Skip to main content
Topic: Status on the DHDA (DolbyA compatible) decoder. (Read 461 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Status on the DHDA (DolbyA compatible) decoder.

Often, when writing (talking) about the DHDA project, I will mention the relative lack of IMD (intermodulation distortion.)  No amount of attack/decay finessing can do the amount of IMD avoidance as done by the DHDA.  I have invented a few gain control algorithms that look significantly different from the normal algorithms, and once the project is completed I'll probably write a paper on the technique.  It is very CPU intensive, and does some REAL IMD reduction (even though part of the algorithm hides IMD like some previous patents -- the DHDA algorithm really helps to decrease IMD.)   There is also an aspect of 'hiding' also -- both techniques are useful.  (The 'hiding' technique is described by patent: US6,205,225).   If I set the algorithm to primarily 'hide' instead of optimized reduction, then the sound isn't quite as good.  So, the mode being normally used (settable by command line) is to maximize reduction.  (The DHDA technique looks NOTHING like the Orban patent, but does use similar mathematical techniques.  Also, the DHDA method probably isn't possible in hardware, and all together the decoder barely runs realtime multi-threaded at 96k on a 4core Haswell CPU.)   Without IMD reduction, the decoder can run approx 5-10X faster than realtime (but also has a less desirable sound.)

I have attached two examples:  one decoded by the DHDA and the other the very best that  I can find (probably DolbyA decoded) commercial version of the material.   One version does sound cleaner and more distinct than the other.  It is ALWAYS best to measure -- this kind of measurement of dynamic IMD is almost impossible without a major research project -- so, I won't formally make a claim or assertion as to which is better.  You can decide it for yourself, however there is a qualitative difference.   The DHDA version is MeI-DHDA.mp3, and the commercial version is MeI-Dist.mp3.  The difference is so very extreme that it is easy to hear even on mp3.  I have done absolutely ZERO adulteration of the distribution or DHDA versions -- only normalized the signal levels -- period.

One more comment -- the DHDA distortion reduction is not normally visible in HD measurements -- in fact, for HD the DHDA looks a little worse than a DolbyA.  The kind of dynamic IMD -- which is endemic to fast gain control devices -- is coincident with the slew of the gain control -- which then splats distortion sidebands all over the place (evil.)  That kind of distortion cannot be removed very effectively by expanding (decoding) after the compression (encoding) phase.  (The LF distortion IS mitigated to a large extent, but as soon as the gain slew rate becomes close to the signal frequency, then all bets are off.)  On INTENSE material, the dynamic IMD is much more noticeable than HD in most cases.  The 'splatted' distortion on the traditional DolbyA decoding sounds like a fuzz or slightly unclear sound (in some cases, similar to tape distortion but with less smoothness.)

Do your own A/B if you are interested.  If anyone knows how to effectively measure dynamic IMD as created by fast compressors/expanders, let me know...  I have been trying to think of a mechanism that measures accurately.   (Simple IMD measurement wont' really detect the dynamic IMD very well.)


Re: Status on the DHDA (DolbyA compatible) decoder.

Reply #1
I have a full quality copy of the DHDA version of the example on my listening repo -- just fyi:

(There is also the default version -- very slightly lower quality -- 44.1k/16 bit because of space limits.)

(not arguing 44.1k/16 vs 48k/24 on flac, but the differences are much greater than can be resolved by the flac differences.)

* I have upgraded the 'dist' version to be full 48k/24 quality *


Re: Status on the DHDA (DolbyA compatible) decoder.

Reply #2
I found another example (same song), and I just upgraded the DHDA to be more precise and get more details from the music.  The math is VERY intense and requires precise timing to gather the details.

I am posting 3 examples (all three are flac for reasons of precision.)  It requires very perceptive hearing to tell the difference between the digital versions.  The vinyl version is posted for reference.

In the digital versions (the commercial 'Dist' version vs. the 'DHDA' version) -- listen for the chorus that starts approx 1-2 seconds into this example.  You'll hear some detail in the vocals on the DHDA version, while the 'Dist' version glosses over the vocals a little more.  (I call it a 'blob', but that term is too derogatory, because actually, the 'Dist' version is pretty good.) 

For 'fun', I included the vinyl version also -- kind of disappointing when compared with the two digital versions..   Even if a listener doesn't like the 'greater' detail of the DHDA (and truly lower distortion), the sound of the DHDA is quite similar to a real DolbyA.  (Part of the greater detail of the DHDA is the lower intermod -- intermod fuzzes detail, and less intermod tends to improve detail.)

