There's a lot of commentary out there about imaging disks. Why add another one? Because this is what I wish I knew when I started. This is mostly going to assume you're on Windows, though most can be adapted to Linux without too much issue.

Imaging Floppies

Why?

The naive way to back up a floppy is of course to just copy the files off of it, and if the floppy is just some old school projects or similar that may well be all you need. But for actual software there's two words that throw a wrench into this: copy protection. Software from the 80s and 90s often did weird things to prevent normal software from just reading the disk straight up so you couldn't just duplicate the disk and give a copy to your friend. Since computers of that era let you do whatever you wanted to the low-level hardware, a program could instruct the drive to seek to a sector that appeared not to exist or was in the wrong place or do other similar tricks. None of this proved more than a speedbump for the pirates, but it's a problem now if you want to make an equivalent copy, especially since floppies aren't going to last forever.

Flux Imaging

The solution to this is to instead record the raw magnetic flux, which is the transitions seen by the floppy drive's read head(s) as the disk rotates. The basic theory is this: a file on a floppy exists within that floppy's filesystem. The filesystem in turn is stored on the disk's sector format. Most copy protection involves changes to the sector format. That sector format is stored on the disk's raw magnetic values (transitions, really). Flux imaging just reads that magnetic data directly rather than trying to parse it into a filesystem or even a sector format, so it doesn't matter what you did to any layer above it, or even what it's formatted as. Mostly. There's a few notable exceptions later. The only real downside is the resulting files can be much larger than the formatted data, often 10-20x the size of the disk as you'd normally think about it. On the other hand, it's not like floppies were that large to start with, and disk space these days is comparatively cheap. I've archived probably a thousand floppies at this point and my total archive is only around 12GB.

To do this you need a floppy drive compatible with the disk(s) in question, a drive controller that can do the flux imaging, a cable to connect them, and a way to power the drive. For now I'm going to stick to the most common case of PC-style drives of either 1.44MB (3.5") or 1.2MB (5.25") and I'll cover some of the other cases later. For a floppy cable you just need to know if you need a 34-pin rectangular "IDC" connector (3.5" drives) or a 34-pin card edge connector (5.25" drives); the latter is more annoying to find but still not too hard as of this writing (mid-2026). For a controller I highly recommend and will assume you are using a Greaseweazle. You can buy the physical hardware from various sellers on eBay or equivalents. The Greaseweazle can power a 3.5" drive itself, though you may need an uncommon adapter to connect your drive to it. 5.25" drives aren't guaranteed to work, but may. If you don't have the right adapter or your drive is drawing too much power or otherwise can't be powered by the Greaseweazle then I'd recommend picking up one of those USB to IDE/SATA adapters that are all over Amazon; they almost all have a "molex" power output that you then just need to connect to your drive. I own a "Cinglink" but I'm sure there's a hundred more that will do just fine. You can also reuse this to read CDs and DVDs as is covered below.

The Drive(s)

But what drive to buy?

First, some terminology. I promise it'll make sense why we're going over this first when we get there.

A floppy is divided into heads (sides), tracks/cylinders (radial distance), and sectors (rotational distance). These three values allow you to locate a specific location a disk, and in fact hard drives used this same notation for some time until they grew too large for it to remain practical. If you've ever seen a reference to "CHS" addressing, possibly in an old BIOS, this is that; "Cylinder, Head, Sector". A floppy drive thus has the following stats that matter:

The drive's data rate controls how many sectors it can have but this should never be an issue unless you're working with pre-PC drives. You want a drive with two heads and the highest TPI count you're going to archive. For PC 5.25" drives, 1.2MB drives are 96TPI and 360KB drives are 48TPI, so a 96TPI drive can archive 48TPI disks but not vice versa--get the 1.2MB drive. For 3.5" drives 720KB drives can't read 1.44MB disks but 720KB drives are fairly uncommon whereas 1.44MB drives are everywhere so this shouldn't be much of an issue as you're unlikely to encounter a 720KB drive unless you specifically seek one out. 1-head drives should only exist on the earliest PC-compatible 5.25" drives, so if you're buying a 1.2MB drive you're set. (Again, there's exceptions for other situations; see below.)

