15 Nov
2007
15 Nov
'07
10:49 p.m.
Subject: Re: Tarbell is making me insane
From: ard at p850ug1.demon.co.uk (Tony Duell)
Date: Thu, 15 Nov 2007 20:50:34 +0000 (GMT)
To: cctalk at classiccmp.org
As I understand it, FM at 125bps has (only) pulse spacings that can occur
in MFM at 250kbps. The latter has some other spacing too, of course, but
the point is that unless you're trying to be awkward, it's impossible to
make a drive that will correctly handle any MFM data at 250kbps that will
not also correctly handle FM data at 125kbps. (About the only way you
could do it is if the drive tries to decode the data back to user bytes
and then re-encodes them onto the disk in a totally different way. I know
of know floppy drive that ever tried something like that!).
There should be no problem at all in using a 3.5" DD floppy drive with
the FM data stream that uou'd have sent to an SA400.
Doesn't completely work that way. If it were bandwidth only maybe.
The problem is it's timing and that causes peak shifts that increase
with jamming those bits in a more confined space (flux tranistions
per inch) and that makes 3.5" track 000 an approximation of SA400L
track 41ish. That doesnt' factor the lower S/S+N ratio of 135 tpi
against 48TPI. Will it work, I may but if it doesn't work well
it may not be broken. Reason is the DD FDCs (both 765 and 1793/2793)
have additional hardware to add a timing shift based on pattern written
and better feature extraction for read. This is not present on 1771
based boards.
Allison
This
faster signal might be better for the disk drive?
Potentially. Remember FM floppy recording was passe' by the early
80s with every one needed far greater space. It want even a
consideration buy time PC 3.5" floppies hit the street. -tony
16 Nov
16 Nov
4:51 p.m.
On 15 Nov 2007 at 17:49, Allison wrote:
Reason is the DD FDCs (both 765 and 1793/2793)
have additional hardware to add a timing shift based on pattern written
and better feature extraction for read.
To recapitulate what I posted a couple of days ago...
IIRC, neither the 765 nor the 179x/279x employs write precomp when
writing FM, only MFM. (Just checked the 279x app notes to make
sure). Write compensating FM recording based on an MFM algorithm
makes no sense and probably is counter-productive.
"Bit shift" is a bit of a misnomer. What happens with oxide-coated
media is that flux transitions occurring close together tend to push
the edges of the domain apart. That is, the previous domain boundary
is actually pushed slightly back in time as the current one is being
written.
Write precompensation circuits examine the data stream as it's being
written and make decisions based on the last bit written, the current
bit being written and the next bit to be written. It sounds
complicated, but it's nothing more than a shift register and some
gates.
This "bit shift" is peculiar to coated media--it doesn't occur to
nearly the same extent with plated media. The problem is especially
acute with MFM as compared to FM, because while the roughly the same
number of flux transitions is used in MFM to record twice the data as
FM, the position of the flux transitions is more critical. Just
another case of TAANSTAFL.
The biggest problem with the Tarbell controller is that it uses a
simple counter comprised of a couple of 74LS161s and a few bits of
"glue"to do its job. While this is fine for conservatively-recorded
media such as 8" diskettes--and requires no adjustment by the user--
it falls on its face when smaller media with more demanding needs are
employed. An analogue phase-locked-loop is far and away the better
solution and can accommodate all sorts of timing irregularities, but
it requires adjustment for proper operation. You pays your money and
takes your choice.
The WD9216 is a counter-type separator with a lock ratio of about +/-
30 percent of the nominal center frequency, which is quite a bit
better than the very simple Tarbell circuit.
One could probably program a 20MHz 8-pin PIC microcontroller nowadays
and get very good data separation for less money than either a not-
easy-to-find 9216 or an analogue PLL.
Cheers,
Chuck
6261
days inactive
6262
days old
1 comments
2 participants
participants (2)
-
ajp166＠bellatlantic.net -
cclist＠sydex.com