• Please review our updated Terms and Rules here

Safe to open up an mfm hdd to try and fix it?

These drives do not have nearly as close tolerances as anything modern. I've had my KC-20B open several times and I still have no further bad blocks than the factory defect list of 2. Don't leave fingerprints or oil on anything, don't do it in a wood shop, and don't smoke while you're doing it and things will generally be fine.

Don't be blatantly stupid. These drives do not have a particulate filter. Do not operate the drive with the cover off, as the airflow is not sufficient towards the edges of the platters and the heads will "crash" so to speak.
If you go the glove route, get POWDER FREE gloves, non-latex preferably.

The KC-20B does not have a rubber bump stop, it is a metal peg affixed permanently to the lower half of the chassis and there is an adjustable set screw to limit outward motion. This screw should NEVER need adjustment during normal operation, if you do turn it DO NOT allow the heads to move further out, once they're off the surface of the disk it's over.

You should be able to rotate the stepper gently with the tip of your finger without even opening the drive, the other end of the shaft is visible from the bottom. You will feel it rotate in one direction and not in the other if it is against it's stop. One or two drops of machine oil down the shaft of the stepper is more than enough. These drives have an exercise mode, place a jumper on the block labelled "W1" (below drive select jumpers) and operate the drive for a few minutes. It should work the oil in on it's own. If the stepper doesn't move at all with this, you can manually rotate it a few times (do not do this too much as it can damage the heads or platters) to get it started.

The seek test for Kyocera KC-20B is to single step to the track 0 sensor, fast seek to cylinder 615, back to cylinder 10 and single step back to track 0. This tests the track 0 sensor and that the stepper does not miss any steps. If the drive reaches track 0 during the fast seek, the stepper is missing steps and is probably just stiff with low oil. If the drive never leaves track 0, the stepper is assumed to be damaged or failed. If the drive never reaches track 0 the sensor is assumed to be failed or out of adjustment.

There is a track 0 line on the "control" section of the ST412 interface (the larger, 34 pin edge connector). Attach a spare cable and jump the drive select that the disk is configured for. This should "always" select the drive and it's access lamp should stay illuminated when powered. With a scope attached to the track 0 line, you should see it strobe once or twice during seek test and then stay high at the end.

Hope this helps, sorry I didn't see this thread sooner. These are very reliable drives, despite being horribly slow (60ms+ access time!).

EDIT: The correct jumper is labelled "W1" for the butterfly seek self test, not TP1.
 
Last edited:
Get a cheap large clear bin from walmart, cut 2 holes on one side and duct tape on some rubber gloves, like dish washing gloves. And in the rear, or top lid, throw a 2" opening for a vaccum cleaner/shop vac hose. Viola, you have a tiny clean room for less then 10 bux.

Anywho, if you are just trying to dump what is on the drive, it will work long enough to do the job. Just don't expect it to be a long lasting drive afterward if you dont use a cleanroom.
this is actually a laminar flow hood, not a clean room - but good advice nonetheless!
 
These drives do not have nearly as close tolerances as anything modern. I've had my KC-20B open several times and I still have no further bad blocks than the factory defect list of 2. Don't leave fingerprints or oil on anything, don't do it in a wood shop, and don't smoke while you're doing it and things will generally be fine.

Don't be blatantly stupid. These drives do not have a particulate filter. Do not operate the drive with the cover off, as the airflow is not sufficient towards the edges of the platters and the heads will "crash" so to speak.
If you go the glove route, get POWDER FREE gloves, non-latex preferably.

The KC-20B does not have a rubber bump stop, it is a metal peg affixed permanently to the lower half of the chassis and there is an adjustable set screw to limit outward motion. This screw should NEVER need adjustment during normal operation, if you do turn it DO NOT allow the heads to move further out, once they're off the surface of the disk it's over.

You should be able to rotate the stepper gently with the tip of your finger without even opening the drive, the other end of the shaft is visible from the bottom. You will feel it rotate in one direction and not in the other if it is against it's stop. One or two drops of machine oil down the shaft of the stepper is more than enough. These drives have an exercise mode, place a jumper on the block labelled "W1" (below drive select jumpers) and operate the drive for a few minutes. It should work the oil in on it's own. If the stepper doesn't move at all with this, you can manually rotate it a few times (do not do this too much as it can damage the heads or platters) to get it started.

