The ZX Microdrive: Budget Data Storage, 1980s Style

An enduring memory for most who used the 8-bit home computers of the early 1980s is the use of cassette tapes for program storage. Only the extremely well-heeled could afford a disk drive, so if you didn’t fancy the idea of waiting an eternity for your code to load then you were out of luck. If you had a Sinclair Spectrum though, by 1983 you had another option in the form of the unique Sinclair ZX Microdrive.

This was a format developed in-house by Sinclair Research that was essentially a miniaturized version of the endless-loop tape carts which had appeared as 8-track Hi-Fi cartridges in the previous decade, and promised lightning fast load times of within a few seconds along with a relatively huge storage capacity of over 80 kB. Sinclair owners could take their place alongside the Big Boys of the home computer world, and they could do so without breaking the bank too much.

When 80 kB Of Storage Was A Big Deal

The ZX Microdrive unit in all its glory
The ZX Microdrive unit in all its glory

As a traveler returning from a continental hacker camp the UK government requires me to sit out two weeks quarantine due to the pandemic, which I’m doing as the guest of Claire, a friend of mine who also happens to be a fount of knowledge on and prolific collector of 8-bit Sinclair hardware and software. Idly chatting about the Microdrive, she bought out examples of not only some of the drives and software, but also the interface system and an original boxed Microdrive kit. This gives me the opportunity to examine and tear down the system, and provide a fascinating insight for readers into this most unusual of peripherals.

Picking up a Microdrive, it’s a unit about 80 mm by 90 mm by 50 mm weighing just under 200 g, and it follows the same Rich Dickinson styling cues as the original rubber-key Spectrum. On the front is an opening about 32 mm by 7 mm for the Microdrive cartridges, and on each side at the rear is a 14-way PCB edge connector for connecting to the Spectrum and daisy-chaining to another Microdrive via a custom serial bus over supplied ribbon cables and connectors. A maximum of eight drives could be connected in this way.

The ZX Interface 1 unit, which screwed to the bottom of the Spectrum and engaged with its expansion port.
The ZX Interface 1 unit, which screwed to the bottom of the Spectrum and engaged with its expansion port.

The Spectrum was an amazing machine for its price in the early 1980s, but this was achieved at the expense of very little in the way of built-in hardware interfaces beyond its video and cassette ports. On its rear was an edge connector that essentially exposed the Z80’s various buses, , leaving any further interfaces to be connected via expansion modules. A typical Spectrum owner would probably own a Kempston joystick adapter in this manner, to name the most obvious example. A Spectrum certainly didn’t come with a Microdrive connector, so the Microdrive had its own interface. The Sinclair ZX Interface 1 was a wedge-shaped unit that engaged with the edge connector on the Spectrum and screwed to the bottom of the computer, providing the Microdrive interface, an RS-232 serial port, a simple LAN interface that used 3.5 mm jack connectors, and a replication of the Sinclair edge connector into which further interfaces could be plugged. This interface contained a ROM that mapped itself over the Spectrum’s internal ROM, which as we noted when the prototype Spectrum emerged at the Centre for Computing History in Cambridge, was famously shipped unfinished and with some of its intended capabilities unimplemented.

Let The Teardown Commence!

It’s interesting to talk about the hardware, but of course, this is Hackaday. You don’t just want to see it, you want to see how it works. It’s time for a teardown, which we’ll start by opening up the Microdrive unit itself. Just like the Spectrum, the top of the unit is covered by a stuck-on black aluminium sheet bearing the iconic Spectrum logo, this has to be carefully teased away against the remaining force of 1980s adhesive to reveal the two screws securing the top half of the case. As with the Spectrum it’s difficult to do this without bending the aluminium, so some finesse is required.

Lifting the top half clear and disengaging the drive LED, the mechanism and board come into view. Immediately the seasoned reader will notice that resemblance to the much larger 8-track audio cartridges, and though this is not a derivative of that system it works in a very similar way. The mechanism itself is extremely simple, on the right is a microswitch to sense when a cartridge has its write-protect tab removed and on the left is the motor shaft with a capstan roller. At the business end of the cartridge is a tape head that looks very similar to that you might find in a cassette deck but with narrower tape guides.

