25

Many home computers in the 1970s and 1980s had two floppy disk drives, or owners bought a second drive, to reduce having to play disk jockey. A second drive was especially popular with Commodore Amiga owners, as many games around 1990s came with several disks (and hard drives were uncommon for that system).

However, disk drives were not cheap.

Given that it was rare that a system or program or game would actually need to read from both disk drives at the same time (I can't think of a single example, except for disk copy tools, and even those can easily use the RAM as a buffer), I have been wondering lately if a "jukebox" approach would have been feasible.

I assume the read-head was the most sophisticated and expensive component of disk drives. So imagine a "jukebox" device that has several slots where you can put disks in, and the read head would mechanically move from slot to slot and read from different disks.

Sure it would require some sort of quite sophisticated mechanics and controller logic, but it might still have been less expensive than a second drive. Also such a device could have 5, 10, or even more slots and thus be eventually really "economical", even serving as some sort of "poor man's hard drive", albeit much slower of course.

So did such devices actually exist, would they have been feasible, and should I get a time machine and invent those? :)

scrØllbær
  • 1,109
  • 7
  • 12
  • 14
    My gut reaction, not supported by data (so not an actual answer): Such a drive would cost more money, not less, because moving parts are the most expensive parts of machines, and you'll need a lot more moving parts. Reliability will also be an issue. You are essentially creating a hard drive but with removable floppy platters with variable and unknown quality. – Wayne Conrad Aug 31 '22 at 16:44
  • Hard drives don't move the head from platter to platter, they have multiple read heads, one per platter. I have no recollection of such a device, but I only go back to the early 80's in computer experience. – Jon Custer Aug 31 '22 at 16:45
  • 1
    Such things did exist for CDs as well as tapes, from small 6-disc changers to massive warehouse-sized tape arrays. In the mid-80s both CDs and tapes could hold ~1000x as much data as a floppy disc. This made the mechanical parts for a changer much more cost-effective. I agree with the other comment: a disc changer holding 10 MB of floppies would not have been very cost-competitive against a 10 MB hard drive. It'd definitely be less reliable. – RETRAC Aug 31 '22 at 17:15
  • 4
    If you didn't have a harddisk you really wanted two floppy drives, because you had the system (MS-DOS, CP/M whatever) on the disk in the first drive and could then use the second drive for data. You could also copy easily between disks. For most people that was quite enough. Floppies were spacious enough to hold quite a bit of typing. – Thorbjørn Ravn Andersen Aug 31 '22 at 17:23
  • 1
    @RETRAC and quite a bit slower too. Floppies had a transfer rate last time I measured around 30 kb/second. – Thorbjørn Ravn Andersen Aug 31 '22 at 17:24
  • @JonCuster At least one seminal early hard drive moved the head from platter to platter, but as you correctly recall, the norm became one head per side. – Wayne Conrad Aug 31 '22 at 17:28
  • @ThorbjørnRavnAndersen: Single-density 5.25" floppies had a raw transfer rate of 250kbits/sec, but were limited in the number of consecutive zero bits they could handle, limiting practical burst rates to about 23kbytes/second. MFM drives used a raw data rate of 500kbits/sec, but required at least 1 zero bit between any pair of 1 bits, limiting burst data rates to 31.25kbytes/sec. The PC could achieve a practical data rate of about 23kbytes/sec; something like the Apple could have managed a practical comparable to that using the Prince of Persia disk I/O routines, but... – supercat Aug 31 '22 at 18:07
  • ...normal DOS and even ProDOS routines were much slower. – supercat Aug 31 '22 at 18:08
  • 1
    Two drive systems were pretty common in the area from say CP/M to PC (one for system disk, one for user disk). A single drive with several slots with one read head moving from disk to disk would be pretty hard to make (you need to spin all disks anyway, you'll have to get the read head completely out between disks before you can move it), so probably not cheapr than two single drivers. Twin disk drives (e.g. Apple Duodisk) did exist. – dirkt Aug 31 '22 at 19:00
  • 2
    “Quite sophisticated mechanics and controller logic” is what made floppy drive so expensive. This would make them Much Much More expensive. Home users on a budget would have manually switched disks themselves, while businesses would have bought a haed drive and QIC “floppy tape” drive for backups. – RonJohn Aug 31 '22 at 20:02
  • 2
