A new, perfected DolbyA (well, not Dolby endorsed) compatible decoder.

John Dyson

Member
Jul 2, 2018
41
1
13
First -- this project DOES decode DolbyA encoded material in a superior way, but isn't *really* Dolby related (they said that they weren't interested in the market anymore -- surprise!!! So, the project is now complete -- source code has been protected and pretty much stashed away. It matches original HW or is actually better wrt various kinds of distortion. Frequency response is a bit deficient (+-0.35dB wrt a real unit.)
Got some information below:

We now have a compatible decoder that does a good job on DolbyA encoded material (we call it DHNRDS) that really excels -- with all of the anti-IMD work (similar to the Orban patent US6205225, but more encompassing -- the results are outstanding. I have some published material that is DolbyA encoded and have produced comparisons and open loop examples. Finally, the decoder has been tested against DolbyA HW, and also before/after on actual recordings. It tends to do much better on cymbals, multiple vocals and any time that there is a transient of some kind. The general modulation distortion is much less (listen to the ABBA examples, with the IMD veil on the DolbyA versions -- that is NOT a noise modulation -- it is IMD.)

*THIS THING IS REAL -- and now have a recording engineer working with me* The professional side of the project is moving forward, but I am the technical side, so discussing something actually real, with a license manager, etc. This will NOT be profit making -- it is for improving the quality the reaches the consumer (and for archives -- but that isn't operative in this discussion.)

Previous versions were 'interesting', but this version matches the actual DolbyA down to the smallesst attack/release -- except runs in the digital domain (no conversion) and for the DolbyA HW quality levels (actually a bit better) runs in about 5X faster than realtime. In the highest quality modes (like for MFSL) runs about 1.1-1.2X realtime for a 4core machine, and proportionally faster for more cores.

The comparisons are on the following sites (and might be the most informative). These include some 'undecoded' examples from commercial releases, some 'decoded' versions (with the decoded or DHNRDS designations) of the same material. These are to show the 'undecoded' state of the material and some improvement from the decoding. There are also some commercial decoded copies (in the case of the ABBA examples, Polar Music releases.) Also, there are a few vinyl examples.

All demo copies are commercially available -- nothing under-the-table. The consumer level DolbyA material IS real.

Note that the DA decoder (shorthand name) provides less 'hash' around signals/vocals, brings chorus vocals out more cleanly, and also transients (cymbals) are much more clear, and don't get blunted like a true DolbyA unit.) I apologize for the ..mp3s, but .flac doesn't play on Dropbox. If anyone wants more detail, I can provide flac examples -- they are definitely better in most cases. All examples are 'snippets' and not full recordings.

https://www.dropbox.com/sh/mduka8faqv1nva7/AAATBBBRIFDht8pVsDN5Dv7Aa?dl=0

Also, an isolated example for a recording from the 'Cars', showing the original 'undecoded' state, and the properly 'decoded' state using the DHNRDS. Note the improvement of the compressed/flat stereo image back to what it should be.

Undecoded:
https://www.dropbox.com/s/y8czwibps2chg1n/Cars-06. Moving in Stereo-undecoded.mp3?dl=0
Decoded:
https://www.dropbox.com/s/bfertqaeabt5dnd/Cars-06. Moving in Stereo-decoded.mp3?dl=0

Also, I put together decoded copies from the 'Cars' 'Complete Greatest Hits' which are apparently still DolbyA encoded without decoding. One might notice that there is a LOT more detail than the original -- I also have another 'Greatest Hits' type CD from the Cars -- the additional detail on these is very significant -- EVEN with the .mp3 versions. I also included .flac versions to provide significantly more detail.


https://www.dropbox.com/sh/vdq7vh9j7cy1usc/AAA6jiENb65J3c9eMrr-zAT5a?dl=0

I am hoping that the distributors might some day stop distributing undecoded versions. I am NEVER going to make money from the project (maybe a few $100, if that), but the goal is to improve the sound quality of music from between the late 1960s through the early 1990s'.
 

astrotoy

VIP/Donor
May 24, 2010
1,547
1,017
1,715
SF Bay Area
John, thanks for all that information about CD's sometimes not decoding Dolby A. I can see why people working from digital files which have the dolby A included don't want to go to the analogue domain and decode in real time, particularly when for many reissues, the mastering engineer is not paid much for the remastering.

From time to time I get the opportunity to buy dolby A encoded tapes (copies of masters, etc). I don't have a dolby A decoder, so I have never bought one. I do have a custom dolby B decoder which does very well with the dolby B encoded tapes. Will your dolby A decoder work with an analogue input (like a dolby A encoded analogue tape) or does it need to have a digital file to do the decoding? I can rip to 192/24 aiff or flac (with Pyramix and my PM Model Two or a lower rate if needed.

