Re: Did Commodore cheat with the quad density floppies?

From: afachat_at_gmx.de
Date: Fri, 04 Jan 2019 22:32:24 +0100
Message-ID: <19338081.tYgx3adm6I@euler>
On Freitag, 4. Januar 2019 22:08:33 CET Francesco Messineo wrote:
> On Fri, Jan 4, 2019 at 8:19 PM <afachat@gmx.de> wrote:
> > Thanks for pointing that out, that is my understanding as well. And thus
> > my
> > question, why the drives still work with 5900bpi media....
> > 
> > Probably, as mentioned in another post here, quality of the media improved
> > over time so more bpi were possible.
> > 
> > > Again, I suggest reading the fine informations contained here:
> > > http://www.retrotechnology.com/herbs_stuff/drive.html
> > 
> > This does unfortunately not answer my question...
> 
> well well.... difficult questions need studying, I don't think in this
> list there's any of the original 8x50 disk drives designers, otherwise
> that was the quickest method to answer your question.

Yes, that would be the preferred solution. If anyone knows someone....?

> In that link, one can read a lot of things, and reason over the data
> that don't directly answer your question, but probably can give some
> clue to reach an educated guess.

Yes, it is a lot of great stuff.

> I will point out a few facts (or at least, they seem hard enough facts to
> me). 300 oersted media was used on most 8-inch floppy disks and on all the
> non-HD 5 1/4-inch floppy disks (SD/DD/QD 48 and 96tpi certified
> media).
> BPI on 300 oersted media varies between 8 and 5 1/4 inch drives, also
> it's *always* quoted either for FM or MFM modulations, never for any
> other, like GCR. It seems to me that it must depend
> on the modulation/bit encoding and it's not a magnetic-media intrinsic
> limit. 

The BPI to me now seems like a pure marketing instrument. I have added a table 
in https://extrapages.de/archives/20190102-Floppy-notes.html just now where I 
compared the various encoding formats (FM, MFM, GCR Commodore/Apple), and 
_all_ of them used a maximum of about 5900 flux transitions per inch on the 
disk. No difference.

Only if you factor in the length of an encoded byte, you get something like 
"bit per inch" values. But they have nothing to do with the media quality, 
only with the encoding method.

> 8 inch drives have another interesting characteristic: they need to
> reduce the write current when writing on the upper half (tracks 43-77) of
> the media. 8-inch floppies are huge, linear velocity of the actual media
> under the R/W head varies a lot between track 0 and track 77, so (I'm
> assuming) using the same current when the linear velocity of the media gets
> too slow (on inner tracks 43-77) produced probably too
> large flux areas and needed to be reduced in order to maintain small
> enough magnetic areas. If someone has an official explanation why the
> 8-inch drives needed to reduce the write current, I'd appreciate to
> know it.

That is indeed my understanding, but it's also only an educated guess. The 
8050 drives actually did use write pre-compensation, in that they shifted the 
transitions slightly depending on whether it was the first or last bit in a 
row of two or more consecutive 1-bits (I can send the diagram with schematics 
and timing I did if you want).

The 8050 actually does something completely weird on the read timing, that I 
am unable to understand. If you look at the schematics here http://
www.zimmers.net/anonftp/pub/cbm/schematics/drives/old/8050/8050-27.gif in the 
bottom there is a lot of CMOS logic (4066 switches) and op-amps that in the 
end create the read clock... (and that does not exist on the equivalent page 
on the 4040, that also does not have the write-precompensation)
 
> The limit obviously seem to depend on head dimension/construction AND
> rotational speed, otherwise 8 inch drives (360 rpm, 48 tpi) would have
> the same BPI figures as 5 1/4 inch drives (300 rpm, 48 tpi) but
> they are different, and the limit is also dependent on the write
> current, albeit in a rather difficult to estimate (without official
> documents) way.

It also depends on at what radius the innermost track is, thus what 
circumference the track has (I'm sure it's on herb's page but to lazy to look 
up the difference between 8" and %.25" right now)

> Other facts: 96tpi MFM formats merely doubled the number of tracks per
> side, but sector count per track is the same, so MFM 40 tracks 48 TPI
> has 360Kbytes formatted capacity and MFM 80 tracks 96 TPI goes to
> 720Kbytes (double side of course). This seems to me a bad choice, but
> probably was forced because of a limitation perhaps of the standard
> FDD controllers.

The standard FDD controllers did not have zone bit recording indeed. That they 
did not introduce it when going to 80 tracks or even HD is a pity. I guess the 
problem was that DOS was unable to cope with different numbers of sectors on 
different tracks, so it would have been wasted anyway.

> CBM engineers seem to have had a different idea here, but I start
> guessing from now on: if the R/W head on a 100 tpi mechanism is
> narrower (and of course it is in the "track" dimension), maybe it's
> also thinner in the other dimension (or maybe they just verified it
> really is smaller in the other direction). So why not trying to use a
> higher flux transition frequency and see what happens with these 100
> tpi mechanics regarding the bits per track on different zones?
> The head difference clearly allowed much more higher flux transition
> rate, otherwise they had used higher frequencies also on the 4040/1541
> format. 4040 format also was slightly reduced as bytes per track on
> one zone (as we all know) respect the original 2040 format, and I
> guess it has been reduced probably because not all the 48 TPI drives
> that CBM was using were reliable enough on that particular bit density
> in that zone.
> I guess (and it seems reasonable) that CBM engineers used the highest
> amount of bytes per tracks in the various zones that allowed reliable
> (by a large margin I'd say) operations. I can still read most of the
> disks that were last written around 1989 or so.
> If we agree that they found the optimal sector per tracks in the
> various zone by probably experimenting (and the 2040/4040 format
> difference seem to suggest that), then I don't see why the same
> engineers would not try the same method using the 100 tpi mechanics:
> see what they can do, maybe tweak a bit the head write current and/or
> pulse shape and see what can be reliably done on the same 300 oersted
> magnetic media, just with better mechanics and smaller heads.

Yes, that is probably the case.

> I never had the luck to have a 100 tpi drive, so I don't know if the
> disks written around 198x are still readable nowadays, but if the
> failed disks have the same percentage of the ones written with 48 tpi
> mechanics, then probably the reliability of the write format was as
> solid as the one used on 48 tpi drives. If, on the other hand, 100
> tpi-written disks are much less readable (and by a large amount, I'd
> say) then we could conclude that CBM engineers went too far pushing
> the limits of the 300 oersted media and the 100 tpi mechanics
> combination.

I don't have many 100tpi disks, so I cannot test that. But maybe someone else 
on the list can. 

André
Received on 2019-01-04 23:00:30

Archive generated by hypermail 2.2.0.