• Please review our updated Terms and Rules here

Computer Memories Inc hard disk collection (Incomplete)

Does that matter? I haven't worked on a servo motor drive but it looks for DC servo motors is driven by a bipolar DC voltage which isn't that different from a voice coil. I don't see the power amp in the schematic. Is that off board? Looks like there is an op amp chain from the encoder that you can inject an offset into which should shift the head position like I did with voice coil.
That much I don't know. I don't deal with the electronics side of thing, but I do know that a small rotary electric motor is substantially different from a voice coil deflection system. I would imagine it is driven by an H bridge or a dedicated motor controller IC somewhere on the board, but I don't know. These drives use servo bursts on the disk to center the heads over tracks, and an embedded track 0. This is in addition to the optical feedback unit.
 
Is it possible to run open loop?

Sorry, I had forgotten the details. Commercial recovery services actually built external motor controllers that would plug onto the bowl, and their own read chain
re. embedded servos from another comment. They also used commercial spin stands from companies like Guzik https://www.guzik.com/products/head-and-media-disk-drive-test/spinstands/
Earlier servo controlled actuators used a separate platter for the servo tracks. Embedded came later. Most drives do not have optical coarse position sensors,
the info all comes from the platter.
I went down a research rathole late last year.
Lots of fun evolution during the 90s described in papers on http://bitsavers.org/pdf/santaClaraUniversity/IIST/ and I acquired a bunch of internal
drive vendor documentation around then as well that I uploaded under the vendor's directories.
 
Hi, what do you mean "No stiction on the heads or platters"? These drives use oxide media and twinarm heads, they are not susceptible to stiction like metallic media drives.

Regardless of media or heads, spindles and bearings fail, and rubber degrades. In this case, both the platters and head arms move freely on their bearings. Not so for other CMI drives I have worked on. Spindle stiction can affect nearly any hard drive of this age, especially if stored improperly, no?

What data do you intend to recover from a 30MB hard drive from 1984? If you are doing it "for fun," there is nothing of interest, and the likelihood of recovering valid information from a drive like this is exceptionally low. Unless you have the exact original controller it was paired with (Not "another WD," the specific card), it's contents will never be legible.

This drive is out of a custom NS32016-based system and contains an unusual copy of early BSD Genix. It was operational as recently as a couple of years ago. While many of the files were able to be extracted while the drive was still running, critically, vmunix and other binaries were not, nor was a binary dump of the drive taken - meaning a working system cannot be assembled without additional data recovery, or cobbling together from other sources online.

It's not "just another WD", nor is it just from some old IBM PC with a copy of MS-DOS 3.30 and Alley Cat on it. And yes, I have the original controller, a NatSemi-rebranded SMS FWD8006 Multibus device (BLC-8232), with full schematics. (Scans coming.)

This behavior is related to a failure to locate the embedded track zero. It's possible that the glass interrupter wheel inside the optical feedback unit has been cracked, shattered, or otherwise misaligned with the shaft of the DC servo motor. You can attempt to open this and check, but without the original alignment equipment, it is unlikely that you will get it correct enough to actually read data - Though the drive may at least become operable after this, the heads probably will not line up with existing track data on the platters anyways.

Thankfully, the optical encoder disc is still in tact, and the optoelectronics are still functional. There is a factory-made groove showing the original alignment between shaft and encoder that implies neither has slipped, or if so, it's only by a step or two at most.

You can attempt to probe the head output test point and compare it with the waveforms posted here earlier, but I don't recall seeing one for the embedded track 0 position. Your drive probably has media rot and cannot locate it's track 0 marks on the disk, rendering it unusable.

The waveforms shown do seem to match, and with @Al Kossow 's helpfully uploaded information, I understand the principle of operation and the expected approach for head alignment. At the moment I suspect a failed DG211, but I need more time in the lab to confirm, I was rushing through my last measurements yesterday before dinner. (I'm more used to 405x chips of this era failing, but all analog switch ICs from early 80s process nodes are susceptible to the same sorts of failures.)

I expect I'll be able to get @djg 's device to work - I have been working with his devices for years now and have good experiences with them.

