Date: Thu, 3 Jan 2008 21:07:02 +0000 (GMT)
From: ard at p850ug1.demon.co.uk (Tony Duell)
If by 'commodity solutions' you mean the
WD1010 or whatever, then I have
to disagree with you. Plenty of manufactuers used their own ASICs in the
hard idsk controller, and admittedly most of them did do a fairly normal
MFM or RLL2,7 encoidng. But I'll bet at least one didn't!
The latter is what I said--and meant. "Unique" encoding schemes,
such as one finds on floppies (zoned recording, CLV, various flavors
of GCR, FM, whatever you want to call the Apple II flavor of GCR,
"mixed" formats, such as the RX02, etc.) were comparitively rare.
MFM and RLL 2,7 were overwhelmingly the rule on the ST506 and ST412
class drives. I mentioned the Perstor with ARLL, but it was hardly
the rule.
And I've got machiens that use a 8x305 in the hard
disk controller.
And there are plenty of systems that use the 8x30x and the WD1000
series chipsets (the WD1001 being one of the boards). Still MFM.
A kludgeoard was added (and the CPU microcode device
diivers changed)
to talk to a Micropolis 1203, and finally a different kludgeboard and
CPU microcode were used to talk to ST412 drives. I would not want to
bet that the encoding on that was normal, at least not without doing a
lot of tests.
I'd be willing to wager that it was MFM or M2FM of some flavor,
however. I've got an SA4000 and controller that's implemented in TTL
with nothing more complex than a few 74LS181s and a few ROMs. It's
still MFM recording.
My point was that with a decent data separator for the common
modulation methods and a simple deserializer, you'd have all you'd
need to cover the overwhelming majority of ST412-type drives. No
need for fancy transition-time recording as in the Catweasel (where
one often can spend a great amount time staring at a histogram and
pulse-time list trying to figure out how clever the designers of a
particular floppy format were.)
Cheers,
Chuck