EARLY DATA ANALYSIS WITH HESSI
After a month in orbit, HESSI is operating well. Dozens of events have been well-observed to date, including several M-class x-ray events.
The purpose of this report is to briefly outline the hardware and software issues that users will face during early analyses of HESSI data. Note that software improvements discussed below will be fully applicable to all data acquired since the beginning of the mission.
POINTING AND ASPECT
The spacecraft is rotating at 15 rpm with its spin axis pointed to Sun center within programmed limits of 3 to 6 arcminutes. The solar panels have been adjusted so that the imaging axis is aligned to the spin axis to about 4 arcminutes. An apparent orbital variation in this alignment has been noted, but this has no significant effect on imaging.
Off-line analysis of Solar Aspect System (SAS) data indicated that it easily meets the (0.4 arcseconds rms) requirement for 2 arcsecond imaging. Using a simpler algorithm, the current SAS analysis software is constrained to a 1 or 2 arcseconds rms solution, which is adequate for imaging with grids 3 through 9. The on-line version of the higher resolution algorithm is currently under test. Users are also cautioned that for now, there may be intermittent periods of higher aspect uncertainty shortly after sunrise.
In the first few weeks of the mission, aspect diagnostics required rebooting the aspect computer twice near the middle of each daylight orbit. Such reboots introduced ~20 second data gaps during which imaging is not possible. Until software filters are installed, users can use hsi_aspect_solution() to verify that their imaging does not extend into or across these gaps.
The primary roll aspect system (RAS) is operating, but its performance and analysis are complicated by a higher-than-expected number of false triggers and by issues related to earth albedo. We are hopeful that these matters will be resolved and that RAS software with suitable software filters will be available within the next few weeks.
Meanwhile, the backup phototube-based roll aspect system (PMTRAS) is operating nominally and is providing robust and accurate data. The spin period has been found to exhibit a significant (for imaging) orbital variation (2 ms peak-to-peak). As described in an earlier memo, the pmtras_analysis routine can be used to determine the roll period that is then input into the imaging routines. Absolute roll aspect can also be determined at present, but at an unacceptable level of inconvenience. Integration of the PMTRAS output into the aspect solution is very close to completion, at which time a complete, user-transparent, absolute aspect solution will be available.
Overall, the spectrometer is operating well and has yielded splendid high-resolution continuum spectra of flares. Background lines are also readily seen and so nuclear spectroscopy awaits only solar cooperation.
The cooler is operating well at about 50 watts with low vibration levels (~10 mG), keeping the detectors in the temperature range 72 – 75 K.
Seven of the nine germanium detectors are operating perfectly with the nominal 1-keV resolution and 3-keV threshold levels.
Detector 7 is operating as it did during integration with 3 keV resolution and a threshold of 7 to 8 keV.
After the first week or so, detector 2 has been operated as a single large detector (viz, no segmentation) that provides a 25 keV threshold and 10 keV resolution. An on-board software configuration issue inadvertently suppressed events from detector 2 from Feb 26 through March 8.
Detector 8 operates normally, except for demonstrating a susceptibility to RFI from the rear transmitter. The transmitter is on for about 50% of the time during ground contacts. This represents an average of about 6 5-minute sunlit periods per day. Suitable software filters will be implemented, but in the meantime, users should disregard data from detector 8 during these intervals.
Accurate energy calibrations of all detectors have been achieved using known lines in the background spectrum. Compared to prelaunch predictions, background levels range from expected levels at equatorial latitudes to ~x2 higher at high geomagnetic latitudes. When background levels are an issue, a significant improvement in background reduction for front segment events can be obtained by considering only events in anticoincidence with rear segment events.
The front segments of all detectors are showing substantial periods of dead time (data gaps) that last from a few milliseconds to several tenths of a second. This is believed to be due to large amounts of charge deposited by hi-Z cosmic rays. The most important immediate consequences are that uncorrected, high-time-resolution, single-detector light curves exhibit obvious data gaps; and that image quality is degraded, particularly for integration times less than ~30 seconds. The good news is that a reliable RESET-based signature of such occurrences has been identified and the live-time algorithm is being adapted accordingly. When these corrections are implemented, the effects of the data gaps on imaging and composite light curves should be largely eliminated, except for an increase in statistical uncertainties.
At present, while the software and hardware supports science-quality continuum spectroscopy, early spectroscopy will require considerable care. In the coming weeks, users can anticipate improvements in background subtraction algorithms and convenience. During their initial analyses, users should be aware that the instrument response is still uncertain at energies below 12 keV.
With some important caveats, 7-arcsecond imaging with subcollimators 3 through 9 is now possible. Image quality is currently constrained by the data gap issue discussed above. Until the software incorporates this into the dead time, image quality will be limited, although useful results have been obtained with CLEAN and temporary versions of MEM-SATO and PIXONS. Phase calibration for grids 3 through 9 is preliminary but usable and will be improved over the next few weeks. Phase calibration for grids 1 and 2 will be done when suitable compact flares have been identified, and the high-resolution aspect solution becomes available.
Both the thick and thin attenuators have been exercised successfully. Data has been acquired in attenuator states 0 and 1 (‘both-out’ and ‘thin-in’ respectively). With both attenuators open, significant steady-state solar count rates have been observed, along with flares as weak as the GOES B2 level. The on-board algorithm for automated attenuator operation is currently under test.
For the most part, operations are running smoothly. Memory management strategies (e.g. event decimation models) continue to be evaluated. Data are being converted to the level-0 format using automated algorithms with a latency of a day or two. Catalog information will be available shortly, at which time it will be added to previously processed data. In particular, flare locations await the integration of the absolute roll aspect data.
Direct comparisons with complementary observations - GOES, SOHO (EIT and MDI), TRACE, radio (Owens Valley, Nancay, and Nobeyama), and H-alpha (Big Bear) - are readily accomplished thanks to the mirroring of these data sets and their accessibility through the HESSI software. One identified weak point in the complementary observations is the paucity of high-time resolution (~5 s) imaging data in optical, EUV or X-ray wavelengths.
This list of issues indicates that analyzing HESSI data will be more challenging (but of course more rewarding) than working with the simulated data. At first, it will not be easy. But as we move up learning curve and as the software and calibrations are adapted to meet both our needs and the demands of HESSI’s real-world performance, it will get easier. At the same time, we can expect the quality of our images, spectra, etc to improve significantly. Because of this, users doing early analyses may want to ‘do the easy stuff first’ and avoid the temptation to immediately push the HESSI data to its limits in terms of image quality, energy range, statistics, etc.
Documentation concerning data access and software can be found at
Please send bug reports to email@example.com
This site last updated November 10, 2008.