    Another application for two floppy drives being used simultaneously: Overlays. If a program was too large to fit in memory completely, it would need to swap in parts of itself from disk. So you might well have a system where one floppy drive would contain the program and its overlays, and the other would contain the data being processed. – john_e Aug 31 '22 at 20:08
  • 1
    There were of course floppy duplicators that robotically/mechanically would run an entire tray of floppys through a drive. And there were tape libraries that would robotically/mechanically mount/dismount 7/9 track tapes from drives so they could be read/written. But I never heard of a similar kind of floppy changer/library for R/W "random" access to a bunch of floppys. – davidbak Aug 31 '22 at 20:34
  • @RonJohn - Just to add QIC at every stage had much higher capacity than floppys. And I have this vague idea that QIC changers did exist but can't find evidence of that on the 2022 interweb. (Though they would have helped avoid tape-juggling by an operator, they wouldn't really be random access devices - as each tape would be strictly serial access...) – davidbak Aug 31 '22 at 20:41
  • If you want to access a lot of floppies then lots of drives are the way to go - see for example the Floppotron - though to be precise there isn't actually any data in those particular floppy drives .... but the idea would work ...) – davidbak Aug 31 '22 at 20:43
  • 1
    @davidbak cartridge tapes have existed since the 1960s. They made robotic tape libraries practical. I can't even imagine how open reel tapes could be reliably moved around, loaded and by mechanical arms. – RonJohn Aug 31 '22 at 23:09
  • 5
    I used to have a 5-disk random-access changer for an Apple II, although that was too long ago for me to remember the name of it. It used 5.25" floppies modified by the addition of a square hole on the leading edge of the jacket, which the drive used to pull the disks out of the cartridge and into the drive mechanism. DOS 3.3 only, I seem to recall looking into the possibility of writing a ProDOS driver for it but never got anywhere. – jasonharper Sep 01 '22 at 02:30
  • Here is one use for two floppy disk drives: backing up the hard disk. The program I used (sorry, I've forgotten the name) prompted you to load backup disks 1 & 2 in the two drives. Once disk 1 had been written, the user was prompted to load disk 3 into drive A, while B was being written on drive B, etc. So a sort of jukebox, implemented using wetware. – Simon Crase Sep 01 '22 at 03:19
  • @WayneConrad "one seminal early hard drive moved the head from platter to platter" - it would be quite valueable if you have any reference to that. – Raffzahn Sep 01 '22 at 15:13
  • @davidbak IBM used QIC for the original 5100 PC. Floppies were only available ad later add on (and for the 5110). – Raffzahn Sep 01 '22 at 15:15
  • @Raffzahn IBM 350 disk storage unit for the IBM RAMAC 305 computer, c. 1956 – Wayne Conrad Sep 01 '22 at 15:20
  • 1
    @Raffzahn - I know. I programmed a 5100 in my first "independent consulting" job (moonlighting for a very forward-thinking accountant). My god was that system slow! Even for its time! Though I did enjoy APL mode when I had the time to play. (All the accountant's work I did in Basic - including an implementation of XModem for the 5100!) – davidbak Sep 01 '22 at 17:05
  • 1
    (TBH I got more use out of his IBM Selectric Model 50. Used it to type professional-looking (proportional spaced! multiple fonts!) resumes for job searches ...) – davidbak Sep 01 '22 at 17:10
  • 1
    @supercat This was just empirical evidence for what I remember measuring back then. Thank you for listing why. Regarding slow loading routines on an Apple II I remember having a program named "OILER" which caused the "BLOAD" ("BRUN"?) command to be much, MUCH faster. I think it was typed in from a magazine - Nibble magazine perhaps? (https://en.wikipedia.org/wiki/Nibble_(magazine) ) – Thorbjørn Ravn Andersen Sep 01 '22 at 20:45
  • Software didn’t start coming on more than one or two disks until around the time hard drives became affordable for home users. So for the early 90s you would buy a new game that came on 4-9 floppy disks and spend a while installing it on your hard drive, inserting disk 2, etc. and pressing any key when prompted. But after install it didn’t matter. Then CD-ROMs arrived in time to prevent most people from having to shuffle more than ten disks on installation of software. – Todd Wilcox Sep 02 '22 at 04:41
  • I remember there being software titles for Apple in the early eighties that would use two floppy drives, if you had them, to reduce or eliminate the need to swap disks. I recall early titles in the Wizardry series being among them. And of course, there is always disk copying software, which worked *much* better with two drives than with only one. – John Bollinger Sep 02 '22 at 19:58
  • Not quite what you asked, but mass duplicators not only did exist, but still exist! See https://www.awp1.com/st.html for example. – abligh Sep 03 '22 at 11:48

