A few weeks ago a Seagate ST-225 drive more or less accidentally landed at the OS/2 Museum. The drive is nearly 30 years old (manufactured in late 1986) and rather interestingly, this particular unit comes with a label declaring it as “Manufactured for IBM Corporation, Armonk, New York”.
Seagate ST-225 20 MB hard disk
It is a standard 20 MB PC/AT disk, CMOS type 2. It uses the ST-506 MFM interface. 5¼” half-height, loud, 3,600 RPM, 5 Mbit/s transfer rate, two platters, stepper motor, no auto-park… essentially first generation PC hard disk technology, but refined for mass production. There is some chance the drive started its life in an IBM XT Model 286.
The drive’s functionality was unknown and doubtful but could not be immediately tested because I was missing one little piece—a pair of ST-506 style cables. After sourcing the cables I hooked up the drive to a 25 MHz 386 board through a true blue IBM fixed disk and floppy controller (68X3815, Western Digital chips). Was it going to work?
The drive spun up but did not boot. It did however get as far as printing a boot error from the MBR. So not completely dead. An examination with Norton Diskedit showed that the drive still contained data (PC DOS 3.3 and a few applications), but lots of sectors were marginal (mostly but not entirely unreadable) and the 4th head could not read anything whatsoever.
The ST-225 lid
Of course the trouble with MFM hard disks is that the data path is entirely analog and data written by one controller is not necessarily readable by another controller. And I had no idea what controller this drive was previously used with. Clearly one which used the same general format (otherwise nothing would be readable) but perhaps not quite compatible with mine.
Still, there was just enough readable data that with a bit of luck I could read a sufficient amount of the FATs and root directory to look around the drive. It had been last used around 1992 judging from the timestamps. With the 4th head delivering no data there wasn’t much that could be recovered, and the FAT and root directory sectors were flaky enough that even ‘dir c:’ only worked some of the time. The data was too far gone for SpinRite to do anything.
After a while I gave up on recovering the data from the drive, especially because I was sufficiently confident that there wasn’t anything uniquely interesting. I ran the Seagate format utility on the drive, changing the interleave from 3:1 to 2:1 in the process.
After the format I did a quick verify, simply reading back all sectors. Zero bad sectors were reported! The drive also came with an empty defect list sticker, but this was still surprising. Next I ran much a more aggressive verification in the Seagate format utility which tries 9 different data patterns (and takes far longer). After good 8 hours, the results were in: zero defects.
I installed DOS and a few utilities on the disk and the next day ran SpinRite for a good measure. Again, no problems found. Amazing.
The Seagate ST-225 has a bit of a reputation for uncommonly high reliability. And that is something which can be judged after 30+ years. Zero defects after 30 years is impressive, and that’s a fairly dumb drive with no sector remapping or anything of that sort. It was just that well made.
Seagate ST-225 label
This is especially interesting (ironic?) when the ST-225 is compared with the original CMI (Computer Memories Inc) 20 MB drive built into the early PC/AT units. Those drives failed so often that CMI went out of business and IBM’s reputation took a serious hit.
It would be tempting to say “they don’t make them like that anymore” about the ST-225, but that’s something we’ll perhaps be able to say in another 30 years.
Could this even be reformatted with RLL and interleave 1:1 with a better HD controller adapter?
1:1 interleave definitely, but that’s a function of the controller/system. With my current setup 2:1 is the optimal interleave.
As for RLL, that depends on who you ask. Seagate says no (of course), but at least some people successfully used it. With a drive made in 1986 it’s possible that there’s no real difference from the corresponding RLL model. I don’t have an RLL controller at hand so can’t test. My guess is that it would work.
Did you try SpinRite before you low leveled it?
In my experience the ST-225 didn’t do well when formatted as an RLL drive, but it may have worked better with the SMART RLL controllers that had their own BIOS.
The 286 and 386 PC’s we built that used an RLL controller that relied on the computers BIOS drive geometry table. They were quite a bit less expensive than the smart controllers with a BIOS that allowed the user to enter their own values.
We used to copy the motherboard BIOS and modify the drive geometry table to fit the newer drives we had. This was both to change heads and cylinder counts and change from the 17 sectors for MFM to the 27 sectors for RLL. We had to regenerate the ROM checksum after making the change. OnTrack had software with some drives that allowed us to access drives with a mismatched table, but I think we had issues with that with Novell servers and disk utilities so we avoided it.