That 1.2MB 5.25" drive, connected to your Greaseweazle, can image practically anything: IBM PC, BBC Micro, TRS-80, whatever, it'll handle it. Likewise a 1.44MB 3.5" drive can handle IBM PC disks, Mac 1.44MB disks, Amiga disks, and so on. My recommendation for 5.25" drives is to get a TEAC FD-55GFR, but to specifically look for one with a second solenoid that raises and lowers the heads. This will be at the back left corner of the drive when viewed from the above-front and that spot will be empty if not present. These models seem to be much rarer; it was probably removed as a cost-saving measure and they don't have a different model number, though the earlier revision FD-55GF may be more likely to have this. If you can find one of these it can improve your odds of reading marginal disks by reducing the disk's exposure to the heads when you aren't doing a read. If you can't find one of these, don't sweat it--probably you won't have the same degree of fun I had trying to archive the Studio Software floppies that formed this preference (my record was one disk that took sixteen passes to get a full read of).

Specific drive models haven't meant much to me in my experience. I do most of my 5.25" imaging with TEAC FD-55GFRs, but that's mostly because they were made in large numbers so they're easy to get ahold of. For 3.5", basically any 1440KB drive is equivalent to any other unless you want to read Mac 400KB or 800KB disks; see below.

Drive Special Cases

Other situations to be aware of:

Using the Greaseweazle

I'm not going to spend much time on the actual Greaseweazle CLI as its own wiki covers that pretty well. Instead I want to cover some best practices I've learned using it.

#1: Always read as flux (--raw). Always. This has three benefits. First, if the disk has some copy protection weirdness or similar, flux will capture that and allow you to rewrite it. Second, if the disk is damaged and has bad sectors, flux will preserve everything it can and potentially allow you to reconstruct the damaged data. Third, if you have to abort a read because a disk is marginal or damaged (see below) and you need to stop to clean the heads or just to avoid damage, the Greaseweazle software will not write out a partial file, but flux is written per-track so you can get keep the already completed tracks.

#2: When possible, read as flux with simultaneous decode (--raw --format=...). If you have decaying disks and need to worry about stopping to clean the heads, this will allow you to see when it starts encountering read errors. Reading flux by itself will not present errors as it has no idea if what it read is valid or not. Depending on the exact disk format this may not be possible, but if you can, you should. From my experience, getting the occasional "Unknown mark" warning or the occasional track retry isn't worth worrying about, but seeing both on the same track means you should probably stop and check things.

#3: Inspect disks before reading them. Be very careful imaging disks that look unusual in any way. As of this writing there is an ebay seller who sells 3D printed disk cleaning rigs that allow you to spin a disk inside its sleeve; consider getting these if you plan to do serious archiving.

#4: If you see the rate of read errors increasing as a disk is read, stop and inspect it and the drive's heads. This can be a sign that the heads are accumuating material that is interfering with the read.

#5: If you hear any suspicious metallic noises, stop and eject the disk immediately. You may be hearing a head crash, either already or imminently. Clean the heads before proceeding and consider that disk high risk.

#6: In general, your ears are a surprisingly good tool for knowing when something's up. You'll need to learn how your drive works which will take a few passes, but by this point I can tell how much trouble a disk is going to be just by how it sounds when I first put it in and the drive spins it to check for an index pulse. Quiet disks are great, loud ones will probably be difficult, and if makes a noise that restarts and fades every time it moves to a new track you're almost certainly going to have to stop to clean the drive at least once.

Marginal and Damaged Disks

I am not an expert on this by any means, so take this with a grain of salt. This is just what I've learned and my current understanding of the problem. This does not affect all disks, but it can affect any. If you haven't ever used a given disk before, don't trust it until you know it works.

Physically, a floppy disk consists of a plastic carrier disc which is coated with some kind of adhesive and then has a ferrous material applied to that adhesive. The outer layer is a plastic (or equivalent) shell with a fabric liner and then some kind of lubricant applied. There's several ways these components can fail:

