10 Apr
2007
10 Apr
'07
5:47 p.m.
On 10 Apr 2007 at 3:44, Jules Richardson wrote:
If the numbers work out, then I wonder if an improved
format would be to have
variable-length counts: say three bits of "header", with the first bit
recording presence/absence of index pulse (vital, as you say), then two bits
which state how many bits make up the count value. That gives four possible
lengths of count which could be encoded....
I don't think that would be profitable. Consider MFM recording--you
have pulses that can occur at N, 1.5N and 2N, and for address marks
with missing clocks 2.5N intervals. Given the right clock selection,
there's no real need for a lot of precision. Consider that most
data separator circuits can tolerate about a 20 (+/- 10) percent
variation in speed quite easily. A floppy drive is not a precision
device, in my experience.
Part of data recovery is relying on the categorization of recording
methods. So, for instance if, on an MFM diskette, you find a pulse
that occurs less than about 3N/4, you're safe discarding it.
Similarly, if you find a gap of 4N, you probably should assume a
pulse is missing. And this is really where the "gold" is. You can
be a lot smarter with software than a "dumb" data separator can be
when it comes to figuring out how to recover from an error.
Where are you typically finding such formats - other
than the well-known
machines that used GCR? Is it more common in some of the factory process
control stuff etc. that you look at, rather than home/business systems?
Some older word processors use it and you can find it on some early
personal/business computers as well as some industrial equiopment
(e.g. Future Data development systems).
Cheers,
Chuck