I think there's a superimposition of info from different programs here. I just played a bit with HD-Copy 2.0 (by Oliver Fromme, see here: https://web.archive.org/web/20160304075734/http://www.asiawide.or.jp/HD_COPY.htm ) and the disk images files it produces are definitely not the one described here. It starts with a constant 3 bytes header FF 18 17, then 11 bytes for the name of the disk images, and then space from some other data. The real image data starts at offset 0xBB. Doesn't seems to be compressed, but the are probably just the actually used sectors, and the file size is variable.