FM, MFM, and GCR channel codes (was Re: stepping machanism of Apple Disk ][ drive)

From: Richard Erlacher <edick_at_idcomm.com>
Date: Sat Apr 17 16:37:13 1999

Yes, . . . at least I think I can.

When data is written to magnetic media, there is a maximal bit density which
can be written which is a function of the head and media characteristics as
well as the velocity with which the data stream passes the heads. each
transition is detected (read) as a pulse of a given width and amplitude.
Changes in the sense of the data are recorded by reversing the magnetic flux
written on the medium. If two flux reversals are written close together,
they can overlap slightly. The consequence of this overlap is that their
peaks are effectively pushed apart by the summing of the adjacent flux
reversals, since the two amplitudes are never zero but are simply reversed
while still having a non-negative value. The pulses resulting from the
recovery of these reversals in flux are added together. Since they overlap,
the insignificant portion close to the crossover are nonzero, so in the sum,
the rising waveform of the second pulse is subtractively combined with the
falling waveform of the first. This means that neither pulse reaches its
maximum amplitude but its apparent peak appears at the point on the rising
waveform at which the difference between the two pulses is reached, which
makes the first pulse peak earlier and the second pulse peak later. In
order to compensate for this, the pulse which would logically be detected
"too early" is written late, and the pulse which would be detected late is
written early, and as a result, the resulting pulse train is slightly
reduced in amplitude, but the peaks are detected at the proper times, which
is what matters.

Dick

-----Original Message-----
From: Mark <mark_k_at_iname.com>
To: Discussion re-collecting of classic computers
<classiccmp_at_u.washington.edu>
Date: Saturday, April 17, 1999 3:20 PM
Subject: Re: FM, MFM, and GCR channel codes (was Re: stepping machanism of
Apple Disk ][ drive)


>On 10 Apr 1999 Richard Erlacher wrote:
>>YES! and that's exactly why the pulse overlap was correctable with
>>write-precompensation at least in hard disks at 10x the data rate. In
>>general the amplitude of the pulses was sufficient to be detected by the
>>usual circuitry, but because the timing was quite far off due to the peak
>>shift introduced by the summing effect of the head/media combination.
>>Precompensation, which was not needed for FM, was between 188 and 125
nsec,
>>depending on the drives in use. On the older drives, 188 was pretty
common.
>>The "bit-shift" was mitigated somewhat by the reduction of write-current
on
>>the inner tracks.
>
>Can someone explain exactly what precompensation is?
>
>I remember reading about this in the Amiga hardware manual; in the ADKCON
>register there are two bits which set the precompensation (none, 140ns,
280ns
>or 560ns), and also one that selects GCR precompensation or MFM
>precompensation. Note that the encoding method is implemented completely in
>software, so you are not limited to MFM and GCR.
>
>(There's also a control to set the data rate, 2us or 4us per bit cell. 2us
is
>used for MFM, apparently for GCR 4us is necessary.)
>
>
>
>-- Mark
>
Received on Sat Apr 17 1999 - 16:37:13 BST