43

180rpm to 360 rpm, unlike 12000rpm on optical discs, was how fast floppy disks got.

I am not sure, whether all drives had the same speed, but 360rpm is not close to the physical stress limitations of floppy disks.

Why were floppy disks not any faster? Because of a technical limitation?

Their speed of 18×512×(rpm/60) bytes per second was easily sufficient for plain/hypertext documents without multimedia attached to it.

https://retrocomputing.stackexchange.com/a/5458/7641

  • 4
    One problem is that floppy drives used a low-level interface. So a drive running at a non-standard speed would need a controller running at the same non-standard speed. – Peter Green Jan 16 '18 at 20:26
  • 2
    Actually I had a LS-120 drive which could also read floppy disks, at a much faster rate than the standard floppy disk drive. However, I think it used a laser to read the disk. However, as soon as USB stick came out, they all went into the trash. I actually had a SCSI LS-120 with like 8mb or more of RAM, and it was really fast because all actions were buffered against RAM. That in addition to a vastly superior native read/write speeds. – cybernard Jan 16 '18 at 22:28
  • 4
    USB drives came out in 2000, by which time the internet, Zip drives, and rewritable CD/DVDs had already put serious inroads into floppy drives. – Acccumulation Jan 16 '18 at 22:59
  • 10
    With only 1.44 Mb capacity, 50kb/sec is enough to read the whole disk in 30 sec. – Acccumulation Jan 16 '18 at 23:06
  • 5
    There was, at one point, a high-speed floppy drive called the X10 Accelerated Floppy Drive. The drive came with a special controller card with its own processor and memory, and it basically read both sides of the disk concurrently and at four times the spindle speed, allowing it to load the whole disk in about five seconds. The memory was used as a write-back cache to speed up writes to the disk. – bwDraco Jan 17 '18 at 22:49
39

As you mentioned, speed was considered sufficient for the typical applications at the time they were introduced. Also, any significant rotational speed increase would have meant:

  • More complex data separator circuitry (due to higher bandwidth needed), probably better shielding on the r/w head wiring and drive-to-controller cabling -> cost, development effort

  • Higher risk of drive/media damage if there is some imperfection in the media, or if any foreign matter gets into the drive or between media and sleeve (observable with hard drives. Dust in a hard drive can wreck it.) -> reliability

  • More wear from media-to-sleeve friction - 5 1/4" and 8" discs typically touched the sleeve, which was lined with soft material -> wear, noise

  • The disc didn't always get perfectly centered by the loading mechanism, leading to rotational imbalance -> vibration, noise, wear

  • Probably also time needed to accelerate the medium to that speed without causing stress -> perceived speed, and/or cost due to much more complex electronics to read before medium is up to speed. If media is clamped with the motor on (as done in many existing floppy systems), mind the stress to the material if that is done at high speed!

  • Designing media so that it kept structurally sound even if punctured, driven against the read/write head by vibration/shock, suddenly braked in case something jams (observable with CD-ROM - high speed drives can rip damaged media to pieces!)

Also mind that very early (8 inch era) drives did not use controllable BLDC motors but synchronous AC motors that might have posed some limitations (and made these limitations into specs).

60

That answer is somewhat trivial:

Hard disks and optical drives are contact-less technologies - Nothing touches (or is even allowed to touch) the media while it spins. On a hard disk or CD-ROM, the heads touching the media would end up with catastrophic effects.

Floppies and tape drives are different. The heads are constantly touching the media directly while it moves. You obviously want this to happen at a much lower relative speed to avoid wear on both the heads and media - Even if floppy disks were specifically teflon-coated to remove friction.

"Really floppy" disks of the 8" and 5-1/4" type also have constant friction between inner sleeve and media while rotating. That has somewhat improved with hard-case technologies of the 3" and 3-1/2" type.

In a sense, floppies are like hard disks, but designed to survive a constant head crash.

  • 6
    The 3.5" plastic floppy disk also has a "quilted" liner paper on both sides of the magnetic disk that touch it. I always thought this was to keep the media clean and to keep it straight while being read, but I have no sources for that. My point is that the 3.5" disks also had "constant friction between inner sleeve", though probably not as much as the 5.25" disks. – JPhi1618 Jan 16 '18 at 16:31
  • 4
    @JPhi1618 The inner jacket lining serves two purposes: in addition to reduce dirt buildup the main purpose is to provide a soft surface for the disk to run on - the hard jacket would quickly scratch it. – Zac67 Jan 16 '18 at 19:50
  • 2
    People underestimate the energies involved in hard disk technologies. When the read/write head touches a hard disk, it's called a head crash because the read/write head has literally crashed into the disk. I have a platter at home from the removable hard disk of a Unisys A10 that suffered a head crash. There is a circular groove in the metal that goes all the way around it. – JeremyP Jan 17 '18 at 9:56
  • 1
    @jpaugh The answer after the colon. – tofro Jan 17 '18 at 19:30
  • 1
    @JeremyP I've also done post-mortems on drives that suffered head crashes, including a 10krpm drive -- it gouged the drive badly, and apparently subsequent trauma from someone trying to perform further recovery actually resulted in one of the (glass) platters shattering. I can only imagine the sound it made when it did that... – Doktor J Jan 18 '18 at 17:24
32

I used to work for a company called ‘Rimage’ who manufactured robotic duplicators for floppy disks, the company still exists and today manufactures robotic equipment to publish/duplicate cd, dvd and Blu-ray.

