Michael J. Mahon wrote:
save all the data that a Disk ][ Controller can read from each quarter track (including its length, determined by scanning for repeats), together with its angular alignment WRT track 0, sector 0, and a map of all bits that are "soft" (read differently on different read attempts).
Okay.* The only hope for cycle-accurate reading without extra hardware seems to be Stephen Thomas's code. It can be modified to flag only 10-bit sync for a few disk revolutions, and then all sync for another few; track length analysis could be done ignoring the high (sync flag) bit, and then the high bits from successive revolutions could be compared to distinguish 9 from 10-bit sync.
* How many revolutions do we need to accurately detect soft bits? And incidentally, why do we need to detect them? Is there software that depends on their having different values on different revolutions?
* Syncing to track 0 sector 0 is easy if it exists, but what if it doesn't? There are non-bootable copy-protected disks. I see no reason to expect any hard bits on track 0 at all.
* Should track length and soft-bit analysis be done by the 6502, or on a modern machine? It depends on how much harder it is to transfer larger amounts of raw data between computers, which depends on each individual user's setup. Do we have to implement the analysis twice?
* The imaging process will be very slow, and a single disk image will be too large to fit on an 800K disk, even if track length analysis is done on the 6502. How many people are going to jump through the necessary hoops to image their disks? It seems very inconvenient.
-- Ben