Best of luck.

I have a good head start, thanks to the work of others. @djg or @Al Kossow 's work has been fundamental in recovering and emulating thousands of hard drives and machines at large, and hundreds of my own clients' drives and systems. Thank you for your time.

Does that matter? I haven't worked on a servo motor drive but it looks for DC servo motors is driven by a bipolar DC voltage which isn't that different from a voice coil. I don't see the power amp in the schematic. Is that off board? Looks like there is an op amp chain from the encoder that you can inject an offset into which should shift the head position like I did with voice coil.

Hey David! The power amp is indeed on a second board. I can draw mine up later today if it'd help, but it's about what you'd expect.

You're right about injecting an offset. After the initial track zero hunt fails, the drive locks the head, asserts write fault, and refuses to obey any further commands. I can override the head lock (remove or disable the solenoid) and work out a means to inject an appropriate control signal. Happy to take this to your mailing list for further work!

Sorry, I had forgotten the details. Commercial recovery services actually built external motor controllers that would plug onto the bowl, and their own read chain
re. embedded servos from another comment. They also used commercial spin stands from companies like Guzik https://www.guzik.com/products/head-and-media-disk-drive-test/spinstands/
Earlier servo controlled actuators used a separate platter for the servo tracks. Embedded came later. Most drives do not have optical coarse position sensors,
the info all comes from the platter.
I went down a research rathole late last year.
Lots of fun evolution during the 90s described in papers on http://bitsavers.org/pdf/santaClaraUniversity/IIST/ and I acquired a bunch of internal
drive vendor documentation around then as well that I uploaded under the vendor's directories.

Thanks Al! Sadly the local shop that had a full rig for this junked their gear a few years back; they weren't getting as much call for it and needed the space back. I wish I'd known, I would have rescued it! A local musician looking to recover data from a synthesizer's hard drive ended up having more success with David's design than with that group.
 
Last edited:
Hey David! The power amp is indeed on a second board. I can draw mine up later today if it'd help, but it's about what you'd expect.
Not needed. Was looking at the schematics on bitsavers and didn't think an op amp was going to drive a servo motor. Hopefully you will fix the underlying problem and don't have to go down any the the paths discussed. If you do you can create a thread where you like to discuss further.

A local musician looking to recover data from a synthesizer's hard drive ended up having more success with David's design than with that group.
Was that the McLeyvier? They were quite resourceful on that drive recovery.
Spindle stiction can affect nearly any hard drive of this age, especially if stored improperly, no?
Odd terminology. By spindle stiction are you referring to what I call dried grease/bad bearing where the spindle doesn't want to turn freely but its not due to the heads? Stiction I'm familiar with is an issue with certain drives where the heads stick to the disk and prevent the drive from spinning up. Like previously stated I think it showed up with the smoother plated media. Remember it being a thing for a while until it was understood and addressed. If you free the heads the drive will the operate fine until they stick again when sitting.
 
Not needed. Was looking at the schematics on bitsavers and didn't think an op amp was going to drive a servo motor. Hopefully you will fix the underlying problem and don't have to go down any the the paths discussed. If you do you can create a thread where you like to discuss further.
I just noticed he has a version of the motor control posted, too.

I'll keep debugging and let you know where I end up - if it turns out I need to go down the road of manual stepping, I'll give you a shout.

Was that the McLeyvier? They were quite resourceful on that drive recovery.
Yup!

Odd terminology. By spindle stiction are you referring to what I call dried grease/bad bearing where the spindle doesn't want to turn freely but its not due to the heads? Stiction I'm familiar with is an issue with certain drives where the heads stick to the disk and prevent the drive from spinning up. Like previously stated I think it showed up with the smoother plated media. Remember it being a thing for a while until it was understood and addressed. If you free the heads the drive will the operate fine until they stick again when sitting.
Sorry about the terminology confusion - yes, the spindle issue, not the drives adhering to the platters. In this case the drive is unaffected.
 