There are two PCBs, on the back of the tape head is one holding a 24-pin custom ULA (Uncommitted Logic Array, in effect a 1970s precursor to CPLDs and thus FPGAs) that selects and operates the drive, and another attached to the bottom half of the case that holds the two interface connectors and motor switching electronics.

The cartridges are 43 mm by 7 mm by 30 mm, containing 5 metres of 1.9 mm self-lubricating magnetic tape in a continuous loop. I don’t blame Clare for not letting me lever open one of her vintage cartridges, but thankfully Wikipedia supplies us with a picture of a cartridge with the top off. Immediately the resemblance to the 8-track cartridges becomes apparent, the capstan roller may be to one side but the same loop of tape feeds back into the center of the single reel.

The ZX microdrive manual optimistically claims that each cartridge could hold 100 kB of data, but the reality was that they held about 85 kB rising to over 90 kB once they had stretched a little with some use. It’s fair to say that they weren’t the most reliable of media, with the tape eventually stretching to the point at which it could no longer be read and even the Sinclair manual advising backups of frequently used tapes.

The final component of the system to receive the teardown treatment is the Interface 1 itself. Unusually for a Sinclair product it doesn’t have any screws hidden under rubber feet, so aside from a delicate maneuver to detach the top of the case from the Spectrum edge connector it’s an easy teardown. Inside are three chips, a Texas Instruments ROM, a General Instrument ULA rather than the Ferranti item used in the Spectrum itself, and a bit of 74 logic. The ULA contains all the circuitry aside from discretes for driving the serial buses for RS-232, Microdrive, and Network. Sinclair ULAs are notorious for overheating and cooking themselves, and this is one of the most vulnerable. The interface here can’t have been used much because it hasn’t been fitted with a ULA heatsink and there are no heat marks in the case or around it.

A last word in this teardown should go to the manual, a characteristically well-written slim volume which gives an insight into the system and how it integrated into the BASIC interpreter. The networking ability is particularly fascinating as it was rare to see it in use, it relied on each Spectrum in the network issuing a command to assign itself a number upon start-up because there was no Flash or similar memory onboard. This would have been incorporated to target the schools market as a competitor to Acorn’s Econet, it’s not surprising to see why the BBC Micro won the Government supported schools contract instead of the Sinclair machine.

Why Are We Not All Using Little Tape Cartridges For Storage?

From 2020 it’s fascinating to go back and examine this forgotten piece of computing technology, and take a look at a world in which a 100 kB storage medium that loaded in about 8 seconds rather than the many minutes a cassette would have done was such a big deal. It’s baffling that Interface 1 doesn’t incorporate a parallel printer interface because looking at the complete Spectrum system, it’s not too difficult to see it becoming an adequate home office productivity computer for its day and certainly for its price. Sinclair did sell their own thermal printer, but even the most starry-eyed Sinclair enthusiast would find it difficult to claim the ZX printer as anything but a novelty.

The truth is that like everything Sinclair it was a victim of Sir Clive’s legendary cost-cutting and ingenious ability to create the impossible out of unexpected components. The Microdrive was completely developed in-house at Sinclair, but perhaps it was just too little, too unreliable, and too late. The first Apple Macintosh with its proper floppy drive arrived in early 1984 as a contemporary of the ZX Microdrive, and though the little cartridges found their way into Sinclair’s ill-fated 16-bit machine the QL, the result was a commercial flop. There would be a Spectrum with a 3-inch floppy from Amstrad once they had bought Sinclair’s assets, but by then the Sinclair micros were marketed purely as games machines. It’s been an interesting teardown, but perhaps it’s best to leave with happy memories of 1984.

I am heavily indebted to Claire for the use of the hardware featured here. In case you are wondering, the photographs above show a variety of different components both working and non-functional, in particular the Microdrive unit subjected to the full teardown is one that has failed. We prefer not to unnecessarily harm retrocomputing hardware here at Hackaday.

