AT&T 3B2 floppy format

[Seth Morabito]

I finally have my own AT&T 3B2/300, and I'm having a heck of a time
getting disk images transferred to physical media.

I have here a set of AT&T SVR3.2 diskette images, apparently made (not
by me) using dd. I would like to transfer them to physical media in
such a way that they're usable by the 3B2/300.

Here's what I know so far:

  * 3B2 diskettes are 720KB, Double Sided Quad Density (DSQD) 96tpi
  * Each side is 80 tracks, 9 sectors per track, 512KB per sector
  * Sectors use 3:1 interleave
  * Physical media should be good quality DSDD
  * The 3B2 fdc is a TMS2797 (WD 2797 compatible)
  * The 3B2 floppy drive is a CDC 9429

On my PC, I'm using a venerable TEAC FD55-GV with the "I" jumper in
place, so at double density it should be spinning at 300RPM.
ImageDisk claims that reading and writing at 300kbps is successful.

I have been using ImageDisk to translate the BIN files I've downloaded
into IMD files with the following commands:

D:\> BIN2IMD DISK1.BIN TMP.IMD /2 /U N=80 DM=4 SS=512 SM=1-9
D:\> IMDU TMP.IMD DISK1.IMD IL=3

(The two-step translation is necessary because BIN2IMD cannot directly
write 3:1 interleaved data unless it's interleaved in the BIN image,
so you have to use IMDU to reshuffle things... it's complicated!)

Anyway, after doing this, what I end up with is a disk that is
_almost_ usable. I can boot off of it, but it fails shortly after
loading the UNIX kernel. I can run the 3B2's "dgmon" floppy
diagnostics on it, and they almost pass, but fail to reliably read and
write during the R/W test.

Now, here's the thing: If I use the exact same media and low level
format it _on the 3B2 itself_, the disks are 100% readable on the 3B2
and pass all floppy diagnostics with flying colors.

So I'm trying to pin down what about my setup is not right.

My pet theory right now is that the R/W gap and Format gap are wrong.
The default values for the gaps when calculated by ImageDisk are
24/64. I've played with 22/32, 34/62, and 42/80, all based on reading
the datasheet and/or old Linux "fdprm" settings, but nothing seems to
make the disks 100% reliable on the 3B2 when written on the PC.

Does anyone have any insight into the gap lengths used by the 3B2? Or,
have you successfully written 3B2 floppies from disk image before?

-Seth
-- 
Seth Morabito