It’s no big secret that I just love old hardware. In some senses, it comes down to nostalgia, and in other cases, it comes down to being able to play with the expensive hardware we could never afford for ourselves at the time. It’s also sometimes quite fascinating to think of how far the envelope was being pushed given the state of the art at the time.
Early in the month, I spotted a friend on OCAU who was giving away an old AT case with two full height SCSI drives. After asking him whether he was willing to post, and parting with a little cash to cover the postage, I managed to become the owner of two Fujitsu M2266SA drives, condition unknown.
The Fujitsu M2266SA is a 5.25″ full height 1.08Gb SCSI hard drives. I haven’t ever laid my hands on a 5.25″ full height drive in my past. The closest I got was a Quantum Bigfoot, which is a 5.25″ half-height drive, so there is definitely some novelty in it.
M 2 2 6 6 S A FUJITSU NO MORE PRODUCED Native| Translation ------+-----+-----+----- Form 5.25"/FH Cylinders 1658| | | Capacity form/unform 1079/ 1266 MB Heads 15| | | Seek time / track 14.5/ 4.0 ms Sector/track 85| | | Controller SCSI2 SINGLE-ENDED Precompensation Cache/Buffer 256 KB FIFO BUFFER Landing Zone Data transfer rate 3.050 MB/S int Bytes/Sector 512 4.800 MB/S ext SYNC Recording method RLL 1/7 operating | non-operating -------------+-------------- Supply voltage 5/12 V Temperature *C 5 45 | -40 60 Power: sleep W Humidity % 20 80 | 2 90 standby W Altitude km 3.000| 12.000 idle W Shock g 2 | 20 seek W Rotation RPM 3600 read/write 30.0 W Acoustic dBA 45 spin-up W ECC Bit 56 MTBF h 200000 Warranty Month Lift/Lock/Park YES Certificates
Already, we can see that the drive is 1079Mb formatted, and 1266Mb unformatted (which is really irrelevant to users as there’s no way to use it unformatted). The drive is a SCSI-2 drive, with a 256kB buffer. The recording method is a very “archaic” RLL 1/7. It consumes a monstrous 30W of power which is a lot considering it’s a 3600rpm drive with a 14.5ms seek time. It’s also decently loud at 45dBA.
The drives that I was getting sent had been weighed at 3.375kg a piece, which is about the same weight as over four Western Digital 4Tb Black drives. They were dated 1992 and 1993, which puts them at 22 and 21 years old respectively. However, I did find this page which shows one of the same model dated 1991. Apparently they were part of a bulk lot purchased at auction and have not been used since.
So, what was storage like in 1991? Well to my knowledge, my family was still using a Miniscribe 40Mb MFM hard drive right up past 1995 (mainly because we could). The machines that I have salvaged have only had, a Maxtor 3.5″ IDE hard drive with 120Mb in 1993, or the most lucky machines, may have seen even 211Mb around 1992.
As a result, a 1079Mb drive in 1991 would felt like having a 16Tb drive today, while everyone else whizzes by on 4Tb drives. That’s a massive gulf in terms of storage size!
How much would one of these have cost? A lot. In fact, even in 1994, according to this snippet from PC Mag through Google Books says that the drive itself cost US$1,295. At that time, judging from the emergence of the Seagate Wren series drives which had faster performance, larger capacities, lower prices and lower power consumption, it would have been on the “way down” for Fujitsu.
It is rather ambitious that they claim a 5-year manufacturer warranty for these drives, especially when the MTBF of 200,000 hours is noted. But I suppose it is enterprise grade …
For a better idea of what it cost when new, I had to go hunting. There was an old archived document from NASA which listed some computing hardware assets which gives us a much bigger figure:
That’s a whole US$4,560 for a hard drive. It makes even the most expensive of SSDs look cheap.
Say Hello to the Drives …
After an involuntary blood sacrifice by the original owner in extracting the drives from the case, and a few days of handling by Australia Post, I managed to get both drives safely within my hands. All (nearly) 7kg of drive. Never have I met such behemoths!
The drives themselves have face-plates intended to face outwards on a computer casing. It’s pretty much the same height as four modern CD-ROM drives. There are vented slots for airflow, as these drives put out a decent amount of heat, as well as a hole for an activity LED. The top drive is the older drive, with a proper 5mm LED, whereas the bottom drive seems to have done away with it for a surface mount LED instead.
From the top, the two drives look pretty much identical, with a button-shaped breather hole at the top. The cover is mostly fastened with Philips head screws as with many of the Fujitsu drives, along with a list of patents in the bottom right corner.
The older drive, on the left, is serial B032510 dated April 1992 with revision B3. The later drive, on the right, is serial B050464 dated January 1993 with revision B8. I wonder what the differences in revision are?
Already, on the rear, you can see some of the configuration ID jumpers, the SCSI 50-pin connection and the molex power connection. As is common with older drives, the molex is oriented upside down (bevels down), so I had to be very careful. The SCSI connector also has retainer clips, one of which was snapped off prior to receiving it on the older drive. There is also a grounding lug present.
