In another opportune moment, I had the chance to play around with a SyQuest SQ5110 88Mb removable hard disk system as part of a little data recovery challenge pitched to me by (one particular) Robert at UNSW.

This particular drive actually dates back quite a bit (roughly 1991 as far as I can tell), and I had never had the opportunity to use one in my youth, but the SyQuest name was legendary amongst some “older” computing folk, and the chance to get my hands on it and actually use one, even if it was for a bit of a job, was really exciting. SyQuest themselves have been defunct for a very long time now, and removable hard disk cartridges have been out of fashion for a long time as well owing to the difficulties in keeping a clean dust-free disk environment, and precision alignment in denser media.

But before Zip drives were introduced in 1994, there were several removable media solutions – namely the Bernoulli Box (another Iomega invention), the SyQuest removable hard disks and the floptical drive. The people looking for speed and value really only had one place to go – the SyQuest as the Flopticals were small (21Mb) and unreliable, and the Bernoulli Box was using “floppy” media which was pretty much outclassed by “hard” media.

By using familiar hard-disk platter technology, they had all the benefits of hard disks in terms of access time and speed. By having the platters removable (similar to earlier mainframe disk packs), they could supply cheap platters enclosed in cartridges for capacity expansion while keeping the expensive mechanisms and heads inside the drive – thus one set of expensive heads would service many platters. And by having a decent capacity, beginning with 5Mb (and through time releasing several products up to 4.7Gb) when internal hard disks weren’t much bigger, they were really attractive.

One would think that such a system would be impossible to realize – and I was personally skeptical about this at first, because I knew that the head flying heights are so small that any contamination of the surface could lead to a head crash and loss of data or a damaged head. But it seems that it actually had a reputation quite the opposite – being reputed to be very speedy as well as reliable. People swore by their SyQuest drives, and by virtue of its SCSI connection, some even booted from SyQuest drives.

Ultimately, SyQuest didn’t end up surviving, going bankrupt in 1998 and in a twist of irony, was bought out by its arch rival – Iomega! According to the Wikipedia article, quality issues and reliability issues contributed to the death of SyQuest – and so did the Zip Disk, which was ultimately cheaper and more ubiquitous. Once the manufacturer of hardware goes under, support does as well, and so there are limited amounts of information still available online – some utilities are now no longer available to my knowledge.

It would seem that the legacy of the Syquest cartridge system went to inspire Iomega to produce their Jaz drive and Castlewood to produce the Orb drive, all of which went to see limited levels of adoption for high speed, large capacity, cheap storage.

Given that it’s now 2013, what am I doing with an 88Mb Syquest drive and cartridge set?

Well, lets just say that buried somewhere deep inside the University of New South Wales, lies a few (four, so far) old Syquest 88Mb cartridges (SQ800, 5.25″ Alta Series), and a Syquest 88Mb drive (SQ5110) in an external SCSI enclosure, and having been discovered in 2013, has some data which may be worth recovering. The challenge was put to me to recover as much as I could – then I could potentially own the drive and the disks. I couldn’t pass this challenge up. Some sites did have some discouraging news though – and these drives are beginning to be quite rare.

Getting the Parts Together

So, how hard can it be to make a Syquest co-operate? Well, logically, one would think it’s easy but with a few subtle twists. The first challenge was to find an older single-ended SCSI Controller which will be able to talk to the Syquest. While I did have a few SCSI controllers of my own, I wasn’t too sure about using them just yet, so I left the challenge dormant for a few months until …

… Electrical Engineering came to the rescue. Back again in their rubbish (aka salvage) pile, I spot an older machine with a few expansion cards. From the back, a DB-25 port emerged, easily mistakable for a parallel printer port. But I took a closer look and saw the familiar SCSI logo – bingo! Better still, it was a decent adapter – one from Adaptec. Everyone knows that if you want to be serious about SCSI, you run Adaptec.