53 thoughts on “The ZX Microdrive: Budget Data Storage, 1980s Style

  1. I used a Sinclair QL for more than seven years, and I have to say that their microdrives weren’t as fragile as people say. I used them regularly for school work and more and never missed a file. But it’s true that presumably there were some “modern” units much more reliable than the original ones.

    About the Interface I, it was quite odd in electric design. The serial port was only a level adapter, and the RS-232 protocol was implemented in software. This resulted in problems when receiving data, because the machine had only the time of the stop bits to do whatever it needed to do with the data.

    Also, reading from the tape was funny: you have an IO port for that, but if you read from it, the Interface I will halt the processor until a complete byte has been read from the tape (which means that, if you forgot to turn on the tape motor, the computer will hang). This allows to easily synchronize the processor and the tape, and it was a must because the video generation produces “random” halts of the processor (up to 6 clock cycles) when accessing the second 16K memory block (the first one has the ROM, and the third and fourth ones have the extra memory of the 48K model), and since the microdrive buffer is precisely in that zone, it is not possible to just use a timed loop. If Sinclair had used an access method like the one used in the Inves Spectrum (which allowed both the video circuitry and the processor to access the video RAM without penalty, like in the Apple ][), the circuits of the Interface I could have been much simpler.

    1. The Spectrum has as much time as it wants to process the bytes received, provided the device at the other end correctly implements hardware flow control (which is *NOT* the case for some (all?) motherboard “SuperIO” chips. I wasted days debugging before realising that and switching to an old prolific USB-Serial adapter which to my amazement Just Worked worked first time)

    2. About the RS232. I got out 115k with error correction and 57k reliably without error correcting protocol with bit banging. The secret was to keep accepting up to 16 bytes after dropping the CTS. Original ROM code didn’t do that and couldn’t talk to “modern” UARTs.

    1. The wikipedia says 120 kbit / sec. About the specific protocol, I have no idea, but I know that it used an stereo tape head and the bits were stored “not aligned”. I don’t know how to explain it in english… the bit in one track starts in the middle of the bit of the opposite track.

      But doing a quick search I found this page, where an user attached an oscilloscope to the data signals, and it seems a FM modulation. But it is the QL ones, which are incompatible with the Spectrum ones.

        1. Yes, but remember that in the link talks about the microdrives for the Sinclair QL: although they were physically identical, the format used was incompatible, so a QL can’t read an Spectrum-formated cartridge and vice-versa.

          1. The bits are aligned. The bytes are interleaved between track 1 and 2. It’s biphase coding. A type of fm, commonly found on credit cards. The interface reassembles the bytes in hardware and the computer just reads bytes. Raw data rate is 80kbps per track or 160kbps for both. Performance was similar to floppies of the era.

        1. It’s not that strange that the Vdrive is slower. It has to do with the fact that the entire cartridge must be loaded to build the file tables, rather than just one sector and the VDrive has more sectors available than a microdrive cartridge. I even worked with someone who was building a virtual microdrive and I managed to build a faster format that was capable of storing 640K on a microdrive, which is the practical limit before the ROM must be reprogrammed. If you replace both the drive and the Interface1 decoder, then you can make the drive around 8 times faster before the INIR opcode used to transfer the bytes is running without delay.

          If just loading a single file, the VDrive can be faster as the file can be optimized, but it depends still on the whole virtual cartridge capacity.

    2. I don’t know, but there were a couple of articles back then about saturation recording. Audio tones were needed in order to use an existing cassette recorder. But if you modified one to directly access the tape heads, you coukd feed DC directly to them, and have a Schmitt trigger directly connected for playback. So it was just the serial signal feeding the tape head. You could get much faster speeds, and no fussing with playback levels.

      It was definitely used in the “mainframe” world. I always assumed that it was used in some of the small computer schemes like the “stringy floppy” but don’t know.

      Wikipedia has entries for both.