Eventually the BIOS was harder to change and the price of the smart RLL controllers had dropped and we used the smart controllers.
I did try SpinRite before reformatting, but it wouldn’t work because it basically couldn’t even read the root directory on the drive.
ST-238R is the ST-225 set for RLL usage. Same exact drive mechanism, better QC: no bad sectors on the first couple of tracks and fewer weak spots on the rest of the drive. RLL took the open loop stepper design to its limits and made the drive susceptible to failure from heat or bearing wear. What I remember is some buyers would go the other way, purchasing the RLL qualified drive variant and placing it on a MFM controller for maximum reliability.
If your ST-225 works the same as mine did, the drive after reformatting will continue to work for a few years before failing again. There will a number of progressively shorter failure followed by reformat cycles until the drive finally fails for good.
Rich Shealer: Yes, older Ontrack drivers based in loading overlay from Config.sys had troubles with non aware OS disk drivers. These problems were fixed in newer Ontrack versions based in INT13H overlay loading via MBR, and updated disk drivers for these OSs (i remember drisk driver updates available for NT, OS/2 and Novell Netware). Ontrack itself comes with an updated Win3.1 FastDisk driver version for generic ST/IDE/ESDI controllers that is Ontrack overlay software aware.
What is Norton Diskedit and what can you do with it? Can you create a drive image with it?
It’s what the name suggests, a disk editor. I think you can create a drive image with it but that’s not really what it was intended for.
I had a few ST-225’s in various XT clones and a real 5150. Most required the data to be re written so that it stayed stable. They had issues with temperature change because the heads were attached to the stepper motor by a metal band that shrank or lengthened and the heads wouldn’t read or write to the exact spot the data was stored on the platters.
That’s good to know. Makes sense, the whole thing is a giant floppy drive in a box, a completely dumb device.
I guess that’s why SpinRite was popular back in the day.
The first hard drive I bought for my 286 clone in 1990 when I was 15. At some point, the drive stopped working on occasion, and I noticed the top of the hard drive was excessively hot. I bought a DC cooling fan from Radio Shack, hooked it up to a spare power connector and situated it so it blew over the top of the drive. Sure enough, the failures went away!
Heat is definitely one of the biggest enemies of electronics, together with flaky power supplies.
I had a different problem with a much newer Seagate drive (2GB size, I think; it was around 1996). It had a tendency to develop bad sectors (CRC errors), but rewriting the sector made it good again. After thinking about it I suspect the drive was either marginal in some way or perhaps power outages while accessing the disk might have caused it. It never looked like media errors, simply corrupted data written to the disk.
I recall with these drives that since they didn’t have a dedicated servo tracking platter or anything like that, that over time the track location on the disk would get out of sync with the track location that the heads would seek to, as mentioned previously. You just got in the habit of periodic backup/reformat/restore, or using SpinRite which would do a trackwise reformat and save you the hassle of the backup and restore. I didn’t have SpinRite back then, so… piles of floppies and a Saturday afternoon.
hi sir
good day
pls give us aqoute
hard disk drive(st225) quantity2
floopy disk drive (55GFR) q uantity 2
I won’t, but perhaps someone else might.
Free 20 mb seagate st 225 drive. S/NO 694250 I just need the data. Anyone have an old AT pc to hook it up to. I think the computer was an Everex
Free 20 mb seagate st 225 drive. S/NO 694250 I just need the data. Anyone have an old AT pc to hook it up to. I think the computer was an Everex
Mark: Different controller manufacturers encoded the data differently, so you need to either find out which controller you used, or try a bunch.
MFM controller – KT-102 –
Tested with Seagate ST225 and ST251
Can these older Seagate drives be powered up without any data/control cables to see if they spin up? I don’t recall the drives waiting for any “OK” signal from a controller before spinning … but wanted to make sure. Applying 12V should spin up the drive and applying 5V should power the voice coil for moving the heads.
I don’t see anything about requiring control signals for spin-up in the ST-225 OEM manual. The drive can draw up to 2.2 amps on the 12V rail when spinning up, so the power supply needs to be able to deal with that.
I’m 99.99999999999999999999999% sure they spin up without any controller attached.
