eisapc wrote on 2023-01-12, 09:36:

The model 30 uses slightly different memory layout than standard PC. Here is an explanation and a conversion instruction:
http://john.ccac.rwth-aachen.de:8000/misc/ps2cache/

That document refers to the IBM PS/2 Model 30 286. I checked the schematics of the IBM PS/2 Model 30 (non-286), and those schematics clearly indicate that standard PS/2 modules are supposed to work in that machine. Pins 26/29 as well as Pins 2/28 are bridged on the mainboard, so there is no need to install these bodges on the SIMM. Furthermore, I know that the Model 30 (non-286) BIOS probes whether memory works instead of relying on any kind of presence detect information. So I am confident that the issue is not caused by non-IBM 30-pin SIMMs.

The MB is covered by a thick layer of dust which indicates the fan has been working with those modules. The whole machine is in pretty bad shape (lots of rust around but not on PCBs) so it feels reasonable that those modules need a clean.

However, I'll bear in mind the info above - which are important as I was not aware about IBM modules!

I'll resume that project soon!

Thanks!

Back to the PS/2 - I found a bad capacitor on the speed control circuit which would cause the motor to spin at 600RPM instead of 300. Once that was replaced, the drive worked-ish. It always manages to read the DOS "Boot error" message, sometimes it boots up, sometimes it doesn't. When it boots, sometimes it gives read error. I managed to run Checkit from it - so the drive more or less works - but the floppy test gives a lot of R/W errors, mainly head 1.

Notes
1. The mechanism is a bit crusty. I added some lubricant to help. That said, I don't think it's so crusty to cause all those errors
2. I am using 1.44MB floppies. I know, it's not the best. But I really don't have any proper 720KB and the ones I have are mouldy so probably not worth the hassle. Been looking for brand new 720K ones for ages.

Since the drive managed to load software, I'd say it generally works. Any ideas to make it a bit better?

tony359 wrote on 2023-03-01, 17:51:

Since the drive managed to load software, I'd say it generally works. Any ideas to make it a bit better?

Try to clean the head, for example using some kind of solvent (e.g. heptane or isopropyl alcohol) on a cotton swab.

Forgot to mention: I cleaned the heads many times with isopropyl 😀

little update: I found some faulty logic ICs on the floppy PCB, I'm in the process of replacing them. Finger crossed.

tony359 wrote on 2023-03-07, 10:31:

little update: I found some faulty logic ICs on the floppy PCB, I'm in the process of replacing them. Finger crossed.

Out of curiosity - what ICs are those, and how did you figure they are faulty? These are pretty modern drives already with just 1-2 ICs, all SMD too. I don't think there's much to go wrong with these, they either work or not.

Also, lubricating parts of the drive is usually a short-term bandaid more than a solution to a problem. Most of these things are supposed to run dry, except the parts must be clean. Dirt, and especially any oily and sticky residue like tobacco smoke will interfere with head positioning in pretty short order. And can't be just blown clean with compressed air either, in these cases you have to take things aparat and do proper cleaning with water, soap and IPA. This can usually be done without having to realign the drive, but sometimes (if repairs are needed or there's gunk you can't get to otherwise) even the head assembly and motors have to come off.

I found that the drive would work/not work when using the hot air/cold spray method and narrowed down to two logic ICs. I removed then both and they tested bad in my TL866.

I hear you on the mechanic but it seems to be good enough. After replacing the AND gate the head finally started moving as expected. Now I hope the rest of the functionality is restored when I replace the flip-flop 🙂

Any indication on how to disassemble without having to re-align?

Thanks.

tony359 wrote on 2023-03-07, 12:00:

I found that the drive would work/not work when using the hot air/cold spray method and narrowed down to two logic ICs. I removed then both and they tested bad in my TL866.

Oh so it does use some older tech after all? Note that some of these chips (usualy special amps, not logic gates) tend to run on the hot side. Not so hot as to burn the finger but more than just warm. That's normal but if the chip does act up with cold spray then it is suspect.

tony359 wrote on 2023-03-07, 12:00:

Any indication on how to disassemble without having to re-align?

