🚀 go-pugleaf

To: GTHEALL@PENNDRLS.UPENN.EDU
Subject: Re: Rainbow EchoMail Digest
Newsgroups: comp.sys.dec.micro
In-Reply-To: <9103230131.AA04966@remote.dccs.upenn.edu>
Organization: Columbia University
Cc: 
Bcc: 
From: Charles Lasner (floppy maven)

A little misinformation to dispell:

    Original non-double (regular?) density disks are (let's use Maxell
numbering system) are MD1 or MD2 if double-sided.  90K on a side (K=1024,
*not* 1000) achieved at most by 9 sectors x 256 bytes/sector x 40 tracks
(48 tpi for 5/6"). Recording speed is 250 KHz, but double-frequency
recording method is used, not MFM as in all later stuff.  Each clock buys
you only .5 bits, since every other clock is a sync pulse, not data.
Used in early machines like Osbornes.  The absolute original 5.25" floppy
known to be scoffed at by 8" people who had bigger disk capacity at the
time because they were double-frequency, but at 500 KHz (albeit at 20%
faster rotation speed, so they were only 1.6 times as dense, not double).
Possibly was only 80K on a side due to belief that drive speed tolerance
couldn't reliably use more than 8 sectors/track with wide guard zones
between sectors.  Larger amounts could be had by using less/bigger
sectors.

    Double-density disks were championed by IBM because they were
"double" the Osborne-type pathetic creatures.  They had 160K per side
because they were MFM, thus each bit has data.  The information is when
precisely does the clock happen: early or late relative to a mythical
middle time point that never happens.  Each bit decision extrapolates the
time for the next bit to arrive late or early against.  Cannot easily
distinguish long streams of ones from long streams of zeroes, but is
otherwise reliable, and is industry standard for floppies.  These need
Maxell MD1D or MD2D media.  Eventually, IBM admitted that they were too
conservative, even with the junky drives they sold on PCs, so they upped
it to 180K/side with 9 sectors of 512 bytes each.  Future programs would
come out which up this further to 10 sectors/track, although without
IBM's blessing.  We still have 40 tracks and 250 KHZ, so the head gap is
relatively large.  All of these low-density floppies require the
high-saturation oxide that goes with 250 KHz territory.  Some
manufacturers make this mix reddish-brown, with the high-density (500 KHz
stuff) darker, almost black.  Actually, there are no hard and fast rules,
so be wary of color comparisons.  For all practical purposes, the MD1 and
MD1D, and MD2 and MD2D are the same media.  Some PC oldtimers perhaps
have early Maxell boxes marked MD1 without the trailing D.  The drives
ability to not jitter the MFM was the real difference, not the media or
its official "certification" as acceptable for ssdd or dsdd usage.
Because of the various IBM quirks, disks exist calling themselves 160K,
180K, 320K, and 360K.  Various formatting programs exist to extend this
further.  Using 10 sectors/track and 41 tracks (yes, most drives have the
extra track; 80 track-type drives mostly have 82 tracks!) and two sides
gets you to 410K.

    With the double-density label attached to IBM's stuff, someone dreamt
up the name "quad-density" for the disks used by the DEC RX50 machines
and the Victor 9000.  These disks are identical to the MD1D and MD2D with
one important difference:  they originally were honed smoother for use
with somewhat thinner heads (not to be confused with even thinner heads
later known as thin-film heads).  These drives get another doubling over
the double-density because there are now 80 tracks (96 TPI), not 40
tracks (48 TPI), but the frequency, etc., is *not* changed.  Many of the
early 96 TPI drives had head registration problems - the media could not
be guaranteed to seat reliably in the same position when remounting the
disk.  Actually, the amount of variation isn't much over the former
double-density, rather the tolerance is about 1/10 of what you used to be
able to get away with.  On some drives, the hub ring protectors actually
backfire and cause the registration problem.  The solution is to
eliminate the hub rings.  In theory the media has a lower saturation
formula, but only slightly, so hub-ringless versions of the MD1D and MD2D
are available as MD1DD and MD2DD.  For preformatted DEC media, get
MD1DD-RX50.  This actually outperforms DEC's RX50 brand media!  Today's
better floppy manufacturing means that all floppies are well-honed, so
it's dubious what the difference is between MD1D and MD1DD or MD2D and
MD2DD anymore.  Formulations which have wider saturation ranges are now
used, so today's cheap floppies are better than older "premium" grades.
I imagine that Maxell and the other "name-brand" floppy people might
insist there is a difference, but I think it's more of a minor
"fine-tuning" of the oxide mix, not a major formulation difference, since
there is now a lot of overlap in the requirements of the manufacturing.