A closer look at the system revealed the SCSI card appeared to be bundled with a Sony CD burner, and was connected by an internal 50-pin ribbon (also worth salvaging). The card itself was a AVA-29xx series PCI controller with active termination – so no need to worry about using resistor packs (and having to find them, plug them in and hoping that you don’t bend them) at least on the controller side. This was definitely a key find!

Next was a DB-25 (from the card) to Centronics-50 cable (to the external case housing the Syquest drive). Another happy coincidence was that the previous lot of stuff I got donated on OCAU contained an old HP Scanner cable – for a HP SCSI scanner. And it was just the right type!

The final piece of the puzzle was the terminator for the back of the Syquest external case – and luckily enough, this piece was already installed. We were pretty much set – minus a semi-modern computer which we have easy access to.

The Drive and Media

At the beginning of the article, I pictured the drive – which I will now show again:

The drive is a SQ5510 88Mb drive inside a Polaris enclosure, as you can see. There is a cartridge loaded into the drive at the moment (and it’s busy working hard)! Looking at the drive, you see two LED indicators – this allows the drive to signal its activity, as well as its status in case of errors. Despite this being an old device, the error code reference does suggest this is quite an intelligent microprocessor controlled device with blink codes for spindle speed problems, run-out problems, servo problems and power problems.

Just in the centre is the eject button, and to the right is the load/eject lever. The cartridge is seated at the moment and the lever is on the left, indicating it is loaded (as does the LED lights). Pushing the eject button lifts the lever slightly from its loaded position. If the drive is not busy and is not locked, it will begin to spin down whilst blinking the orange indicator. Once that ceases, you can pull the lever to the right eject position, where the cartridge will be ejected. Removing the cartridge results in a spring loaded plastic dust flap closing over the opening – quite similar to how early caddy based CD-ROMs operated.

All the media came nicely enclosed in a “fold together” style plastic library case (of the sort where they used plastic shells with air inside, not solid plastic). The original insert was quite colourful – after all, what’s inside isn’t nearly half as exciting … it’s a disk … but it comes with many cautionary warnings about how one should treat such a fragile piece of equipment.

You can see they also had some indents in the case to absorb pressure and shock. The disks themselves were brown translucent solid plastic, screwed together, with a rotating write-protection system in the corner. They were labelled on their front, with a barcode label as well. The orange arrow tells you which way to insert it.

Looking at the bottom of the cartridge, you can see the spindle interface which is a flat metal plate with screw-ends where the spindle interface would be clamped onto the disk, as well as a hole with some “springy metal feelers” which would centre the disk on the drive spindle. The top right of the image shows the cut-out where a sliding door which “rotates around” opens up to allow the head of the drive to be loaded onto the media. There are also some cut-outs in the plastic shell of the disk so as not to obstruct the head.

The Install and Power Up

In a rather unusual move, we decided that we should attempt the recovery, but at UNSW rather than in the comfort of my own house. No problems – after all, I had bought in all the above bits and they were already collected at UNSW. Robert had a few spare boxes which weren’t doing much and were pretty much ready to be used for recovery.

The first step was to see that we had a working machine. Luck of the draw meant the first machine we picked just so happened to be working (aside from needing the CMOS configuration redone since the battery had failed). The machine itself was an old Pentium 4 with Windows 2000 Professional. What was important was that it had a PCI slot – ready for the SCSI card.

We slipped the side panel off and got the card in. Plugged the power back in and that was it. We had a SCSI bus operating after Windows 2000’s PnP process had completed. Part of the reason it was so easy was due to the popularity of Adaptec SCSI cards, especially the one I salvaged which was already seen to have drivers included in the Windows 98 boot floppy. Quite an achievement indeed for your hardware to be that important that they merit inclusion in a boot floppy where space is at a premium. The inclusion of active termination means we don’t have to worry about termination at the controller end.

The drive itself was already in a Polaris branded external SCSI enclosure. The drive itself was detected as a Syquest SQ5110 drive – from this, we can tell that it is an early 88Mb model. In fact, this particular model was quite annoying for those with older 44Mb cartridges as it would be able to read, but not write to, those older cartridges. The backlash led to the introduction of its successor, the SQ5110C where the C, I assume, stood for Compatible – which could read and write to both sorts of cartridges.

