foo_dmftest

foo_dmftest is a plugin that allows playback of DMF modules in the foobar2000 audio player. While playback will not be identical to that in X-Tracker, the program used to create DMF modules, it should be close enough for most modules to be listenable.

• Downloads

  The current version is 0.3.3, released on 2004-12-31. Older versions should not be needed as the new versions are likely to be faster/more stable/more accurate, but they can be provided on request if it's really necessary.

  The foo_dmftest.dll plugin itself should be copied into your foobar2000 components directory, which is most likely to be C:\Program Files\foobar2000\components but could be different on your machine. If you do not have the .NET 2003 framework installed, you will need to download and copy msvcr71.dll into the same directory as foobar2000.exe or your computer will not be able to load foo_dmftest.

  • SSE-optimized binary. Uses Visual Studio .NET 2003's MMX and SSE optimizations to produce a noticeable increase in decoding speed. If your CPU can support the added SSE instructions (I think it's Pentium 3 or newer from Intel and Athlon XP or newer from AMD), this is the plugin to download.
  • Vanilla binary. If your computer does not have the necessary MMX and SSE support, use this build instead. It will be slower, but at least it will actually run.
  • msvcr71.dll. If your computer does not have the .NET 2003 framework installed on it, you'll need to copy this DLL into the same directory as foobar2000.exe. If you get an error message when you start foobar2000 stating that foo_dmftest.dll could not be loaded, it's probably because you're missing msvcr71.dll.
  • Source code, with project and solution files for Microsoft Visual Studio .NET 2003. You'll need the foobar2000 SDK installed and you'll need to change around the dependency and include directories in the foo_dmftest project, but it shouldn't be too hard to get it to compile other than that.

• Miscellaneous features

  • Linear and cubic interpolation.
  • Volume ramping that almost makes an audible difference.
  • Floating-point output. If you scan your DMF files with ReplayGain and store the information in foobar2000's database, you shouldn't ever have to worry about clipping.
  • No seeking. Wait, that's a misfeature.
  • I haven't seen it crash recently when dealing with a valid DMF file and it seems to be pretty good with handling garbage renamed to .DMF. I haven't tried to deliberately break it, though.
  • Underlying testdmflib library should be relatively portable; there may be a few glitches where I messed something up outside of an #ifdef _MSC_VER and there could be a couple places where GCC disagrees with Visual C++, but making a plugin for a Linux audio player with this shouldn't be any harder than adding big-endian support to the loader and finding an audio player for Linux with an easy-to-understand, well-documented API.

• To-do list

  The following things still mostly suck:

  • My German is pretty bad, so I'm not sure if this is true or not, but the DMF format documentation makes some reference to updating vibrato, arpeggio, and retrig a constant number of times per second rather than updating it at a specific interval each tick. Right now, testdmflib uses the latter method, which means that the arpeggio can sound very off at times and that vibrato, retrig, and tremolo, aren't completely right. There may be other effects tied to this frequency, too... I'm honestly not that sure about it. The bottom line is that arpeggios are going to be most noticeably off and other effects will be less so.
  • The triangular vibrato waveform is pretty off. It's not a simple triangle as the docs imply; I think it may be more like the sum of a sawtooth and square wave, but I need to test things a bit more before deciding.
  • Because the mixing is entirely done with floating-point values, it's pretty slow. Okay, really slow.
  • Big-endian support in loader isn't there. Anyone care to donate a big-endian machine?
  • Forward seeking is possible with testdmflib by rendering into a null buffer, but seeking in reverse has not been implemented yet.
  • Volume ramping algorithm distorts the waveforms without eliminating much of the popping. Yay.
  • Pattern delay is almost certainly messed-up.
  • Effects that don't get used very often like scratch to note and sample delay may be buggy.

• Acknowledgements

  • D-Lusion (site now defunct), the demogroup responsible for X-Tracker and the DMF format, which did some kind of interesting things like categorizing effects and allowing for them to happen simultaneously.
  • Radical Rhythms, a group of demoscene musicians who got a lot of use out of X-Tracker in the mid-90's and made enough quality music for me to have an interest in coding this plugin.
  • Peter Pawlowski, author of the foobar2000 audio player. Best audio player/audio player SDK ever.
  • Modland, a HUGE collection of module files of all sorts of formats, including some DMF files.

• Contact information

  If you have comments, questions, or would like to report a bug I haven't that already indicated I know about (see above), contact me at singergr ATE msu DOTTT edu.

Back to Potatoriffic

Copyright 2004-2005 __sdfg. All rights reserved until I get bored and put a Creative Commons license on the page.