  2. I had a QL with 2 micro drives and it’s true, at least the QL ones were more reliable than people said. I had a ZX Spectrum, but no micro drives (though I did want them). The nearest I got was to do some cross development where I used a QL as a Text editor and transferred files over serial to a Spectrum which assembled the files (I was writing a printer driver for a ZX Spectrum PCB Designer program which would upscale and interpolate pixels to a 216ppi resolution so that tracks wouldn’t be jagged).

    It’s amazing what people did with fairly crude technology.

    1. I loved my QL and It’s bundled software, but got to hate it’s microdrives. After hours of work often getting the “BAD OR CHANGED MEDIUM” ERROR. Frustrating and non reliable.

      1. I wrote my Computer Science BSc dissertation on my 128Kb QL. Quill could only store about 4 pages and I never dared to overflow ram because it would start thrashing the micro drives and that error would pop up soon after.

        I was always concerned enough about Microdrive reliability to back up every editing session on two Microdrive tapes. However, after one long day of writing, I accidentally saved my new chapter with the old chapter’s name thus overwriting my previous day’s work.

        “That’s ok I thought, at least I have the backup!”; switched tapes, remembered that I should still save today’s work on the backup and promptly overwrote the previous day’s there too!

        Aaaaarggghhhh! It was a very late night so I guess it served me right.

        I still have my QL, I was able to actually successfully use a 30-35 year old microdrive cartridge to save and load on it about a year ago :-)

  3. This takes me back. I hacked everything back then. It took me a week to get Elite onto Microdrive, and break LensLok to always be the characters AA. Elite loaded in 9 seconds. Took over a minute on an Amiga! Basically a memory dump. I used an interrupt routine which monitored int 31(?) for kempston joystick fire. LensLok used interrupts for keyboard input, so I just had to squeeze in the code, and have it self-disable. Elite only left around 200 bytes unused, and the shadow mapping of Interface 1 ate my interrupts when I saved it with *”m”,1! Wow. 36 years ago.

    1. I kinda cheated…
      I had a Discovery Opus 1 3.5″ floppy on my Speccy, and I found, thanks to a happy accident on day when Elite crashed on loading, I could save Elite to floppy.. and it being the 128 version, no lenslok! result!

      ok I still had to save my Commander to tape tho…

  4. It’s fun to see that, ~40 years later, the floppy disks are quite dead, ans tapes still lives :)
    PS: I work with tape libraries, 18 drives each, able to deliver 350 MB/s per drive ;)

    1. The head is quite similar, if not the same (but in the schematics there is an “eraser head” which should be integrated), but the tape in a microdrive is narrower, so you would have to build a new tape guide.

    1. I remember that a PlusD + disk drive + power adapter cost, in 1990, about 33.900 pesetas (about 203 euros). With inflation it would be now 433 euros (512 dollars). It was about the same cost of the complete computer.

      1. As best I can remember in 1984 a C64 cost $200 and a 1541 cost $230 (actually more than the computer but not surprising considering it has its own 6502). Both of those plus a cheap TV was still less than 1/4 the price of an Apple II. The diskettes were $15 for a box of 10, but that decreased over the years.

        I read something that said by 1985 90% of U.S. C64 owners had disk drives.

  5. before i retired I used to use an excellent mechanical design and manufacturing company north of Cambridge (UK) that made all the machines to manufacture the cartridges for the Microdrives.

  6. I don’t think the lack of a centronics compatible parallel port was such a huge deal in the early 80s, serial printers were still common. Besides, Uncle Clive wanted to sell you the ZX FireHazard… umm Printer. Endless buzzing and ozone odors as it zapped it’s way down the silver coated paper.

    Microdrives, I had abysmal luck with, I lusted after them when they came out, but it wasn’t until a few years later I started picking bits up second hand cheaply that I got any hardware. I ended up with 2 interface 1s, 6 microdrives, some random used carts and a box of 30 brand new 3rd party ones, and buggered if I could get any of them to work, in any of the 2×6 combinations. Chiefly, they would appear to fail to format. Never figured it out, even with help of newsgroups as soon as I could get online in the early 90s. However, having “real” computers by now I did have the serial ports working so saved stuff to them over null modem cable and ran some dumb terminal stuff.