Sorry, I had forgotten the details. Commercial recovery services actually built external motor controllers that would plug onto the bowl, and their own read chain
re. embedded servos from another comment. They also used commercial spin stands from companies like Guzik https://www.guzik.com/products/head-and-media-disk-drive-test/spinstands/
Yep, I've actually seen a few of those show up on eBay. I don't think I've ever seen a 5.25" one, though, nor one that claims to interact with oxide media.
Earlier servo controlled actuators used a separate platter for the servo tracks. Embedded came later. Most drives do not have optical coarse position sensors,
the info all comes from the platter.
This drive is motor actuated, not voice coil. Drives of this era were usually using stepper motors in open loop - The CMI CM6000 series is one of the most technologically advanced consumer hard drives of it's era, with thermal compensation, dual feedback closed loop positioning, and I believe a complete algorithmic seek curve. They use essentially a standard DC motor with an optical encoder on the back for coarse track-based positioning (Useful for very fast seeks), and on disk feedback in the form of an embedded track zero (Servo bursts at cylinder -1 to indicate to the microcontroller that it is at the outer limit of the disks) and more broadly a servo burst track alignment system. After the coarse position is reached, the heads can "hear" bursts of noise from the inner or outer directions. To my understanding, an analog circuit balances these two levels to center the heads on a given track. Because this information is on disk, this compensates for the thermal expansion of the disks and head stack, meaning their manufacturing process could have far greater tolerances in those components, a big cost saving for the time, which led IBM to contract them for the original AT.
I went down a research rathole late last year.
Lots of fun evolution during the 90s described in papers on http://bitsavers.org/pdf/santaClaraUniversity/IIST/ and I acquired a bunch of internal
drive vendor documentation around then as well that I uploaded under the vendor's directories.
Indeed, I have a rather sizeable collection of vintage hard disk drives spanning about 1979 to 1994. The engineering advancements and production techniques during that time are absolutely fascinating!
 
Regardless of media or heads, spindles and bearings fail, and rubber degrades. In this case, both the platters and head arms move freely on their bearings. Not so for other CMI drives I have worked on. Spindle stiction can affect nearly any hard drive of this age, especially if stored improperly, no?
Stiction is specifically an interaction between the heads and the media of the disk, which does not happen for oxide drives. Even if it did, it is not typically a visible symptom. The observation would be "I need to manually jolt the drive to start the spindle." This can also happen as a result of a weak permanent magnet motor (Not common in 5.25" drives), a failing spindle driver (very common) or dried grease in the spindle bearings (very common).
This drive is out of a custom NS32016-based system and contains an unusual copy of early BSD Genix. It was operational as recently as a couple of years ago. While many of the files were able to be extracted while the drive was still running, critically, vmunix and other binaries were not, nor was a binary dump of the drive taken - meaning a working system cannot be assembled without additional data recovery, or cobbling together from other sources online.

It's not "just another WD", nor is it just from some old IBM PC with a copy of MS-DOS 3.30 and Alley Cat on it. And yes, I have the original controller, a NatSemi-rebranded SMS FWD8006 Multibus device (BLC-8232), with full schematics. (Scans coming.)
I see, the drives I imaged at request were from a Convergent Technologies AWS workstation, specifically two drives in this very post, the cylinder limited CM5410C and CM5616C. Both drives showed the correct format, and one included some lost industrial software, which to my understanding was not very interesting, and the other drive has become misaligned with it's inner cylinders over time, and as such any data currently on the drive is illegible past about cylinder 12. The CM5000 uses completely open loop steppers. I think they might use embedded track 0, but that amount of automation would not really help in this situation anyways. After the failed reads, I wound up LLF'ing the drive on a PC and it is currently working, so it was not a fault of the drive.

