That you can touch a cap to get a speed change tells something... Have you tried replacing the caps? Many do go bad with time.

If you can get it to the point where it can "read a bit", you could use ImageDisk 'T'estRpm function to determine the actual speed.
It does the minimal read possible (ReadID) which means it does have to be able to read something from the drive.

Failing that, if you have an oscilloscope (or possible low-range frequency counter), you could read the index sensor to
get the RPM.

DaveDDS wrote on 2025-12-24, 12:30:

That you can touch a cap to get a speed change tells something... Have you tried replacing the caps? Many do go bad with time. […]
Show full quote

That you can touch a cap to get a speed change tells something... Have you tried replacing the caps? Many do go bad with time.

If you can get it to the point where it can "read a bit", you could use ImageDisk 'T'estRpm function to determine the actual speed.
It does the minimal read possible (ReadID) which means it does have to be able to read something from the drive.

Failing that, if you have an oscilloscope (or possible low-range frequency counter), you could read the index sensor to
get the RPM.

Thanks Dave. I went ahead and replaced the "touchy" cap, and indeed, the motor started behaving properly again and I can measure 360RPM perfectly in IMD.

However, the joy didn't last for too long. As I started to extend my tests further, I noticed the following:
Drive fully reads perfectly without issues.
But it fails to write using head 0 (the bottom one). If I attempt to write anything with it, the disk gets corrupted and can't be read anymore, until I reformat the disk in some other drive, then it becomes readable again.

In IMD, I can read all tracks cleanly. But when writing an image to disk, IMD reports all write attempts using H0 as error NoAdrMark on all sectors of the track. Writes using H1 go fine without issues.

After writing is done, if I attempt to read the disk again, IMD stops right after it attempts to analyze T0 H0, and says "0 sectors" and stops the read process. So it seems the writing erased the sectors, but couldn't write new ones.

Once again, I swapped the main board between this and its working clone, alas, it didn't make a difference. The board from the non-working drive operated properly on the working drive and I was able to write disks without issues there. And the board from the working drive didn't fix the issue when installed in the non-working drive.

This leads me to believe it's an issue with the heads or their cables, not the PCBs.

Each head gets 5 wires.. not sure what is the purpose of each and which one I should focus on. Too bad the problematic head is H0, which requires me to fully disassembly the heads to get to it.. which means realignment most probably... ugh

I measured resistance across the set of 5 wires from the good head, and compared them against the 5 wire resistances of the bad head.. both match. I believe this rules out broken connections in head 0?

Update: more info, the harness connecting the head cables to the PCB has two sets of 5 wires. One set for each head. I identified the green line to be responsible for this not-writing issue. I disconnected H1's green wire and now both heads can't write in IMD with the same error (NoAdrMark). So now I really now which cable carries the write signal. But I'm still not sure if it's a H0 cable fault or coil fault. The cable measured 11Ohm resistance to its neightbor red wire, so it's not severed, I guess?

wbahnassi wrote on 2025-12-28, 18:38:

Update: more info, the harness connecting the head cables to the PCB has two sets of 5 wires. One set for each head. I identified the green line to be responsible for this not-writing issue. I disconnected H1's green wire and now both heads can't write in IMD with the same error (NoAdrMark).

That error is kind of strange. The process of writing a sector (not formatting a track, which works differently) is reading until an index address mark is encountered which marks the sector that should be overwritten. Then the floppy controller switches over to writing the data synchronization sequence, the 512 byte sector payload and the CRC, and then goes back to reading again. The complete logic that generates the write enable signal (and its duration) is only present once in the floppy controller, not the drive. So if you can successfully write using head 1, this means the detection of the position to write and the length of the write enable signal is properly generated by the floppy controller.

As reading from head 0 works, it also is unlikely that head 0 is permanently erasing the data on the disk, or unable to read properly. I suspect the first sector write attempt still finds the "Address Mark", because only that will turn on the write enable signal. But the consequence of writing a sector seems to be that no further sector can be found, i.e. obviously a longer part of the track is tampered than the data area of the sector to write. If the subsequent index address mark is overwritten, trying to write that sector will fail with the error "address mark not found" until the track is reformatted.

I don't have any idea what damage a head could have that it does not overwrite data in read mode, but overwrites excessive data on a write attempt. As you already tried swapping the drive PCB, it is unlikely that turning off the erase coil of head 0 is too late with both of the PCBs - otherwise that would be a plausible explanation of your observed behaviour.

Replace the electrolytic capacitors on the spindle motor PCB. These often either leak or dry out (or both) and the resulting excessive noise from power lines and feedback circuit will cause the motor control chip to misbehave. Note some of these caps are of the non-polarized type and might be difficult to source - especially considering their size, there usually isn't space there to install a bigger one, not by much anyway.

As for the weird behavior I would suspect slight head misalignment. These heads do both writing and trimming at the same pass so if the track is just a bit off the head will still pick it up and read it (modern head amps are way better at this) but during writing it might "trim" the actual address marks and won't be able to write more than the first sector. Which wont' be readable afterwards anyway. One way to test that theory is to do a full format of the floppy on this particular drive - it will lay its own tracks and that should be writable now, but might (should actually) exhibit the same behavior when the floppy is moved to another drive. That is, it'll be readable but not writable.

You might need to format with /U or scramble the first track at least with a magnet for DOS to properly format the floppy, or just use IMD for that. If you still can't write to a floppy that was formatted in that drive then I would suspect head to be somehow damaged. But that seems unlikely if the resistance measurements show the same values between the two.

Deunan wrote on 2025-12-30, 10:18:

Note some of these caps are of the non-polarized type and might be difficult to source - especially considering their size, there usually isn't space there to install a bigger one, not by much anyway.

I know this is a controversial topic, by my stance is that replacing electrolytics below 10μF, even if used for DC decoupling in 5V-powered circuits by modern ceramic caps is preferable due to their better durability. The loss of capacitance at elevated voltages is usually not that relevant if a 35V or 50V ceramic cap is DC biased at 2.5V. I especially prefer ceramics if the original caps were non-polarized, as the ceramic caps are more common nowadays, and thus easier to obtain. If you can get electrolytics, including possibly non-polarized ones, at reasonable price, shipping cost and delivery time, and you don't mind swapping them again 40 years later, there is nothing wrong with using electrolytic caps as well.

Deunan wrote on 2025-12-30, 10:18:

As for the weird behavior I would suspect slight head misalignment. These heads do both writing and trimming at the same pass so if the track is just a bit off the head will still pick it up and read it (modern head amps are way better at this) but during writing it might "trim" the actual address marks and won't be able to write more than the first sector.

I doubt this theory is viable. As far as I know, the trimming is supposed to only happen while the write gate is open, so trimming is supposed to stop after the data of the first sector has been written, and should not damage the address mark of the subsequent sector. Is there any reason to assume that trimming takes a longer time to turn off than the actual writing, like time for the AC erasure oscillator to decay, and thus stop generating the trimming signal after the write gate has been closed?

Anyway, assuming the theory is correct, the test to format a disc in the suspected drive is a very good idea, and it doesn't take much time to perform, so I don't recommend to skip that test. And if that test shows the behaviour you describe, the OP learned that it's likely an alignment problem and I learned that some drives obviously "over-trim".

mkarcher wrote on 2025-12-30, 11:15:

As far as I know, the trimming is supposed to only happen while the write gate is open, so trimming is supposed to stop after the data of the first sector has been written, and should not damage the address mark of the subsequent sector.

Yes, but every digital system is in fact an analog system with well defined thresholds. However it's one thing to define and another to actually implement. There is always some overwriting involved due to rotation speed differences and the fact that erasing part of the head is following the write part and not as the same place - and that's why there are post and pre-gaps. Problem is with enough timing differences and possible off-track writing there might be enough damage to the gap areas done for the drive to not be able to resync the PLL properly. So the address mark itself might still be there but it just can't be read, same result.

Well it's my theory anyway, a format should confirm or deny it.

