On 26/02/12 19:44, Tony Duell wrote:
A number of hard drives are like that. The worst one
I've seen is the
SA4000 (Shugart 14" winchester) where there's an aceptable speed range
for 'buffered seeks' and a slower range for track-at-a-timne seeks. But
if you get between them, the drive will mis-step and end up on the wrong
cylinder. With no warning.
Oh, this one has a great failure mode.
If you seek at 300us/step, the drive sits there as dead as a doornail
until you power-cycle it... INDEX pulses, but READY and SEEK_COMPLETE
are stuck inactive...
I suspect it'll probably reset if you do a deselect-and-reselect, though
I haven't tried that yet. What I'd really like is a copy of the ST277R
OEM manual (aka Product Manual)... I bet I'm doing something that it
doesn't like.
The 'seek wraparound' feature is a nice touch. Seek past the maximum
cylinder, and it seeks back to track 0. Bleh. The CDC Wren and Alps 20MB
(rebadged as Amstrad) just flag a seek error and wait for a
deselect/reselect cycle.
Is the problem in providng a greater reage of times
the length of the
divider or the numbero f bits needed to sleect the time? If the latter, I
mighty be tempted to have a doubling series onece toy get to, say, 10ms
(that is, have 10, 20, 40, 80 ms only). I can't think that would cause
any problems.
The main issue is getting enough resolution to specify the 'buffered
seek' step rates while still being able to specify the
tens-of-milliseconds values needed for slower drives.
--
Phil.
classiccmp at philpem.me.uk
http://www.philpem.me.uk/