The seek test for Kyocera KC-20B is to single step to the track 0 sensor, fast seek to cylinder 615, back to cylinder 10 and single step back to track 0. This tests the track 0 sensor and that the stepper does not miss any steps. If the drive reaches track 0 during the fast seek, the stepper is missing steps and is probably just stiff with low oil. If the drive never leaves track 0, the stepper is assumed to be damaged or failed. If the drive never reaches track 0 the sensor is assumed to be failed or out of adjustment.

There is a track 0 line on the "control" section of the ST412 interface (the larger, 34 pin edge connector). Attach a spare cable and jump the drive select that the disk is configured for. This should "always" select the drive and it's access lamp should stay illuminated when powered. With a scope attached to the track 0 line, you should see it strobe once or twice during seek test and then stay high at the end.

Hope this helps, sorry I didn't see this thread sooner. These are very reliable drives, despite being horribly slow (60ms+ access time!).

EDIT: The correct jumper is labelled "W1" for the butterfly seek self test, not TP1.
Thank you for all this detailed info! I had set it aside but now I am curious to see if at least one of my drives acts like you describe
Cheers
 
Thank you for all this detailed info! I had set it aside but now I am curious to see if at least one of my drives acts like you describe
Cheers
Well one of my drives does the butterfly test just fine.

The other drive fails. It seems to bottom out hard and repeatedly at track zero like it is hammering against the stop. I assume this is a failed track zero sensor.

So probably just one of my drives has a shot at life.
 
Knocking like it's hitting the stop? Track zero sensor has failed or is out of adjustment. The drive can't get any worse than it is now - It will never ever work again without something getting done. Good to experiment on.
The one that passes the butterfly test (A full test takes about 2 minutes. You will hear it single step back to track zero and start again when it has completed a full round.) is at least in mechanically good health.

Since swapping controllers did nothing, I doubt there is an electronic problem stopping the other drive from working, but it may be a good idea to check the track 0 sensor if you want to try your hand at repairing it.

Please note before you continue that Kyocera explicitly states in their documentation to NEVER "mount" the drive with the PCB facing up. I am uncertain if this applies to operating the drive upside-down temporarily, but I have not tried. I cannot imagine anything particularly bad will happen, but let's stick to one problem at a time. Kyocera does list that it is acceptable for the drive to be on it's left or right side, though not the face plate end. For testing, it might be best to place the drive with the "Kyocera" label (the silver and blue one on the side) facing down and prop it up against a stack of books so you can probe the pins without accidentally tipping it over. If you have a spare control cable or a floppy cable WITHOUT a twist, it may be easier to do this with the drive in it's right side up configuration. A PC-2000 will do these tests for you, if you have one or access to one, though it is not required by any means.

With the drive's connector side facing you mechanism-side down, the "control" connector, the larger 34-pin edge connector, should be in the center. The notch should be to your left, towards the molex power connector.
The side facing you, on the component side, is the even pins. The underside is the odd pins, which are all grounds. The first pin on the far left, the edge with the notch, is pin number "2". My drive has this labelled but yours may not. The first pin after the notch is pin 6 (third from the left total).

The third pin from the left after the notch (fifth pin total) is pin 10, which is the output of the track zero sensor. The drive will NOT assert this signal without being selected. The jumper just behind the control connector, on mine labelled "TP1 S5 S1 S2 S3 S4" is the drive select jumper. Place the jumper on "S4" to make the drive respond to drive select four. Ensure a terminator resistor is in place to the left of the jumper block. Pin one is likewise to the far left, towards the molex connector. Pin 1 of the terminator resistor goes directly to the ground pin on the molex power connector. Without this in place the drive may respond erratically.
"Drive select four" is pin 32 on the control connector, the second from the far right edge. Ground this pin to "select" the drive.
Remove the butterfly seek jumper from "W1" for now. Connect the drive to power, and have a multimeter handy. When power is applied, the drive should slow step to track 0, fast seek twice, and slow seek back to track 0 before becoming ready. This should take at most 20 seconds.

With your multimeter or a 5 volt test lamp, probe pin 10 of the control connector. The lamp should be illuminated, indicating the drive is at track 0. If it is, move your probe to pin 22, which should be directly beneath the word "TP1," or 7 positions from the right edge of the connector. If the lamp is illuminated, the drive is indicating that it is ready. If it did not complete a seek or is not at track 0, it should NOT be ready.


Depending on your results, this is how I would continue:
*If you get "READY" but no "T0" then the drive is not conforming to the ST506/ST412 control interface for some reason. This should not be the case.

