Good news -- new DolbyA decoder available.

[johndyson10]

This new version of the decoder is very good at removing the harshness from before 1990's recordings on digital media. It doesn't work on every recording, but does work on perhaps 1/4 of them -- in my collection is is about 1/2. Recordings from MFSL will NOT benefit, because they do their stuff correctly. However, many of the quick playout type CDs and digital downloads haven't been properly mastered, and that is where the DolbyA decoder is useful. This decoder will work on almost all DolbyA material (everything that I have tested.)

Rather than getting into all kinds of discusisons, just refer to my repository -- and listen to the before and after examples (the before examples have a prefix of 'ORIG-' in the name.) Always, listen to the cleaned up versions first (the files without ORIG- in the names),becuase ears can get used to the harshness pretty quickly. However, the decoding IS quite accurate -- and is within about 1dB of a real DolbyA cat22/360, the attacks/releases are also very close to identical and the spectographs are almost identical, except for the lack of distortion products (can have a long discussion about the heroics done to avoid distortion in the decoder, but the work has been successful.
The best way to determine if something is DolbyA encoded is to try the decoder, and the lousy results are obvious if the material is not encoded.
Again, later on, I'll get into the technical details (lots of really *new* stuff in the decoder),but the quality is supurb now.
The decoder resides in a zip file, and when unpacking you need to keep the dlls with the program so that the libraries are available.
The decoder is NOT nagware, has no ability to make internet connections, has no timeouts and is free to use.
The examples reside in the same directory as the decoder, and you can play the examples online if you wish.
Again -- DO NOT PLAY THE 'ORIG-' files first, or sometimes ears can get used ot the intense high end and hiss.

Location: https://spaces.hightail.com/space/tjUm4ywtDR

Let me know if you have troubles. There is a help file called 'DecoderA.pdf', which might help you get started.

The program is a command line 'stdin/stdout' type program, and uses input and output redirection if you wish, but also
can use --inf=infile and --outf=outfile type swtiches if you want. It works with 16bit, 24bit and 32 bit floating point little endian .wav files at 48k -> 96k ideally, but works moderately well at 44.1k and somewhat better than 44.1k at 192k. The decoder is much slower at 192k, and the quality hits a limit at about 72ksamples/sec. So, I suggest using sox to convert to/from 72k if you want, or just use it at 48k or 96k for best results.

Also, there is a necessary --thresh=xx.xx parameter in the command line. That corresponds to the 'DolbyA' tone level. If you use the --info switch and watch the levels, when you measure the input level during a DolbyA tone, that is 1dB higher (that is, subtract 1dB) for the threshold. If you don't know the threshold because of no-tone on the recording, really good ears can find the correct value that needs to be within about 0.25dB or perhaps down to 0.10dB errorr. However most commercial recordings seem to require a threshold between -14.5 and -16.0 dB, where -15.50 is a good start. I have one recording that seems to need -16.5dB, but that is one out of several hundred. If the threshold is incorrect, it won't sound horrible, and will still sound better than the original, but won't sound ideal. I have taught my ears be able to get the threshold correct, but there are numerous little things that need to be considered. For now, the details aren't important.

The improvement can be amazing, and might even help you forget vinyl -- because this bridges a big part of the difference between the 'harsh' digital sound and the lush sound of vinyl.

PS -- the program works on 64 bit Windows10 (probably Windows8 also),recent CPUS like Haswell (might work with 3000's, but definitely 4000''s or better.) Should also work on recent AMDs. Also will work on recent ATOMS (a special version that runs quite a bit slower, but it works.) I can build a version for machines as early as P4, but I have had no demand. If there are any requests, I'll look at it -- I built P4 versions a few months ago -- shouldn't be any trouble if needed. 32bit versions are also possible and have been built before. I am not smart enough to merge all of the versions into one binary -- actually, just haven't had time to do it yet.

John

 

[Irvrobinson]

I've said this before, and I'll say it again, John. You really need to write an AES paper about this. The improvement is easily audible, and makes a strong case for your supposition. I find the differences between your before and after files are most pronounced with headphones.

 

[johndyson10]

I've said this before, and I'll say it again, John. You really need to write an AES paper about this. The improvement is easily audible, and makes a strong case for your supposition. I find the differences between your before and after files are most pronounced with headphones.

