On Tue, 19 Feb 2019, Paul Koning wrote:
I' reminded of the Bordynuik tape reading machine that uses an MR head. Capturing analog flux levels at, say, 10x the nominal flux change density means all the rest is simply digital signal processing. That can be done in real time if you must, but much more easily in post processing using whatever tools and languages you like.
As Al pointed out, at least an initial stage of analysis should be done real-time, or close to that, to determine whether the read seemed to have been successful. 'Course, that doesn't preclude later stages from also requesting a new attempt, if the data doesn't look good on further analysis.
And, of course, using multiple machines, or just multiple processes, one system could begin processing, while another is reading more tracks. The reliability encountered would determine how much testing should be done before moving on, either to the next track? or the next disk? If errors are RARE, then there would be no cause to hesitate and hand off processing while moving on to the next JOB. If errors are FREQUENT (Al, Chuck, and I dealt with some that required dozens of attempts to get a successful read of a track), then that would caall for MORE processing before stepping to the next track or disk.
FDC does a CRC "real time", and compares that with the value stored in the sector header.
