Wobbly Floppy Drive?

Over the weekend I tried to revive a PS/2 Model 80 whose CMOS battery died (after 25 years!). Finding a replacement battery fortunately isn’t difficult (6V lithium “photo” battery), but recreating the configuration information of course requires a reference floppy and the .ADF files for any non-IBM adapters installed in the system.

The early PS/2 implementation of Plug and Play leaves something to be desired. Keeping a reference diskette with a system for many years turns out to be difficult in practice. Newer PS/2 systems with a reference partition do not suffer from this problem but still need floppies if the hard disk dies.

Finding a Model 80 reference floppy image or the ADFs is not difficult, but the system just kept refusing to boot from my floppy. The diskette was readable in other systems and I tried several of them, both original IBM floppies and self-made copies. None of them worked. The drive was spinning, did a few seeks, and then the system showed an error.

I tried and tried many times, and just when I was giving up, the system booted up from the floppy! Of course I was missing an ADF or two, so I had to modify the floppy—it doesn’t help that the reference disk only says that configuration information for adapter in slot so-and-so is missing, and doesn’t indicate what device is in that slot. But the number of MCA adapters is limited and identifying the cards is not too hard.

Anyway, after updating the floppy, I had the same problem booting from it. This time I took the drive out and noticed that when it spins up, there is a noticeable wobble. I assume the drive was not spinning at a constant speed and that caused reads to fail.

After retrying many times, the drive somehow settled down, completely stopped wobbling, and started reading without any errors.

It’s not clear to me what the problem might be. At first I thought it was a mechanical issue but now I’m convinced it is not. More likely it’s a problem with the drive’s motor which causes it to run unevenly. I don’t know if it needs to warm up or something but the drive always eventually started working, and after power cycling the system it was wobbly and failing to read again.

Is this some sort of typical floppy drive problem? The PS/2 floppy drives unfortunately use non-standard connectors and cables and can’t be replaced with standard PC/AT style parts. So the drive would be worth repairing… if I knew how.

Update: Forgot to mention that the floppy drive does not report any errors during POST. That indicates there’s no problem with the stepper motor/head movement.

This entry was posted in Floppies, IBM, Plug and Play, PS/2. Bookmark the permalink.

30 Responses to Wobbly Floppy Drive?

  1. Chris M. says:

    The early PS/2 floppy drives suffer from leaking SMD capacitors on the control board. The drive slowly beginning to work after warming up screams capacitors.


    This thread reports some success with repairs, plus a link on how to modify a standard drive to work on a PS/2: http://www.vcfed.org/forum/showthread.php?50816-Again-a-faulty-PS-2-floppy-drive

  2. raijinzrael says:

    I would recommend you go to HxC forum and ask about a way to replace your floppy, preferably with a floppy hardware emulator. Even if comes the case you don’t want to replace it, and prefer a fix, these guys know a lot about floppies (sometimes even they can provide manuals, pinouts and schematics) which can be useful to trace and fix the problem.

  3. raijinzrael says:

    And yeah, their knowledge includes non-standard interfaces like your PS/2 drive.

  4. Michal Necasek says:

    Are there PS/2 floppy emulators? The interface isn’t quite the same. Also I’m not sure I’d want one for a historic piece of hardware like a Model 80 🙂

  5. Michal Necasek says:

    That’s useful stuff, thanks. My drive is an ALPS but I’ll check it for obvious problems (in a few weeks when I can get back to it).

    I should have added that the FDD does not fail POST so there’s no problem with a stuck head or suchlike.

  6. raijinzrael says:

    “After retrying many times, the drive somehow settled down, completely stopped wobbling, and started reading without any errors.”

    Chris M. would be right, this sounds like a failing/leaking caps case. When them start to fail, generally they start in “bad” condition, and once them heat up, them seem to start working normally. Ofc this can only be repeated a number of times, until them don’t want to work anymore, or what is worse, getting shorted so them end frying the rest of the components.

    A good recap of the drive would be in this case a good starting point to check.

  7. Bill S says:

    Sometimes I think the newer machines are harder to keep alive than the older ones. I have a Compaq Portable III with a 3.5″ standard PC floppy drive hacked in it because the original quarter height 5.25″ floppy drive is almost impossible to find. I have a PS/2 model 30 sitting around unused because I don’t want to spend the money on the goofy floppy drive they use. Don’t get me started on PS/2 hard drives… Model 50 with an adaptec scsi card because the original drive died and it naturally the card I got my hands on isn’t supported by OS/2 1.x.

  8. Michal Necasek says:

    I think it’s more standard vs. non-standard. PS/2 machines were far less widespread and it’s correspondingly harder/more expensive to find replacement parts. I was lucky that I got the Model 80 at a time when shipping was the biggest factor of the cost. And that it has standard 50-pin SCSI hard disks which are easy to replace.

