By the way, I have now got a working 1.2mb drive, so can use that for comparisons to this failed one. It is a completely different model and brand though. At least that proves though that my floppy controller on motherboard is fine, and so is the cable.

Deunan wrote on 2025-06-28, 10:30:

If you are up to some more scope measurements, probe pins 1, 2 and 3 of the servo chip. Might tell us more about that suspect capacitor.

I am not seeing any waves on these pins, though not sure if I have the right settings. What kind of frequency are we expecting here?

One strange thing I notice though. When I put the scope on pin 2 the motor sounds like it's speeding up, and when I probe pin 3 the motor sounds like it's vibrating. Is that expected? My scope is grounded on the molex connector before the one attached to the drive. It's a portable scope, so there can't be a connection through Earth on it.

I tacked on the 1uF cap, so there's now two in parallel. The disk speed is now being reported at 9hz now, sometimes saying 10hz briefly. So that seems to have made it worse, as we had 8hz before. Is that what would happen if the original cap is still fine?

RetroPCCupboard wrote on 2025-07-06, 07:02:

I am not seeing any waves on these pins, though not sure if I have the right settings. What kind of frequency are we expecting here?

That's very odd, there should be small pulses there from the speed sensor. Frequency will be low, 100Hz, but the pulses might be pretty narrow. This is where 10:1 probe starts to matter, the pulses must be there if you can hear the motor change speed with the probe touching the pin. But quite weak if that's enough to affect it. Set the channel to lower voltage, even these protable scopes should have someting like 10mV/div or so.

As for the speed getting worse with additional cap, maybe the old one is not dried out but very unformed and leaky. In any case, if this didn't help then I would remove the extra cap and desolder the old one completly. Then you can again just tack in a replacement, or just do a proper replacement job if you are sure the new cap is good. You can try to measure both caps once out of circuit to compare them.

EDIT: It might be even less than 100Hz, like 10 or so, I'm starting to think the datasheet is confusing Hz and rpm in the equations they give. 100Hz would require 10 gaps in the magnet shield to count this many pulses per second, but that might be the case here.

Do we know for sure that the rotation speed sensor is actually producing a signal and that signal is making it though
to the controller?

The signal from the sensor is going to be "weak analog" and what looks like a valid signal might be just noise... You really need to confirm
that it's representative of the actual speed of the platter (adjusted for how ever many sensor triggers occur in a rotation on this design)

If you wan to figure out how fast the platter is actually rotating, using a SoftSec diskette (1 index hole) look at the index sensor
with the disk spinning - you should see one pulse per rotation. Observed Hz x60 would give you RPM.

Keep in mind that your scope may not directly give you super accurate measurements in fractional Hz - probably best to
measure in ms and convert to hz : (1000/ms)*60=rpm

DaveDDS wrote on 2025-07-06, 12:53:

Do we know for sure that the rotation speed sensor is actually producing a signal and that signal is making it though to the con […]
Show full quote

Do we know for sure that the rotation speed sensor is actually producing a signal and that signal is making it though
to the controller?

The signal from the sensor is going to be "weak analog" and what looks like a valid signal might be just noise... You really need to confirm
that it's representative of the actual speed of the platter (adjusted for how ever many sensor triggers occur in a rotation on this design)

If you wan to figure out how fast the platter is actually rotating, using a SoftSec diskette (1 index hole) look at the index sensor
with the disk spinning - you should see one pulse per rotation. Observed Hz x60 would give you RPM.

Keep in mind that your scope may not directly give you super accurate measurements in fractional Hz - probably best to
measure in ms and convert to hz : (1000/ms)*60=rpm

I can't see an easy way to probe that sensor, as the wires have heatshrink on. But the Index pulse, measured from the edge connector that the floppy cable uses, seemed to have a consistently higher speed than 360rpm. It didnt look like noise.. Is that the same thing as measuring the sensor?

Sorry for the delay in responding. I have had some work and family stuff to deal with hopefully will be able to get back to looking at this drive soon.

Deunan wrote on 2025-07-06, 10:22:

That's very odd, there should be small pulses there from the speed sensor. Frequency will be low, 100Hz, but the pulses might be pretty narrow. This is where 10:1 probe starts to matter, the pulses must be there if you can hear the motor change speed with the probe touching the pin. But quite weak if that's enough to affect it. Set the channel to lower voltage, even these protable scopes should have someting like 10mV/div or so.

As for the speed getting worse with additional cap, maybe the old one is not dried out but very unformed and leaky. In any case, if this didn't help then I would remove the extra cap and desolder the old one completly. Then you can again just tack in a replacement, or just do a proper replacement job if you are sure the new cap is good. You can try to measure both caps once out of circuit to compare them.