Thanks, Larry
 

John Dyson

Member
Jul 2, 2018
41
1
13
My DolbyA compatible decoder (it is called DHNRDS -- there is a www site for pros) is digital only -- takes .wav files specifically 16bit, 24bit and floating point. It always outputs 24bit or floating point as appropriate. Also, it creates/uses the BEXT and RF64 stuff as needed. (It always creates a BEXT even if one doesn't already exist.) The manual is only available through my project partner, but I can make time limited versions (generous) free for consumers -- it will NEVER be a consumer product for obvious reasons (pain in the b*tt customer support issues.) I can make a 1-2page quick reference for a consumer who really wants this thing (but I only mildly encourage it -- not for the faint hearted -- command line only, even on Windows!!!) The decoder does 96k and 192k at professional quality, and works reasonably well at 48k. I have to say that it sucks at 44.1k because the bandwidth isn't wide enough to deal with the gain control sidebands. At 44.1k it sounds metallic.

Regarding using flac or other formats -- I use Linux for all of my stuff, but Windows can be similar. I use sox to convert -- so my command lines sometimes look exactly like this (partially copied from a recent command line -- the command line can be MUCH MORE simple, but this is what I do):

> sox -v 0.50 inputfile.flac --type=wav --encoding=floating-point - treble 3.0 3k 0.707q | ~/ap/nrs/dabuild/da-avx --ingain=6 --outgain=-1.0 --info=2 --tone=-20.00 --finalp | sox - outputfile.flac

* Precisely what the above command does:

First, use SOX to convert
decode "inputfile.flac", but cut the gain in half at first (-6dB),
convert whatever format (flac) to a wav file floating point.
undo the -3dB@3k/Q=0.707 EQ that is sometimes done on leaked recordings -- so give it a boost to 'undo' the previous de-emphasis.

decode (with da-avx),
an input gain of 6dB (makes up for previous 6dB cut), output gain of -1.0dB, calibration tone of -20.0dB
(the --info command gives a running log of the input/output/gain control)
The mode is '--finalp' for '--finalizeplus'

Last, convert back with SOX.

It publically has 3 modes, but there are actually 6 useful modes, where I only talk about 4 of them. The first two (lowest) modes are very conservative, both are similar to DolbyA HW quality. The lowest mode (--basic) is absolute bare bones, and is the fully functioning equivalent of DolbyA HW. There is also a mode (--basicx) that does attack/release trajectory shaping (still the same times -- fully compatible), that mitigates some IMD, and 99% of the time sounds better than anything else available (the distortion veil, the vocal separation, etc) The lowest modes run about 5X faster than real-time on a 4 core Haswell Intel (prob works on higher end AMD also), but can run no faster if there are more cores.

The next two modes are do heroic optimizations, along with the trajectory shaping for decreasing IMD -- it actually does quadrature/hilbert tricks to aggressively miitigate IMD. (IMD is the big evil from fast gain control systems -- incl DolbyA.) One mode is about 2X faster than realtime on a quad-core, but can be up to 5X faster than realtime if lots of cores (maybe 8). The highest mode is approx realtime - (a bit slower on a Windows box, a bit faster on Linux box), but is something that one does ONCE (other than calibration tweaks to get absolutely the lowest distortion -- tedious, but worth it.) I *always* use the highest quality mode unless the material is fairly dirty (the fuzz from a normal DolbyA compatible mode helps to hide evil), or I am testing. (--finalize is the second best, probably most used mode -- then --finalizeplus is the *really good* mode.)

SO, the decoder can run 5X faster, 2X faster or approx realtime depending on the quality desired. The 'calibration' is equiv to matching the tone levels, and the decoder does provide the 1st order information when reading the tones (IF you have tones) to set the calibration close. But since the DHNRDS (DA for short) is capable of very high performance, the curves and timing must match to get the very lowest distortion -- so instead of perhaps +-0.75dB accuracy needed on a real DolbyA, the DHNRDS can benefit all the way down to 0.1dB -- so it CAN be tedious, but the results can almost be a religious experience (refer to the surprising Cars examples -- really surprised me!!!)

I'll probably keep the decoder available forever -- so if you do decide to get a DolbyA tape -- contact me (send me a private message, and I can send contact information.) If you are a serious professional, I'll direct you to my project partner (he is a moderately well known pro in the field) -- he can supply a better manual. The professional price is a nuisance fee -- the decoder is VERY inexpensive considering what it can do. A new professional windows license list price is probably more expensive than the decoder. (I know the price, but I don't sell things -- esp on someone else's forum.)

I have probably told a longer story than desired -- but this is a good first step to explain a little about the REAL program.
 

astrotoy

VIP/Donor
May 24, 2010
1,547
1,017
1,715
SF Bay Area
John, thanks again. I am definitely an amateur. I will contact you privately if I get hold of any dolby A tapes. Much appreciated. Larry
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing