3.1.3.8 Bias Telemetry Choices

The means of telemetering the bias values will depend on the choice of event recognition and duration of time that a single bias calculation remains valid.

In bright mode, where telemetry bandwidth is at a premium, the events are graded and biases subtracted internal to the on-board processing. Only the total event size and event grade (as reconstructed on-board) is telemetered on an event-by-event basis; no bias information is telemetered with the events.

In faint mode, where more telemetry is available for non-event transmission, the user can elect to have the local 3x3 bias values telemetered along with the 3x3 event pixel values. If the same bias remains valid for an extended period, this method is an inefficient use of telemetry, as the same bias values get telemetered with succeeding samplings of the same locations (which occurs frequently due to the narrow point spread function of the AXAF mirrors). To avoid this inefficiency, faint mode can also select not to contain bias data for each event.

When bias is not telemetered on an event-by-event basis, then the entire bias frame must be sent to the ground. Complete bias maps are compressed using a Huffman first-difference algorithm, with re-loadable compression tables. Each row is compressed separately and interleaved in the downlink telemetry by a BEP processing task called the `bias thief'. This interleaving allows ACIS to resume observations after a bias calculation, without having to wait for the bias telemetry to be complete. (We estimate that it will require a maximum of 24 minutes to downlink all six bias maps if the full 24 kbs science telemetry is free. At calibration the bias maps were telemetered much more quickly than these estimates due to higher than expected compression efficency resulting from the high degree of uniformity of the bias maps. In this case bias telemetry only took 11 minutes with the full telemetry stream.)