If you have a scope and some patience, and a known good floppy, it's possible to realign the drive. It's just the first time, and on some drives, it can be rather time consuming and frustrating before you get it right.
But if you wish to avoid that the rules are simple: the stepper motor and the zero track sensor must remain in place. On drives with worm screw the head assembly can usually be taken out completly (unless the signal cables are soldered to the PCB instead of using connectors), and the sliding parts (polished rods, brash bushings) cleaned to perfection. On drives with metal belt the head can be moved about a bit but can't be removed completly, though usually there's enough freedom to clean everything.

Removing zero track sensor is easiest to fix afterwards, with just software like IMD and no scope. Moving the stepper motor requires full procedure afterwards. Disassembling upper head from lower one is the procedure x 2, plus possibly some other steps like azimuth correction - which can't really be done without a proper floppy with test signals but it's not impossible to get it good enough without it. But I would not do that on the first realign ever.

Those ICs definitely test bad in the TL866 - and replacements don't so I am hopeful! When I replaced the AND gate, the head finally started moving as you'd expect and the drive would seek at every boot (previously it would act erratically). Flip-Flop coming tomorrow, finger crossed 😀

Gotcha about disassembling. I will never get a "known good floppy" so I'm happy to keep the stepper in place. I did loosen the plastic screw at the back though (my bad) but it seems it can only go back in one position, it's basically used to remove the worm gear mechanical play: too tight and the stepper doesn't have enough torque to move, too loose and the gear has mechanical play. I only need to remove a rusted plate which sits in front of the spindle but that seems not to be connected to anything critical so it should be ok.

Thanks for your input! 😀

tony359 wrote on 2023-03-07, 13:31:

I will never get a "known good floppy" so I'm happy to keep the stepper in place.

I tend to use OEM software floppies, these were written using well calibrated duplicating machines. Well, should be anyway, it's best to use known software (like MS install floppies) than some no-name card drivers for example. Having more than one floppy like that helps too, for verification. Even pre-formatted blank floppy fresh from the box (never used) can be used for this purpose.
If none of that is available then a floppy written to by a working drive can be used as well, though with the wear these drives have seen you can no longer be sure it's close to perfect. Better then having a non-functional drive though.

tony359 wrote on 2023-03-07, 13:31:

I did loosen the plastic screw at the back though (my bad) but it seems it can only go back in one position, it's basically used to remove the worm gear mechanical play: too tight and the stepper doesn't have enough torque to move, too loose and the gear has mechanical play. I only need to remove a rusted plate which sits in front of the spindle but that seems not to be connected to anything critical so it should be ok.

So long you didn't move (shift, rotate) the motor itself it should be fine. Still, it if gets worse (reading is flaky or completly fails but the drive can format and R/W floppies after that) it might mean the tension it had was actually helping to correct the positon, but I would expect the aligment to be mostly OK even in that case. Note that usually the worm screw has a "ball bearing" at the end opposite the motor. Just one tiny ball actually that the tip of the shaft rest on, it also provides correct centering. If you have to remove stepper motor do not loose that ball.

It's a 720K drive, I'm struggling to find good empty floppies!
But let's see what happens next. At the time being the drive refuses to format disks, IBM own's utility says "no index pulse". Unsurprisingly the index pulse is coming in the AND gate I replaced - and I guess it goes through the flip flop as well.

Now "checkit" says "drive changed" when testing. IMD just doesn't read anything.
Let me get to a point where the drive appears to work correctly and then I'll think about alignment 😀 I'll keep you posted of course!

Replacing the second logic chip, a Flip Flip, worked... for 20 minutes. Then the drive started misbehaving again.

Or better: the head seems to work better - I scoped those logic IC and indeed those logic ICs look after the direction of the stepper motor and so on.
The behaviour is erratic. I can boot from any floppies, I run Checkit and it works for <insert a random time here>. At some point Checkit would report either "drive not ready" or "disk changed". If I reset the PC with CTRL-ALT-DELETE the head does not seek and I get an error 601- Diskette error. However sometimes I can run multiple runs of full Read and Write cycles of Checkit and there are no errors whatsoever. Right now I am at the 5th read cycle and it's working 100%...

If I run IBM diagnostic, I get a "Error 645: no index pulse" to whatever I am trying to do.

if I try to format I get a timeout error message from DOS and same error 645 from IBM utility.

I think the "index pulse" is the little sensor attached to the flywheel, where a magnet triggers a sensor. I can scope that and it looks ok.
I can also scope the IR "track zero" sensor and all is well.