4 Answers4

27

There was a product, called I think "Juke Box 5" designed for the Macintosh, which had a hopper on the top that could hold a IIRC about 20 floppies, and was designed to be placed in front of the computer.

Each time the machine ejected a floppy, a battery-powered motor on the device would pull the floppy from the machine and insert the next one from the hopper. If one were either mass-producing floppy disks or were running a backup program, the device would only need to be serviced when necessary to empty the space below the drive and reload the hopper.

I don't know of any devices that were designed for random access, but especially given the name I think Juke Box 5 could be considered a "floppy disk juke box" of sorts.

As for the mechanical complexity of switching disks versus having more mechanisms, I would expect that the cheapest way to design a "two-disk" drive would be to have one head or pair of heads per disk, but use a single spindle motor and a single stepper motor to operate all of them. The biggest limitation with this approach is that inserting or removing a disk while another disk was being accessed would be likely to cause data corruption.

If the drawers that held disks could be moved up and down, it might be possible to design a head-movement mechanism with enough mechanical range to pull the head completely away from the disk, so as to allow the drawer assembly to be raised and lowered to select a different disk, but one of the big reasons for using two drives was to allow the system to switch quickly between accessing the contents of the disks on them. If the time required for a mechanism to switch disks would exceed the time for a human to do likewise, the device wouldn't be very useful.

supercat
  • 35,993
  • 3
  • 63
  • 159
  • 4
    My first thought on reading this was "you're kidding!". But of course I know your answers here at retro at totally serious and on topic. So it must have happened. But ... wow! – davidbak Aug 31 '22 at 20:30
  • 4
    @davidbak: I never saw one in person, but I saw ads for the devices back in the day and it seemed to be a real product. The fact that the Apple Macintosh drives had a built-in eject mechanism would make such a device easier to construct for those machines than for something like a PC. – supercat Aug 31 '22 at 23:30
  • 2
    @supercat What you describe as "one head or pair of heads per disk, but use a single spindle motor and a single stepper motor to operate all of them" is exactly what the Persci 270 drive was. – Raffzahn Aug 31 '22 at 23:32
  • Don't forget that the Macintosh had auto-inject capability In its floppy drive as well. I'm sure that helped make the Juke Box changer simpler and more reliable. – Coxy Sep 01 '22 at 07:08
  • I once used a 3½" disc duplicator that sounded very similar - but writing was all it could do, at least as far as I remember – Chris H Sep 01 '22 at 12:20
  • 16
    I had one of these units! We distributed software on floppy and copying was a time sink so we got one. It was absolutely useless. It would jam at least one time in three, but you could improve this slightly by loading fewer disks which defeated the entire purpose. – Maury Markowitz Sep 01 '22 at 14:04
  • @MauryMarkowitz: Can you find anything on line about them? I did a google search and came up empty. – supercat Sep 01 '22 at 14:43
  • 1
    @supercat - Not offhand, but we found it through either develop magazine or MacUser (this being before the era of MacWeek). – Maury Markowitz Sep 01 '22 at 17:35
  • 4
    I just DuckDuckGoed it and got this software for it from an archive and this video showing its use as the first 2 results. I then tried Google, and indeed, I didn't get anything about it. Interesting experience. – JoL Sep 01 '22 at 20:16
  • 1
    @MauryMarkowitz My experience as well. After a couple of weeks of messing with it we threw it in the bin and had one of the interns do the disk shuffling. That was more reliable and actually had better overall throughput in the whole copy process. – Tonny Sep 02 '22 at 12:29
  • @Tonny: I wonder if reliability could have been improved if the feeder designer were to buy bare drive mechanisms without housings, and build around the bare drives, thus having access to the electrical signals connected to the internal switches and motor, as well as a much better way of establishing mechanical alignment. – supercat Sep 02 '22 at 15:26
21