    At this point in the story enters the 3.5" disk as used in PCs with
720K disks.  These are in theory the same as 360K disks, just that the
stiffer media and thinner heads are now in a newer, smaller package.
That 5/6 of an inch is getting 80 (up to 82 actually!) tracks on it
"officially" now.  Notice that RX50 format always used 10 sector disks,
so DEC gets 400K/side, while IBM only gets 360K/side (but has two
sides!).  There are formatting programs to put 10 sectors/side, and use
82 tracks, so you get 410K/side or 820K/disk.  Since this is still 250
KHz territory, Maxell calls this stuff MF1DD or MF2DD for two sides.
Note also that virtually all disks are made double-sided today, so they
probably have dropped the MF1DD entirely.  Notice that there is totally
consistency here: the name isn't MF2D, it is MF2DD; a quad density
notation because there is 80 tracks at 96 TPI.

    When the AT came out, IBM used the miniaturized 8" design that was
started as far back as 1982 by Mitsubishi and others.  They would have
you believe they invented high-density disks, but that's pure bull.  I
used the Mitsubishi drive on the PDP-8 two whole years before IBM decided
to replace the XT.  These drives run at 360 RPM and 500 KHz clocks, just
like the 8" stuff; there is even an adaptor cable to work in the 50-pin
environment with real 8" drives if you insist.  My OMTI 5400 controller
for hard disks and high-density floppies even calls itself an 8" floppy
controller.  These drives have thin-film heads, and all are 96 TPI.  They
get more capacity than the 8" drives' 77 tracks; all claim 80 and most do
82.  I have some early Maxell MD2HD media with pure white labels
imprinted "Maxell MD2HD" on them and nothing else.  All of the other
Maxell disks had the (then) familiar red labels.  This formulation takes
into account the shorter head gap, and attendent lower saturation levels.
Attempts to use this media on the low-density drives results in
unreliable operation because the media overloads and distorts the bits'
flux changes.  Since these drives rotate 20% faster, the capacity is only
160% of double-density, not 200% as otherwise would be predicted.
High-density disks always come in double-sided versions, so there is no
media MD1HD.  Some versions of the drives, such as TEAC FD55GFR are
capable of being run at 300 RPM for retro compatibility in some systems.
ATs use a third clock speed of 300 KHz which is 20% faster to match the
20% rotation speed increase when reading double-density floppies.  Some
add-on cards such as the Compaticard series will assert the speed change
bit instead, so a TEAC FD55GFR or FD55GFV or equivalent is mandatory.
There are formatting programs that can up the capacity of this disk to as
much as 1.48 MBytes, but there are trade-offs:  You must use a 2-1
interleave, so sloppy programs might actually run faster, but good ones
will definitely run slower.  The number of sectors, normally 15, is upped
to 18, because you can now use the "in-between" sectors as the "cool-off"
guard zone instead of using "dead space" for this purpose.  This refers
to the relatively lengthy time zone where the head read amplifier is
still unable to read the media to find the next header after recently
writing the end of the last sector's data.  There is no writing going on
anymore, but the read amp is overloaded because the write current was
saturating the input circuit.  (Remember, only one head to read *or*
write, so they are in common.)  To allow normal 1-1 interleave, you must
leave dead zones past the end of the written data, so the head amp can
recover and then start reading the next sector's header record (which is
not written on) before re-enabling writing of the new sector's data area.
This allows 5.25" disks to be capacity-identical with 3.5" disks' 1.44
MBytes, and even more if you use the 82 track variation most drives
allow (1.48 MBytes).