  9. Chris M. says:

    PS/2 machines used to be dime a dozen around here as “nobody got fired for buying IBM”. Most of them were cast offs of the local IBOC and were further cycled into use at various public service agencies before being retired in the late 90s. They weren’t “rare”, but being proprietary (ironic considering its an IBM system!) doesn’t help any system in the long run as parts supplies dry up. On that note, I just sold off the last of my MCA “direct bus” (think IDE, but the Microchannel version) drives pulled out of various Model 50s and 70s.

    I got rid of the Model 80s I had, they outlived their usefulness as table legs 😉 They didn’t have SCSI though. Mine had giant 5.25″ ESDI drives, which was IBM’s usual standard configuration.

  10. Michal Necasek says:

    PS/2s certainly weren’t rare… but they were never sold in numbers anything like AT compatibles. The prices are definitely going up.

    I have one of those ESDI drives too. Sounds like a jet plane taking off when you power it on. 105MB I think, and every MB weighs about a pound 🙂

    One of these days I should find a Model 30 because it was such a weird hybrid half-AT, half-PS/2, in many ways close to latter-day AT compatibles.

  11. S Duggan says:

    Early PS/2 drives were notoriously prone to failure. Usually, the drive belt perished/stretched, which led to groaning and whirring and a failure to read any disk inserted. I have a P70 and a P55S both of which share the same unique ALPS drive, which lacks a dust cover thus encouraging dirt ingress, which also kills the drive over time. Mine are both dead.
    You can potentially replace drives with current standard models. You just need to rewire the cables as PS/2 drives carry the power through one of the cable wires rather than a separate power cable.
    The caps are definitely an issue on the 80.

    Some pointers:




  12. Yuhong Bao says:

    I should also mention the PS/1 model 2121. I found out about UNIXBOOT.COM. It is pretty obvious that OS/2 2.0 has special code for it.

  13. Mr Martin says:

    Hi now the subject is on IBM PS/x machines I have a quick Q. This has bothered me for years. My first PC (which replaced my C64) was a PS/1 2133 something (tower model, 486 SX 25MHz 4MB RAM, Cirrus Logic graphics with 512KB RAM). I live in Denmark and when we received the unit, all the BIOS configuration menus were in Danish (the system was bought in Denmark and the Windows version etc installed was Danish). However, when I upgraded to MS DOS 6.2 (using a special upgrade disk made by IBM) the BIOS configuration menus had turned into english and so they stayed until one day I did a system restore using the recovery disks. These disks did a complete wipe of the system, complete with “scripting” of FDISK to do the partitioning, installation MS-DOS, running DBLSPACE etc. However, after this whole process had run, the BIOS language was once again Danish! I then upgraded to MS-DOS 6.2 again and the BIOS turned into English.

    It is of course no major issue the language became English, but I have always wondered what was the trigger. Unfortuantely I didn’t have the skill to investigate at the time. I would guess it would be some kind of hint on the HD maybe in the bootsector or some of the sectors in the first cylinders that indicated the language to the BIOS. Somehow the MS-DOS upgrade managed to destroy this hint. Another theory could be that a whole BIOS overlay was stored on disk (perhaps with other differences than language, but also bug fixes) but again that this overlay got destroyed by the MS-DOS upgrade. It could also be the MS-DOS installer somehow messed with CMOS settings and that this is where the language indicator was stored.

    Anyone who has better theories?

  14. Michal Necasek says:

    I’ve never really seen a PS/1 up close. But I’m somewhat familiar with PS/2 machines and the second generation systems had an “IML partition” (Initial Machine Load) which contained things like updated BIOS, copy of a reference diskette, or adapter-specific configuration and diagnostic files. These systems had good old ROM, not flash (early 1990s) and the IML was among other things used to update the BIOS because the ROM only needed to read the rest of the BIOS from disk.

    So I don’t know if your theory is about zapping something on the disk correct but it sounds plausible. DOS messing with CMOS data is on the other hand extremely unlikely.

  15. Mr Martin says:

    Michal: OK yes that makes it more plausible (i.e. the PS/2 having a similar “overlay” feature). I guess those machines were just few years before flash BIOS started to become common, so it was a feasible way to make BIOS upgrades possible.

    By the way, that PS/1 model also supported the “Rapid Resume” feature which worked akin to hibernate (saving all RAM and various other state to disk). This was saved to a file in the FAT file system. This might suggest (it is at least a possibility) the BIOS was “FAT-aware” to some degree and so it is also plausible that it could be looking for an updated file on disk.

  16. techfury90 says:

    I have a 2121 myself- there’s no IML partition. It’s pretty much your typical OEM 486 system. I’m guessing the recovery media just flips a flag in the CMOS.