  7. Did anyone write a program to do ‘pre-stretching’ of the tapes by running them around the loop a bunch before attempting to format them?

    It would be nice if hard drives and SSDs increased capacity by 11.76% simply by using them!

  8. I seem to recall the printer was electrostatic, not thermal….I may be wrong. The guy I worked for developing embedded software in the late 80’s had one of these tape drives plugged into a Speccy with an EPROM programmer plugged into the port at the back. To say it was a bastard to use would be an understatement.

    1. It’s neither. The paper is coated in a thin layer of metal which the printer drags a metal stylus across. High voltage pulses are generated to ablate the metal coating anywhere a black pixel is required.

  9. Actually, Microdrives were firmly out of my (minimal) budget. Nobody i knew had one until i met this guy who sold pirated games LOL. In hindsight I should have bought the Interface 1 and some of the ROM games. Rare as hen’s teeth.

    1. Did you manage to examine the tape head? It would be interesting to know if one could do a virtual tape. It would be generating signals, like a virtual cassette tape that you use to play CD in older cars. I did not find such tiny tape heads available anywhere. And could not get the dimensions. The head within mdrive unit is huge and would not fit into the cartridge. You’d need a very slim counter head, compatible with the one inside the unit.

  10. I had an Atari in those days and was using commercial audio cassettes.

    I never saw the microdrive tapes, but I assume they would have been more expensive considering audio tapes were a commodity in those days.

    What made these microdrives better than the audio tapes used by other systems? I have to say the atari was very unreliable as well. So many times I would hear the playback going wrong (ERROR – 21) and having to start again. It would play back the interpreted code as well as the actual audio and you could clearly hear it lose the track as the interpreted sounds would change.

    Oh and how slow they were…… Awful. But the 5.25″ floppy drive cost more than the computer itself (and in fact *was* pretty much a computer on its own with a very similar CPU to the computer itself). It was massive.

    1. Oh one thing I forgot to mention: If you had a *good* dubbing deck you could copy Atari tapes without tying up your computer, which was pretty nice. It wouldn’t work with the cheap “Yoko” crap of the 80s but a good double tape deck like Sony or Philips could do it. You could even use high-speed dubbing but this caused mixed results so I usually avoided it.

      Also, some radio programs would broadcast programs which you could record and play on your comouter. All pretty nice benefits of using cheap audio tapes. The mechanics of the computer tape drives were also pretty much off-the-shelf. So I’m wondering what the reason was to go for this separate standard. The continuous looping thing looks pretty great for one, but I wonder how this was used as a benefit in practice.

      1. Yep, TROS online It would drive my parents bonkers listening to the screeching noise. I still love the idea of using radio to transmit code. Why not use shortwave to create some sort of wireless network of sorts? Those freqs are free now anyway…

  11. LOL, brings back some mems here… Although I dumped my Sinclair ZX asap for a Comodore 64 as soon as I saved enough, never looked back after that. I hated the “dead meat” keys and their size. So sorry to be slagging the Sinclair off here peeps. It was cheap, yes, I will give it that. I still have a working C64 + PSU +1541 lying around somewhere. So if anyone wants it, just comment with an email below. I will try and get in touch.

  12. Believe it or not … I used a Spectrum Microdrive to do serious scientific work (“Rule induction in the discrimination of virulent and avirulent isolates of the plant bacterium Erwinia amylovora, from fatty acid profiles”. Chemometrics Intell.Lab.Syst., 53-63. 1988.) Hugely tedious but IT WORKED! I remain pleased and astonished…

  13. Sinclair micro computer we used up at Naval Training Center Great Lakes Illinois in our free time to play Zork, Dungeons & Dragons, and Trek. The unit was very compact, easy to set up and use, and easy to learn.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.