Many home computers in the 1970s and 1980s had two floppy disk drives

Home computers had any number of drives from zero to 6 or 8, all depending on budget. Even rather restricted systems like a C64 had by default addresses reserved for 4 drives.

However, disk drives were not cheap.

And price is why there were single drives. Home computers are all about getting a low price point

Given that it was rare that a system or program or game would actually need to read from both disk drives at the same time

It was extremely common to have programs using two or more drives. One holding the program disk, so overlays could be loaded according what was to be done (remember, there were no multi megabyte memories holding everything), plus a data disk holding your document(s). Of course there's always the chance to swap disks whenever access to either was needed.

In fact, it's the very reason why the DOS on the IBM-PC emulates a drive B if there is only a single drive available, by managing them as logical drives and checking disk IDs with every access. Doing so allows to handle two different disks without interaction from the program, while avoiding overwriting the wrong one.

except for disk copy tools, and even those can easily use the RAM as a buffer)

I have been wondering lately if a "jukebox" approach would have been feasible.

Sure, and they have been built and offered.

I assume the read-head was the most sophisticated and expensive component of disk drives.

Not really, especially early on the mechanicals were the most expensive part. This is also the reason why Processor Technology designed their Helios II drive around he Persci 270 drive - a drive having two disk slots, each with two heads, but only one frame, one drive motor and one stepper motor. The idea was to offer the ability of two drives but cutting cost. Would have worked fine, except the floppy became such a success that, by sheer scale of volume, the price for two single drives fell below what the otherwise simpler mechanicals for the Persci 270 would cost.

Another example was given by Apple. They only bought the bare mechanism and head mount, but designed everything else at a lower price themself.

Sure it would require some sort of quite sophisticated mechanics and controller logic,

Controller wouldn't be a big thing, it's the mechanics. it would need at least another motor, able to stop exactly at multiple positions and rather fine mechanisms to move the disk pack as well as the disks in and out. Such a setup would be way more expensive than a second drive, while at the same time not being able to access two disks concurrently - at least not without several seconds between each change.

So did such devices actually exist, would they have been feasible, and should I get a time machine and invent those? :)

No need to warm the Flux Capacitor, such drives have been made, but not for home computers, but professional systems. In fact, they have been around longer than home computers. IBM offered in 1973 the IBM 3540 Diskette Input/Output Unit for their mainframes which could operate up to 20 8" diskettes at once per drive and two drives in maximum configuration (B2). Then again, floppies were already back then way too slow, so IBM had as well IBM 3747 Data Converter, which could autofeed floppies to be copied to tape, which then was transported to the mainframe - they were part of the 3740 Data Entry System.

Similar attempts have been made in the 1980s for 5.25 drives, but were never really successful. It was simply less expensive to go ahead and buy an HDD instead - or, for most home-computer users to go ahead and buy a drive with higher capacity.


The only niche market for auto loader systems were disk copy stations used to manufacture software. But their purpose was to feed a stack of disks one by one, not to select between different disks of that stack. Equally important, they were way outside the price range of home computer users (several thousand USD per station), often self contained and usually HDD based.

