There's probably an easier way to look at the issue of data rate versus
density on magnetic media. First of all, you must realize that every
head/media/rotation-rate combination has a maximal flux reversal density
with respect to time. Once you know that, you need merely understand how
many flux-reversals the heads must induce on the media in order to record
what will be recovered as a single bit.
When you look at the manuals which come with various drives and controllers,
etc, you get the whole shmear in timing diagrams. Unfortunately, these are
marginally misleading. The drive takes the data stream it receives, in most
cases, and divides it by two, using the complementary outputs of a flip-flop
to drive the circuitry which drives the read/write head. Each time it
toggles, it produces, effectively, a single flux reversal on the medium.
When the data is read back, it produces a waveform more closely resembling
what's in the "books" in that the flux change is "seen" as a pulse by the
read/write head. This feeds filters and timing circuits and conditioning
circuits which make it into precisely what's expected.
However, not all modulation schemes cause the same effect at the head. The
FM technique, with a clock always written "on schedule" (except for address
marks) and transitions written only for "ones" one can clearly see that FM
is an F/2F code, wherein modulation is at one of two rates, either the clock
rate, or twice that rate. MFM improves on that by encoding the clock into
its data by reversing the flux at more or less the same rate all the time,
except that it shifts phase, always positively, in order to avoid exceeding
the flux-reversal-density limitations of the head/media combination.
(That's the reason the first zero of each string of zeroes is omitted, and
that a zero between two ones is not written.)
For the reasons above it's not resonable to try to compare FM and MFM. The
data rate which has been used with 8" floppy drives has always been (AFAIK)
250 Kbps. The 125 Kbps rate was used with 5.25" floppies, but they came on
after the adoption of MFM as the "smart" modulation scheme, so they were
normally (except with RADIO SHACK computers) used with MFM. The way in
which this modulation technique was applied was not the smartest way in
which it could be done, but it did work well, fell well within the limits of
the the-available technology, and provided a substantial improvement over
what was previously done. The high cost of MFM hardware was what motivated
WOZ, at APPLE, to figure out a way to process the data himself into a scheme
which used both cheaper hardware and gave, effectively, density equivalent
to "double" density, thereby defeating the critics.
Dick
-----Original Message-----
From: Tony Duell <ard_at_p850ug1.demon.co.uk>
To: Discussion re-collecting of classic computers
<classiccmp_at_u.washington.edu>
Date: Friday, July 30, 1999 4:42 PM
Subject: Re: Cromemco 4FDC, How do you format a disk?
>>
>> <
>> <But from my measurements, I think that it's very unlikely that a drive
>> <can tell FM from MFM. I will try it sometime to confirm this.
>>
>> True, but the timing is still important.
>>
>> The SPEC I have for a Sony MP-F17W-70D (11/1988) makes it pretty clear
>> that the acceptable rates are 250/500kBITS/Sec MFM with no discussion
>> of plain FM. So it's probable the 8" SD rate of 250kbits/sec would
>> work. The 5.25" floppy single density rate of 125kBITS/sec clearly would
>> fall below the spec. There are limits to the write pulse width as well
>
>You are still missing the point, totally...
>
>125kbps FM is _NOT_ 125kpulses/sec at the disk interface, as you seem to
>be implying.
>
>Let's consider what's actually written on the disk, using standard 5.25"
>data rates.
>
>Single Density, FM, 125 kbps.
>-----------------------------
>A 'bit cell' is 8us long, so 125000 of them per second. Each bit cell
>starts with a (clock) pulse. There will be a (data) pulse in the middle
>of the bit cell if the bit is a '1'.
>
>So, allowable pulse separation times are 4us (between the 2 pulses in the
>same bit cell, and between the data pulse in one bit cell and the clock
>pulse at the start of the next one) and 8us (between clock pulses if '0's
>are written to the disk).
>
>Double Density, MFM, 250kbps.
>-----------------------------
>A bit cell is 4us long, so 250000 of them per second. There is a (data)
>pulse in the middle of a bit cell written as a '1', no (data) pulse in
>the middle of a '0'. There is a (clock) pulse written at the start of a
>bit cell if there is no (data) pulse in both this bit cell and the
>preceeding one.
>
>OK, consider some data patterns. Here a ';' singals the start of a new
>bit cell, a ',' separates parts of the same bit cell :
>
>11 -> data pulse, 1/2 bit cell gap; 1/2 bit cell gap, data pulse. Space
> between pulses = 4us
>
>101 -> data pulse, 1/2 bit cell gap; 1 bit cell gap ; 1/2 bit cell gap,
>data pulse. Space between pulses = 8us
>
>1001 -> data pulse, 1/2 bit cell gap; 1 bit cell gap ; clock pulse, 1 bit
>cell gap; 1/2 bit cell gap, data pulse. Space between pulses = 6us
>
>So,
>
>For FM, you need to be able to record/reproduce pulses with separations
>of 4us and 8us.
>
>For MFM you need to be able to record/reproduce pulses with separations
>of 4us, 6us, 8us.
>
>Since the drive can't tell data pulses from clock pulses, there's no way
>it can distinguish continual '1's at FM from the same at MFM, or
>continual 0's at FM from 1010... at MFM. Therefore, if it can do MFM
>correctly, it can also do FM _at half the user data rate_.
>
>> Keep in mond most all of the floppies I know of do have a minimum due
>> to the read amps bandwidth (there is a banpass filter!) and the
>
>Oh, absolutely. But the FM at half the data rate meets that spec. It has
>to. I am not suggesting you could feed 300 baud data into a disk drive
>and expect it to work - it almost certainly won't.
>
>-tony
>
Received on Fri Jul 30 1999 - 18:41:56 BST