Figure 11. The variation of in-scan residuals (dy, in arcminutes), with respect to reference stars, as a function of time (T, in seconds) measured from Point-D for all packets of the overlap region.
• Figure 12. The variation of cross-scan residuals (dx, in arcminutes), with respect to reference stars, as a function of time (T, in seconds) measured from Point-D for all packets of the overlap region.
• Figure 13. The distribution of estimated dispersions in RA (sigma_RA, in arcseconds), DEC (sigma_DEC, in arcseconds) and magnitude (sigma_Mext) from the STS source extractor; the Figures are based on the data for stars seen more than 3 times during the first 2400 seconds after Point-D.
• Figure 14. The distribution of in-scan residuals (dy, in arcminutes) with respect to reference stars for the time period 900 seconds to2450 seconds following Point-D for all packets of the overlap region.
• Figure 15. The distribution of cross-scan residuals (dx, in arcminutes) with respect to reference stars for the time period 900 seconds to 2450 seconds following Point-D for all packets of the overlap region.