20

I have found inexpensive (~$15) 3.5in floppy to USB adapter cables but have been unable to find a similar adapter for 5.25 floppy drives. Does anybody make such a thing?

Are the pinouts of a 3.5 the same as 5.25 so that I could just make a cable end converter for the 3.5 adapter?

enter image description here

chicks
  • 397
  • 1
  • 4
  • 15
jwzumwalt
  • 4,479
  • 6
  • 24
  • 47
  • 9
    Note that the USB-floppy standard officially only supports 3.5" drives, so while the cable pinout is the same, it might still nor work (timing differences etc.). If you get it to work, please add an answer of your own (which you can also accept), because that would be interesting to other people as well. – dirkt Mar 03 '19 at 18:49
  • Wouldn't the 5.25 floppy use a standard IDE cable like a 5.25 DVD-ROM? Can you take a picture of the 5.25 drive's plugs? – Brythan Mar 04 '19 at 15:54
  • 1
    Whatever happens, please do come back and give an update on what worked for you. I still have both my 5.25" and 3.5" HDD floppy drives from 1989. – Criggie Mar 04 '19 at 18:45
  • 3
    @Brythan - The 3.5 and 5.25 were usually on the same cable with different connectors. When I wrote the question, I forgot I had a dual floppy cable in my parts box. – jwzumwalt Mar 04 '19 at 21:10
  • @Criggie - Due to the lack of r/w floppy interface, I am now considering making one... perhaps using a PIC micro-controler (many PICS have chip USB) to interface to a host computer. The research I have done apparently shows the floppy drive only responds to sector (block) r/w requests and does not know or interpret filo io. This makes it more of a software problem than hardware issue. Todays computers have plenty of memory space so I am thinking of copying the entire drive to a mem buffer, have software do its thing then save the entire disk when the software ops are done. – jwzumwalt Mar 04 '19 at 21:19
  • 2
    I know it's not cheap, but Kryoflux already solved this for most vintage disk formats. – Brian H Apr 16 '19 at 15:00
  • @Brythan 5.25" FDDs predated IDE by at least several years, even just on the IBM. – user Apr 17 '19 at 14:25
  • @jwzumwalt: I've also been playing with raw floppy I/O, though in my case it's for interfacing with an Apple //c. A floppy drive will at minimum have inputs to select it, move the drive head, enable writing, and supply data to be written; and outputs for the write-protect sensor and data read off the disk. PC floppy drives will also have outputs that detect the index hole and the insertion/removal of a disk. If a drive is selected and write enable is active, whatever pattern of wiggles is present on the data line will overwrite whatever at the current spot on the current track. – supercat Oct 09 '20 at 19:23
  • Otherwise, if a drive is selected and write enable isn't active, the data output will reproduce, more or less, the pattern of wiggles used when writing the disk, subject to four deviations: (1) Wiggles that are too short may disappear; (2) If too much time elapses without the drive encountering any wiggles, the drive may report wiggles where none exist; (3) variations in motor speed and other factors may cause a certain amount of timing uncertainty in wiggles occur; (4) some drives like the Apple expect the computer to supply a signal that changes level for each wiggle... – supercat Oct 09 '20 at 19:26
  • ...but when reading back data they output an arbitrary-duration high pulse (about 1 microsecond on the Apple's drives) whenever the written data changes state. If all you want to do is read data, moving the head to the outside of the disk and then capturing the patterns of wiggles from the drive heads while periodically advancing the head to the next track should be an easy way to do it. – supercat Oct 09 '20 at 19:28

6 Answers6

16

The data pin-outs are the same, assuming a reasonably “new” 5.25″ drive, not an early ‘80s model.

It was a common upgrade to replace the second 5.25″ drive on an older machine with a 3.5″ one, so you could have a choice of format depending on your needs. They were interchangeable.

You might have an issue with the power though. I don’t remember seeing a 5.25″ drive using the compact power connector pictured there. There might be, and I just don’t remember. There was another older connector type that was wider, with four tubular thick pins in a line. I’m learning that it is commonly referred to as “Molex” but that’s a company name. Power supplies typically provided both sets for choice of device. You will probably need an adapter cable for power. The electric characteristics are the same, it’s just a pin converter.

