Apple Hard Disk 20 repair: A success story

[Dog Cow]

Could someone make a sound recording for me of what the HD 20 sounds like when it's not connected to a Mac and is then powered on, and then a second recording of what it sounds like when it is connected to a Mac, powered on, and the Mac boots from it?

 

[Dog Cow]

So, since November I now have another HD 20, this one is working so I have something to compare against.* I also finished a somewhat comprehensive writeup on the history and development of the Apple Hard Disk 20 in my Mac 512K blog. There's quite a bit more to explore, but that blog article was getting too long, so other topics will have to wait for later blog articles. The long and short of it is that I still have some more avenues to explore with Nisha and René, and also the PCPC WSI replacement board that transformed the HD20 into a SCSI device.

* Fun fact, not mentioned in the article: several of the applications on the HD 20 are infected with nVir!

 

[james_w]

That article looks very extensive and interesting - instapapered! thank you

 

[Dog Cow]

Could someone make a sound recording for me of what the HD 20 sounds like when it's not connected to a Mac and is then powered on, and then a second recording of what it sounds like when it is connected to a Mac, powered on, and the Mac boots from it?

Turns out this is a useful diagnostic, as you can hear the difference between my faulty HD20 powering on and my working HD20 powering on.

Stamped on the white Rodime label are the track (cylinder) locations of the two spare tables. Every drive I've seen has these at cylinders 204 and 104. In the case of my HD20, it could not find either spare table (they were corrupted) so it was failing the selftest.

Part of the selftest is the drive starts at cylinder 304 and seeks every 8 cylinders down to 0 or 8 to find the spare tables. They are always at head 0, sector 0.

I copied the spare table from my working HD20, using the HD Diag, but there is still something else that needs to be changed within the table for this drive to work.

 

[techknight]

Oooo hopefully you figure this one out, that would be awesome. Your deeper with the drive than I have ever been. 

I assume your talking about the P-List table? I dont know if the drive is new enough to have both a P_list or G_List. 

 

[Dog Cow]

It turns out that the documentation and source code for the Apple Widget, Nisha, and the Hard Disk 20 are all very closely related so I was reading the Widget documentation on bit-savers just now, and I've learned more about how the spare table works. The keyword is "f0783c1e"

On my non-working drive, I cannot reliably write to track 104-0-0, which is normally where the second spare table is located. So I think I'll need to put that spare table on a spare block! Imagine that.

There are 76 SpareBlocks on each Widget/Nisha/Hard Disk 20, spaced 256 blocks apart on a 10MB drive, 512 blocks apart on a 20MB drive...

This is what a spare table looks like, several examples.

When I get this all figured out I'll post a blog entry on it.

 

[Dog Cow]

Success!

Last night after several hours I got the HD20 to mount on the desktop and I can boot from it too. It was a long road getting here, and it did involve replacing the spare table at 104-00-00 to pass the self-test diagnostic. But I'm not out of the woods yet. There are still several unreliable tracks/sectors and as a temporary work-around I have the HD20 formatted as a 400K MFS volume!

Perverse, yes I know, but it's better than nothing for now (and shows that the HD20 can be MFS-formatted; but why wouldn't it? It's just a block device that shouldn't care what file system is on it)

The full story and run-down will be posted in the Mac 512K Blog

 

[Dog Cow]

I made a tweak to the MFS Volume Header this afternoon and doubled my storage, so now the HD20 looks like a 800K MFS volume. I'm probably the only one in the world with an HD20 formatted like it's an 800K MFS disk! It boots too.

But I can't format the drive using the Finder or any other application that I've used so far, because there's still unreadable tracks. Hand-building an MFS volume (which is what I'm doing via HD Diag) is so much easier than HFS.

 

[Alex]

So now, I have a perfectly working and mint looking HD20. I'm the happiest man on earth right now. I actually mended an old hard drive. I had previously tried to revive an old sony but that ended catastrophically...

Damn, this is just crazy stuff. So cool!

 

[Dog Cow]

I hand-built an HFS volume this weekend using HD Diag on my formerly non-working HD20. I manually typed in the MDB, alternate MDB, volume bitmap, catalog file, and extents file. This task was highly unpleasant but the result is that I now have the disk booting Finder 5.x and about 2 MB of applications and stuff copied from floppies on to it. There's still a crazy high number of unusable sectors, and I don't know if there is any practical fix. I've used the Format Track option in HD Diag dozens of times on these problem tracks, and while it works in some cases, I still can't get all of them reliably formatted.

Whenever the HD20 finds an IO error because of a bad block, I manually patch the volume bitmap (we're up to block 4 now) to mark those blocks as used. So I'm wasting disk space, but that's a small price to pay I think.

If I knew then what I know now I probably could have saved a fair amount of data from the drive, and probably have gotten it to mount in the Finder too. But it's too late now. I've blown away too many tracks using HD Diag.

 

[techknight]

I wonder if its because of the platter itself loosing magnetic permeability. Thoughts? 