EDIT: It might be even less than 100Hz, like 10 or so, I'm starting to think the datasheet is confusing Hz and rpm in the equations they give. 100Hz would require 10 gaps in the magnet shield to count this many pulses per second, but that might be the case here.

Sorry for the delay in responding. I havent had a chance to look at the drive again. I have bought a box of capacitors of various values. I am thinking of just replacing them all and see what happens. There's only 5 of them. I was planning to leave the ceramic ones alone for now, as I think they are less likely to die?

I will start with just the cap you suggest first though. Then will checl RPM to see if anything changed.

After you said you didn't see any signals on pins 1-3 I'm somewhat worried that it might be the speed sensor coil that went open, but with just 5 caps a repair attempt won't cost that much anyway. I'd start with C11 and see if it helps, or you can just replace all for the soldering experience.

Deunan wrote on 2025-07-17, 11:13:

After you said you didn't see any signals on pins 1-3 I'm somewhat worried that it might be the speed sensor coil that went open, but with just 5 caps a repair attempt won't cost that much anyway. I'd start with C11 and see if it helps, or you can just replace all for the soldering experience.

Ok. Thanks again. I will let you know how I get on. Hopefully will have time this weekend.

RetroPCCupboard wrote on 2025-07-17, 10:35:

DaveDDS wrote on 2025-07-06, 12:53:

Do we know for sure that the rotation speed sensor is actually producing a signal ...

I can't see an easy way to probe that sensor, as the wires have heatshrink on. But the Index pulse, measured from the edge connector that the floppy cable uses, seemed to have a consistently higher speed than 360rpm. It didnt look like noise.. Is that the same thing as measuring the sensor?

If you can figure out where it if connectd to the motor board, perhaps you can probe it there?
I've not been above using pins to go down inside heatshrink or into wires to make test points.

The index pulse should be once for rev - so 360hz for 360rpm. Unless you have somehow gotten hard-sector diskettes - which would be
11 or 17 pulses/rpm.

Another option - ... tape a little wire that will move in a circle to the motor spindle and be accessible -
you should be able to get it to touch your probe once/rev ... the signal will be very noise at the moment of contact, but
you should see "bits" of noise for every revolution.

And no, the motor speed sensor detect the motor, but a diskette - without a diskette where would be no signal, yet you
want the motor to always be the right speed ... using the method I suggested above (or the diskette index sensor)
will at least let you see if the motor or spinning at the right speed but to debug the motor controller, you need to access
its speed sensor.

DaveDDS wrote on 2025-07-17, 15:53:

The index pulse should be once for rev - so 360hz for 360rpm. Unless you have somehow gotten hard-sector diskettes - which would be
11 or 17 pulses/rpm.

Er, 360Hz would bit a tad too fast for a floppy drive. Even helium filled HDDs don't spin that fast if the pulse was once per revolution.

Anyway, we don't know if the floppy index sensor works properly but the output was pretty clean, and the higher speed more or less matched the 500k rate but for floppies using 250k. So it's a bit too consistent for a simple sensor glitch. I think it works just fine and it's the speed sensor in the motor that's suspect. How that one works depends on the driving chip. Some might require 1ppr, so for 360rpm that'd be 6pps or 6Hz. This one seems to want 60pps but that's for the 492kHz resonator from the datasheet. Which we know is changed to 819kHz so I would expect to see 100Hz on the pins I mentioned earlier.

Deunan wrote on 2025-07-17, 20:50:

Er, 360Hz would bit a tad too fast for a floppy drive. Even helium filled HDDs don't spin that fast if the pulse was once per revolution.

My apology ... I was thing in Cycles/Minite ... it just came out wromg.

Anyway, we don't know if the floppy index sensor works properly but the output was pretty clean, and the higher speed more or less matched the 500k rate but for floppies using 250k. So it's a bit too consistent for a simple sensor glitch. I think it works just fine and it's the speed sensor in the motor that's suspect. How that one works depends on the driving chip. Some might require 1ppr, so for 360rpm that'd be 6pps or 6Hz. This one seems to want 60pps but that's for the 492kHz resonator from the datasheet. Which we know is changed to 819kHz so I would expect to see 100Hz on the pins I mentioned earlier.

I should also have mentioned that the motor RPM sensor is probably much faster than 360 hm...
You really need to examine the motor and it's sensor to figure out how it works.

Deunan wrote on 2025-07-17, 11:13:

After you said you didn't see any signals on pins 1-3 I'm somewhat worried that it might be the speed sensor coil that went open, but with just 5 caps a repair attempt won't cost that much anyway.