  1. The lubricant (or adhesive) generates mold. This is an annoying problem, but usually means the data is still intact, the head just can't get to it. You will need to clean the disk. What's worked for me is cottonbuds dipped in distilled water and then rubbed along the disk, always inside to outside, discarding the cottonbud after both sides have been used. I recommend you align the index hole before you start so you can tell when you've made a full rotation. It takes me about 30-40 minutes to do each disk like this. It's incredibly boring and consumes a ludicrous amount of cottonbuds, but it's saved every disk I've tried it on. Some people will say to use diluted dish soap, but I haven't yet had that be required. Definitely do not use IPA, which can destroy the adhesive (it is a solvent, after all) and ruin the disk. You'll also want to let your floppies dry before reading them. What a sentence.
  2. The disk just gets dirt on it. Same basic solution as the mold, really, but you'll need to be gentler.
  3. The disk is mechanically warped. I don't really have a solution to this one yet, as it's only happened to me once and it was for a copy of dBase III that's already very well preserved so I didn't think it was worth trying. In theory you'd probably want to cut the disk open and extract the actual inner disk (or "cookie") and insert it into a donor sleeve, but again, I haven't tried that yet.
  4. The adhesive loses its adhesion. This one is very bad--you will only get one or two reads off the disk before it disintegrates and renders itself unreadable, and in the meantime it's going to be shedding its magnetic material directly onto the read head(s). If you clean the heads with a cottonbud and they come back looking clean, your problem is probably mold. If there's visible black spots, you may have this problem. If you suspect this is happening, stop the instant the drive does anything suspicious and clean the heads before resuming. If you wait too long, enough magnetic dust will build up that the head collides with the disk and starts gouging it, making a horrific noise while destroying the information on the disk.
  5. The heads stick to the disk. I've only seen this once and I honestly have no idea what was going on there, but it was repeatable on that disk. If you have a disk that feels like it's stuck in the drive, don't yank on it.

Cleaning heads can be done in one of two ways. Either you get a special cleaning disk that's just some felt in a standard sleve and then drip IPA onto it and insert the disk (the Greaseweazle software has a special clean command to try to do this as well as possible), or you put some IPA onto a cottonbud and very very gently run it across the head(s). The latter will require you to move one of the heads to get proper access, make sure you're very gentle with this as you can theoretically misalign a head doing this though I haven't yet had that happen to me and I've run hundreds of cottonbuds through my drives this way. The former I've heard mixed reports of it also being bad for the heads when done frequently since it's more abrasive, but if done rarely it's harmless. Since neither solution is perfect, my personal stance is cleaning disks for preventative cleaning and cottonbuds for reactive cleaning. This also helps preserve the much rarer cleaning disks.

There is a trick done with archiving tape media that involves "baking" the tape in a moisture-free environment to try to restore the adhesive (not just a regular oven, there's specific devices for this). I've heard that this can work for floppies too, but only as a rumor-of-a-rumor, I've yet to even hear of anyone doing it themselves. If you have disks that otherwise appear to be a lost cause this may be worth looking into, but it will require removing the actual disc from its protective casing so this is a destructive operation, and I can't offer you any real advice on the details.

After the Read

Once you have your .raw files, you'll want to convert them to something more useful. Greaseweazle provides a "convert" command to do this; most times you'll want the ".img" file format though you need to again tell it which disk format that is (such as --format ibm.1200) as an img file can be anything. These among others can be loaded into 86box. If your disk has copy protection, try .hfe, which can encode more advanced information (but is larger and less broadly compatible). For non-PC platforms you'll probably need to look up what formats are standard there; for example Amigas use .adf files. Most of these were defined in the 90s before any kind of standard existed so there's a lot of very similar file formats out there. Greaseweazle can read almost everything and write to most of them, and even if Greaseweazle can't write to whatever format you need there's probably some other utility that can do it, probably by converting an img to whatever platform-specific format.

If you just want the files, 7-Zip can open most PC disks that are stored as .img. It will have issues with disks written by DOS 1.x which used a very different disk header, but seemingly so does everything else, I've found practically nothing that can read those aside from mounting them in actual DOS in 86box so their contents can be copied to a "new" floppy image. HxC (see below) can also extract files from disks, but doesn't preserve modification dates. Disks made by CP/M or UCSD p-System are functionally unreadable like DOS 1.x, at least as of the last time I tried. Disks from other platforms will probably need tools specific to that platform.

Other Tools

There's two other pieces of software worth mentioning that perform similar or adjacent functions:

HxC is a floppy drive emulator and can present a visual view of your disk, highlighting bad reads. It can also export files from PC floppies (though it seems much picker about disks than 7-Zip) and show you a very low level view of a disk's data from flux reads. I've found it invaluable for examining reads of marginal disks to determine which sectors I want to try again as well as when reading disks of unknown format as you can use it to determine the layout of a disk. For example I used it to verify a disk was 96TPI, 10 sectors, FM encoding (I haven't covered encodings here, but there's only a few and they control how data is stored as magnetic flux) and a brief visit to Wikipedia later confirmed that that's a format used by Acorn. I've also had HxC manage to recover sectors that Greaseweazle said were bad when there were only a few bitflips in the flux. On the negative's side HxC's CLI is nowhere near as friendly as it could be in my opinion. For example: do you want to extract all the files on the disk? You can't. You can extract one at a time, and you can get it to generate you a list, but the list is not in a format that can be fed back in to the get file command, and even once you work all that out it doesn't have an option for the output path for the file, it just always puts it in the working directory.

