Beautiful Data

Downlink, or, It's All Downhill from Here

The last step in the process was the hand-off to the downlink manager in the spacecraft flight software. Some science instruments could simply pass their data to downlink and be done with it, but because of the large amount of data and the use of packetization, the ICS ended up doing a lot of downlink preprocessing on its own.

For the JPEG data, this meant handling each of the compressed segments individually. The first and last segments in a sequence always included a full-sized image data header. The intermediate segments got a smaller form of the header data, which included an image ID code and a sequence number. As each segment was read from the image slot, the header data was applied. The use of a sliding window form of readout allowed the segments to be packed end-to-end while assembling a flash-sized packet. This in turn allowed the downlink handler to maximize the use of the temporary flash storage space, because some of the compressed segments could be smaller than others if the part of the image corresponding to a segment had a low entropy. In fact, it was common to see compressed segment sizes vary widely, so packing them end-to-end avoided wasting any of the on-board flash memory.

Because the data consisted of uniquely identified segments, the loss of a downlink packet wouldn't consign the entire image to the garbage. The reconstruction and decompression software back on the ground at JPL could figure out what segments were missing ...