The signal of the speed sensor coil is quite low, see Re: Let's fix some PS/2 floppy drives! about how I entirely missed it at my first attempt of verifying the signal. And if you have a broken cap on that signal, the broken cap might block the signal, so you better measure on both sides.

Also be aware that some motor controllers (like the one in that thread) have separate feedback for rotation speed (with many pulses per rotation) and the BLDC hall sensors (at as many pulses per rotation as required by the motor design, likely less than the feedback signal). The index signal derived from the index hole sensor is likely entirely separate, as it is useless for both speed control purposes (feedback too slow at just one pulse per rotation) and motor control (phase relating to the motor rotor position is unknown in 5.25 drives).

Deunan wrote on 2025-07-17, 11:13:

After you said you didn't see any signals on pins 1-3 I'm somewhat worried that it might be the speed sensor coil that went open, but with just 5 caps a repair attempt won't cost that much anyway. I'd start with C11 and see if it helps, or you can just replace all for the soldering experience.

I took out C11. My tester seems to say it is 7000pf. So well off from the 1uf it should be.

I replaced that cap with a new 1uf cap. Much bigger physically, but still 1uf 50v. I had to bend the pins and lay it sideways.

The temptation was great to replace all of them, given how far off this one was. But I resisted the urge and put the drive back together to test it.

To my delight, it was able to read a 1.2mb formattted disk! I tried reading and writing from that disk, and all was good.

I used my scope on the index sensor pin of edge connector and it said 6hz (compared to 8-9 it said before). IMD said its 360rpm also:

This is what alignment test reports:

I have tried 360k disks also and they read fine too.

Do you think it worth replacing the other caps, or should I quit while I am ahead? (Given that I am not very good at soldering). Two of the caps dont appear to have obvious markings on them for which is negative terminal.

That's great news! Bigger cap is not a problem, in fact bigger is better - if it fits. The 50V rating is not really required here, it's probably the cheapest they could source. As for the rest of them, C11 was the only critical one that was electrolytic (now that I think about it, replacing it with foil or MLCC type is tempting, maybe even tantalum since there doesn't seem to be any DC bias here). The ones that really matter are ceramic or foil, but considering how far this one was gone it might be a good idea to swap the rest of the electrolytics too. It shouldn't make any difference to how the drive operates but might prevent damage to the controller chip from voltage spikes on the stator coils, and prevent back-feeding that noise to the drive main PCB and the rest of the PC. I suppose it depends on how confident you feel soldering, if you have any doubts about the PCB integrity (these single layer PCBs can have the copper peel off quite easily because of age, I've had that) then you can skip it.

Deunan wrote on 2025-07-19, 09:45:

That's great news! Bigger cap is not a problem, in fact bigger is better - if it fits. The 50V rating is not really required here, it's probably the cheapest they could source. As for the rest of them, C11 was the only critical one that was electrolytic (now that I think about it, replacing it with foil or MLCC type is tempting, maybe even tantalum since there doesn't seem to be any DC bias here). The ones that really matter are ceramic or foil, but considering how far this one was gone it might be a good idea to swap the rest of the electrolytics too. It shouldn't make any difference to how the drive operates but might prevent damage to the controller chip from voltage spikes on the stator coils, and prevent back-feeding that noise to the drive main PCB and the rest of the PC. I suppose it depends on how confident you feel soldering, if you have any doubts about the PCB integrity (these single layer PCBs can have the copper peel off quite easily because of age, I've had that) then you can skip it.

The capacitor I used came from a kit box. 670 capacitors of varying capacitance for £14.99. Was this a bad choice? Since you mention MLCC, foil and tantalum, I googled them. There seem to be a variety of sizes and shapes, and much more expensive than this kit I bought. I am not really sure what to buy.

The C11 cap came off fairly easily and the traces didn't get damaged as far as I see. I think perhaps I should replace the other caps too then. Question is, with what. Are the cheap ones I bought going to fail quickly? The measured capacitance was very close to the stated value.

RetroPCCupboard wrote on 2025-07-19, 10:44:

The capacitor I used came from a kit box. 670 capacitors of varying capacitance for £14.99. Was this a bad choice? Since you mention MLCC, foil and tantalum, I googled them. There seem to be a variety of sizes and shapes, and much more expensive than this kit I bought. I am not really sure what to buy.

No, no, leave it be. It's good. Even if it degrades again, it'll take 30+ years. By that time you won't care, or just replace it again.

