Edited by Herb Johnson, (c) 2005; last changes Feb 20 2005
http://retrotechnology.com/herbs_stuff/8inch_3.5inch.html
Old computer newsgroups often discuss converting old CP/M systems from 8-inch and 5.25-inch floppy drives, to 3.5" drives and media. Issues raised included support of FM or single density; finding old media; choices of formats. I decided these discusssions were worth preserving, as these issues come up over and over. For more technical information on floppy drives and floppy disks, go to our technical page on floppy drives and media. For specific discussions about converting from 8-inch, 5.25 inch, and 3.5 inch, look for links on this section of that page.
Jay Maynard posted in comp.os.cpm on Jan 22 2005:
"Okkay...so if I wanted to hook up a 3-1/2 inch drive to my S-100 system instead of the SA-800s that are on it now, would the WD 1771 talk to them and work? Yes, IBM 3740 format would be Just Weird on a 3-1/2 inch floppy, and I don't expect to read the data with a normal PC without special software, but this would represent one way to get data from the 8-inch system into, say, a P112 (with a little bit of progamming)."
This led to a discussion thread about creating standard formats for the use of newer drives on older systems. It's an old topic, but along the way there was informative discussion about the technical issues of single, double and high density media and drives; floppy controllers; and formatting. It also led Randy McLaughlin to propose some standard formats and to discuss tradeoffs and expectations about reliability and compatibility with several other comp.os.cpm regulars. Randy also moved the discussion to the "classic computer" maillists at cccomp.org.
I've compiled, sorted and edited the discussion for ease of reading and future reference. The problem with discussion threads is that they rarely lead to either results or a useful permanent reference; and quick discussion can gloss over problems. I hope my editing and archiving efforts support both study and results.
My editorial changes and additions are either in []'s for edits and add-ins; or in italics (except the above paragraph) to name authors. While I retain copyright to this document, the individuals named have copyright to their posted as quoted, and have given me permission to quote them. Changes, corrections, and more info is appreciated. But, you may want to post in the appropriate maillist or newgroup any responses in the original thread. This is an archive and reference, not a replacement for those discussions.
See also the corresponding edited discussion on the same subject, from the "cctech" and "cctalk" maillists of classiccmp.com, on my site in the file named: 8inch_cctech.txt
Randy McLaughlin is continuing this discussion, and starting an effort to create some standards, on a "disk drives" page on his Web site. (Note: As of mid-2006 Randy does not appear to be active.)
Herb Johnson
http://retrotechnology.com/herbs_stuff/s100.html
Randy McLaughlin comp.os.cpm Jan 27 2005: I also believe in 8" disks as a great media. My point is and always has been that keeping 8" drives working is a dead end road. Sooner or later something fails that can not be repaired or replaced. I take two roads to preserve my CP/M systems: Buy spare drives when I can find them for the right price. Use 3.5" drives. I find that 3.5" drives have several benefits: Easier power requirements, cheap, plentiful, easy to transfer data to PC's (I have a PC that I have an 8" drive on but 3.5" is easier than dragging out another PC), takes less space. As I said I like 8" drives and still have several, some are not functional but I keep them for parts or to repair later. The real point is 3.5" works and we must bring our systems into the 21st century. > Herb johnson posted: > However, I'll say up front that writing a unique floppy disk format is > a TERRIBLE idea. Why create a disk that only one system in the world > can read? A search of this newsgroup's archive will inform you of the > troubles owners have, just with disks that only one BRAND OR MODEL of > system can read! Jay Maynard replies: [using a 3.5" drive on my 8-inch based system] ... is intended as a temporary expedient until my P112 arrives, and as a supplement to my 8-inch drives on my S-100 system. I'm having great difficulty with keeping them running, as they seem to be hrowing CRC errors more often than not on the same disks that, other times, read correctly. I've also got one drive that seems to insist on dropping ready at the most inconvenient times. Yes, unique formats suck. They suck less than not running at all.
Randy McLaughlin, comp.os.cpm Jan 27 2005:
As Herb stated we must not all go in a thousand directions.
We should come up with a standard.
I sent an email to Howard Harte recomending changing the disk format for
the [IMSAI] SuperIO. The SuperIO boots from Flash ROM and as such
needs no reserved tracks, I recomended reserving two system tracks to be
compatable with systems that boot from floppy.
There are three physical formatting standards I recommend:
32 128 byte FM sectors for 3.5" HD media.
9 512 byte MFM sectors for 3.5" DD media.
18 512 byte MFM sectors for 3.5" HD media.
These are simple standards easily used by CP/M systems.
I personally have disks going back to the 70's and have had a lot more than
2% failure to read (most are readable but maybe 20% are not).
