Rotating memory data recovery

[Gordon JC Pearce]

John Foust wrote:
> I like this approach in theory.  I think it has great long-term
> potential for data recovery and preservation for our hobby.

Certainly, if the original hardware can't reliably recover data stored 
on disks or drums, this is a good way to go.  The more information you 
have, the better chance you hace.

> Anyone care to give a thumbnail sketch of what sampling rates
> would be necessary to digitize the signals on recording media 
> (3 1/2", 5 1/4", 8" floppy, hard disks, drums, disk packs, etc.) 
> at sufficient rates to preserve the encoding for post-processing
> in other tools?  
> 
> I saw one reference online that said 3 1/2" and 5 1/4" floppies
> were 250-500 kbit/s.  If we were digitizing at a much higher 
> rate than the original simple circuitry that might've only
> detected flux traversals, couldn't we recover more data?

Well, 300kb/s would suggest that if it went 10101010 then you'd have a 
150kHz signal.  Now, possibly it uses some kind of self-clocking or NRZ 
code, so it may be 300kHz.  I'd say that 8 bits of resolution would be 
plenty, and perhaps four times the data rate, so a sample rate of 1.2MHz 
might well be adequate.  That would be well within the range of 
(relatively) cheap A-D converters.

Obviously, the higher the resolution, the slower the conversion.  Most 
A-D converters (and I'm sure many, if not most of you know this) use 
"successive approximation" to derive the correct value - get a D-A 
converter and a comparator, then loop through "toggle the MSB, is it too 
high or too low?  Toggle the next bit, too high or too low?" all the way 
down.  You get "flash" converters for video (or you used to) which were 
32 or 64 individual comparators and voltage references, and some 
encoding logic to give you a binary output.  Overkill really, but very, 
very fast.  If you found a fast 6-bit flash converter cheaply, that may 
well be enough resolution for your track scanner.  We're not going for 
Dolby Digital quality here...

Gordon.