Be careful to get the right “direction”: you need a male 3.5″ style to “molex 4 pin” female. The converter might be a bit hard to find because most people who needed a converter needed the opposite thing (plug a 3.5″ drive on a 5.25″-only power supply). Cabling is not my area, so others might provide you with better information.

user3840170
  • 23,072
  • 4
  • 91
  • 150
Euro Micelli
  • 2,685
  • 3
  • 16
  • 24
  • 5
    Looking at the photos, those Molex connectors seem to be the "regular" Molex connectors that have been in use forever for IDE drives, they shouldn't be difficult to source (you can just take them from any broken ATX power supply). – Matteo Italia Mar 03 '19 at 23:03
  • 2
    Those connectors are more generically known as "peripheral power connectors", though most people do just call them Molex. – Bob Mar 04 '19 at 00:37
  • 2
    I also hate the "Molex" name. I use a lot of Molex connectors in my projects, neither is the 4-pin power connector. – pipe Mar 04 '19 at 12:44
  • While the data pinout is the same, the connector is different (card edge connector on 5,25" and dual inline pin header on 3,5"), so an adapter will be needed (as alephzero's answer states). I never saw a 5,25" PC drive with the dual inline pin header, even 1993. Furthermore, to read 5,25" DD disks in a 5,25" HD drive, you would need to support a data rate of 300kbps (see the answer by user12037) and use double stepping. You don't need it for 3,5" drives, so 3,5" USB-to-floppy solutions are likely to omit this data rate and the double stepping feature. – Michael Karcher Apr 17 '19 at 10:48
8

The connections on the 34-pin ribbon cables are identical, but the size of the connectors are different.

You can get ready made cables with both 5.25 and 3.5 connectors, for example http://www.cablesonline.com/36unflopdriv.html (and of course from Ebay also). These are likely to be more reliable than trying to wire a 5.25 connector onto a 3.5-sized cable by hand. These cables used IDC (insulation displacement) connectors which are crimped, not soldered, and they only work as designed with the correct size ribbon cable.

Rewiring the 4-pin power cable is a much more feasible proposition than rewiring the ribbon cable, if you can't find what you need ready-made. It should be possible to release the individual wires from the connector by pressing down the spring clip that locks the wire in place (visible through the slots in one side of the connector) with a thin screwdriver blade or something similar, while pulling on the wire. The wire itself is soldered onto the metal "pin and spring clip" part which goes into the plastic housing.

alephzero
  • 6,646
  • 3
  • 29
  • 34
  • Actually, all floppy power connectors I have ever seen (since the early 1990s) had crimped connectors. I have never encountered a single one that was soldered. – user149408 Mar 03 '19 at 21:49
  • The pins on a typical 5.25 drive's power connection are nowhere near the ones on the 3.5 connector, so releasing the terminators isn't going to do much in that you'd have to cut them off anyway. It's generally much simpler to cut the wires behind the connector and then solder\shrink-tube the wires themselves. – Comintern Mar 04 '19 at 04:26
  • Power is not a problem, my motherboard has the correct "Molex" 5pin plug. – jwzumwalt Mar 04 '19 at 11:35
5

I am somewhat sceptical that you'll get this to work for high density. The basics are that normal (single or double density) 5¼" floppies are turning at a rate of 300rpm while high density 5¼" floppies mimic 8" double density floppies with a rate of 360rpm. In contrast, 3½" floppies always use 300rpm. Double density on either have data rates of 250mbps in MFM (double density) mode while high density has 500mpbs. The slower rotational speed of 3½" HD disks is what results in the higher capacity.

It's probably safe to assume that only MFM will be supported and it would be rather surprising if both 250mbps and 500mbps data rate and/or 300rpm and 360rpm would be since those generally have to come from the drive. In theory, the "drive" could decide to vary those based on the size/kind of images stored on the USB stick but at least for double density, the size of 3½" and 5¼" images would be the same. Note that essentially it is the drive that "detects" the format of the floppies.

