B00ze wrote:
On 2018-04-30 07:33, J. P. Gilliver (John) wrote:

[snip]

Yeah, well, if the failed head has started magnetizing everywhere it
goes, then it's too late now ;-) If it's burnt something on the PCB,
then it will burn it right away again on the replacement PCB. I should
really replace both the head stack and the PCB (then I would not need
to mock around with the calibration chip.) But I've never done that
before; chances are high I can screw something up...

Well, not calibration, but you'd still need whichever chip holds the
list of bad sectors and which ones have been swapped in their place.

Ahhh, yes, now that's annoying; if calibration and bad sectors go into
the same chip I'm kinda stuck. But I guess it's OK, I can loose some
files, I just want the bulk of them...


The PCB I looked at, the outboard chip had too small of a capacity
to hold a spare sector table.

One IBM drive, it was claimed it was using 1MB of cache RAM to
hold the spares table for that drive. The little 8 pin chip
I looked at, was only 64KB of storage.

Whatever is in there, is smaller. Maybe it's an add-on
code module. They could have gone smaller, to a 2KB config
EEPROM if they wanted, and save some money. That suggests
64KB was selected for a reason, and there's actually close
to 64KB of stuff in it. They wouldn't buy a 64KB chip, if
a 2KB chip could identify the number of platters and the capacity.

The service area on the platter, is normally where bulk information
is stored. You'd only resort to an external chip, if the main
chip needed to be "patched" to be able to finish the access routine
to get to the SA. Maybe you could store the entire bootstrap
in the 64KB chip, and not bother with a level 1 metal ROM inside
the controller SOC. But then, they wouldn't need that nine digit
part number on the controller, if it wasn't "custom". The main
chip would have a shorter part number if it was generic.

The Maxtor that died on me, if it cannot read the SA, the controller
defaults to "declaring itself as a 10GB drive". It only changed
the ID string, when it sees the SA and then it knows "this is a
40GB drive with four platters" or whatever. The controllers used
to be smart enough, to handle several model variants, with
different platter counts. You could do that, say, by always
accessing platter 0 to get the SA (platter 0 would always
be populated in the stack).

As drive capacities go up, the odds of holding a spares
table in that external EEPROM go down.

Paul