• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

The DA/FA decoder is finally working nearly perfectly (really)

John Dyson

Active Member
Joined
Feb 18, 2020
Messages
172
Likes
90
The DA/FA decoder has been project-hell, but finally got independent of my hearing and bad headphones (ended up with the wonderfully transparent & usefully intense DT990 special edtion to avoid grey market.) I KNOW that there are a lot of skeptics out there, but trust just for a little while. I have some stuff, already prepared, that is quite beautiful. Frankly, I am just to tired right now to do everything for a release!!! The current version of the decoder is V9.0B, unlikely to change before the weekend release, because it works so well now. It is great in FA AND DA mode, finally!!!

DA: DolbyA decoder
FA: "FeralA" decoder (all normal consumer recordings are processed with this)

Firstly the original FA design was about 2/3 correct, AND WRONG EQ, but also I forgot to kneel to the 'descrambler god'. It is a monster, but reorganizes the audio. It is not a follow-on expander but is a bunch of discriminators that mathematically reorganizes the audio. The release of the masterpiece, including source, will be coming in days. I gave due contributions to the 'descrambler god' and he eventually blessed the project with his grace :). Also, I did figure out how to validate the freq response, very accurately. There does appear to be a 0.5dB peak at 6kHz, but I am still not sure. It might be correct, but my measurement method does seem to indicate correctness. (It can also partially be validated by a hack response measurement, but kind of yukky. Sine waves don't work.)

We all know what the DA decoder might be: it decodes DolbyA tapes, infinitely better than ever. 'FA' is 'Feral A', something that should have never gotten out of the lab. In fact, the ABBA snippet demo that I am going to provide today is BOTH FA and DA decoded. FA is done by an insane combination of DA units and EQ, and descrambler, which actually does the output EQ. FA encoding, the 1980's version of MQA, is what you are hearing, even on something that you think might be 'digital'. DA units are NEVER digital, until my software.

What is this FA thing really? Almost every recording, except specific boutique stuff is so encoded. It is something that used to be called 'digital sound' back in the 1980s, and is the major reason why I quit hifi listening in about 1990... The audio from consumer recordings is/was NOT hifi by my standards. (I have exacting standards, ignoring my variably bad 66yr old hearing -- back when I could hear, the standards come from orchestral recording experience.)

What is this demo? Have you ever heard 1992 ABBA Gold, or 1993 More ABBA Gold? Notice something wrong with it? It is DolbyA compressed -- THEN the normal FA processing used on almost every consumer recording. What am I demonstrating in this demo? A snippet of a decoded version of 'I am the City' from ABBA. It is NOT manipulated other than by decoding and necessary EQ -- mastering engineer apparently did a bit of tweaking. Same thing on the 1970 Carpenters album. (I can tell you what they did -- but isn't important here.) 'I am the city' sure appears to sound like a true master (not a distribution master.)

THIS EXAMPLE DOES NOT SOUND LIKE NORMAL ABBA -- IT IS DEFINITELY CORRECT. Uncompressed pop can be a bit intense. Just hasn't been compressed and homogenized... Also, put your headphones on -- stereo image!!! BTW: The necessary EQ comes from a mastering trick -- it isn't like what a normal consumer might know about... Learned by hard knocks. One more thing: you are hearing something that probably needs a slight rolloff at 9kHz or so for listening -- I am showing you the technically correct version for subsequent mastering.

Also showing a cut from an ONJ album...

ABBA Gold: https://www.dropbox.com/s/zr0gf3wsnfbzhwm/16-I Am The City.flac?dl=0
ONJ: https://www.dropbox.com/s/vamxafamapnjj0p/(16) [Olivia Newton-John] What Is Life-DECODED.flac?dl=0

Have fun -- hope to announce the decoder/demos some time this weekend, after some needed rest...
PS: Don't look for a GUI -- the decoder was already too much to do. Command line only, sorry, but VERY easy to use, almost never (really) any tweaking anymore.

project-hell is almost coming to an end!!!

PS: the decoder is FREE, and source is FREE. The only restrictions are those that others have put onto some of the components. Nothing is GPL like. The code has sections of unreadable tweaks, is imponderably complex, but is amazingly precise at the 0.001dB level or better. It is probably the single most complex piece of audio software that you would likely use.

PPS: the online player sucks badly. PLEASE DOWNLOAD!!! High quality material sounds WORSE on the Dropbox player.



John
 
Last edited:
OP
J

John Dyson

Active Member
Joined
Feb 18, 2020
Messages
172
Likes
90
Can you post a link to the FA decoder?
Very likely will do so this weekend. I am in 'pure recreation mode' right now, pestering people on forums. I promise, the decoder is coming out, I'd like late Friday night USA EST time. If you run Linux, the binary package is already made and can send directly to you now. I still need to build the Windows version, update a little usage sheet, test the Windows version, verify the demos and do the uploads. The user docs suck, and the only people who REALLY know how to use it really are those who have been helping for months or years. It is best to wait for a few days when it is actually ready.