An early example for such a mechanic is documented in this patent application of 1982. It feeds a stack of disks from the top and and selects the output into two bins below. Most likely to separate successful writes from failures detected in verify. Some other similar patents from the 1980s

An exception to these production orientated devices might have been some auto loaders for the Mac. The most well known (in the US) maybe the Jukebox Five Automated Disk Changer (video). It allowed to sequence thru a stack of disks, accessing them in sequence, but not random. Useful for installing software, or small scale production. It wasn't software controlled, but relied on the auto eject function of the Mac. whenever a disk was spilled, it was dropped below and the next in line was inserted. The handler was introduced at the 1991 MacWorld Expo by Fifth Generation Systems, a company focusing on backup solution - which benefited a lot from such a device. There were other similar systems as well


The only disk technology where changer systems had a tiny niche with end users were CD changers during the mid 1990s. Like the well renowned Pioneer 602X/604X/624X series, or the later priced Nakamichi MJ-4.8s (4 disks) and MCD-1020 (7 disks). Still, they cost >4 times the price of a single drive professional (read, already expensive) drive. Here again it was a matter of scale - in this case the fact that these drives were originally designed to be used in high end car audio systems, which produced some volume.


Of course stuff like that gets reinvented over and over again, so here a video of a nice arduino based hack.

Raffzahn
  • 222,541
  • 22
  • 631
  • 918
  • How did you find the Jukebox Five video? When I tried searching for Jukebox Five, I got all sorts of stuff about actual or on-line juke boxes, but nothing about the Macintosh disk swapper. – supercat Sep 01 '22 at 16:21
  • @supercat No idea. Came early up, among the top 15 or so. After all,who knows how google fu really works? – Raffzahn Sep 01 '22 at 18:33
  • I think the key is adding "Fifth-generation systems" to the search query, since I just tried that and the YouTube link came up first. Until I read other answers here, though, I didn't remember who made it (though Fifth Generation Systems, makers of the FastBack software that advertised "bulletproof" reliability (it could recover data from a disk with a .22lr bullet hole in it) would have been a logical choice given the Five in the name. – supercat Sep 01 '22 at 18:45
  • @supercat There were others as well. – Raffzahn Sep 01 '22 at 19:10
  • 3
    "their Helios II drive" - to quote wedge from star wars, "look at the size of that thing!" – Maury Markowitz Sep 02 '22 at 19:10
10

There are photos of the "JUKEBOX FIVE" 15 (3 1/2") floppy disk changer from "FIFTH GENERATION SYSTEMS" ("The Publishers of Suitcase II" font software) at reddit with a with a Macintosh SE/30. There is an advertisement on page 185 in the October 1991 edition of [MacUser] on archive.org. This even claims that "...unattended disk backups are a snap."
enter image description here

Mojo Techno
  • 101
  • 3
  • Backups. I'd forgotten what a chore those were. Those were time-consuming and you had to be there feeding the stupid floppys. Later on you had to be there feeding the stupid CDs. – davidbak Sep 02 '22 at 16:50
  • 1
    Mind to format that in a somewhat readable way? – Raffzahn Sep 02 '22 at 19:15
  • In mid to late 90s cd drives with a built in changers became available where you could have each disc as it's own drive if you wanted to which was fine as long as you didn't access 2 at the same time. This jukebox five seems to be more like an old style record player thing that just plays a stack of records through.

    The cdrom-changers were reasonably priced, as I understood how ours worked was similar as cd changers for cars had(and cheap). Of course radio stations and such had cd changers that could hold hundreds of discs and changers with tens of discs existed for Karaoke for LaserDisc.

    – Lassi Kinnunen Sep 04 '22 at 16:01
4

I was one of those who spent a seeming stupendous amount of money to get twin floppy drives for a 1978 designed "LNW-80" (essentially a TRS-80 model 3 clone). The pair of drives cost more than the computer did.

The difference between one drive and two was roughly a factor of ten, because with only one you had to keep inserting the "system" disk to re-load parts of the OS that had been "swapped out".

Using only one drive to copy a single 400kB disk took 20+ disk changes (10 each for the source & destination disks), assuming one could use 40 kB for the copy buffer and 8kB for the copy program & OS.

While true concurrent access wasn't possible, the main speed limits were mechanical latency:

  1. rotational acceleration & speed stabilisation
  2. head stepping
  3. head settling
  4. rotational skew

For a true "jukebox" with only a single drive axle clamp, this would have been about the same, but for an economical mechanism, the disk change time would likely have been worse than a skilled manual operator.