FluxEngine is an alternative to Greaseweazle that performs a similar function though in my opinion it's not as intuitive. I've had some luck using it to read Mac formatted 3.5" disks as it seems to be slightly better at that, or at least better at providing feedback. Like HxC I think its CLI is more awkward to use than Greaseweazle's so I only use it when I'm already manually intervening, but it is still far better than HxC's.

Putting It All Together

At this point I've written my own tooling to automate most of this, but reading a disk works like this for me:

  1. Read track 0 and open it in HxC's track viewer to identify format
  2. Read the rest of the disk using --raw --format=... as above, in as many passes as it takes depending on the condition of the disk
  3. Assemble the individual reads into one good read, or at least the best possible
  4. Convert the flux files into .img and .hfe
  5. Use HxC to export the flux read as an image (-conv:PNG_DISK_IMAGE)
  6. Extract the contents of the disk

Reading a floppy then consists of the following steps, at least for me:

This gets you a canonical copy of the disk in raw flux, a usable copy of the disk in img, and a folder full of the files on the disk if applicable. If you ever need to write the disk back to a real floppy, write using the flux as the source data.

Don't Use Floppies

This may seem a strange note to end on, but there's a reason a decent chunk of this has been about dying floppies. Dying floppies can destroy a floppy drive, and neither are being made anymore. While the aesthetic of swapping disks is fantastic, in my opinion it's not worth the risk. Just get a Gotek emulator (but get one of the ones with the OLED displays, they're way easier to use) and flash it to FlashFloppy if it didn't come that way. It won't feel as cool, but it'll make actually using your retro system much easier and won't make you have to wonder if a 30 year old disk is going to still work this time. If you have old disks, image them now before they have even more chances to go bad, and then use those images rather than the physical disks. As of this writing a Gotek is about $30-60 depending on model and seller, and that plus a flash drive is much cheaper than replacing a damaged 5.25" drive and competitive with a 3.5" drive.

Imaging Optical Media

The "why bother doing it this way" of optical media is the same as that for floppies, really: copy protection. Just copying the files off the disk is not a guarantee you'll be able to use your backup when the original inevitably fails. Software like redumper will instead read the disk as if it was audio which permits much lower level storage. It also takes up more space, though not to nearly the degree that flux does for floppies. Most of my own CDs have wound up around twice their original size when archived this way (though on the other hand doubling the size of a CD consumes much more space than 10x of a floppy). The gold standard for this process is redump.org, who have an entire standardized process.

The optical disk version of this process is much more straightforward and much more annoying at the same time compared to the floppy version. To actually archive an optical disk I highly recommend you check out MPF which will walk you through the whole process (Windows only, at least as of this writing, but you can also just control redumper yourself from a CLI if you want and/or you're on Linux). You just plug in your drive and you're off, no worrying about TPI or any of that. The problem is that low level CD access is not nearly as universal as it is for floppies, so you need a drive that redumper can use for this process, and which ones can is effectively completely random. The drive needs to be able to read into the area before the start of a track, and you need to make sure that your drive is sending data in the order that redumper expects, as there's no standard, and some drives either won't send all the optional data or won't respect the commands to perform a raw read of a data track at all. Naturally, known good drives for this are very popular and thus very expensive unless you can find somebody who isn't aware that the drive they're selling is blessed in this way. I got lucky and a random Blu-Ray drive I bought at Microcenter for like $25 back in the day is one of the known good ones. There are compatible drives not on redumper's known good list which are much easier to find for sale, but you'll need to do your own research to figure out what they are because as soon as someone realizes what they have that drive's price goes from $20 to $200.

Generally if you have issues reading a CD or DVD, either try again or try slower (all the way down to 1x if you can wait). Optical media is surprisingly resilient to damage on the bottom layer, though not immune. If your drive is generating a ton of errors on every disk you try you probably have it configured wrong--see redumper's GitHub page for what to try based on how it's failing.

The one downside is that while you can successfully image a copy protected CD, you generally cannot use that image file to pretend to have the same copy protection. If you backup a CD that you own and you want to use that backup to play a copy-protected game, it probably still won't work.