  • Thanks for the additional info. I had forgotten about the rpm differences. – jwzumwalt Mar 04 '19 at 11:36
  • This probably does not represent the floppy disk as a "USB stick" but rather as either USB mass storage Uniform Floppy Interface or something else. So it is not clear that the implementation details are going to be as hidden (and therefore fixed) as you assume. Consider that someone designing silicon for this role would have wanted to make it as universal as possible (and probably did so some time ago when there was still greater demand). So while boards might be specific to 3.5 inch drive connectors it is less likely that chips would be. – Chris Stratton Mar 04 '19 at 15:22
  • OTOH if this is actually a USB enabled MCU with some firmware to operate the drive itself, then it could indeed be quite restricted in what it can do. – Chris Stratton Mar 04 '19 at 15:23
  • 5.25" floppy drives for HD (1.2MB) and double density (360KB) are different. Double density drives are 40 track with wider tracks. HD drives are 80 tracks with thinner tracks, and can usually read double density floppies, but can't write them correctly since the write width is half the size of a double density track. I don't know if any "hybrid" 5.25" inch drives with both sizes of write heads were ever made. Although lower capacity, double density media lasts much longer. I was able to read a "library" of 25+ year old double density media a while back. – rcgldr Mar 04 '19 at 19:08
  • @rcgldr: I think that with a few tweaks to the electronics, it would have been possible to make a high-density drive write 360K disks in a manner that could be reliably read by low-density drives, though doing so efficiently it would have required that DOS understand some special handling requirements and have a minimum of 4.5K of buffering available. The trick would be to have the high-density drives add a few extra bytes between the sector header and a sector, which would get ignored by older drives, but could be used to recognize what type of drive wrote a sector last. – supercat Oct 08 '20 at 16:27
  • @rcgldr: Before writing a sector, the drive should observe which sectors (if any) on the current track were last written with 360K drives. If any were, the drive should read all such sectors, then enable the write head with a constant polarity for the duration of each such sector while the head is quarter-step above the track, and then while the head is a quarter-step below, and then write all such sectors with the head in the center of the track. That would add about a half second for each track were it was necessary to write a sector that was last written with a 360K drive, but... – supercat Oct 08 '20 at 16:31
  • ...I think such a disk should be usable in either kind of drive. Alternatively (and perhaps this might be better), a utility could be used to scan the disk for tracks containing sectors that had last been written with a 1.2MB drive without a cleanup step, and perform a cleanup step on all such tracks. If the sector layouts on the two sides of the disk were skewed by 20 degrees, such a scan could, with proper electronics and software, be done in about 10 seconds plus half a second per track needing cleanup. – supercat Oct 08 '20 at 16:34
  • @rcgldr: The writes above and below each track would set uniform magnetic polarization (erase); those would be followed by write of the actual data in the center of the track. Floppy drives routinely write sectors in place; they're written with a little blank space before and after each sector to accommodate timing uncertainty. – supercat Oct 08 '20 at 20:04
  • @rcgldr: If the erased stripes are regions of constant magnetic polarity, there would be no "timing uncertainty" except at the spots where the head is switched on and off. I don't know if the existing controller can be configured to write a uniformly-polarized stripe, but the drive itself shouldn't have any trouble. Drives have an automatic-gain circuit, so flux transition patterns written in the middle of a track, with uniformly-polarized blank areas to either side, should appear as slightly weaker than normal flux transitions. – supercat Oct 08 '20 at 20:22
  • @supercat - I deleted my prior comments.Take a look at retro technology web site. Not mentioned is that two offset writes to the same sector will not work because the timing (mostly motor speed variations) is not accurate to within one clock time of the signal being written. – rcgldr Oct 08 '20 at 23:47
  • @rcgldr: If one tries to write data both times, that would be true, but that's not what I was proposing. If one wrote data both times, then on readback one would receive an average of the two written data signals. If, by contrast, one writes blankness on all but the last pass, and valid data on the last, what is read will be a weighted average of valid signal and silence. – supercat Oct 09 '20 at 14:42
  • @rcgldr: Imagine that one has a recorder that can write to either of two stereo tracks, and a player that mixes both. If one tries to record the same audio on both but the timing is 100ms off, the result will be garbled, but if one writes silence to one track and records a signal to the other, the playback won't be as loud as if one wrote the same thing to both tracks simultaneously, but would still be intelligible unless it was on the threshold of detectability. – supercat Oct 09 '20 at 14:50
  • @rcgldr: If one were to attempt to record audio onto a type whose domains were saturated in the direction associated with positive voltages in the recorded signal, the tape would respond somewhat more eagerly to rising edges than to falling ones, thus causing audio distortion. I would expect that disk drives (or digital tape drives) would use a high enough write current to saturate the magnetic domains under the head regardless of their previous state. – supercat Oct 09 '20 at 21:11
  • @supercat - erases heads write a high frequency AC. Audio tapes can use permanent magnets for erase (stronger on leading edge, weaker on trailing edge), since there isn't an issue with erasing sector headers. Floppy media is too thick to fully saturate, as outputting a stronger signal results in a larger affected area of the media. – rcgldr Oct 09 '20 at 21:21
3