Harold Bower replied, 1/27/05:
I would recommend a different set of DD formats. I would suggest the
default formats used for the Ampro Little Board, SB-180, SB-180FX,
Yasbec and ON! The same set of formats were used on all for 3.5" DD
drives (same as 80-track 5.25) as 5-1025 byte sectors/track (sectors
numbered 1-5) for single-sided and (sectors numbered 16-20) for
double-sided with skew recorded during format (normally 2) and
sequential readback. for 40-track 5.25" drives, 10-512 byte
sectors/track (sectors numbered 1-10) single-sided and (sectors numbered
16-25) double-sided with the same convention for skew. Two tracks were
reserved for boot tracks in all formats.
I included this format series in mods I did to the P112 as well and
included a format that has been very robust on SB180, SB180FX, YASBEC
and P112 for 3.5" HD drives of 11-1024 byte sectors per track for a
total capacity of 1.76 MB per diskette.
"Randy McLaughlin" wrote, and Hal responded, on Jan 28th & 29th:
>> Randy: I agree that the disks can be formatted to maximize space. The
>> problems, as I see it are:
>>
>> Many systems such as Tarbell do not already have 1K support.
Hal: Good point, but how many Tarbell systems are still operational vs SB180,
SB180FX, Ampro Little Board, On!?
>>I would guess there are many more Tarbell systems running than the total of
>>the above mentioned systems. Tarbell disk controllers were and are
>>extremely popular.
>> On the systems Like Tarbell 1K support can be easily added but on
>> stock CP/M v2.2 systems it reduces the TPA by more than 512 bytes.
Only if the system does not now support 1024-byte sectors. From Cam
Cotrill's any my work on ZSDOS back in 1987/8 and B/P Bios 90-92 work,
we found very few that did not. In any event, would you be willing to
go from, say a 58k CP/M configuration to a 57.5k system to gain 80k more
storage on an 80-track DD disk?
>>Here we are talking about extremely small numbers either way. Any system
>>that supports 1K sectors have more than 512 bytes less TPA than systems
>>which support only 512 byte sectors. The 80K takes into account the entire
>>disk, it is a little less when you take out the system tracks.
>>
>>I personally have no special preference for sector size, 1K is fine with me.
>>The SuperIO is a product of Howard Harte and before making changes in his
>>CP/M implementation I will get his input (I have sent him a private email
>>asking for his opinion).
>> By using 512 byte sectors you only "need" to fill the directory area
>> with 0E5h's to convert a PC disk to a CP/M system.
That is a valid point, but probably not as critical on Linux systems as
DOS/Windows where the OS seems locked heart-'n-soul into 512-byte
sectors.
>> Disk formats can be optimized for each environment but it misses the
>> point of standardizing. All of the newest CP/M implementations such
>> as updates to CP/M-86 & the Imsai series two have been following the
>> PC lead and have been using 9/18 512 byte sectors.
What standard? The only "Standard" is the IBM 3740 which is 8", 77
track single sided, 26-128 byte sectors per track FM. There was a later
modification (forget the number ATT) that added Double-density and
double-sided, but again only for 8" disks. Where is Joe Wright when we
need him? He wrote the BIOS for the Davage DiskMaker II which was the
cadillac of conversion processors (I had one, and hacked the Bios after
talking with him).
If we consider market share as a "defacto standard" then it makes more
sense to look at the operational systems out there, and again, I venture
to suggest that for bootable disks, the Ampro formats would prevail with
emulation being used (either a la the Ampro E-Disk or with the emulation
table in B/P Bios) to use the PC-style 3.5" disks as an exchange medium.
I adapted B/P Bios to the SuperIO and still have it up on the OLD
version of the board, but have not heard anything from Howard since
sending him the image, nor have I seen it available for D/L (so I guess
I'll add it to my page when I update it). In that system, I boot from
the CP/M3 disk, then switch directly to the Ampro-formatted diskettes
from which warm boots and all work can then proceed, including the IBM
format emulation.
>> It is obvious that more discussions such as this need to be expanded.
>> I'd be happy to change the SuperIO BIOS to 1K sector sizes if that is
>> the direction people want to go, it is more disk efficient.
If you were to adopt the Ampro-"standard" formats, the 2-track boot
sector reservation is already there, the skew factor of 2 (normal, but
changeable) is there, and you would have complete interchangeability
with a great many other systems, and an extra 80k bytes per 3.5" DD
disk, and 40k per 40-track DS or 80-track SS disk. The formats have all
proven remarkably robust and have plenty of tolerance slop in them to
accomodate loosey-goosey hardware.
>>Randy:Also referring to a previous post why the odd sector numbering on the second
>>side? I would always prefer to start sector numbering with 1.
>>
>>In truth I was hoping this discussion would have more people involved to get
>>a true consensus. We have an opportunity to direct the transfer to a common
>>format.