We had no idea whether the drive or the power supply were operational – Robert carefully plugged it in and flipped the switch. A quiet whirr of the fan in the enclosure, no smokes, no sparks. Good. Now time to load a cartridge. unpack it from its protective case, orient it in the right direction and push.

After pushing the cartridge into the slot, the load lever was jutting out in an intermediate position and nothing happened. Giving it a slight helping hand towards the loaded position gave a satisfying mechanical clunk as the drive engaged the disk. Then, the familiar whine as the disk was being spun up, click – the head loads onto the disk, and a little while later, the green LED flashes. The disk is ready, idle and with no errors. Wow. We’ve made it this far.

To eject the disk, one has to push the only button and wait while the disk spins down. The orange LED flashes to remind you that it is still spinning down. Once there are no LEDs, you grab the load/eject lever and move it to the right, thus ejecting the cartridge. The mechanical feel is quite rigid, with very little flex in the lever, drive tolerances and cartridges too.

Having ascertained the drive was healthy, we had to power down to connect it to the SCSI bus (not hot plugging capable). Once connected, Windows had no difficulty seeing the drive and showing it as a removable disk drive. No drivers were required since it conforms to the SCSI standards, however, there were some utilities that sent special commands to the drive for low level formatting and diagnostics, but these are no longer available.

The Recovery and Aftermath

It may seem that we’re on the ‘home stretch’ now, but it wasn’t quite as easy as one would have liked. The first problem was that having put a disk in, the system would hang on accessing the disk. After a while, sometimes Windows would complain of not being formatted. Was it low level format damage? Servo damage? Was the data really lost?

So out comes my favourite utility – WinHex, to do some direct disk analysis. Opening the disk with WinHex took a long time, and resulted in many sector read errors which were not encouraging. But it did turn up a very important clue – it recognised an Apple Partition Table and HFS partition. No wonder we couldn’t list its contents on the PC!

But hold on, there’s still sector errors. This (in all likelihood) wouldn’t end well.

Ultimately, I had to pull out Knoppix and dd_rescue to image the disk and then analyse the image. Only under Linux was the system persistent enough to wait for the drive to recover the data – ultimately we spent anywhere from 24-48 hours recovering whatever was readable from 88Mb. Of the two cartridges completed, the data lost was about 1-4Mb, but it took a lot of time. With the raw images written back to a USB key, it was possible to mount the copy, repair the filesystem and inspect the files.

As of writing this, two of the four cartridges have recovered with mostly success. The other two cartridges are more problematic and seem to have some servo issues or media defects which complicate recovery. The SyQuest drives are very persistent at reading the media, and can take a very long time to execute read commands – it is believed that under Windows, it times out and “loses control” of the drive too quickly, meaning that recovery is very futile. Under Linux with dd_rescue, we have seen some success, but these cartridges are even more difficult due to a strange quirk in the SyQuests operation.

It seems that under many circumstances, these difficult cartridges will be so riddled with errors that if you leave it in the drive for about a minute before accessing them, they will come back with errors very quickly for every sector, without even attempting a read. If you boot Linux without a cartridge inserted, it may have difficulty with recognizing the physical media format. The best success has been had by booting with an inserted cartridge, ejecting the cartridge and re-inserting just prior to executing dd_rescue. Provided you don’t execute dd_rescue too soon, you can expect to see the drive work hard and it take anywhere up to half an hour to recover 1Mb from disk. But even then, sometimes it will not be able to complete the full read without the bus “losing” it and timing out, or the drive itself may lock into “giving up by default”.

So we’re still working out how (likely multiple images and gddrescue would come in handy) – but these cartridges will require a bit more manual work to recover – to eject/powercycle the system to recover it from the giving up mode.

[As an aside, sorry for the quality of the images – this post was made while the recovery project was still underway, and so the images were made with a cameraphone, and the drive was being used so couldn’t be disturbed or further inspected.]