I think I used some of those old drives as dummy loads for testing PSUs back in the days. Had a broken one which really used power and after a while the solder on some of the transistors started to melt, so could only be used for shorter PSU tests 😉
I wouldn’t try applying only one of the two voltages though.
I don’t know too much about this drive, but I have one from 1987 that works fantastic. Definitely a cool piece of history. 🙂
The ST225 was a very reliable drive unless you made one mistake, which probably explains why you had to do a low level format.
The mounting screws go directly into the main frame of the drive. When you bolt it down the frame warps, changing the alignment on the read/write heads. Reformat and all is golden — until you unbolt it and the warp reverses itself. Read errors galore — until you do a new low-level format in the new shape.
So: Never fasten more than two of the mounting screws.
Fascinating — I wouldn’t have thought of that. Makes sense though, the drive is fine as long as it stays in place.
I used two HDs MFM, using a OMTI RLL controller. The first was a CMI 21Mb, the second was a ST225. First thing was to use spinrite on the drives, locate any additional errors ( the CMI had an error on track 615 ), but both drives tested with only that one error. OH! The OMTI had built in error correction… so we had to turn that off and retest the drives. Still just the 1 error. I mapped out all the bad tracks anyway, and off to installing the OS. DOS 2.1, and XENIX x86. Both drives were NOT RLL certified, but ran well with RLL controller, having never found another bad sector in the 6 years I ran the system. The CMI was a 6426, which the guy who sold me the OMTI controller, swore that it had 640 cylinders, and that the OMTI was supported by XENIX. He was right on both counts.
I am sure, that Seagate, CDC, Maxtor all ran their own internal tests to certify a drive as RLL, and relabled MFM drives as such, like semi-conductor companies did ‘binning’ so as to increase the value of their inventory.
The manual for the OMTI did have a section on interleave, but no information on how well it would perform on a 10Mhz v20, I think that the Turbo XT had a 8Mhz bus speed, so it ran the bus like a AT 8Mhz, and I remember the manual saying that it could support a 1:1 interleave on even the fastest drives.
Windows 1.0.3 ran well, XENIX was fair, for K&R, but crash and burned with larger projects.
One last thing to note: ( I never knew this, but it explains a lot about how fast dual drive systems were/are )
From Byte Nov 1985:
“One last relevant feature of the AT is that its hard-disk controller supports overlapped (buffered) seeks. Also. the controller board supports one floppy-disk and one hard-disk data transfer at the same time. The fixed-disk interface is ST412. Over-
lapped seeks allow a system with two or more hard disks to overlap operations. This is done by sending multiple step (seek) pulses to a drive that then disconnects and does the stepping without the controller. The controller is then free to work with its remaining drivers).”
So installs, and disk copies would be sped up by this.
“When you bolt it down the frame warps, changing the alignment on the read/write heads.”
This may have been a problem on XTs, but ATs had a rail system, that used plastic rails, the rails had slots, not holes, so you could treat the steel sheets the frames with less stress, but the ST-2xx series had cast iron bodies. I do not think I could warp one of those with a hammer. They did get pretty hot, which is what raised our concern for thermal expansion. It would have been a big problem with Quantum, and also WesternDigital – also prone to having their seal strip being torn off.
XTs and clone XTs also had a slot system, so you had a hole and a slot, on one side.
So installing, you put a screw in the hole, and made it tight, while the slot the frame, you did not tighten as much so it could slip. ( that is what those paper washers were for that came with the cases. )
I did change the orientation of the XT from horizontal to vertical, and it had no effect on the hard disks (MFM -> RLL ).
“the whole thing is a giant floppy drive in a box,”
Old 8″ floppies were FM, 5.25 were MFM, but the heads on a floppy drive actually pushed into the plastic, no floating, and the bottom head was closer to the edge of the drives, but your point is well taken. Bernoulli and Iomega disk drives… temp storage at best.
“they didn’t have a dedicated servo tracking platter or anything like that,”
The drives that did not have a dedicated platter/head for servo tracking, had servo tracking on all the tracks. Then some engineer had the bright idea to have servo marks on a piece of plastic on the head mechanism. Wanna know how that went? The plastic would outgas, or shatter, or fall loose. Times 100 for data recovery.
ST-225 had servo tracking? No, it didn’t, that’s why it had trouble with things getting out of alignment.