I was making the assumption that this drive came from a PC or clone, like my CM6426es and CM6640es all had, which basically all contain utterly boring and useless information which is illegible without the exact original controller.
Since your controller is special, but the machine it is in is working, it may still be recoverable. I have found that sometimes even other examples of the exact same controller (DTC 5150XL for DTC 5150XL or WD1003WAH for WD1003WAH) will still be unable to read each other's drives due to analog timing differences. I have not seen any record of anyone manually adjusting these to remedy this.
Thankfully, the optical encoder disc is still intact, and the optoelectronics are still functional. There is a factory-made groove showing the original alignment between shaft and encoder that implies neither has slipped, or if so, it's only by a step or two at most.
These drives do not use a stepper motor, the actuator motor is actually DC, not a stepping motor. It may be off by some degrees, but there are no "steps" for this motor type.
The waveforms shown do seem to match, and with @Al Kossow 's helpfully uploaded information, I understand the principle of operation and the expected approach for head alignment. At the moment I suspect a failed DG211, but I need more time in the lab to confirm, I was rushing through my last measurements yesterday before dinner. (I'm more used to 405x chips of this era failing, but all analog switch ICs from early 80s process nodes are susceptible to the same sorts of failures.)
Interesting, best of luck.
Hey David! The power amp is indeed on a second board. I can draw mine up later today if it'd help, but it's about what you'd expect.
That board is for powering the spindle motor.
 
Like previously stated I think it showed up with the smoother plated media. Remember it being a thing for a while until it was understood and addressed. If you free the heads the drive will the operate fine until they stick again when sitting.
I have found many early drives with metallic media are horrible about stiction, but recently I was informed by another collector that another common method of failure that can appear as stiction can occur with improper storage of some models of drives - Apparently Epson-Seiko did not properly seal the inside of the spindle motor, and the oil can leak out and leave a film on the surface of the platters nearest it, which causes stiction-like symptoms, but no other ill effects. I'm told that if disassembled, it can be fixed by carefully wiping the oil from the platters with very fine cloth and high purity isopropyl alcohol, but I have not attempted this on either of my examples at this time.
 
Stiction is specifically an interaction between the heads and the media of the disk, which does not happen for oxide drives. Even if it did, it is not typically a visible symptom. The observation would be "I need to manually jolt the drive to start the spindle." This can also happen as a result of a weak permanent magnet motor (Not common in 5.25" drives), a failing spindle driver (very common) or dried grease in the spindle bearings (very common).

Forgive my memory: back in the 90s we called it "stickation" I guess, when it affected the Quantum drives in the SparcStation 1 and 2. It ended up in the comp.sys.sun FAQ:

14) Why did my Quantum 105 megabyte hard disk stop working?

This the now infamous Quantum drive "stickation" problem.

If the drive is allowed to cool down(even for a short period
of time) the drive lubricant will congeal and prevent the disks
platters from rotating.

Before you get a replacement, try lightly tapping
the drive to loosen the lubricant. If this does not work try
shaking and twisting the drive at the same time. One last
thing to try is to lift the system up a couple of inches
and drop it.
Perhaps everyone's been wrong since then, or maybe this issue was actually the heads sticking to the platters, and not spindle "drive lubricant?" Doesn't matter either way.

I see, the drives I imaged at request were from a Convergent Technologies AWS workstation, specifically two drives in this very post, the cylinder limited CM5410C and CM5616C.
Interesting. What did you use to image the drives?

These drives do not use a stepper motor, the actuator motor is actually DC, not a stepping motor. It may be off by some degrees, but there are no "steps" for this motor type.

That board is for powering the spindle motor.

Yes, and there is still a power amp for the actuator motor on that board, driven by the output of Z27, at least on this model. I can't speak for any other drive.
 
Paying this thread another visit since I've been looking at CMI junk again and, it's, well, interesting to see what it devolved into, tone-wise.

Hi, what do you mean "No stiction on the heads or platters"? These drives use oxide media and twinarm heads, they are not susceptible to stiction like metallic media drives.
Eminently incorrect - anecdotally, people have yanked heads right off the flexures on plain old iron oxide drives such as MiniScribe junk and ST-225s/251s, plus many others. Not uncommon, especially depending on the storage conditions. Whatever the mechanism of failure, the heads are stuck to the media, be it like gauge blocks and thus formally stiction or not. Anyway, besides the point.

What data do you intend to recover from a 30MB hard drive from 1984? If you are doing it "for fun," there is nothing of interest, and the likelihood of recovering valid information from a drive like this is exceptionally low. Unless you have the exact original controller it was paired with (Not "another WD," the specific card), it's contents will never be legible.
None of those assumptions were yours to make, if I'm being completely honest. Judging what someone else's data might be worth or not comes across as being rather rude.