The thing is that the drive works for a while before dying. I've reflowed all the joints - I've just done the VIAs. I've checked all passive components and checked with a thermal camera.
I've noticed that when the drive stops working, the logic goes quiet - but I guess it's difficult to decide whether the main logic died or the main logic stopped pulsing because it sensed an issue - it would be compatible with the 601-Diskette error, the main logic senses an issue and stops working until it's power cycled.

A modern Panasonic drive works flawlessly on this motherboard.

There is something marginal on that board but I cannot find it 😀 Any help appreciated! 😀

I truly admire your perseverance with that floppy drive! If there are IBM gods out there somewhere, you surely have proven yourself to them.

It is interesting stuff to read. I just recently fixed a floppy from my model 35SX, an ALPS drive (34 pin, media sense). I’m so glad that I got it fixed with recapping and cleaning, because I think model 35SX requires media sense drive and I couldn’t find any instructions for building an adapter for that. I looked the pinouts for PS/2 34 pin media sense drives from Ardent Tool, and it looks a lot different from standard floppy. Early PS/2s with non-media sense drive is pretty simple.

tony359 wrote on 2023-03-08, 18:44:

Replacing the second logic chip, a Flip Flip, worked... for 20 minutes. Then the drive started misbehaving again. […]
Show full quote

Replacing the second logic chip, a Flip Flip, worked... for 20 minutes. Then the drive started misbehaving again.

Or better: the head seems to work better - I scoped those logic IC and indeed those logic ICs look after the direction of the stepper motor and so on.
The behaviour is erratic. I can boot from any floppies, I run Checkit and it works for <insert a random time here>. At some point Checkit would report either "drive not ready" or "disk changed".

[...]

There is something marginal on that board but I cannot find it 😀 Any help appreciated! 😀

The symptoms sound like it might be an open input to CMOS logic, which happens to be read at the correct logic level directly after power-on, but leakage current over time puts it into the non-working level. You should focus on the logic that drives pin 34 ("diskette change").

I truly admire your perseverance with that floppy drive! If there are IBM gods out there somewhere, you surely have proven yourself to them.

I get those machines to repair them. Once they work they're boring 😀
Still, there is a limit over that I just have to give up. Unfortunately the main ICs of that PCBs are not available - and also no schematics so diagnosing can be done up to some point.

mkarcher wrote on 2023-03-08, 20:14:

The symptoms sound like it might be an open input to CMOS logic, which happens to be read at the correct logic level directly after power-on, but leakage current over time puts it into the non-working level. You should focus on the logic that drives pin 34 ("diskette change").

Ah, fantastic. I was going to check that pinout and then I forgot. I'll trace the signal.
That said, I think I haven't seen the "disk changed" error since I replaced the flip-flop. Now I get "drive not ready" followed by "read error". The head stops moving and to be honest there is nothing wrong with the mechanism, I can exercise it for a long time and it never fails. It's just that the whole logic driving it gets quiet.

UNLESS, the head fails for whatever reason (it could be the main driving IC), the main logic IC realises that and it stops reporting a fatal error to the interface.

I've ordered the remaining two logic ICs - buffer ICs - they test ok on my 866 but in case they are marginal.

Anybody aware of schematics for that floppy?

Thanks for your help!

Edit: I forgot to check IBM's own reference manual. I found this. I suppose the "index" pulse is what the software is complaining about. And also the "diskette change" as mentioned. I'll check those out! I didn't realise the floppy is being driven by the PC - or at least the floppy controller on the mainboard. Lots of signals I can inspect on the connector!

Double edit: I've checked all those signals and indeed the PULSE is erratic. First it was always high, then always low. Then it followed the motor. At some point the pulse actually appeared! But it disappeared by itself after a few seconds. It can't be good for the controller! 😀 I'll trace that line and hopefully find the culprit! Thanks a lot, you gave me many trails!

tony359 wrote on 2023-03-08, 20:37:

Double edit: I've checked all those signals and indeed the PULSE is erratic. First it was always high, then always low. Then it followed the motor. At some point the pulse actually appeared! But it disappeared by itself after a few seconds. It can't be good for the controller! 😀 I'll trace that line and hopefully find the culprit! Thanks a lot, you gave me many trails!