The price tag is certainly higher (~100€), but AFAICT you can use a KryoFlux to read everything (even strange low-level software protections):

  • Works with all major 3.5" and 5.25" drives
  • Works well with selected 3" (e.g. Amstrad FDI-1) drives.
  • Also works with 8" (e.g. Shugart 851; might require additional adapter) drives; other types of drives and media currently under investigation.
lapo
  • 131
  • 3
1

The firmware on all the 3.5" USB floppy drives of relatively modern make I've looked into seems to present the floppy as a LBA device with exactly 2880 sectors (of 512 bytes) to the host computer. That is the very same thing as an USB thumb drive (and in fact, most OSs seem to see such a thumb drive. You can even partition such a floppy).

Replacing the 3.5" drive with a 5.25" drive might work in principle, (however, most recent drives in such floppies don't even have standard floppy connectors anymore, there you couldn't) but there simply is no such drive that could store 2880 sectors.

A Kryoflux as proposed in another answer seems to be your best (and only, in 2020) bet to connect a 5.25" drive to USB.

The main problem with Kryoflux for your purpose is: It doesn't present itself as a floppy drive (not even as a mass storage device) to the host computer, but rather as a commandable streaming data source and sink and will only work with it's own software, that can mainly produce disk image files. Any other floppy disk software that will happily work with a "real" floppy disk controller is not gooing to work with Kryoflux.

tofro
  • 34,832
  • 4
  • 89
  • 170
  • Most 8 inch floppies use/used 50 pin connectors, but there are/were 34 pin to 50 pin adapters. I don't know if MSDOS or Windows supports 8 inch floppies. I have an old CP/M system that can use them. – rcgldr Oct 08 '20 at 20:02
1

Instead of the board you suggested or a KryoFlux, maybe build a FluxEngine.

You just buy the specified FPGA development board, solder one row of pins from a floppy connector onto it (or, in a pinch, a row of pin header), flash it, and you're done.

You'll also need one of those wall-to-Molex power bricks that you see bundled with USB-PATA/SATA adapters, but the FluxEngine itself will cost around the same as that adapter you linked and a lot of 5.25" floppy formats are officially supported.

Another option is the Greaseweazle but be warned that, if you go for the "build it using a Blue Pill" option, it won't work with a counterfeit STM32.

ssokolow
  • 6,746
  • 2
  • 25
  • 40
  • The FluxEngine sounds a lot like something I'm working on, which would be intended to facilitate reading single-sided disks of any sort using an Apple //c, producing Apple II disks from an image, or feeding software directly from a PC to an Apple //c without having to store it to disk first. My approach would require using an Apple //c rather than a stand-alone drive, but would be able to read the back side of a "flippy" disk without needing to see the index hole. – supercat Oct 09 '20 at 22:27