On 10 Jun 2010 at 15:46, Eric Smith wrote:
My experiments with floppy data recovery in software
used a simple
DPLL, and I found that it tracked the data much better than any simple
threshold scheme.
What I was trying to say, but in a rather roundabout way. Given
that most common floppy data separators can accommodate at least a +/-
10% variation (or more) in long-term speed variation, a too-slow
drive can easily write a sector that's way out-of-spec timing-wise on
a track that's otherwise perfectly in spec. It's necessary that any
read algorithm be able to adjust to short-term (sector-to-sector)
variations, as well as long-term (track to track) ones.
Given that you *know* the structure of a track ahead of time, there
should be no reason that your data recovery routines can't be *much*
better than the usual brain-dead floppy controller's efforts.
Ethan Dicks wrote:
Except that Amigas read and write an entire track in
one pass,
unlike most disk schemes.
Heh, there are a couple of other systems that do that (Commodore
didn't invent it). I've got an 8" hard-sectored floppy sitting on my
desk here, where the data for cylinder 9, complete with address ID is
duplicated on cylinder 3. Apparently there's no "look before you
write" in whatever wrote it.
--Chuck