*If you get no "READY" and no "T0" then either the drive has NOT seeked to track 0 or the track 0 sensor is NOT active, implying that the drive did not complete the seek or that the sensor has failed or is out of adjustment. If you hear the drive banging the head stack against the end stop, it has completed the seek as far as it will ever go and the sensor has failed. I believe the sensor is internal to the stepper on these drives, sadly. At this point, lift the top cover on the drive and make sure the head is actually moving at all. These drives are actuated by a rotary band positioner, which is essentially a thin shim of metal being twisted to move the heads. Of course, metal fatigue will eventually set in and destroy this component or stretch it out with the tension on it. It may have fallen off or snapped or become so stretched the drive will no longer operate. With the power off, gently push the heads towards the inside track of the drive and replace the cover. You should not need the screws in place, but make sure it is somewhat lined up. Power the drive. You should hear it "ticking" as it seeks the heads back to track zero and then two "beeps" for the fast seeks. You probably will not hear the final ticks for the slow seek to 0. After about 20 seconds, power the drive back off and remove the lid again. The heads should be in their maximum outward position, all the way against their back stop. If this is the case, the mechanism and stepper are likely functioning. There should be ZERO discernible slack in the rotary band. With the stepper all the way at the full back position, push gently on the head stack and watch the stepper carefully. It should rotate immediately with any pressure - There should be little resistance in the swing arm but there should also be no play between it and and stepper. If there is, the band is worn out and the drive is junk without another one. You can contact a machine shop or try to make one yourself, if you like, if you do attempt either PLEASE post your results as I have many drives in need of them and I would be interested in repairing this problem myself in the future.
If the drive "slams" into it's stop at the end, there is a chance that the adjustable track 0 stop is incorrectly placed. This should ALWAYS be your very last check, as it is set at the factory and not easy to adjust correctly yourself - Unless someone has been inside the drive, it should never be out of adjustment. If everything else is in perfect shape to this point, you can ATTEMPT to thread the set screw out slightly to give the head stack slightly more actuation room. Press gently against the head stack to bring it to the point it is touching the stop, and keep light pressure against it as you adjust the stop by unscrewing the set screw. Adjust in VERY small increments and retest the drive with the lid on for different results. You should not need more than two full turns, ever. As you are adjusting, ensure that the head lifters do not come to close to the edge of the platters! If they fall off, it is extremely difficult to get them back on without damaging either the heads or the platters. If they get too close to the edge, about 2mm or so is "the limit" for most drives, there is a chance that the airflow is not great enough to lift the heads, or that the heads will fall off on their own, as they do have some side to side play in their springs, though not much. Even without this, there is also the possibility that the heads become positioned over an area of the drive never meant to hold data, which will make track 0 appear as having "bad blocks" which would make the drive useless in DOS/Windows and I'm sure some other formats as well. It is best to avoid this. After the drive "successfully" passes a power on seek test, you can reassemble it and try the butterfly seek and interfacing with a computer again.

*If you get "T0" but no "READY" then the sensor is working but the controller has detected some other condition preventing it from becoming READY. This will require further diagnosis. Feel free to reply for this, because this would make this post ridiculously long.

*If you get "READY" and "T0" then the drive thinks it is working. The failed seek test is likely due to a sticking stepper motor. Try lubricating the oil (one or two drops of thin oil and running the exercise mode should fix a stuck stepper). If the oiling does not work, ohm out the stepper and compare to the other drive. If needed, I can ohm out mine as well, but the spec plate on my stepper shows that each coil should be 37.5 ohms. If the stepper is open, you can attempt replacing it, the part number is SANYO 101-4901-0142 with 37.5 ohm coils and 0.9 degrees per step. If you are replacing the stepper anyways, it may be interesting to experiment and replace it with a different unit. Sanyo made MANY steppers for hard disks during this time, of various specs. A common way that drive manufacturers made their drives faster at this time was to replace the stepper with one of lower impedance. SANYO 103-4902-0740 is 22 ohms with the same stepping angle. This is the stepper used on the Kalok Octagon I (KL-320, KL-330, KL-341, KL-343). It is the same physical dimensions and has the same connector. An even lower impedance stepper is SANYO 103-4902-12T1 which is 4.3 ohms and still same stepping angle with the same connector. This is the stepper used on the Kalok Octagon II (KL-380, KL-3100, KL-3120, and the Xebec versions of each).

The other drive that passes the test, what does it do attached to a computer? Does it respond to control signals from the machine at all? Programs such as SpeedStor, Spinrite, can they seek the disk or does nothing happen at all? There's a chance that both of these drives could be revived.
 