RetroPCCupboard wrote on 2025-07-19, 10:44:

The C11 cap came off fairly easily and the traces didn't get damaged as far as I see. I think perhaps I should replace the other caps too then. Question is, with what. Are the cheap ones I bought going to fail quickly? The measured capacitance was very close to the stated value.

The problem with other capacitors is some of them are non-polarised (don't care which lead is + and -) and most electrolytic caps are not like that. Chances that you have some in a standard set are slim. Unless it's a kit specifically for this kind of repair? At the same time replacing these with something solid state would be more difficult due to higher capacity. Any foil based 4.7uF would be way bigger than even NP electrolytic.

So, if you already have the correct capacitors, and don't consider more soldering to be too much work, then go for it - what's the point of having the capacitors if they are not going to be used. Otherwise consider the job done, you are not going to be using that floppy drive daily for years to come.

Deunan wrote on 2025-07-19, 12:18:

No, no, leave it be. It's good. Even if it degrades again, it'll take 30+ years. By that time you won't care, or just replace it […]
Show full quote

RetroPCCupboard wrote on 2025-07-19, 10:44:

The capacitor I used came from a kit box. 670 capacitors of varying capacitance for £14.99. Was this a bad choice? Since you mention MLCC, foil and tantalum, I googled them. There seem to be a variety of sizes and shapes, and much more expensive than this kit I bought. I am not really sure what to buy.

No, no, leave it be. It's good. Even if it degrades again, it'll take 30+ years. By that time you won't care, or just replace it again.

RetroPCCupboard wrote on 2025-07-19, 10:44:

The C11 cap came off fairly easily and the traces didn't get damaged as far as I see. I think perhaps I should replace the other caps too then. Question is, with what. Are the cheap ones I bought going to fail quickly? The measured capacitance was very close to the stated value.

The problem with other capacitors is some of them are non-polarised (don't care which lead is + and -) and most electrolytic caps are not like that. Chances that you have some in a standard set are slim. Unless it's a kit specifically for this kind of repair? At the same time replacing these with something solid state would be more difficult due to higher capacity. Any foil based 4.7uF would be way bigger than even NP electrolytic.

So, if you already have the correct capacitors, and don't consider more soldering to be too much work, then go for it - what's the point of having the capacitors if they are not going to be used. Otherwise consider the job done, you are not going to be using that floppy drive daily for years to come.

I think C9 (47uf) and C3 (33uf) are the only ones I can change. The other three seem to be non-polarised.

I think it is more likely that I will use 3.5" drive, but want the 5.25 just in case I aquire more software / games in that format. Also I love the sound the PC makes when it initialises each drive

Deunan wrote on 2025-07-19, 09:45:

C11 was the only critical one that was electrolytic (now that I think about it, replacing it with foil or MLCC type is tempting, maybe even tantalum since there doesn't seem to be any DC bias here).

Now that we have those ultra-cheap very small high-capacity MLCC SMD capacitors, even their worst grade, the Y7V-type dielectric performs perfectly well for DC decoupling of sub-1V signals at below 10V DC. Look for example at YAGEO CC0603ZRY5V8BB105, which are 0.1€/piece at mouser if you by single capacitors and and 0.01€/piece if you buy 50.000 of them. For this specific application, I am confident they perform at least as good as the electrolytic originally used on that drive, while having a virtually unlimited life time.

Please don't take this as a general recommendation for MLCC caps, and especially not Y5V SMD caps. They earned their bad reputation for a reason. They have many drawbacks, mostly prominently loosing most of their capacitance even below the nominal voltage, and are not generally suitable to substitute electrolytic caps, but I don't see how their downsides could be relevant for DC removal of the tachometer signal.

Deunan wrote on 2025-07-19, 12:18:

So, if you already have the correct capacitors, and don't consider more soldering to be too much work, then go for it - what's the point of having the capacitors if they are not going to be used. Otherwise consider the job done, you are not going to be using that floppy drive daily for years to come.

Ok, so I removed capacitors C3 and C9 and put them in my tester. Both had near enough correct values. C3 was almost exactly 33uf. C9 was slightly over the 0.47uf it should have been, but only slightly at 0.5uf.

In my kit I had a 0.47 capacitor, but bigger physically. The tester said it was 0.47uf, so I replaced the original.

My kit also had 33uf capacitors, but when I tested them they were all reading as 41uf. So, I put the original 33uf capacitor back.

Heres the board front and back after cap replacement:

These are closeups of my attempts at soldering:

Perhaps burned a little, but I dont know.

Anyway, I have put it all back together again, and it still works great! I am very happy!

Thanks again to everyone that helped me. Especially @Deunan for his excellent electronics knowledge and to @DaveDDS for writing IMD, and other helpful comments. This Vogons community rocks!