Remember that back then we were relying on mechanical position registration, not like today where moving parts have position feedback that enables pin-point accuracy under software control. So to get reliable positioning at speed you would have spent a lot of money.

An alternative would have been some sort of "shishkebab" single spindle loaded with many floppy disks as a fixed set, somewhat like the multi-platter hard drives of the time. That would have dramatically improved the mechanical registration and time to swap disks, but would have had significantly worse positioning latency.

The thing that made the drive head expensive was (a) getting it to glide over the surface without scratching it; and (b) getting it accurately positioned.

Both of those would be a lot harder (more expensive/slower/less reliable) with changing between "platters".

It's worth noting that by 1980, multi-platter hard drives always had a separate r/w head per platter, even in low-end equipment.

Toby Speight
  • 1,611
  • 14
  • 31
  • Because of how long it would take a jukebox to swap disks (a human could do it faster), the jukebox is probably only practical for unattended, overnight operation. Otherwise get that second drive or more RAM or a hard drive. – snips-n-snails Sep 01 '22 at 17:50
  • 1
    @snips-n-snails A few years later, CD jukeboxes were a thing, considerably faster than humans. Though in many cases, even being a little slower than a human would also have been fine for some applications, since it'd allow the human to continue interacting with the computer without a disk-swapping distraction. – Dewi Morgan Sep 01 '22 at 20:54
  • @snips-n-snails "more RAM" wasn't an option; it had a 16-bit address bus able to address 64 kB of anything, and in the TRS-80 you could only have 48 kB of RAM because the ROM and IO took up some of the address space. – Martin Kealey Sep 02 '22 at 01:37
  • 2
    @MartinKealey: If a system has any unused area of address space, it's relatively trivial to interface a CPU with a small memory bus to a memory subsystem with a much larger address space. On the C64, for example, which has open address spaces space at $DExx and $Dxx0 with pre-decoded chip selects, adding a two 512K dual-chip-select SRAM (if one happened to have them) would require only a 74LS377 and a 74LS379 as support circuitry. To select a range of RAM to access at $7F00, simply perform an accesses to $DExh and $DEmm, where h and mm are the upper and middle portions of the address. – supercat Sep 02 '22 at 14:53
  • @MartinKealey: Obviously if one doesn't have 512Kx8 SRAM chips available, adding more RAM would require adding more chips and associated chip-select circuitry, but the main limitation with adding more RAM is that unless a computer maker offers particular recommendations for how to connect it, different people faced with the task will likely do so in different and incompatible ways, and RAM expansion will only be useful if one has software to talk to whatever RAM expansion one has. On the other hand, makers of things like RAMdisk or disk-copying software would be likely to... – supercat Sep 02 '22 at 14:58
  • ...coordinate efforts with the makers of RAM expansions since those would simultaneously get the most benefit from expanded RAM and have block-oriented address patterns. If one wanted to copy a range of RAM stored in the aformentioned device upward by 260 bytes, the only practical way to do it would be to alternate between copyring 256 bytes data from the expansion RAM into main RAM, and copying 256 bytes (252 bytes from the page just read, and four bytes from the page before that) from main RAM to expansion RAM. But a RAMdisk read-sector or write-sector routine would work beautifully. – supercat Sep 02 '22 at 15:04
  • 1
    @supercat I was simplifying somewhat. "Bank switching" of course was technically an option, but RAM was prohibitively expensive in 1978, so much so that a TRS-80 typically shipped with only 16 KiB, nowhere near the maximum 48 KiB. Fortunately disks then were only 100 KiB, so it still took only 20-30 swaps to copy a disk. RAM was much cheaper by the time the C64 came around. – Martin Kealey Sep 03 '22 at 10:15
  • @MartinKealey: The cost to add an extra 64K of RAM to a system with a 16-bit address bus wouldn't be hugely different from the cost of adding it to a system with a smaller address bus. RAM was certainly expensive enough that systems with over 64K of RAM weren't terribly common, but I think the fact that many systems filled up their address space without going beyond was more of a mindset issue than technical one. – supercat Sep 04 '22 at 22:12