I apologize in retrospect for the massive wall of text that turned out to be. I tried to format it for better readability but I realized about 3/4 of the way through just how long it was getting and gave up. If you need clarity, feel free to ask. Again my apologies.
 
I'm going to focus on the drive the spins up correctly, and ignore for now the drive that seems to have a problem with power up.

"Place the jumper on "S4" to make the drive respond to drive select four." - roger that.

"Drive select four" is pin 32 on the control connector, the second from the far right edge. Ground this pin to "select" the drive." - roger that.

"With your multimeter or a 5 volt test lamp, probe pin 10 of the control connector. The lamp should be illuminated, indicating the drive is at track 0. If it is, move your probe to pin 22, which should be directly beneath the word "TP1," or 7 positions from the right edge of the connector. If the lamp is illuminated, the drive is indicating that it is ready. If it did not complete a seek or is not at track 0, it should NOT be ready."

just to confirm, pin 10 = 5Vish implies drive is at track0, and pin 22 = 5Vish implies ready.
 
running drive with electronics down seems to make a difference!

I realize now that to test any output signals from the drive, since they are open collector outputs, you need to pull these high. So I have rigged a 1k resistor to +5 on my scope probe/DVM.

Testing as you suggest, I get the following results.

LED flashing
Pin10 = 5V
Pin22 = 5V

..and it certainly does not do this:
"When power is applied, the drive should slow step to track 0, fast seek twice, and slow seek back to track 0 before becoming ready. This should take at most 20 seconds."

Hard to say what it is exactly doing, but maybe

- LED off
- fast stepping sound
- slow stepping sound, maybe bottoming out?
- 2 fast seeks
- slow stepping sound, maybe bottoming out?
- flashing LED


Pin 10 remains high during the entire test, which seems odd. Track 0 should be toggling low I would think.
 
Last edited:
My apologies, both signals are active low! Looks like I'm a bit rusty.
Does the LED appear to flash with any specific pattern? This drive is equipped with a microcontroller. If it's not just on/off flashing then it could be reporting an error code on it's own, which would be helpful.

Track 0 should pulse low and then stay low (active) until the drive is instructed to move.

Here is a video (timestamped) of a KC-20B doing it's seek test, so you can hear and more easily visualize what is happening:
The first "buzzing" sound is just the stepper driver initializing. It is not part of the seek test, technically, but I suppose I should have mentioned it. My bad.
 
OMG big progress!
So, while testing with the 'good' drive I noticed that sometimes when I power cycle it, it actually succeeds and gets ready!
must be some flakiness in the power supply, maybe a recap needed.

Anyhow, I dug out my Z171, plugged in the WD-XT GEN card, connected the drive, power cycled till I thought it sounded like it was "ready".

booted the Z171 to dos, and I am now running the WDXTGEN formatter!!!! the drive is ticking away!
 
thank you for all your help!

and the results are in...

21309440 bytes total disk space
122880 bytes in bad sectors
21186560 bytes available on disk


so this is great! to get my Z171 to work with an external HDD, I had to make an ISA bus adapter, purchase a used untested WD XT-GEN controller, and see if my old drives still worked.

and now I know!

interestingly, as I RE-FORMAT the drive, the # of dead blocks is increasing. So clearly this is not a reliable drive anymore.

still, satisfying to solve the problem and get a working system.
In my specific case, I also made an XT-CF card and of course that is awesome compared to any HDD.
 
Last edited:
trying to debug my "non working" KC-20B...

So, this drive spins up ok, and the heads actuate just fine. But what it seems to do is step the heads out to try and find track 0, but it bumps up against the hard stop repeatedly like it is trying to find track 0 and never does. It eventually just stops.

quote from earlier in this thread: "I believe the sensor is internal to the stepper on these drives, sadly."
I see no obvious sensor for track 0. It can't be the stepper as it only has the standard 4 wires coming off of it that drive the stepper. No logic.

The controller board has 3 connectors to the body of the drive.
1) a stepper connector
2) a connector for driving the drive motor itself
3) a flex cable connector that goes to the read/write arm/heads.

I really see no way to detect track 0 unless track (-1) is some special pattern written on the drive that the heads are meant to detect.

Any suggestions at all as to how this particular MFM drive is meant to find track 0?

Note: when I look at the top platten, I can see a serious defect on one spot near where I think track 0 should be. It is as if the magnetic coating has flaked off for about 1 sq. mm.
 