The fastest rotational speed for a floppy disk was 1440 rpm and the equipment would also write to both sides of the disk at the same time. This meant you could duplicate a 1.4MB disk in less than 20 seconds.

The drives used could be modified versions of consumer drives such as Sony 73W which ran at 600 rpm, but YE Data produced a drive especially for this industry and it ran at 1440 rpm for duplicating.

  • 7
    Invalidating the premise of the question is a valid answer, despite not answering the question. Welcome to the site! – wizzwizz4 Jan 17 '18 at 7:35
  • But, presumably, a disc would be written once at this high speed and then shipped to a customer. For the rest of its life, it would be read in an ordinary speed drive. Would the disc wear out significantly faster if used at 1440 rpm all the time? – David Richerby Jan 18 '18 at 18:33
  • @DavidRicherby: If the disk only spun while it was being accessed, and it wasn't accessed continuously for longer than it would take to read or write the whole thing, wear should be essentially equal since the same number of rotations would be required in either case. Most systems keep drives running for awhile after each disk access, however; if the system writes some data, thinks for a half-second, writes some more data, thinks some more, etc., doubling the rotational speed would double the wear during the time the computer was thinking. – supercat Jan 18 '18 at 19:27
  • 2
    @supercat You're assuming that wear increases linearly with rotational speed. Is that true? I don't know; kinetic energy increases quadratically with rotational speed but I don't know if that's relevant. – David Richerby Jan 18 '18 at 20:27
  • 1
    When a disc is being duplicated, it can be run at high speed, while the heads write data sequentially. With normal, use, they may have to access data randomly. Drive mechanics that allow random access at higher speeds may be very expensive. – TimSparrow Mar 5 '18 at 14:11
4

Many computers had floppy drives that couldn't process data as fast as it came from a disk, even at 300rpm. If the time required to process a sector is twice as long as the time required to read it, reading all the sectors on a track would require that the disk rotate about three times. If the drive were to spin twice as fast, the time required to process each sector would be four times the (now shorter) time to read it, making it necessary for the disk to rotate five times. Doubling the rotational speed of the drive would thus only reduce the required time by 16% rather than by 50%.

To make matters worse, getting good performance out of a floppy generally requires that the disk be in a position to read out a sector just before the computer is ready to receive the data. If the drive is spinning fast enough that the disk rotates past a sector just before the computer is ready for the data, it will generally be necessary to wait an extra revolution. While reading 16 sectors on a normal-speed drive with 3:1 interleave would take about three revolutions at 200ms each (600ms total) reading those same 16 sectors on a "double speed" drive would require 19 revolutions, or about 1.9 seconds. Unless the data was written with a 5:1 interleave (which would add 66% to the time required to read it on a normal-speed drive), the faster drive would read it much more slowly than a slower one.

  • 1
    If this were the reason, why did they invent hard drives after all when computers couldn't even cope with floppies? I think this has it backwards - The electronics didn't need to be faster to operate a floppy disk drive within proper mechanical limits. – tofro Jan 17 '18 at 9:06
  • 2
    @tofro: Inexpensive floppy drive controllers relied upon the main CPU to perform data separation and decoding. Hard drives invariably added hardware for that purpose. More significantly, hard drives would generally be permanently installed into one particular computer. Until hard drives started having on-board caches, getting good performance from a hard drive required that it be formatted with an interleave factor optimized for the speed of computer it was working with. If a computer needed 3:1 interleave with a hard drive, using 4:1 would reduce useful transfer rate by 33%... – supercat Jan 17 '18 at 15:30
  • 1
    ...and using 2:1 would reduce it by an order of magnitude. Once hard drives started having on-board caching, interleave became a non-issue, but I'm unaware of any consumer-grade floppy drive having enough RAM to hold even one track of data (one of my big disappointments with the 1571 back in the day was that it didn't). – supercat Jan 17 '18 at 15:31
  • @tofro: BTW, when not using caching, there would often have been a trade-offs between write performance and read performance. On something like an Apple II, using a certain amount of RAM, one could either write a track in 800ms that would take 800ms to read, or spend four seconds writing a track in a way that could be read in 600ms. If one used more RAM for buffering so as to allow tracks to be written and read in a different sequence, such issues could be avoided, but of course RAM was expensive. – supercat Jan 17 '18 at 15:42
  • Well, as always: Don't expect optimum performance when you haven't optimized for performance. Running with a non-optimized interleave is just wrong, so don't understand why you bring this up. And the 1541 was slow for other reasons than a missing track cache ;). The other thing is: I cannot see why read and write speed on the same drive could differ - It's determined by data density and rotational speed which won't change whether you're reading or writing. – tofro Jan 17 '18 at 15:51
4

There were, in fact, attempts to create faster floppy drives.

Corporate Systems Center once sold a high-speed floppy drive called the X10 Accelerated Floppy Drive. The drive came with a special controller card with its own processor and memory, and it basically read both sides of the disk concurrently and at four times the normal spindle speed, allowing it to load the whole disk in about five seconds. The memory was used as a write-back cache to speed up writes to the disk.

Some floppy-based Sony Mavica digital cameras used a high-speed floppy drive that were capable of reading and writing to disks at two or four times the normal speed using a higher spindle speed. This was particularly useful because it greatly shortened shot-to-shot times; the camera had to write each image to the disk before it could shoot another.

As noted by several other answers, floppy disks weren't really designed to handle these high speeds. I have a Mavica FD-90 with the 4x floppy drive and it was distinctly less reliable than any of the PC floppy drives I've used. I've had disks jam inside the drive, and in one case, a partial disassembly of the camera was required to eject the disk. I would suppose that because of a lack of market interest and cost concerns, nobody made floppy disks specifically for high-speed drives.

I have a video of the FD-90 in action here.

3

The limitations of floppy drives were not mechanical. While they certainly couldn't have stood up to the 15000 RPM that some hard drives today have, they were certainly more than capable of being read much faster than the old hardware would have you believe.

There were a number of things that hindered older hardware, including a slower interface to the motherboard (fewer maximum bits per second), poor programming practices (floppy drivers not aligning DMA buffers to 64KB boundaries), and so on.

Later disk drives were able to perform much better by having an IDE interface. For example, the LS120 was capable of writing the entire contents of a 1.44MB image to a 3.5" floppy in just a couple of seconds, probably at least 15 times faster than older systems. Anecdotally, this was a problem for me; I'd often format a floppy more than once, since it happened so quickly I wasn't sure it'd happened at all.

Even the older drives using the older interface could achieve much higher speeds if programmed with optimal DMA transfers and a sufficiently beefy processor. Floppies were slow back then, but they didn't have to be. I actually found a MS-DOS driver a few decades ago that aligned the DMA buffers to a 64kb boundary, and I was able to write 1.44MB of data to the disk in about 10 seconds, while the MS-DOS driver took almost 33 seconds to transfer the same amount of data.

If you have an older system, you might try doing some research on DMA programming and floppy drive interfaces to figure out if you can write something that runs much faster on older hardware. You'll just need a basic understanding of DMA, ports, memory, and some machine code. The firmware will stop your drive from destroying it and/or the disk, but if you spend some time researching this, you could actually get some serious speed.

  • Probably depends on where you look: (Early) Computers that had floppies as single means of storage definitely had to look into disk transfer performance and did so (CP/M aera). Later computers (PC aera) that could be equipped with hard drives probably didn't, maybe for two reasons - (1) "Floppies are slow anyways" and (2) Who would buy our hard disk if the floppy drives are not much slower?" – tofro Jan 17 '18 at 9:15
  • @tofro There was a point where the bottleneck was certainly the CPU, and the eventual HDD tech that was emerging probably did factor into decisions. My answer was more about the fact that the actual mechanical properties of floppies allow much faster speeds than they were typically used at. I loved my LS120 drive, but I could never seem to find the 120MB 3.5" disks anywhere, since it died off so quickly. – phyrfox Jan 17 '18 at 9:34
  • Later, faster, floppy technologies did definitely leverage improvements in mechanical technologies (like teflon-coating, faster steppers,...), but still traded mechanical wear for speed. – tofro Jan 17 '18 at 11:18
1

The main limitation was the speed with which the computer could process the data from the drive. To keep costs low standard floppy drives were fairly simple devices, relying on the computer to do the decoding and detect things like the start of track markers.

Later faster floppy drives became available. 2x USB floppy drives were not uncommon, and even 4x units existed. However those had their own custom controllers to handle the higher rotation speed and thus data rate.

For example, consider the classic Intel 8272A and NEC UPD765A floppy controllers, which were very common PCs from the 80s.

Both parts operate from an 8MHz clock, relatively fast in the 1980s when CPUs were typically running at 1 to 5MHz. As speeds get above about 10MHz special consideration is needed to make the electronics work reliably, which increases cost and complexity.

