Beautiful Data

To Pack or Not to Pack

As with any highly constrained embedded system, the software needed to meet both its operational requirements and the constraints of its execution environment. As one might expect, these were not always complementary conditions, so trade-off decisions had to be made along the way. Both the SSI and the RAC/OM cameras utilized 12-bit conversion for the pixel data, which led to the first major trade-off decision: data packing. For a general high-level overview of binary data, see the sidebar BINARY DATA.

During the early design phase of the mission, the notion of packing the 12-bit pixel data came up and generated some interesting discussions. Given that only a limited amount of memory was available for image data storage, the concept of packing the 12-bit pixel data into 16-bit memory space was appealing. By packing, I'm referring to storing the 12-bit pixel data contiguously, without any "wasted" bits in between—in effect ignoring 16-bit memory boundaries. But more efficient data storage came at the cost of increased processing time (unpacking, shifting, repacking). A digital image is an array (whether it's treated as a 1-D array or a 2-D array depends on what is being done to it), so any operation on an image involved handling the image data utilizing one or more algorithmic loops working through the array. The amount of data we were planning to push through the RAD6000 was significant, and even at the full-out clock rate of 20 MHz it was going to be painfully ...