The right side of the drive has a C/N sticker and an identification number sticker labelled M2266X/DE. The older drive has C/N A8RG24-A, and B0E5090084805. The newer drive has C/N A8VT2Z-A and B0E5090139146.
Notable is that the drive mounting is via rails which are affixed to the “tub” of the drive via vibration absorbing grommets. The mounting screw holes are different from modern mounting holes, especially from the side, where there are only two holes, however, mounting from the underside appears more traditional although rarely seen today.
The left side of the drive has nothing too special, but keeps the same grommet and rail type mounting system.
The underside of both drives houses the main logic PCB. It’s an amazing thing to behold, as it’s quite sophisticated on its own. Notable is the existence of several configuration jumper areas scattered across the board, and the unpopulated termination IC DIP slot near the SCSI connector. The two drives also have different markings on their EEPROM (older one marked 81202G, newer one marked 81202I).
Along the bottom are a number of power transistors and silicon chips which are mounted to a tab which uses the side drive rail (and by extension, the chassis) as a heat sink. The main PCB also has a number of Omron mechanical relays.
Despite the use of surface mount components throughout, there’s a few large capacitors and resistors which could easily be victim to a careless installer. There is even a fuse in case of termination power shorts.
Many of the chips appear to be Fujitsu ASICs which have been designed for the drive itself, with a Motorola MC68HC000FN12 12.5Mhz CPU controlling the SCSI part of the drive, and an Intel N80C196KB12 12Mhz microcontroller (which is probably an 80196). There are also components from NEC, Motorola(?), Hitachi, TDK, Texas Instruments and Silicon Systems (now TI). Many of them have date codes in 1992, with copyrights dating back to 1986, 1988. There’s also a package made by potting something in some conformal compound, which is pretty interesting. There’s also a 15Mhz and 73.3104Mhz (very unusual) crystal oscillator.
The opposite side of the board is, surprisingly, fully populated with parts too. However, this is where the drives start to have some differences – for example, the older drive has Toshiba SRAM (100ns), Fujitsu SRAM (45ns) and Mosel FPRAM (70ns), whereas the later one uses Sony SRAM (70ns), Fujitsu SRAM (45ns) and Fujitsu FPRAM (80ns).
The underside also seems to house some Analog Devices components as well, and some “greenwiring” – hand-soldered jumper wires to correct for PCB errors or impossible to route connections. One of the free-standing capacitors seems to have been changed too.
That’s not all the circuitry there is external to the platter chamber, however. If we remove the plastic front faceplate, we see that it’s been built with serious shielding in mind.
However, under this plate, is one more PCB.
The older revision, on top, has a full 5mm LED which the bottom one has replaced with a surface mount LED. The top one also has some further greenwiring going on. Interestingly on both, there is an adjustment trimpot – I wonder what that adjusts? Given that this board connects between the heads and the logic board, it might be quite an interesting adjustment.
The white wire that runs under the label marked E is likely the voice coil motor cable for positioning the heads.
The rear side seems to share the same sort of ICs, with slightly different passives and other components.
Once the PCB is removed, we are left with the connector that is mounted to the flexible flat cable leading to the heads inside the platter chamber.
Interestingly, once the main PCB is removed, even the spindle motor seems to have changed between revisions – the older one has a Nidec, while the newer one is a Matsushita.
Initial Power-Up and Run
Given these drives have seen no action for many years, it was uncertain what would happen when power was applied. Needless to say, a drive of this calibre has a very impressive noise, so I did my best with my Zoom H1 Handy Recorder to capture it:
It’s a very lovely noise – very deep and mechanical at its heart. Unfortunately, it’s a noise that bodes poorly for the 1992 drive, as it fails to calibrate and retries several times before giving up and I finally shut it down.
Common to both drives is the relay-based soft-starting procedure, in fact, it’s similar sounding to the DC resistance-based traction control systems used in electric multiple units where resistances are progressively switched out of the circuit as the motor gains speed. This manifests itself as “timed” clicking which sort of kicks in and “boosts” the speed at a particular stage of spinning up.
Under Linux, the 1992 drive was detected as:
[0:0:4:0] disk FUJITSU NOT READY 0020 /dev/sdb
The later 1993 drive was correctly detected:
[5:0:4:0] disk FUJITSU M2266S-512 002C /dev/sdb
Hmm. That’s a start nonetheless. It doesn’t mean the 1993 drive is working perfectly though, although it’s a good sign.
The drives themselves are very much a “historical” item, and a very expensive and rarely seen one at that. Full height 5.25″ drives are things which I have never handled before, and the sheer weight and heft of it left a very lasting impression. The noise it had, while certainly not of the same musical calibre as the MFM stepper drives that preceded it, is still of a deep and mechanical nature that’s rather impressive.
Join me in Part 2 (coming soon), as we work it hard and do some “peeking” under the hood.