 

[PB145B]

Some of the old style (pre IDE) drives I have opened have had oxidation on the platters. But others I have opened had some kind of residue on the platters. Not exactly sure what this “residue” came from though. One of the drives I opened that had this “residue” in it was actually a Rodime RO352 10MB drive very similar in design to the one in the HD20. I still have the drive and it does work but it has quite a few bad sectors. I guess it’s possible that if you removed the platters, cleaned them, reinstalled them and did a low-level format that it might bring a few of the blocks back. 

This is is probably a dumb question and it may have even already been discussed in this thread, but is low-level formatting possible on the HD20? Because I know on MFM/RLL drives that is usually enough to bring them back to perfect health, unless they have physical platter surface damage like I mentioned above.

 

[Gorgonops]

I suspect the residue you find on the platters is either bearing lubricant that slowly worked its way up the shaft and misted out or it's the result of rubber parts like the gaskets breaking down. I gather that the older Quantum 3.5 inch drives found in a lot of early-mid 1990-ish era Macs are prone to the latter.

Have to admit I'm iffy on the prospects of cleaning that stuff off. Finding the right solvent that can take it off without damaging the surface sounds like a tall order; remember, even a *tiny* scratch that you might make while trying to buff them off is likely to raise ridges on either side that'll be of similar size magnitude as the flying altitude of the drive heads. 

 

[techknight]

Back in the very early 2000s I got my hands on an Intel server which had Quantum 4.3GB U320 drives in them that had all seized. Taking them apart revealed some type of "goo" the heads were sitting in, in the parked position. 

Surprisingly I was able to clean up all the drives to the best of my ability and they all fired up and worked, Long enough to backup the data. and many many more years after that. Surprisingly. Culprit was a failed fan that cooled the drive chamber. 

 

[PB145B]

Back in the very early 2000s I got my hands on an Intel server which had Quantum 4.3GB U320 drives in them that had all seized. Taking them apart revealed some type of "goo" the heads were sitting in, in the parked position. 

Surprisingly I was able to clean up all the drives to the best of my ability and they all fired up and worked, Long enough to backup the data. and many many more years after that. Surprisingly. Culprit was a failed fan that cooled the drive chamber. 

Wow! How exactly did you clean it?

 

[Dog Cow]

I wonder if its because of the platter itself loosing magnetic permeability. Thoughts? 

Can't rule it out. It's interesting to note that with a couple of these tracks with errors, that I've examined, it's usually just a handful of sectors that are unreadable, and the rest are fine. This is the Rodime 552 unit that I removed the lid from and watched its operation. I now regret doing that, as it served absolutely no purpose towards restoring the drive, and in fact may have made the problem worse by introducing dust or other contaminants.

This is is probably a dumb question and it may have even already been discussed in this thread, but is low-level formatting possible on the HD20? Because I know on MFM/RLL drives that is usually enough to bring them back to perfect health, unless they have physical platter surface damage like I mentioned above.

Yeah, that's one of the many functions of HD Diag, and it's what I've done to restore a few tracks so far, but it's not working with every track. The Format_Track command lays down new headers on every sector of the track that is currently underneath the head. See Nisha Firmware Specification December 1984, page 32.

 

[PB145B]

This is the Rodime 552 unit that I removed the lid from and watched its operation. I now regret doing that, as it served absolutely no purpose towards restoring the drive, and in fact may have made the problem worse by introducing dust or other contaminants.

The data density is so low on these old drives that dust doesn’t even really bother them. I doubt that opening it even made a difference.

Yeah, that's one of the many functions of HD Diag, and it's what I've done to restore a few tracks so far, but it's not working with every track. The Format_Track command lays down new headers on every sector of the track that is currently underneath the head. See Nisha Firmware Specification December 1984, page 32.

Cool!

 

[Dog Cow]

As promised, here is Part 2 of my adventures in repairing the Hard Disk 20 (HD20). I show how to use HD Diag.

 

[Iesca]

Resurrecting a slightly old topic, but Dog Cow, is their a resource for downloading HD Diag? I found a copy of Scavenger Mac, which I've uploaded to the Garden for safe-keeping, but I recently acquired an HD 20 myself and may want to play around with HD Diag if lubrication doesn't work. Although it does spin up after having manually turned the spindle, it does not seem to want to initialize; I don't know if this is due to the stepper motor or bad blocks or what..

 

[JDW]

As promised, here is Part 2 of my adventures in repairing the Hard Disk 20 (HD20). I show how to use HD Diag.

Did you happen to make a detailed list of the electrolytic capacitors on the Rodime 552 controller PCB that should be replaced?  (Capacitance, rated voltage, Diameter, Height & pin spacing (2.5mm?))  

We know all too well how these fluid-filled caps on our motherboards, analog boards and PSU's need to be replaced.  These HDD controllers are no different.  The need to recap applies to keyboards and mice as well.   If you have not made a detailed list of those capacitors on your Rodime HDD controller board, I will tear apart my HD20 to glean that info, but I wanted to ask first, just in case you wrote that information down.  No use doing the same job if another among us has already done it.

Thanks.