ard at p850ug1.demon.co.uk (Tony Duell) wrote:
I've never seriously used ESDI, but I was under
the impression that, like
ST412, it was a 'raw; interface with the bitstream to/from the heads on
the interface connector.
Mostly true. Not completely raw, the drive incorporates most of the clock
and data recovery circuitry. (Actually this is true on some MFM/RLL
drives too!) In that way a little more like most SMD implementations.
In whcih case there's always the possibility of
the controller foing
something unusual and not, say, recording the same sector headers as
other controllers. Maybe 99% of ESDI controllers did do the same things,
but it's that odd 1% that we're likely to have to deal with :-)
On PC-clones, most all the controllers were using the standard
WD chipset and will have standard sector headers etc. One issue
is that the largest ESDI drives used higher datarates (18MHz or
20MHz or higher) and not all controllers can handle such
data rates.
Outside of PC-clones, and into mini's/mainframes, some really funky
bitstreams/sectoring were being used in drive arrays (where large
numbers of >600MByte drives were sold to). In some cases the
ESDI drives were used to emulate CKD drives, where the concept of
a single fixed sector length doesn't exist at all (replaced by records
and record lengths) and every time a dataset was laid down
it effectively reformatted the track. In the early/mid-90's
large numbers of ESDI drives were coming out of mainframe storage
arrays and hitting the surplus market. Gees, that was 10 or 15 years ago :-).
More modern CKD drive arrays use drives with conventional sectoring
and do the funky stuff by packing up the emulated tracks into the
regular-ish sectors. (I say "regular-ish" because it is still common
to use 4096 or 8192 byte sectors, often plus a few extra bytes
for overhead). This makes much more sense because a modern
drive's track or cylinder will be larger than the whole disk
the CKD array is emulating!
Tim.