Ambient temperature dependant fault (old MFM hard drive)

[helmut]

Hi !

I'm looking for advice because i'm trying to troubleshoot an old 10mb MFM hard drive which has a weird fault.
Device is a BASF 6188R from 1985. It comes from a vintage PC i got this summer and it never worked (the status led gives me a constant 4 flash red light code which means bad spindle motor speed according to the documentation).
Since a few days, temperatures dropped significantly in my area and it seems it made the drive go back to life again !!! Well, it boots and stops flashing this error code for 10 minutes then goes back to error mode again. If i let it cool down for 30 minutes, it works again etc.
My intuition is that there's a "partial" fault on one of the components which control the motor speed. I took the control board apart and looked for components exposed to high temperatures, there's a L298 dual H bridge and two BD436 power transistors mounted on a heatsink. In this hot area, there's also an electrolytic cap across the 12v line which measures 12v after a few seconds at power on.
Anyway, i'm no electronics guru and before replacing things i wondered if there's something obvious i should check and if this kind of temperature-dependant fault rings a bell to you. Of course i have no schematics and taking the control board apart is quite a pain, doing it too many times is risking damaging those fragile flex cables 
Thank you !

[RoGeorge]

I would start by replacing all the old electrolytic capacitors with new ones.

[helmut]

I guess it's a good idea, the one which is touching the heatsink must've had quite a tough life 
Thank you !

[aqibi2000]

Good old tantrum capacitors

[helmut]

Good old tantrum capacitors

Do you mean that this temperature-related failure suggests a tantalum cap failure ? Because i thought i'd replace only the aluminium electrolytics. Usually when i encounter a tantalum cap problem, it fails short and just stays short.

[brabus]

Caps are a good point to start.

To be honest, it could be anything. Changing the caps may compensate for some other issue, which will remain unknown.

I would start with some cold spray: let the drive reach the fault condition and then start cooling down some specific parts of the board or some specific components, until it comes back to life.
You will at least isolate the issue and prepare the field for a proper component-level diagnosis. May be a faulty bonding wire inside an IC as well.

[tunk]

IIRC most (all?) MFM drives do not have dynamic head positioning.
When it heats up, the heads may drift out of alignment. The error
code suggests otherwise though.

[helmut]

I would start with some cold spray: let the drive reach the fault condition and then start cooling down some specific parts of the board or some specific components, until it comes back to life.
You will at least isolate the issue and prepare the field for a proper component-level diagnosis. May be a faulty bonding wire inside an IC as well.

I would love to try this but the "sandwiched" design makes it impossible to access the components while the motor is connected  The only thing i could try was to cool down with a bit of ice the heatsink on which the L298 and the two BD436 are attached. No dice.

Btw, here's a picture of the logic board and the motor board :

[helmut]

IIRC most (all?) MFM drives do not have dynamic head positioning.
When it heats up, the heads may drift out of alignment. The error
code suggests otherwise though.

Interesting ! Actually the exact error from the manual is "Motor speed outside +- 1 %".
When it's cold the drive really works well, i could format it without any errors or weird sounds.
I wonder if a bad spindle motor could cause this issue. Or a bad bearing ... 

[shakalnokturn]

Electrolytics is a good starting point. I'm always weary of the blue Philips axials ones.
A bad motor or bearing would usually mean more noise (they're already loud when normal).
Another thought is if there's an "index" plulse used for the spindle servo based on a magnet and inductive sensor.
Magnets weaken a little with heat, maybe check on distance adjustment if available.

[Gyro]

Dare I mention the grey multiturn speed adjustment pot?

Early MFM drives didn't have crystal referenced spindle speed synchronisation. The whole 'system' motor board, spindle bearings, magnets etc may have drifted slightly so that it is marginal when cold and out of tolerance when warm.

Of course, tweaking the speed adjustment could have unforeseen results (although as the controller seems to recognise when the speed is outside its acceptable range so ought to disable writes). There should to be a spindle speed test point somewhere on the board [Edit, or one phase of the motor drive] that you can check against a frequency counter.

[helmut]

Electrolytics is a good starting point. I'm always weary of the blue Philips axials ones.
A bad motor or bearing would usually mean more noise (they're already loud when normal).
Another thought is if there's an "index" plulse used for the spindle servo based on a magnet and inductive sensor.
Magnets weaken a little with heat, maybe check on distance adjustment if available.

