Subject: daxwav lossless audio compression Date: Mon, 11 Dec 2000 18:10:19 -0400 From: "D.A.Kopf" Organization: DAKX, LLC To: rw@firstpr.com.au Hi, I saw the post in comp.compression about the update of your web site (thanks for the links), and thought you might be interested in testing my W9x version of daxaif, http://dakx.com/aps/daxwav32.zip I have only a W95 box for development, but have been told the program runs on 98 and NT. I've been hesitating to release it because the speed is not representative of the compression method (it's written entirely in C with byte i/o) and also the interface seems kind of unexciting. But the compression factors should be the same as for the Mac version, and the .dax files are cross-compatible with the Mac. You can link to the zip file if you like, and I'd be happy to hear any comments and answer questions you might have. By the way, in your description of DAKX you could say it's kind of the complement of lossy ADPCM such as ima4 - instead of changing the step size within a fixed word size based on the slope of the waveform, it changes the word size for a fixed step size (=1 for lossless) and thus can give an efficient way of tracking small changes in integer size over a large dynamic range. When a lot of values have similar probability (e.g. lower bits are noise) this can do better than entropy encoding. So it's not really a differencing technique as such, although it does use the first or second derivative for audio. There is an implicit repeat count method built in as well, triggered on the 4th repeat in daxwav, which doesn't add much overhead to normal audio but comes into play during periods of silence.