Re: Clock Bits On Disk Question

[mjmahon@aol.com (Michael J. Mahon)]

David Empson wrote:

<snip>
Again, a lovely and complete description of disk encoding--this
should really find its way into the FAQ.  It certainly comes up
regularly.  ;-)

>The problem is that the disk can reliably reproduce consecutive flux
>reversals when read back, but prolonged absence of a flux reversal
>produces unrelabile data readback. 

It may be useful to say why the prolonged absence of flux reversals
is an issue.  Magnetic coatings are applied to the substrate by a
mechanical process that can produce nonuniform thickness and
properties.  As a result, the amplitude of the signal read back from
a disk may vary considerably with rotational position, and these
variations may be relatively short-term.  The drive electronics must
be able to cope with these rapid amplitude variations in order to
correctly recover the magnetic transitions.

The method of coping is to vary the read signal gain to try to keep
the readback levels approximately constant in the face of wide
variation in read signal levels.  This Automatic Gain Control works
by looking at the short-term average of the head signal amplitude.
If the amplitude drops, the AGC increases the gain to compensate,
with a time constant that depends on the anticipated data rates
and media characteristics.

In the Shugart drive which was the basis of the Apple Disk ][,
the AGC time constant was chosen so that the gain would rise
or fall significantly within several bit cells.  As a result, when there
are no magnetic transitions for three or four bit times, or whenever
the average number of transitions within a half dozen bit times falls
below average, the AGC turns up the gain enough that read signal
noise can begin to look like an actual transition signal--a false 1.

This is why both the number of consecutive 0 bits and the frequency
of occurance of multiple 0 bits must be limited to ensure reliable
recovery of the transitions written to the disk.  The AGC must be
kept supplied with a signal so that it keeps the gain properly
adjusted.

There is a design tradeoff between the quality of the anticipated
disk media and the time constant of the AGC.  The choices that
Shugart Associates made were appropriate for the rather low
uniformity of media pravalent in the mid-70s, but coating technology
and composition have improved somewhat since then.

It might be interesting to increase the capacitance of the AGC filter
on the Disk ][ analog card, which would make it less tolerant of
short-term read level fluctuations, but more stable in dealing with
longer strings of 0 bits.  The controller would still need the high bit
set as a "start" bit, but the remaining seven bits might be usable
without further encoding--providing a 7-and-1 encoding scheme. ;-)

-michael

Check out 8-bit Apple sound that will amaze you on my
Home page:  http://members.aol.com/MJMahon/