Hal: The "odd" sector numbering is not just for the second side, but for the
entire disk. The algorithm used in the Ampro Little Board, SB180,
SB180FX, On! etc was something like:
1. Home heads
2. step in two steps
3. Read Sector ID from side 0 (first side .. bottom)
a. Track number was 2, drive was 80-track
if Track number was 1, drive was 40-track, so set flag to
"double-step"
(allows 40 track disk to be used in 80 track drive .. carefully!)
b. If sector number was < 16, then drive was single-sided (1..5,
1..10, 0..10, etc)
else assume drive is double-sided
3.5" drives (DD) were treated the same way as 5.25" (DD) units so this
simple algorithm allowed a wide range of complements to be handled
simply and efficiently. Special handling was often included to allow
Kaypro disks (512-byte sectors numbered 0 .. 9) to be detected along
with the Ampro 1 .. 10 numbers in 40-track drives by doing an explickt
read for Sector 0 if step 3 indicated a 40-track drive.
Others expanded this in various ways. For example, the author of XBios
for the SB180 started sector numbering 80-track double-sided formats
without boot tracks (intended for data-only disks) at 27, neatly
avoiding the highest number normally used in that day of 26 in 8" and
down. In B/P Bios, Cam and I chose an offset of 30H (start at 49
decimal) for 5.25" HD formats and 40H (start at 65) for 80-track HD
formats (where the Bios automatically switches density while logging
disks in) while Tilmann Reh starts HD formats at 1 in 3.5" systems. If
you are going to allow mixed drives in systems without locking users
into specific configurations, then implementing some algorithm and
implementing non-conflicting formats might be a bit more usable.
I too would like to see more people involved in this, but I doubt if
there will be a single "common format". Anyone???
Randy McLaughlin, Jan 22: To replace an 8" drive with a 5.25" HD drive is a snap, to hook up a 3.5" drive requires modifying the BIOS. The data transfer rate are compatible but the rotational speed of the 3.5" drive is slower allowing more data to be written. There are lots of technical differences in the different types of drives. Herb Johnson has been good enough to collect and post some info: [note from Herb: I've changed the link referenced below:] http://retrotechnology.com/herbs_stuff/s_drives.html The modifications to the BIOS/format program are not too great, mainly increase the tracks to 80 use both sides and change the SPT (sectors per track). I've never done it with a 1771 so I am not sure of how many 128 byte FM sectors the average 3.5" drives can handle. To use 5.25" drives is much easier and does not require any software changes. Luckily the BIOS and format program for the Tarbell controller is available and easily modified. After checking with [TEAC FD-235HF 3.5" drive docs at:] http://www.beg-buerkle.de/support/Hersteller/Floppy/FD235HF/FD235HF-A429.pdf I found the 3.5" drives can be formatted FM (1771) as follows: 128 bytes per sector 32 sectors per track 655K per disk. 256 bytes per sector 18 sectors per track 737K per disk. 512 bytes per sector 10 sectors per track 819K per disk. Unless the BIOS is already handling larger sector sizes stay with 128 byte sectors. Additional, from Herb, unposted: [The TEAC document specifically says that an FM data stream at 250Kbits/second on an 80-track HD diskette has a low bit density (8717 bpi) but the same FLUX density (17,434) as MFM - as Randy posted. The document has the format schemes Randy mentions above, for "2MB" HD diskette format. For DD media at "1MB" format, with a flux density of 8,717, the document specifies FM use at 125Kbits/second, about 400Kbyte capacity; and MFM use at 250Kbits/second for about 800Kbytes capacity. Disk rotation speed is specified at 300 RPM. Write precompensation, a kind of hardware adjustment for reading and writing data frequency effects, is listed in the document as +/-125ns for "2MB"; "0 +/-125ns" for "1MB".] CF Falconer replied: > Why not cut the tracks to 26 sectors with wider gaps, and now you > have a track by track clone of SSSD and DSSD 8" drives, with > 256/512 k capacities and 4 extra tracks per side. CP/M can easily > read those extra tracks, but writing them requires a driver. So > they can best be used for mounting r/o files. Randy: No matter what, you need to modify the software even if it is only the format program. [Otherwise,] the tracks can be formatted to 32 sectors and the BIOS will ignore the extras. The format I described works even if only the first 26 sectors are used by the BIOS. With only the slightest change in the BIOS, you can handle near 650K of data (leaving room for the boot tracks). It would be best to test sector interleave and skewing as well as step rates. The format program will usually fail because the 3.5" drive rotates at 300 RPM. I happen to know that [the controller in question] is a Tarbell controller and the source code to everything [in it] is available. [Herb: responses to Randy follow in the middle of the "format" discussion below.]