In terms of the drive gaining bad sectors AND having odd power-on results, these problems are likely linked. Hard disks of this era are almost entirely analog so any noise or fluctuation will essentially stop them in their tracks. Make sure the drive has a good ground and a reasonably stable power supply - if you use the ground tab to ground the drive to the chassis of the machine with the controller, you can use an external power supply to run the drive no problem. I suggest something "modern" with a bunch of power behind it - Even these small 3.5" drives have a pretty high spinup current of like 3 amps on the 12 volt rail. Most of the cheapy external bricks don't do so well with that and sag heavily under load.

You may also choose to run SpeedStor or SpinRite, which can detect bad blocks much more reliably than FORMAT or the WD LLF program. In many cases, they can even be recovered with a reasonable degree of success. And of course, SpeedStor will allow you to do several interesting seek tests to make sure that the drive is in mechanically good shape.
If you plan to run the disk in a machine, SpinRite can actually do some interesting and pretty useful things as well, like undelete and the sector recovery mentioned earlier, *and* that it can change the interleve of the drive to better match the performance of your machine. Eek a couple extra kilobits per second out of it, without losing any data at that, since it will format just one track at a time.

There is no hard-coded information on the platters of these disks. I believe the sensor *is* internal to the stepper, as several other drives from this era with track 0 sensors also use the same stepper with no external sensor. My guess has always been that rather than a dead shunt or a logic signal, it is a switch which makes a little bit more resistance on one winding to the chassis ground to indicate track 0, though I have nothing particularly to back this up with.

With the lid off, the platters should be a pretty even orange, typically reduced in the very center ring (cylinder 617, the landing zone). It will look slightly yellowish green where the oxide coating is worn away. If that is the case, that drive is probably reduced to parts only. This is caused by not parking the drive (it does not park itself) and the head sliders gradually removing the material from the surface over time. With a bad cylinder 0, I do not believe MS-DOS will format the drive successfully. There may be a utility to bypass this, but that is typically an indication of a poorly cared for disk that has failed or is failing.
 
Spinrite is quite cool.
Interleave of 5 nearly quadrupled speed!
Lets see how it does with LLF..

thanks again for all your help famicomaster2.
 
No problem. It's nice to help people get these things working again - Very few of them are truly irreparable! Glad you didn't give up on it.
 
No problem. It's nice to help people get these things working again - Very few of them are truly irreparable! Glad you didn't give up on it.

well, thinking about this non-functioning KC-20B....

I separated the stepper from the housing and the read/write armature, so I could see how it reacts on power up.

The controller steps the stepper through 360degrees, looking for "track 0". It shuts down as soon as it gets a full 360 degrees.
If your assertion is correct then there is something inside the stepper that modifies the current flowing through the stepper to mark track 0. I imagine that it is some kind of mechanical switch on the axle of the stepper. There are only 4 wires on this stepper.. 2 for each coil.

Sanyo-Denki type 103-4901-0142

Have you ever taken a stepper apart to observe the track 0 sensor?

This stepper does not appear to be something that can be opened up. I've also read that opening a stepper pretty much ruins the stepper.
 
The track 0 sensor is external to the stepper motor. It's usually an optical thing, but not always. Your drive appears to have a mechanical stop, after which it appears that track 0 is located on the disk, perhaps on a servo surface or via an embedded servo. Modern 3.5" drives almost always use embedded servo, but I doubt that's the case with your early drive.

Looking at your drive, track 0 seems to be set by the mechanical stop adjustment. Sounds crude, but I seem to recall that some early systems did this with their floppy drives. Bang into the stop and step in a bit to hit track 0.
 
Last edited:
There is no servo surface on the Kyocera KC-20B, and the embedded servo technology was not even invented at this time.

A floppy drive has far fewer tracks per inch than even these 20MB hard disks - when you hit the end stop on a floppy drive it can throw the heads out of alignment, just as with a hard drive, but the head alignment is far less sensitive on a floppy drive than a hard disk. Ramming the heads against the end stop in this case will probably just end up with the heads exiting the platter surface and decapitate the drive.

There is no obvious way to open these steppers, though I have observed some stepping motors (Such as that which is present in the CMI 3412) which does have an internal switch and a cam for the track 0 sensor, hence my assumption.
The liklihood of getting the stepper installed correctly after removal is pretty low, the drive is probably a display piece now.
 
There is no obvious way to open these steppers, though I have observed some stepping motors (Such as that which is present in the CMI 3412) which does have an internal switch and a cam for the track 0 sensor, hence my assumption.

Well, believe what you want to, but I do observe that the end stop is adjustable. In any case, the drive is trash now, so it hardly matters.
 
Back
Top