HEIDI SOLAR ASPECT SYSTEM
Design of the Aspect System
Purpose: To provide aspect data
- With high-bandwidth,
- With high-precision.
- Grid-grid vector to be provided with respect to sun center
- At the time of arrival of each X-ray photon.
Optics:
- Lens/filter assembly mounted on the front grid tray
- Focussed an optical solar image onto a detector assembly mounted
on the rear grid tray.
- Both aspect system and X-ray optics would respond identically to
telescope deformations
Detectors
- Four inear diode arrays
- 1024 x 25 micron pixels
- 25 microns => 1 arc second
Scan Rate
- The output of each array was scanned every 20 ms
- Analog thresholding used to identify the limb pixel.
- The telemetry included up to 4 limb pixel numbers
- Useful number of crossings = 3 or 4
Calibration
- Accurate measurement of the relative locations of the ends of the
array
- Initial measurement error estimated as roughly 25 microns
- Done before and after integration into the caniister using ECDS system
- Inconsistencies of about 60 microns found
- Reconciliation of the two data sets reduced internal inconsistencies to 48 microns
- Inconsistency is due in part to refraction in the sensor cover
19 March 1998 EJS HEIDI_SAS page 1
Microprocessor
- Programmed to obtain a rough aspect 25 Hz solution
- To provide input to the pointing control system
- Resulting aspect solution was satisfactory,
- But software problems with acquisition software,
- So used a Lockheed Intermediate Sun Sensor instead for tracking.
IN-FLIGHT PERFORMANCE
Overall Functioning
- In general the system functioned well,
- Providied arcsecond solar aspect 50 times per second.
- Occasionally (less than 1% of the time) the digital data transfer from
the SAS was corrupted by noise,
- But analysis algorithms easily identified and rejected such occurrences.
Quantitative Evaluation
- Quantitative evaluation of relative aspect performance is based on the
fact that
- Each readout not only locates Sun center,
- But also provides one or two independent measurements of the solar radius,
- Roughly speaking, the former is the half the sum of the two limb
positions,
- The latter is half the difference.
- A histogram of 13 minutes of
Radius data
- can contain up to 17,000 estimates of solar radius
- and provides a good measure of system performance.
- Ground testing indicated that performance was limited by
atmospheric seeing.
- In-flight performance was much better.
- Found discrepancy between pre-integration and post-integration
measurements of SAS positions.
- Using pre-integration SAS positions, Four-limb crossing data yielded:
- 4 determinations of radius for each time, using 3 arms
for each determination
- Initial analysis (including both 3- and 4-crossing data) showed a
scatter of 0.95 arcseconds RMS in the
solar radius histogram.
- Plot of radius R vs inferred sun-center x position (XC)
without using -X arm
- Showed a flat distribution
- With FWHM about 1 arcsec
- Internal parabolic curves--an effect of
quantization
- Plots of R vs XC using other 3-arm sets produced non-flat R vs
XC
,
- For example, the plot of radius
without the +X arm
produced non-flat R vs XC
- The other plots showed similar bumps and curvature.
19 March 1998 EJS HEIDI_SAS page 2
- Attempted resolution of discrepancy between pre-integration and post-integration measurements of SAS positions.
- Used redundancies to optimize the CCD pixel locations
- Varied the estimated x-array position until the R(XC) curve
flattened:
- Optimum locations corresponded to positional errors of 65 microns
RMS
- Concluded that pre-integration measurements were more "correct".
- Summary:
- The -X arm appears to be 'bad'-- inconsistent with other
3-arm data
- RMS radius data was 0.5 asec (except for 3-arm data with
-X arm)
- Optimized SAS (without -X arm) resulted in
3-crossing histograms
with 0.3 arcseconds RMS.
- This can be compared to the value 0.3 arcseconds, determined by
computer simulations,
that might be obtained with
a perfect system with
precise 1-arcsecond (25-micron) sensor pixels.
- It is especially noteworthy that this distribution was very 'clean' in
that typically there were no 'bad solutions' in typical runs of 30000
estimates.
- The residual errors can be attributed in part to random pixel-to-pixel
variations in sensitivity.
- In principle, the effects of such variations can be removed by
statistical analyses given samples of 10^6 or more outputs.
Lessons Learned
-
In terms of design, the use of 4 half-length rather than 2
full-length sensors had two undesirable effects: the effects of the
'gap in the middle' of the sensor were twofold:
- First, it reduced the field of view
- Second, it added 6 additional degrees of freedom to
possible pixel locations.
- Another weakness in the design was
- that since the outputs of successive pixels were
transmitted serially through a lengthy cable before analog
thresholding was performed,
- cable dispersion was an uncertain factor
- which would systematically bias the limb determination.
- Digital thresholding at the sensor would alleviate this problem.
- In retrospect, additional care should have been taken both in
- mounting the sensors (whose positions were constrained only by
normal dip sockets)
- and in prelaunch calibration of the sensor locations.
- Fitting the sensor locations as we did on the basis of inflight data,
has the potential to
- cover up other problems
- and so introduce systematic errors.
-
Once these steps are taken, performance could also be improved by
- interpolation of two or more sensor outputs at each limb,
- thereby overcoming limitations associated with pixel resolution.
- Inflight calibration of the relative sensitivity of the 4 arrays would
also be highly desirable both to
- avoid systematic biasses in the solution associated with sensor to
sensor differences
- and also to provide an alternative means of evaluating pixel-to-pixel variations.
19 March 1998 EJS HEIDI_SAS page 3
Summary
In summary, the SAS performance, both qualitatively and quantitatively,
would have provided
- Excellent support to X-ray observations
- Resolution of a few arcseconds.
- WIth the refinements suggested above, sub-arcsecond performance
should also be possible.
19 March 1998 EJS HEIDI_SAS page 4