Skip to main content

I have just produced a usefully new version of the DolbyA decoder. The behavior is so exacting that it comes very close (if not superior in distortion) to the real DolbyA WRT spectograms and resulting spectrum response. The only caveat is that there is no real LF limit on the current version, so like on pipe-organ material, you might be suprised about the amount of bass (the online spectral example is such material.) However, look at the higher freqs for comparison. That test was done with a few days older verison of the decoder where itmight not quite have had the HF performance of the current one.
The decoder is infinttely more useful than most people think. There is TRULY a significant amount of HF boosted, harsh sounding material from before the 1990s, and often that really is DolbyA encoding. My guess is that at least 25% of the digital copies of the old recordings have DolbyA left on them -- my collection shows over 50%.

The decoder cleans up the sound, removing the hiss, excess HF emphasis and compression, and can help recover the spatial depth of the sound in the older recordings.

This is VERY CLOSE to a real 360 in many regards, but distortion has been minimized to a heroic extent, with even the gain multiplication process being done in a very tedious way to help minimize/cancel distortion products, and the 4 bands are split further to keep the distortion products from multiplying.

The price is free -- no nagware, no timeouts or anything like that. It runs on recent Windows machines, and prefers at least the 3000 series of Intel or also recent AMD chips, but will limp along on recent ATOM chips.

The decoder is a normal windows command line program, actually developed on Linux. It comes in a zipfile, which needs to be unpacked and copied to where you like to keep your binaries. There are dll files that also need to be included. It can be run as-is in the location where it is unpacked for verfication reasons.

The repostiory is available, and there are clean sounding examples (with the before versions given the ORIG- filename prefix.) PLEASE LISTEN TO THE PROCESSED VERSIONS FIRST without the ORIG- prefix, because if you listen ot the undecoded verisons, your ears will get used to the hyped up highs in the undecoded copy.

There is a pdf on the repository that shows how to use it -- but it is generally a standard in/standard out program, while the --inf=infile.wav and --outf=outfile.wav can be used. It works with most standard, little endian, 16bit, 24bit and floating point .wav files, at 44.1k through 192k inclusive, but much prefers 48k though 96k -- and 72k is the theoretically ideal, but not necessary -- 96k is probably the best otherwise, with 48k a close second. The decoder runs directly at these rates, and the code was written very carefully to maintain freq resp and timing for all of the sample rates. The threshold setting is usually between -14.5 and -16dB for commercial recordings, where I had found one at -16.50, but that is seldom. It likes to be set within 0.25dB, while 0.10dB accuracy is sometimes needed. A good initial value that is probably good enough for a quick demo is -15.50. It is possible to find the best level by listening carefully, but the criteria is complex. If there is enough interest, I can write a paragraph or so about the procedure -- but for right now, the -15.50 is good enough. If your material has DolbyA tones, then the threshold is usually best at 1dB less than the tone input level as measured by the program (--info switch.)

I promised myself not to release any more versions until it was perfect, and it is as practically perfect as it can be now (modulo the excess -- not boosted -- LF response.) Absolutely NO digital sound artifacts, and NO aliasing. A lot of software processors aren't written to avoid aliasing, so they impart a bit of a harsh sound -- but there isn't such a problem at the recommended sample rates. 44.1k does have some problems due to the lack of room for the gain control sidebands, so there can be a bit of a metallic sound at that sample rate.

The fileaname of the decoder is something like, and grab the one with the most recent date and the largest A,B,C...
The filename of the help file is DecoderA.pdf.
You'll see examples, but PLEASE do not play the files starting wtih ORIG- first, because they will change your hearing so that a correct sound will sound dull, so it is best to play the processed versions first.


Have fun!

Topic Tags