  17. Michal Necasek says:

    I have some trouble believing that because that would imply DOS twiddled with the CMOS in the first place, and DOS is not known to do that.

  18. Michal Necasek says:

    Many ThinkPads (and now I don’t remember if it applied to the PS/2 TP 700/720, but it definitely did to the later 750/755/76x/770) supported hibernation and yes, the hibernation file was just a regular file on a FAT partition.

  19. Yuhong Bao says:

    The way BIOS configuration is done in the PS/1 model 2011 and 2121 is using customiz.exe and configur.exe. My guess is these programs probably detect the language version of DOS.

  20. techfury90 says:

    I just realized that I meant 2133, not 2121. Whoops, my bad.

  21. Richard Wells says:

    customiz.exe and configure.exe are both single language executables. Too small to have multiple languages. Maybe the upgrade included updated English language versions of both or some other update was also applied. Without the DOS 5 upgrade disks used, this is all conjecture.

  22. Mr Martin says:

    Hi, on the PS/1 model I had (2133 tower model) I entered the BIOS setup by pressing F1 when the Wave pattern was displayed upon boot, before DOS was loaded. However, there was also the configur.exe program which took one to the same BIOS setup screen from within DOS.

    I think if someone could get a hold of a PS/1 2133 ROM, ideally the international version, it should be possible to settle this.

  23. Richard Wells says:

    The 2133 ROM available to me is 128kB and includes only English. For it to work as a multi-language ROM, the size would need to be increased by about 8kB per additional language. Difficult to parse the text completely but I do not believe the BIOS includes the full range of setup options. There is no text defining types of floppies for example.

  24. Mr Martin says:

    Richard Wells, OK – the BIOS could potentially also contain code loading in the language data from an area on disk. Possible even a full BIOS overlay.

    It would actually make sense for IBM to design the BIOS this way (English but with special override data on disk) to avoid having to have a unique BIOS ROM in each geography, the differences being limited to what is stored on disk (which was already nationalized).

    Does the 2133 model you have the BIOS for support Rapid Resume?

  25. Richard Wells says:

    Mr. Martin: I think it has Rapid Resume.

    Actually detailing how things work in the BIOS would take effort on my part. Examining a file to see how much text there is and what language is used was the limits of my interests in the topic.

  26. Mr Martin says:

    Richard Wells, yes I realize that. Would it be possible for you to share the file?

  27. raijinzrael says:

    Just a little update. Seems someone found his way to replace the ps/2 proprietary floppy unit with a cheap standard 3 1/2 floppy drive.

    Not the cleanest method, but works well enough :-P.

  28. Pingback: More Wobbles | OS/2 Museum

  29. Mr Martin says:

    This is an old thread but I don’t know if anyone is still watching 😉 Regarding my question on how the IBM PS/1 model 2133 supported multiple languages: I’ve downloaded some recovery disk sets for various languags. I can see that they as part of the recovery process run a program UPDATEEA.EXE. The program code reads in a file called MRI.BIN which is also supplied. When one opens this file in a text editor it does indeed seem to be BIOS texts and in the relevant language – i.e. the Germany recovery disks contain German BIOS texts etc. I’ve disassembled UPDATEEA.EXE and I’m not an experienced disassembler… but I can see it reads the MRI.BIN file and also does some disk writes using INT 13H. Also, I can see that (by using the PS/1 recovery disks in the PCEM emulator) that the disks create a partition setup where there’s one big primary partition taking all the space and a very small “Non MSDOS” partition at the end. As a final test, I’ve looked at the end of the harddisk image file of the virtual machine, and at the absolute end there’s a copy of the data from the MRI.BIN data file. Unfortunately, the BIOS language didn’t actually change to what was included, but I think the ROM image I used was maybe not an exact match for the model that the German recovery disks were made of. But all in all, I think this points strongly to the IBM PS/1 BIOS having a feature that allows at least the text – and possibly other aspects – to be read in from a partition near the end of the harddrive, and that the UPATEEA.EXE has the function of copying the MRI.BIN data to this partition. I don’t know if the program sets up the partitions itself, because the partition setup is earlier in the recovery phase (based on scripts that calls FDISK).
    It makes sense IBM set it up this way, because it would allow them to ship the same ROM chip for all models and then have the national language customization be controlled by content on the harddrive, which had to be adapted to the languages anyway.
    There must have been something in the step-up program to MSDOS 6.2 that somehow ruined this image or made it impossible for BIOS to recognize the non-MSDOS partition

  30. Michal Necasek says:

    Most likely MS-DOS 6.2 rewrote the MBR and possibly modified the partition table.

    Those PS/1 machines were… interesting. Putting bits of firmware on hard disk was common circa 1990, but went out of fashion when sufficiently large flash ROM became feasible.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.