Ah ! I was wondering if those philips caps are known for being bad or good quality, i will most definitely replace them then.
The motor isn't particularly noisy, except maybe when it stops after power off (i hear like something isn't very well lubricated) but compared to my other mfm drives it's pretty silent.
The index pulse thing is very interesting. There's actually a sensor with adjustable distance which is also used as a brake to slow down the rotation after power off. Originaly i thought the problem was that it was too tight against the motor so i pushed it a little bit away. I will try the other way around to check your theory about magnets and heat !
Thanks !

[Gyro]

Oh, happy days! Taut band stepper motor head positioner, spindle brake to stop the heads bouncing too much in the landing phase. 

That's a fancy one - it's got a brake on the stepper motor to stop the heads from being moved when the platter is stationary too.

[fzabkar]

ISTM that the L298 H-bridge drives the stepper motor for headstack positioning, whereas the spindle motor is driven by the L293B.

https://www.st.com/resource/en/datasheet/l293e.pdf
https://www.sparkfun.com/datasheets/Robotics/L298_H_Bridge.pdf

Speed would be crystal controlled, albeit indirectly via the MAB8048H uC. The uC would sense the spindle speed and control it by pulsing the phase drivers at the appropriate frequency.

Could we see this manual you refer to?

[Gyro]

Speed would be crystal controlled, albeit indirectly via the MAB8048H uC. The uC would sense the spindle speed and control it by pulsing the phase drivers at the appropriate frequency.

Could we see this manual you refer to?

Not guaranteed, I have a Seagate spindle motor and motor controller board from that era which runs in isolation. The spindle motor free runs (within it's own analogue control loop), fine adjusted by a multi-turn trimpot. The only signal to the main board was a hall effect index sensor monitoring the motor hub (on a flying lead to the main board).

Agreed, the manual would be very helpful - and interesting.

[fzabkar]

The manual …

https://www.mirrorservice.org/sites/www.bitsavers.org/pdf/basf/80307-060_BASF_6188_Fixed_Disk_Mar84.pdf

[fzabkar]

Not guaranteed, I have a Seagate spindle motor and motor controller board from that era which runs in isolation. The spindle motor free runs (within it's own analogue control loop), fine adjusted by a multi-turn trimpot. The only signal to the main board was a hall effect index sensor monitoring the motor hub (on a flying lead to the main board).

I see a trimmer and a counter. Maybe the bottom PCB has the rest of the VCO

[helmut]

Dare I mention the grey multiturn speed adjustment pot?

Early MFM drives didn't have crystal referenced spindle speed synchronisation. The whole 'system' motor board, spindle bearings, magnets etc may have drifted slightly so that it is marginal when cold and out of tolerance when warm.

SUCCESS !!!

Actually when i was trying to figure out the motor control pcb, my guess was that it was related to the front panel LED. I guess i got confused with the tracks and thought it was just some kind of brightness control, which is quite stupid.
Then i read your message and realized there's an access hole on the back plate of the hard drive !



Then i hooked it up to a PSU, let the drive heat up til i got the flashing red light and started gently tweaking the screw. After a quarter of turn the blinking stopped !!!

Then i re-installed the SCSI/MFM Xebec bridge board :



And ... would you look at that !!! :



Thanks for helping me !!! Such an easy fix, i can't believe it.
The reason i wanted this drive to work so bad is because this machine (Bull Micral 30, first french IBM PC compatible, historical piece of hardware !) has a very dumb bios, the hard drive configuration is hard coded. I tried various alternatives, like an XTIde card, but the BIOS int13h is set in stone and at POST is seems to handle BIOS extensions in a very non standard way. Even the keyboard controller is weird : same pinout as an XT but with a different baud rate ...
Anyway, this was the last fix for this machine (i already refurbished the PSU) and now it works perfectly. I'm so happy !
Thanks again !

[shakalnokturn]

Phew! Saved!
Did you re-cap all the same?

Any Thomson TO16 or Goupil in your collection?

[Gyro]

Glad it worked.

Edit: Ideally you would measure and set it to exactly 3600RPM but if the LED starts blinking when it's out of spec then maybe you can centre it by eye, making sure that there's enough range that it still works when cold.

[helmut]

Phew! Saved!
Did you re-cap all the same?

Any Thomson TO16 or Goupil in your collection?

Nope, i didn't ! I don't have any spare axial caps but i think i'll do it later !
And it's my first French PC ! I'd LOVE to find a Goupil G5, they look so cool !
Do you collect these too ?

[shakalnokturn]

I don't really go to the trouble of retro-computing collecting on top of all the rest.
I do have a number of old machines around though: Thomson MO5, TO7, TO8, Amstrad CPC464/6128, PPC640, Amiga 500, Atari 512/1024, IBM RS6000, white Mac's...