Fermi Science Support Center

Caveats About Analyzing GBM Spectra

The GBM data delivery system has recently experienced problems owing to high count rates during periods of intense solar activity. The large number of triggers has caused downlink and processing delays. The GBM instrument team has implemented two changes to restore data quality, improve latencies, and ensure consistent delivery:

(i) All trigger algorithms operating on count rates below 50 keV have been disabled.
(ii) The Low-Level Threshold (LLT) has been raised to about 10 keV in the sun-facing detectors (NaI 0 - 5).

We hope this LLT value will be high enough to mitigate the problems, as we are sensitive to the loss of science that will result by our raising energy thresholds. Changes to LLT settings and related instructions to users of GBM data are documented:


The effectiveness of the measures will be reviewed and nominal settings will be restored when possible.

During the period July 4 - July 9 the GBM also exhibited data issues owing to the thermal configuration of the spacecraft while Fermi was pointed at the Crab for a Target of Opportunity observation. The onboard electronics were held in a temperature range in which known timing anomalies became common and were severe enough not to be fixable on the ground. These anomalies resulted in data glitches which manifest themselves as spikes or holes in the lightcurves. The data are reliable for temporal resolutions greater than a few seconds, over which the spikes are smoothed out, but users should be aware that untriggered spikes in CTIME and TTE data are likely not indicative of astrophysical activity on short time-scales.

In analyzing the GBM data, you should be aware of the following:

  • The EXPOSURE keyword in CTIME and CSPEC files was discovered to have a slightly incorrect value because a deadtime-related adjustment to the width of the time bin was incorrect. This problem is corrected for all data files produced from May 16, 2013. The data for GRB 130427324 were fixed and redelivered. For all other CTIME and CSPEC data prior to May 16 2013, the effect of the incorrect exposure on true count rates calculated from this exposure is small. At the boundary between the nominal (4.096/0.256 s) and accelerated (1.024/0.064 s) CSPEC/CTIME accumulations that occur when GBM is triggered, the exposure-adjusted count rates above background can be incorrect by a few percent, up to 10 - 20 % for very sharp rises, but typically 2 - 3 % or less. During triggers, spiky events may result in time bins that are incorrect by a few percent, but typically exposure-adjusted count rates after the initial accumulation discontinuity are incorrect by less than 1 %. The TTE data are unaffected by this bug. The GBM team is evaluating a possible reprocessing of data to correct this error in data delivered prior to this discovery.
  • Shadowing in Detector Response — The LAT, other GBM detectors, the spacecraft body and/or the solar panels may be in the field-of-view of a given GBM detector, and these can both shield the detector and scatter burst photons into it. Therefore, the GBM's response is very dependent on the localization of a burst on the sky. It is important to have responses generated for the most up-to-date source location, especially if this can be obtained from the IPN or a mission such as Swift. The response matrices will be routinely updated by the GBM team when improved localizations can be obtained.
  • Detector Response Time Dependence — Fermi will most likely change orientation during the burst, with a possibly significant change in GBM response, particularly for long bursts. Most bursts will be detected while Fermi is in survey mode, with the Fermi pointing moving at ~4 degrees per minute. In addition, Fermi may autonomously repoint towards the location of a strong burst on the timescale of a long burst. The response file provided with the GBM data will include a series of response matrices (in separate FITS extensions). If only one matrix is found in the file, it was calculated for Fermi's orientation at the time of the GBM trigger. If multiple matrices are included, then the analysis software should use the appropriate file for the designated analysis.
  • Calibration — The Iodine K-edge at 33.17 keV, above which an incident photon results in the release of an X-ray from K-shell transitions, is seen as a discontinuity in the response of the NaI detectors. For bright bursts, this can result in large residuals around this energy when spectral fits are performed with the NaI detectors. These residuals contribute to fits that are judged poor in terms of the chosen statistic values for a particular fit model but do not appear to cause a bias in the fit parameters. The user may wish to exclude energy channels at and around the K-edge in order to better assess the quality of the model they are fitting to the data. A description of the ground calibration performed on the GBM detectors is given in Bissaldi et al. Experimental Astronomy, Volume 24, Issue 1-3, pp. 47-88.
  • Low Level Threshold Changes — In order to support Fermi pointed observations without loss of data quality or excessive data volume, the GBM team is occasionally adjusting the level of the discriminators for the sun-facing NaI detectors (NaI 0 - 5) so that their energy threshold is higher than in nominal mode. During these periods of non-standard Low Level Threshold (LLT) operation, photons with energies below this threshold are not recorded. Data above this threshold energy are nominal, as are all data from the BGO detectors and from NaI detectors 6-11. The energy thresholds of the affected detectors are between 18 - 30 keV, depending on the nature of the pointing. See the LLT Settings Page for more information.
  • Timing Glitches — GBM Time-Tagged Event (TTE) data suffer from timing glitches arising from rare conditions in the FPGA logic that produces GBM science data onboard. Every effort is made on the ground to correct these glitches but some are not cleanly reparable using pipeline software logic, and the TTE data files occasionally show the effects of these glitches, which can be seen in the TTE lightcurve. More specifically, TTE data have a fine-time clock that rolls over every tenth of a second. In order to reconstruct the event times in Mission Elapsed Time, a coarse time word is issued every .10 s; the full event time is the sum of each event's fine time and the previous coarse time. Due to hardware limitations, there are cases where the TTE coarse time word is either missing, inserted late or not updated. Most of these cases are detected and corrected by the TTE processing software. However, there are still cases in the TTE event time reconstruction where a time word error is undetected and the event data get shifted in time by .10 s. In that case, it is easily detected in time-binned TTE data, where the problem is manifested as an adjacent pair of bins where one bin has a low rate and the other high. This can be completely corrected by combining the two bins into one. The raw event data are more difficult to correct, since it will not be clear which events correspond to those missing a .10 s (plus or minus) correction.

As the GBM instrument team continues to test the detectors' in-flight reponse, the detector response functions will undoubtedly evolve and these caveats will change.

The GBM team is in the process of developing solutions to these known issues.

Please contact the FSSC helpdesk for assistance.