Don't forget that the floppy interface is a bus that can connect to multiple drives. A floppy drive may send it output signals to the floppy cable only when it is selected (that's how you call addressing on the floppy interface bus). On PCs, drives are generally set to respond to DS1 (not DS0), i.e. they act as the second drive. The A: drive gets the signals from the PC after a twist in the cable that swaps DS0 and DS1, as well as Motor On for drive 0 and drive 1. IIRC the PS/2 is PC compatible regarding drive selects, so you shouldn't worry about missing index pulses, unless /DS1 is not asserted (low).

Thanks.

What is exactly the "diskette index" pulse? What is its purpose? Is it possible the IBM diagnostic tool is looking for that pulse and throw out an error - while other software don't care? Could that be responsible for the failure in formatting? And could it be that IMD is also looking for that?

I'm not sure I understand the relationship between the index pulse and the /DS1 signal.

I should check with the Panasonic drive. But the index pulse was just totally erratic. I could make head and tail of the other signals but the Index was more or less random - unless I'm just misunderstanding the way it works. When the pulse appeared, it looked like the one coming from the flywheel indeed.

tony359 wrote on 2023-03-08, 21:38:

What is exactly the "diskette index" pulse? What is its purpose?

It's for the FDC to be able to tell where sector 1 should begin. One could make a system without it, based on simple timers (if the drive spindle is stable in rpms) but it makes things easier. You know the drive has a disk in it and it's spinning with the pulse, you can tell how fast it's spinning, and early FDCs required this signal to make sure the format you are trying to apply to the disk does not have more sectors than the track can hold.

In general, no index signal (or erratic) and the drive will not work. Might even refuse to work already at the chip level on the drive side, some of the smarter chips need to see the index pulsing to tell a valid floppy is inserted, also to stop the initial motor spin on insertion (which is it help center the disk properly). A drive that outputs READY signal must see the pulse, and it must fall within pretty narrow timing window, for that signal to go active.

On 5.25" and 8" floppies the index come from the hole near the center, on 3.5" the spindle motor has a sensor for that. Older system could get dirty and stop working, the magnetic sensor pretty much never fails - it's a coil that picks up magnetic field from part of the wheel which is unmasked (no metal cover over it). Sometimes the sensor is SMD, on the PCB under the magnet in the wheel. Corroded leads, traces or cracked solder (or the flex cable between spindle and main PCBs) can mess up that signal.

One more thing you should check - floppy drives have O/C ouputs and need pull-up resistors for each signal. This is usually a resistor pack, often called a terminator although it's not what it's for (but it serves that purpose too). Older drives have 150 ohm pack and only the last drive on the cable should have one installed. That pack is removable, there is a socket for it. Newer drives are "auto-terminated", which is fancy name for a permanently soldered pack of 680 ohm to a few kohm resistors. This limits the cable length to some 50 cm but that's still more than what typical modern PC will have, plus you can have the pack on each drive and not have to mess with it.
Point here is - check those pull-ups, any problems with those (too high resistance, broken connection, etc) will give you semi-random readings. Also do not mix new and old tech, there can be only one low-value pack on the cable. If you have 2 drives and one is auto-terminated and the other has a removable pack, the pack needs to go. Having the pack installed not at the end of the cable might give you some more R/W errors but nothing as bas as you have now.

Thank you.

I did check the terminator resistor pack and it was fine but I did not check continuity with +5V and ground, I shall do that. For now though the signals I am observing at the connector look clean.

I understand about the index pulse. On my drive it's a small magnet glued to the flywheel and there is a coil at the edge. I see this pulse on various ICs on the PCB but it does not seem to reach the connector - or at least not all the time. So that is a path to follow.
However I am a bit confused: shouldn't the controller on the motherboard complain that the index pulse is not present all the time? How come it works and only the IBM diagnostic utility complains about that?

I am not sure I understand the "some of the smarter chips need to see the index pulsing to tell a valid floppy is inserted, also to stop the initial motor spin on insertion" - are you referring to the 5.25" disks where the pulse was coming from the hole on the disk? On this drive the pulse comes from the flywheel (or maybe I've been calling it wrong and it's the rotor, or stator?) so it's always there. I think on this drive the item which tells the logic there is a disk inserted is a microswitch which gets pressed when the disk is inside. Indeed I've noticed this triggers the "disk changed" pin.

This is all very interesting. If you could clarify those couple of points above, that would help a lot. But I will definitely follow the "erratic index pulse" trace later this week!

Thanks!