Note that at 8MHz both those parts are limited to single and double density disks. They can't even handle 1.5MB high density disks at this speed. And this is dedicated controller hardware, earlier systems that used the CPU to control the drive directly in order to reduce costs were limited to single density (350kB) 3.5" disks or 5.25" disks.

  • 1
    Same comment as on other answer: If this were the reason, why did they invent hard drives after all when computers couldn't even cope with floppies? I think this has it backwards - The electronics didn't need to be faster to operate a floppy disk drive within proper mechanical limits. And faster drives simply traded mechanical wear for speed. – tofro Jan 17 '18 at 9:31
  • Early hard drives were very slow for this reason. Later ones added their own dedicated decoding hardware that was able to handle higher frequency signals. That's one of the reasons why hard drives were so much more expensive - the electronics were much more complex and high performance. – user Jan 17 '18 at 11:19
  • Agree with first sentence, but: They were never as slow as floppy disk drives. – tofro Jan 17 '18 at 11:25
  • Yes, some hard drives were as slow as floppy disk drives back in the day (around 40kB/sec for double density, 20kB/sec for single density). You had to pay a lot more for a fast drive. Of course, when machines only had kilobytes of RAM it wasn't an issue. – user Jan 17 '18 at 11:30
  • 1
    @user: I don't know of any hard drives that didn't do hardware data separation, but a number of computers connected hard drives using electronics that were designed for connecting floppies. – supercat Jan 17 '18 at 15:35
0

It is a limitation of the media -- which operates at lower audio frequencies. Magnetic media has a similar limitation across the board in all technology. In other words, it is a purely analog medium.

  • 2
    Hard disks are magnetic media. So is LTO tape. A floppy disk can be approximated as a magnetic tape with the "tape" having a different shape. – a CVn Jan 17 '18 at 21:49
  • Does the medium have a frequency limit (units of 1/seconds), or does it have a density limit (units of 1/meters, i.e., inverse wavelength)? These two quantities are related by the speed at which the medium moves past the read-write head. – Ben Crowell Jan 18 '18 at 16:02
  • @BenCrowell: The inductance of magnetic heads would impose a frequency limit, while the quality of the media surface would limit physical density. The spot I find these issues most interesting is not with disks, but with cassettes. For machines that used custom cassette drives, doubling the motor speed would have been a trivial mod, and most machines would have had enough CPU power to easily double the rate at which they load and store information to/from the tape. – supercat Jan 18 '18 at 17:15
  • @supercat -- I spent a few years working on and with AT&T Model 40 Teletypes. They had the type of drive that you are referring to -- however, the CPU unit was not so great. I made a converter, hooked one up to a 386 PC and got better data rates, but the length of my cable and the patch panel introduced additional RCL components that surely limited it somewhat. – jinzai Jan 22 '18 at 16:50
0

The 400kb and later double-sided 800kb and bigger Classic Macintosh floppies did run faster, depending on head location.

One of the Mac innovations was variable-speed drives see http://lowendmac.com/2016/floppy-disk-compatibility-and-incompatibility-in-the-mac-world/

These drives had a variable speed motor that allowed the Mac to pack 400K into a disk that would only hold 360 KB on a fixed-speed drive. (This is also the reason non-Mac computers can’t mount 400K and 800K Mac floppy disks.)

My first Mac was a 1985 512KB "Fat Mac" with the internal 400kb drive, for which I purchased a third party 800kb double-sided. They definitely varied in speed - you could hear it.

I wonder if varying in speed also allows the drive to maintain a more uniform density of writing over the increased circumference of the tracks.

I just dug out my paper copy of Inside Macintosh Vol IV which on page 251 says With double-sided disks a single mechanism positions two read/write heads - one above the disk and one below - so that the drive can access two tracks simultaneously...This lets the drive read or write two complete tracks of information before it has to move the heads....The double-sided drive controls its own motor speed, ignoring the speed signal (PWM) from the Analog Signal Generator (ASG).

  • Can you add some more information to your answer about variable speed drives? Adding a link is helpful, but elaborating on the contents of the link will really add value to this answer. – JAL Jan 19 '18 at 17:31
  • This answers the question, but minimally. It could be made much more useful if you included some information from the link into your answer. – wizzwizz4 Jan 19 '18 at 20:43
  • This seems like an odd way around of doing it: Commodore's mixed-density GCR just alters the data input/output rate while spinning at the same speed. I know the Woz disk controller tightly couples data speed and processor clock rate though, so maybe the options for zone recording were more limited? Either way, it'd be interesting to know whether Apple sped up or slowed down the drive relative to the normative 3.5" speed of 300RPM. – Tommy Jan 22 '18 at 15:45
  • @Tommy: Actually, I would think the Woz Machine could fairly easily accommodate different data rates simply by using a somewhat larger ROM and giving the main CPU control over the upper address bits. – supercat Jan 22 '18 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.