Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time



This project started in 2015. I had 300 floppy disks that I kept from my Amiga 500 and Amiga 1200 period. I planned to buy a second hand Amiga after I sold them for a new PC. That of course didn't happen. The data stored on the floppy disks was mostly my personal files, videos, music modules, painted pictures, 3D objects and renderings etc. And a complete backup of my system and data files. I wanted to reconstruct the files that I used to have.

However, the Amiga disk format is not something any PC could read, at least not in Windows in a simple way. The disks were in a poor state with fungus and dust. So I set out to clean them. The 3D printable Floppy cleaning kit was a result of this.

This is the stategy I followed to get the most data without the risk of losing it:

  • The most valuable data will be recovered once I can be certain the recovery proces is sound
  • No captured data is thrown away. The first capture could be the best one because the disk may disintegrate/scratch during the capture process due to the disk surface letting loose.
  • Extensive analysis is done to make sure that every opportunity is taken to get the data. This includes doing oscilloscope captures of the analoque flux signal, error correct and looking for duplicate data for reconstruction
  • Low cost, off the shelve parts
  • Windows application to control the hardware (C#)


This is the first commit of the project. I will call it version 1.0. It's capable of capturing Amiga and PC DOS disks and some variants of both. MSX disks for example are also supported. All disks can be captured, though not all disks can be decoded to a disk image yet. You could try adding your own.


The first version I made was based on a PIC16F1938 overclocked at 80MHz. An USB2RS232 donlge with a PL2303 chip was used at 5mbit/s. The second (current) version uses a standard Arduino Due. I've designed a shield for it so it can easily be built. The native usb is used which eliminates the USB2RS232 dongle and makes things a lot more stable.

Arduino Due shield pcb


The real magic happens in the software that converts the flux transitions (or periods) into MFM, then into data and finally into sectors and disk images.

The following disk formats are supported:

  • Double density and high density (720KB and 1.44MB) 3.5" disks
  • 360KB 5.25" PCDOS disks.
  • M2 special format for storing larger amounts of data
  • AmigaDOS and Amiga DiskSpare disk formats. PFS is also supported since the format is similar to AmigaDOS.

File formats that can be imported:

  • .bin files produces by FloppyControl App, a simple stream dump file format. Each byte contains a period of 2, 4, 6 or 8us. Timer tick frequency may vary.
  • .SCP files can be imported
  • .raw files from Kryoflux can be imported.

There are some experimental error correction functions in FloppyControl that allows you to do some analysis and limited error correction. It's still a very time consuming and manual process.

Other features include:

  • Oscilloscope analogue flux transition capture support. At this point only the Rigol DS1054Z is supported and tested. Other scopes may work but it's not tested. Please let me know if it works. A modified floppy drive is needed for this. The oscilloscope should have at least 20MB storage to store a full track in memory.
  • Oscilloscope captures can be edited so that errors can be corrected manually. A waveform editor is included.
  • The waveform can be converted into sector data using the processing capabilities (differential filtering, time based filtering, period conversion and data conversion).
  • A sectormap shows you exactly the state of the capture and conversion. Bad sectors can be recaptured and existing data can be reprocessed with different settings, allowing you to 'scan' for good sectors in the dataset.
  • All the graphs have some kind of interaction with them.
    • The histogram can be clicked, this will activate automatic peak detection for the MFM processing. This will find the optimum settings for you.
    • The scatterplot will zoom in and out using the scroll wheel and you can drag to move the view to another location. For PC data, the periods can be edited (enable editing to avoid accidental changes).
    • The Oscilloscope graphs can likewise be moved and zoomed. It has some advanced scaling routines to display the waveforms like an oscilloscope does, so you can see subpixel detail to find weak spots in the signal easier and faster.
    • The error correction map can be used to locate problem areas by overlaying two captures of the same sector (if the header is crc ok). This may help pinpointing the problem area and allows you to try automatic error correction (using guestimated brute force).
    • Error corrected sectors will be copied to the sector map.
    • There's some rudimentary multi threading support but it breaks in most cases. Try if it works, it'll speed up processing up to 3x on a quad core.
    • FloppyControl uses a base folder in which the captures are stored. The Base filename is used to create a folder and naming the files inside that folder using a counter. It will never overwrite data (for .bin, .prj and dsk/adf files).
    • bin files with bad sectors can be saved as bad sectors only. This makes them much smaller for reprocessing and finding that illusive good sector.
    • You can choose 'bad sectors only' during processing. This does the same as saving just the bad sectors, it'll reduce the amount of data processed (all good sectors are skipped, they are found already so no point in reprocessing them).

FloppyControl app

The sources are released as GPL v3.0.

The firmware can be compiled using Arduino Studio v1.8.8.

The project makes use of DPLL class which is based on US patent 4,780,844 and written by DrCoolzic (Jean Louis-Guérin) and gracefully allowed me to inlcude it in this project. See the file DPLL.cs for more information.

The FloppyControl App can be compiled using Visual Studio 2015 or 2017 community edition. You may need to install VC redistributables to run the software as stand alone.

The PCB for the PIC and Arduino Due shield version is created in Eagle v7.3. Newer versions of eagle (free/community edition should work).

Please let me know if you have any questions. Contributions to the project is appreciated. Please send the pull requests, I'll see if I can integrate them.

More information

See my blog for more information:

See videos about the use of FloppyControl App: