Interesting drive failure...

[Christian Corti]

On Tue, 21 Apr 2015, Josh Dersch wrote:
> Fujitsu MK2322) -- I'd been running it off of a spare, because I didn't want 
[...]
> pretty poor.  There are lots of bad sectors.  In particular surfaces 2 and 3 
> are completely unreadable for all 823 cylinders, surfaces 4 and 8 are 
> marginal most of the time.  After a lot of retries I've managed to get most 
> of everything (except for surfaces 2 and 3 which yield no data at all).

I have several of that exact drive in different machines. They have all 
been formatted a long time ago and reveal *many* bad blocks. For 
example, one is in a PCD Cadmus (QBus 68020 machine) and one in our Zilog 
S8000. Especially the Zilog's drive is valuable as it contains the IMO 
only existing copy of the ZEUS operating system for the Model 32 (I'd 
*LOVE* to get hold of an image of the tape cartidges).
My experience with reading out the contents of those drive (with my 
own-built simple ISA SMD interface) is that one has to do 1. many retries 
and 2. use all means that the drive provides to successfully read a track 
(i.e. offset+ and offset- commands).

> So I figured that heads 2 and 3 are toast -- perhaps bad analog logic, maybe 
> something in the digital logic that selects the head, maybe even bad heads, 
> who knows.

I don't think so. Old drives simply tend to get "weak". I suppose that the 
alignment of several heads changes slightly over the years.

Christian