Thanks for the suggestions guys. For formatting, I first start by using IMD's Erase function, then head back to DOS and FORMAT /U /F:1.2 . On this drive the format immediatelly fails reporting that track 0 cannot be written and the disk is damaged.
If I take the same disk again to the good drive, it can reformat it successfully.

Furthermore, in IMD, I take an image and attempt to write it with the bad drive.. Right away, I get NoAdrMark on each sector of each track on head zero. Head one writes without any errors.

I doubt the heads are misaligned off of each other, as the drive really looks factory clean and never tampered with.

Anyways, I went ahead and took out the entire head assembly since there's nothing much to lose. Looking at the cable and wires below that feed into the bottom head. All look good and pristine. All lines have full continuity going to the head. The head is clean and floats nicely flat. If I touch it lightly with the tweezer I can see it tilts and then goes back to its good position, so the springs are good too.

I have now reassembled it and made sure to put it back on the same alignment as original factory. So far no changes... it still reads perfectly fine, until I write anything with it.. then the disk will be ruined.

I retried the the IMD write test, but this time I set the number of retries to 0 to see if any track by chance would succeed on H0 and not spend an eternity to get there.. I noticed that sometimes the writes fail with "NoAdrMark", and sometimes with "Data" instead.. Not sure if this could mean anything.

Could it be a ferrite issue? I'm certain the problem is in the lower head as I've swapped out everything by now and all parts and PCBs work great when put on the good drive, except the head assembly.

IIRC IMD will format each track before writing to it (in case it had different format to what's in the image file) so there is one more possibility here, and that is some mechanical head damage. Still good enough to read the data but writing causes the signal to be too weak or somehow too washed out and it's just not being picked up.

So try one more experiment, format the floppy in another good drive and then try writing to it (on this faulty drive) but not through IMD but just plain DOS. It might look like it works but should fail during verify read - if you attempt one. Doesn't really matter if DOS detects the failure or not though.

Then once some write attempt has been tried run IMD and try reading the entire head 0 surface. You should get data/crc errors at least on the FAT area and where the files ended up. But not address mark errors since DOS will not touch those during simple sector writes. See what you get.

Alright. I started with a working disk. First I read it fully in IMD under the faulty drive amd it read totally fine.

Next, I went back to DOS and copied just one small file to it (copy FLIC.COM A:\"). The copy command executed smoothly (1 files copied successfully).

Next I went back to IMD directly to read the same disk. All this on the same faulty drive.

Now IMD reports "Data - Unavailable" on sectors 6 and 13 of track 0 head 0, and one more "NoAdrMark" error on track 48 sector 1 head 0.
Everything else is still good.

I believe the two sectors at the beginning are the FAT updates, and track 48 held the actual file data.

I've also re-read the disk on the good drive and it reported the same errors at the exact locations.

wbahnassi wrote on 2025-12-30, 18:06:

I believe the two sectors at the beginning are the FAT updates, and track 48 held the actual file data.

Hm, the no addr mark error is kinda unexpected. A weak write should not produce such error, so possibly it's not something like a shorted turn in the head coils (or partially broken wire in the head cable) but rather a crack in the ferrite material that produced a second gap. This gap will also radiate magnetic field during write that might be just strong enough to flip some magnetic domains on the media. Considering such gap would also be off-center it might just be "overwriting" the gap/sync parts. To truly know this you'd need something like Greaseweazle or Kryoflux and dump the flux directly.

Point is though the head is most likely toast. Since you already had the upper one removed once you can do another inspection if you want. Note that small cracks might be next to impossible to see, even the actual gap is tiny and hard to spot. Make sure the head is level and there is no dirt buildup on the surface that might lift the surface of the media away from the head. Even tiny air gaps can cause weird problems that would start with writing/erasing before affecting reading.

If there is nothing obvious then I would just assume the head is no good, unless you only need reading, so not a total loss but such a drive has little practical use.

Thanks Deunan. Indeed, I feel I've reached the end with this poor thing. Too bad.. I'll keep it stashed in case I get replacement parts from another bad drive. YE-Data was a well-known brand, so there is a chance I get another one for parts and I can combine the two for a good one...

Thanks all for the help