    The final popular format is the 3.5" high-density version.  This is
designed to be 500 KHz *and* 300 RPM, so performance is truly exactly
four times double-density.  There are 18 sectors (double the nine) and 80
tracks (double the 40).  Unlike the 5.25" high-density drives, these
extra-thin film heads can run at 300 RPM and still achieve good
performance at 500 KHz recording rate.  Formatting programs exist to take
advantage of the 82 tracks and 2-1 interleave, which puts 21 sectors on
each track for a total of 1.72 MBytes.  (Same trade-offs apply as above.)
Apparently, formulation for these inherently thinner heads is so much
like the MF2DD version, that MF2HD disks are identical to their DD
counterparts, except there is a hardware problem: an extra notch hole is
needed to allow the high-density identification to most drives.  There
are third-party hole-punchers on the market to get around this.  Be wary
of off-brand media though, they may be still making their diskettes only
to MF2DD spec, not the wider-range of saturation formulation of the
current crop of name-brand disks.  There are also reports of
contamination from errant pieces of plastic winding up strewn about the
media after punching out the hole.  There are also some drive
modifications to avoid media mutilation.

    So, a DEC-oriented summary:

    Do use any double-density media made today; there ain't much
difference.  If the media gets around, don't use hub rings; they'll "get"
you eventually.  There are some vendors, like Fuji, that use thicker hub
areas, not rings, so these are recommended.  Apparently, exclusive use on
RX50s doesn't cause much registration drift using name-brand
double-density media (D, not DD), but beware mixed usage with other
people's machines.  Remember, RX50 disks wind up on AT-clones using
RX50DRVR.  The AT-clone may have a hub-eater in it!  To play it safe, pay
more for MD1DD or even MD1DD-RX50 if you are format-lazy (or only have a
VAX, DECmate, or PRO).  MD1DD can be formatted fine on an AT for RX50
usage.  Just specify 10 sectors, 80 tracks, and 1 side.  Unfortunately,
RX50s really have only 80 tracks, no extras. :-(  The resulting disks are
totally legal for DOS usage on the PC, but not for any other purpose, so
don't forget to use prevailing directory initializers on the DEC
machines; this is just the "low-level" physical formatting that cannot be
done on any DEC RX50 machine except for the Rainbow.

    Do *not* use HD media.  It will be more expensive, and it won't work
reliably because it is designed for more "delicate" heads.

A side note:

    All formats can gain transfer efficiency by including a stagger
factor in the formatting process like the software mapping of RT-11, etc.
The idea is that it takes a number of milliseconds to get to the next
track when you seek, so you won't see the first sector on the next track
after transferring the last sector on this track, rather some higher
number depending on track-track seek speed.  RT-11 uses a software
mapping scheme, but PCs, etc., just do it straight linear.  So, the
formatter can compensate by staggering (aka sliding) up the order of the
sectors as they appear on the disk.  A good number for RX50 format is 2.
This means that track zero is formatted 1,2,3,4,5,6,7,8,9,10, and track
one is formatted 3,4,5,6,7,8,9,10,1,2 and so forth.  These are all stated
relative to the index mark.

cjl (lasner@watsun.cc.columbia.edu home of KERMIT-12 and other fine KERMITs)

In article <1991Mar23.080919.25327@cunixf.cc.columbia.edu>, lasner@shibuya.cc.columbia.edu (Charles Lasner) writes:
>     With the double-density label attached to IBM's stuff, someone dreamt
> up the name "quad-density" for the disks used by the DEC RX50 machines
> and the Victor 9000.

I first saw this notation with the Superbrain, predating both the RX50 and
the Victor 9000 by a number of years.

>     All formats can gain transfer efficiency by including a stagger
> factor in the formatting process like the software mapping of RT-11, etc.
> ...  RT-11 uses a software
> mapping scheme...

Sure wish they'd used a hardware mapping scheme. All floppy controllers
these days can be told which order to lay the sectors out in, so there's
really no excuse for doing it with software. The problem with software is
that if you want to read the disk, you have to know the software mapping
scheme. But if they had just scrambled the order of the sectors on the 
disk you wouldn't have to. Sigh.
-- 
===============================================================================
Roger Ivie

35 S 300 W
Logan, Ut.  84321
(801) 752-8633
===============================================================================