Basically, it is always like this (except for certain, very very few recordings):
(for the avx version -- the code is complex, but also primitive).

da-avx --info=1 --fa --xpp --input="infilie.wav' --output="outfile.wav"

You can also pipe in and out like this:

sox infile.flac --type=wav - |da -avx --info=1 --fa --xpp | sox - outfile.flac

There is a da-avx, da-avx512 or da-sse. Windows is similar.
The info=1 switch gives you 'dot progress'. --info=2 gives you lots of cool gain info.
--fa means 'feral A' mode. If you don''t use that, the program will complain that for decoding DolbyA materais, you must use --forceda
--xpp is magic. It removes distortion sidebands (DolbyA fog). With the higher modes, can actually reach back and remove fog from previous NR uses in the studio.

IMPORTANT RULE and DO NOT VIOLATE -- NO GAIN MODIFICATIONS ON INPUT BETWEEN THE MEDIA AND THE DECODER. The level is important, but because of certain design characteristics, is not as quirky as DolbyA alone.

I just (really) need to rest, and will make it ready for early this weekend (USA EST time.)

THANKS!!!
John
 

Curvature

Major Contributor
Joined
May 20, 2022
Messages
1,074
Likes
1,353
Very likely will do so this weekend. I am in 'pure recreation mode' right now, pestering people on forums. I promise, the decoder is coming out, I'd like late Friday night USA EST time. If you run Linux, the binary package is already made and can send directly to you now. I still need to build the Windows version, update a little usage sheet, test the Windows version, verify the demos and do the uploads. The user docs suck, and the only people who REALLY know how to use it really are those who have been helping for months or years. It is best to wait for a few days when it is actually ready.

Basically, it is always like this (except for certain, very very few recordings):
(for the avx version -- the code is complex, but also primitive).

da-avx --info=1 --fa --xpp --input="infilie.wav' --output="outfile.wav"

You can also pipe in and out like this:

sox infile.flac --type=wav - |da -avx --info=1 --fa --xpp | sox - outfile.flac

There is a da-avx, da-avx512 or da-sse. Windows is similar.
The info=1 switch gives you 'dot progress'. --info=2 gives you lots of cool gain info.
--fa means 'feral A' mode. If you don''t use that, the program will complain that for decoding DolbyA materais, you must use --forceda
--xpp is magic. It removes distortion sidebands (DolbyA fog). With the higher modes, can actually reach back and remove fog from previous NR uses in the studio.

IMPORTANT RULE and DO NOT VIOLATE -- NO GAIN MODIFICATIONS ON INPUT BETWEEN THE MEDIA AND THE DECODER. The level is important, but because of certain design characteristics, is not as quirky as DolbyA alone.

I just (really) need to rest, and will make it ready for early this weekend (USA EST time.)

THANKS!!!
John
Will wait for the Windows version to come out.
 
OP
J

John Dyson

Active Member
Joined
Feb 18, 2020
Messages
172
Likes
90
I feel kind of guilty about the 'City' demo. IT IS VALID, but ABBA Gold is one of the very very very few recordings that needs FA & DA decoding. Of course, the DA/FA decoder can do it, but like I wrote above: DA decoding is not for the faint of heart. It is VERY VERY difficult to do when there are no calibration tones and the source DolbyA units have calibration varying between the Left/Right channels AND the 4 bands. I am one of the few people who can hear the effects of calibration errors and correct them fairly quickly. I mean, correct them to the +-0.001dB level or just a little better. Honestly, it is true, but the effects of calibration errors are somewhat profound.

The 'I am the City' demo was done basically src -> FA -> DA -> output. It was not fully optimized, but is close to correct. All 19 songs on ABBA Gold work *pretty well*, but More ABBA Gold is astounding. Sombrero is okay except Dreamworld, which appears to be damaged. The results for Dreamworld are slightly improved, but not the profound impact of the others songs. (Honestly, I haven't tried Sombrero yet, but I can tell that it will be okay by listening.)

I WILL publish the command sequences for ABBA Gold 1992, More ABBA Gold 1993 and Sombrero in the usage docs. This complexity is seldom needed, very seldom. There are a few recordings that need calibration shifts, usually --coff=-3 (which means minus 3dB.) Some Dire Straits, Supertramp, at least one Carly Simon need this mod. I have an archive of well over 100 albums, and only very few need this.

John
 

MaxwellsEq

Major Contributor
Joined
Aug 18, 2020
Messages
1,628
Likes
2,427
What is this demo? Have you ever heard 1992 ABBA Gold, or 1993 More ABBA Gold? Notice something wrong with it? It is DolbyA compressed -- THEN the normal FA processing used on almost every consumer recording. What am I demonstrating in this demo? A snippet of a decoded version of 'I am the City' from ABBA. It is NOT manipulated other than by decoding and necessary EQ -- mastering engineer apparently did a bit of tweaking. Same thing on the 1970 Carpenters album. (I can tell you what they did -- but isn't important here.) 'I am the city' sure appears to sound like a true master (not a distribution master.)
I don't really understand.

The music was Dolby A encoded onto tape and then decoded when the LPs were mastered. It was not normally possible to buy LPs that still had the encoding on them.

However, the process was dependent on correct setup of the encoder, recording machine, playback machine and decoder. Setup errors were magnified by Dolby or DBX.
 
Top Bottom