Thank you -- and I agree that the matter should be publically exposed. I have been so limited until apprx now. The decoder has gone through so many iterations -- where it feels like I would get so very close to correct, and then not be able to get there. I have been FULLY occupied with the decoder -- even though I already knew quite a bit of DSP, I have had to learn and study much more. This has NOT been trivial.

I hope (and pray) that the decoder is now settled down to where I can actually think more strategically, rather than the simple tactical matters of getting it working perfectly.

At his point, I can deeply and honestly (without testosterone or adrenalin) say that the decoder is so close to a real DolbyA in performance, that it is interchangeable for the decoding process, even for pro applications.
Before, it mostly just made things better, but always missed the mark in one way or another. YOU WOULD NOT BELIEVE MY MISTAKES!!!

Once I could get spectograms and spectrums that nearly match (I mean within 1dB),and also be able to tune out (remove) the most minor encoding artifact -- I am now pretty much satisfied. The only mismatch is that the decoder currently has only one rolloff for LF at 10Hz, and three poles at 4Hz. This means that some kinds of material with lots of bass will REALLY have lots of super low bass (e.g. pipe organs.) But the sound is tremendously good even in that application. (The temporary fix is a single pole LF rolloff at 44Hz on the input. Most tape decks are not accurate at that frequency anyway.)

The paper would be a good idea, but not to 'sell' or 'tout' the decoder, but to educate as to the massive amount of DolbyA encoded material out there. (I do find it to be hit or miss -- in my most recent 3 aquisitions, two were NOT DolbyA encoded. But, more than half of my current total library IS encoded.) My decoder should NOT be the only decoder out there, and entrepreneurs will certainly find ways to make the decoders more efficient, yet still maintain quality. I am just first in the line (there is another DolbyA that does both encoded/decode from Satin -- but it doesn't really sound like a real DolbyA decode.)
Again, Thanks. My biggest impediment to writting a paper now is the request that I write a DolbySR decoder. The difference between DolbyA and DolbySR is like the difference between using Neosporin on a cut vs. brain surgery... (Not quite that bad -- but it is really complex.)

I think that I will deeply consider now starting a paper (not really an expose, but rather a kind of research paper describing what I have found on commercial recordings, and my potential solutoin to the problem.)

John

 

[Irvrobinson]

The paper would be a good idea, but not to 'sell' or 'tout' the decoder, but to educate as to the massive amount of DolbyA encoded material out there.

I wasn't suggesting that you would use the paper as a vehicle to tout the decoder, just to share the knowledge. IMO, your conjecture falls into that forehead-slapping why-didn't-I-think-of-that category.

I was just listening to my old (1987) CD of Paul Simon's Graceland album, and it sounds suspiciously bright now, like one of your examples before decoder processing...

 

[johndyson10]

I apologize for implying or even misthinking -- I do NOT think that you might have been implying anything wrong no matter which way you meant. I so much appreciate your comments -- whether or not I end up being a crackpot. I took you comments with the greatest respect, and didn't mean to be contrary in any way!!!

You and other insightful people see that I am trying to present a new idea -- history has shown me acceptable skepticism, but also total denial and sometimes rude personal rejection. It is important to avoid being pushy so that the 'wall' of disagreement doesn't become strong. I so appreciate your kindness, and I respectfully accept anybody's kind skepticism also. People with some insight likely realize that my ego doesn't need agreement, just acceptance that I have an idea, and simply appreciate a bit of 'support' of some kind (e.g. maybe write a paper? type of question.)

There is a fine line of helping to inform that many CDs might very well be DolbyA encoded without appearing to be an irritating evangelist I have always been skeptical about the alleged reasons for the 'digital sound.' Maybe we finally found the answer after all of these long years? Maybe DolbyA is at least part of the reason why I quit listening? This project started as an investigation, and has grown into something that REALLY seems to work.

I still have an audio collection -- but it was more of a collectors habit instead of being an audiophile any more. If the DolbyA thing becomes an accepted fact (even I don't claim 100% fact), maybe there will eventually be a better infrastructure to support listening to all of those old CDs? My program REALLY isn't an audiophile caliber tool, and not really suitably packaged for professional use. However, I am willing to encourage others to make something more useful - if this ends up being a big thing.

If there is a big acceptance, I expect to (very inexpensively -- probably free) license my code for review and to possibly include into products that are better packaged than my own software. I know that I'll never make much (if any) $$$ from this endeavor, but it would be nice to participate in something good again!!! (The methods needed to do the decoder are probably not obvious even to a DSP expert. I can/will give assistance if requested.)

Thanks again
John