This behavior is related to a failure to locate the embedded track zero. It's possible that the glass interrupter wheel inside the optical feedback unit has been cracked, shattered, or otherwise misaligned with the shaft of the DC servo motor. You can attempt to open this and check, but without the original alignment equipment, it is unlikely that you will get it correct enough to actually read data - Though the drive may at least become operable after this, the heads probably will not line up with existing track data on the platters anyways.
It'd be easy to check just by looking at the encoder waveforms, no HDA opening required. Which you suggest in the next paragraph break for a different circuit.

Signing off, raw imaging on a CMI6k is totally possible and not hard to do. Seek out, look at encoder, seek one encoder step, look for identifiable servo burst by looking at head waveform, seek in, one encoder step at at time. Anyone who is doing this kind of stuff (i.e. directly driving the seek mechanism, be it a stepper, DC motor or voice coil servomech) is clearly clued-in enough to figure that out by themselves and adapt to the case at hand. Saying "it can't be done like that because this is not a voice coil drive" is a bit myopic.
 
Forgive my memory: back in the 90s we called it "stickation" I guess, when it affected the Quantum drives in the SparcStation 1 and 2. It ended up in the comp.sys.sun FAQ:
Perhaps everyone's been wrong since then, or maybe this issue was actually the heads sticking to the platters, and not spindle "drive lubricant?" Doesn't matter either way.
Quantum's drives have rubber bumpers inside which become sticky as they age and can cause the heads to become stuck in place. The drive will spin but not seek as a result.
Interesting. What did you use to image the drives?
A borrowed pdp8 emulator, which was loaned to me by the gentleman who requested that I image the drives to begin with.
Yes, and there is still a power amp for the actuator motor on that board, driven by the output of Z27, at least on this model. I can't speak for any other drive.
I see, I don't remember the actuator motor connecting to that board, but it has been a while since I had to disassemble one of these CMI CM6000 drives.
Paying this thread another visit since I've been looking at CMI junk again and, it's, well, interesting to see what it devolved into, tone-wise.
I'm not sure what you mean.
Eminently incorrect - anecdotally, people have yanked heads right off the flexures on plain old iron oxide drives such as MiniScribe junk and ST-225s/251s, plus many others. Not uncommon, especially depending on the storage conditions. Whatever the mechanism of failure, the heads are stuck to the media, be it like gauge blocks and thus formally stiction or not. Anyway, besides the point.
Miniscribe offered plated metallic media as an option on all of the MS-III series (3006/3012/3212/3212PLUS/3425/3425PLUS/3438/3650/3675). The ST-225 was surely all iron oxide, but the later ST-251 and ST-277R are a lubricated oxide media, which I suppose can have this issue, but I've never heard of it personally.
None of those assumptions were yours to make, if I'm being completely honest. Judging what someone else's data might be worth or not comes across as being rather rude.
Hijacking my thread for data recovery is just as rude, I would argue. Additionally, the vast majority of people I have been asked to help recover data from these drives expect some type of USB adapter to read it under Windows 11 and get their dad's MP3s, as if such a thing were even possible.
Signing off, raw imaging on a CMI6k is totally possible and not hard to do. Seek out, look at encoder, seek one encoder step, look for identifiable servo burst by looking at head waveform, seek in, one encoder step at at time.
The value of doing this is highly debatable, and arguably entirely pointless. If the drive has lost it's embedded track 0 anyways, the liklihood of any valid data still being legible is extremely slim.
Saying "it can't be done like that because this is not a voice coil drive" is a bit myopic.
The suggestion I was denying was "You can't wire this up like a speaker to a DAC," because the drive directly given in the example was a Micropolis rotary voice coil drive recovered in this way. It is a fundamentally different process.
 
Hey folks! New member just signing in to say hello. I just thought I'd let everyone know that I saw this thread and have a cm 5619 and some other older drives. I I would be interested in selling them. If this isn't the proper way of extending this invitation I will try to follow the rules just reading this post so far has made me glad I joined and I'm interested in this kind of stuff but I'm not any kind of expert. Good day to all and I will monitor the post for any interest in my hardware thank you.
Quantum's drives have rubber bumpers inside which become sticky as they age and can cause the heads to become stuck in place. The drive will spin but not seek as a result.

A borrowed pdp8 emulator, which was loaned to me by the gentleman who requested that I image the drives to begin with.

I see, I don't remember the actuator motor connecting to that board, but it has been a while since I had to disassemble one of these CMI CM6000 drives.

I'm not sure what you mean.

Miniscribe offered plated metallic media as an option on all of the MS-III series (3006/3012/3212/3212PLUS/3425/3425PLUS/3438/3650/3675). The ST-225 was surely all iron oxide, but the later ST-251 and ST-277R are a lubricated oxide media, which I suppose can have this issue, but I've never heard of it personally.

Hijacking my thread for data recovery is just as rude, I would argue. Additionally, the vast majority of people I have been asked to help recover data from these drives expect some type of USB adapter to read it under Windows 11 and get their dad's MP3s, as if such a thing were even possible.

The value of doing this is highly debatable, and arguably entirely pointless. If the drive has lost it's embedded track 0 anyways, the liklihood of any valid data still being legible is extremely slim.

The suggestion I was denying was "You can't wire this up like a speaker to a DAC," because the drive directly given in the example was a Micropolis rotary voice coil drive recovered in this way. It is a fundamentally different process.
 
Hey folks! New member just signing in to say hello. I just thought I'd let everyone know that I saw this thread and have a cm 5619 and some other older drives. I I would be interested in selling them. If this isn't the proper way of extending this invitation I will try to follow the rules just reading this post so far has made me glad I joined and I'm interested in this kind of stuff but I'm not any kind of expert. Good day to all and I will monitor the post for any interest in my hardware thank you.
Sent you a direct message.
 
Hello,these pictures are a reply to famicomaster2 requesting to see what I have to offer. Still learning to navigate the forum but here it comes!
 

Attachments

  • IMG_20240602_154206269.jpg
    IMG_20240602_154206269.jpg
    1.2 MB · Views: 1
  • IMG_20240602_154216303.jpg
    IMG_20240602_154216303.jpg
    2.4 MB · Views: 1
  • IMG_20240602_154219152.jpg
    IMG_20240602_154219152.jpg
    2.6 MB · Views: 1
Hello,these pictures are a reply to famicomaster2 requesting to see what I have to offer. Still learning to navigate the forum but here it comes!
 

Attachments

  • IMG_20240602_154148672.jpg
    IMG_20240602_154148672.jpg
    1,012 KB · Views: 2
  • IMG_20240602_154155404.jpg
    IMG_20240602_154155404.jpg
    1.2 MB · Views: 2
  • IMG_20240602_154058905.jpg
    IMG_20240602_154058905.jpg
    3.4 MB · Views: 2
Hi kids! More pics
 

Attachments

  • IMG_20240606_163433789.jpg
    IMG_20240606_163433789.jpg
    2.7 MB · Views: 2
  • IMG_20240529_131448353_HDR.jpg
    IMG_20240529_131448353_HDR.jpg
    1.4 MB · Views: 2
  • IMG_20240529_131329814.jpg
    IMG_20240529_131329814.jpg
    2.4 MB · Views: 2
  • IMG_20240606_163419664_HDR.jpg
    IMG_20240606_163419664_HDR.jpg
    3.5 MB · Views: 2
  • IMG_20240606_163359723.jpg
    IMG_20240606_163359723.jpg
    2.5 MB · Views: 2

Attachments

  • IMG_20240529_133617992.jpg
    IMG_20240529_133617992.jpg
    1.3 MB · Views: 1
  • IMG_20240529_133507350.jpg
    IMG_20240529_133507350.jpg
    2.3 MB · Views: 1
  • IMG_20240529_133607648.jpg
    IMG_20240529_133607648.jpg
    2.2 MB · Views: 1
  • IMG_20240529_133128102.jpg
    IMG_20240529_133128102.jpg
    3.2 MB · Views: 1
Back
Top