I sure hope that the industry starts using a superior way of decoding the 'good old' material over and above simple de-emphasis, or even the HW DolbyA.  It is so much more convienient to use digital decoding when the material is stored digitally in DolbyA encoded form.

ALL THIS SAID -- RAY DOLBY was a genius -- and considering that he designed the DolbyA in approx 1965, it is amazing that he did it so well!!!



Re: Status on the DHDA (DolbyA compatible) decoder.

Reply #3
Who uses Dolby A encoding, and keeps it in that format?   In other words, what's the use of a decoder when no one (like me) has encoded material?  And if you have the matching encoder,  what's the use when I can record digital and not care about needing noise reduction?

If this is a continuation of another forum topic, I missed it.

Re: Status on the DHDA (DolbyA compatible) decoder.

Reply #4
Who uses Dolby A encoding, and keeps it in that format?   In other words, what's the use of a decoder when no one (like me) has encoded material?  And if you have the matching encoder,  what's the use when I can record digital and not care about needing noise reduction?

If this is a continuation of another forum topic, I missed it.
Well, there is a large amount of 'not quite right' sounding material which was mastered up to the early 1990's that is left mistakenly (or cheaply) DolbyA encoded.  (I was waiting for that question :-)). 

On the repo below, I have some original digital ABBA distributions, some vinyl copies, and some decoded copies from other (leaked) digital distributions.  I have found that the leaked DolbyA sometimes seems to be more common than normal, properly decoded material on digital (I had to look high and low for properly decoded digital material!!!)  I have found that sometimes the normal digital releases have been 'enhanced', so I tended to make my results sound closer to the 1970's vinyl (often a little more dead sounding.)    Geesh, the average consumer has adapted to that travesty!!!

PS: I also have Carpenters, Eric Clapton, Anne Murray and many others.   I bought a Carpenters album from HDtracks -- it was DolbyA encoded also.

In an hour or so,  I'll put up an undecoded copy also.  It sometimes sounds somewhat similar to properly decoded -- except for excess HF compression (not always lots of harshness) and sometimes they are EQed down if needed (sometimes -3dB or -6dB at 3kHz.)   The distributors often do their mastering 'on the cheap' and don't do the proper, rather time consuming decoding operation -- the material is already on digital, why bother with the expense of setting up the DolbyA decode and playing it through the DolbyA in realtime only?  It is easier to do a bit of EQ -- the customers dont' realize what is going on.  (Some engineers in the industry have started realizing/talking about what what used to happen also.)

The files with Orig are digital copies, those with DHDA are DolbyA decoded, and those with vinyl are from a vinyl rip.   When I get around to it, I'll put a short example that is DolbyA ENCODED, and add 'encoded' to the filename.

It is tricky to discern DolbyA copies, because on compressed material it is kind of stealthy.   They ARE commonly available for the older recordings.  There are two ways to tell pretty positively -- the noise veil on a spectogram, and the high frequency compresison.  If you try to decode something that is not encoded, it typically has troubles with HF and also can have a gating sound if it is moderately low levels.

So -- yes, high end consumers who have realized that there were problems with their CDs, and even comment that EQ doesn't quite fix the problem MIGHT be able to benefit from it.  Often, the problem is DolbyA encoding that hasn't been decoded.  And, contrary to those pontificating who do not know -- DolbyA IS listenable -- depending on the material -- it is just not quite correct sounding...

BTW -- why hasn't this been detected a long time ago?  Few DolbyA decoders have been tried with the older material -- once my very very very sound alike (but better) decoder came into existence, then it has become easy to check out the possibilities.

Here is the repo -- and I'll put up more examples a little later on:



Re: Status on the DHDA (DolbyA compatible) decoder.

Reply #5
One more things about my repository -- I have two decoded copies from Carpenters (A Kind of a Hush), one is flac -- only 44.1k/16bits, and the other is mp3/insane mode of lame (320k).  If you listen very carefully to the delayed voice in Karen's vocals -- the delayed voice is somewhat obscured on the mp3 version.  This is one of the big bugaboos about mp3 that I have noticed -- and one reason why I cannot use it.  (I think that the delayed vocal might be related to the old hybrid quad stuff like SQ.  I have my own SQ decoder that I have written, and it seems like the old Carpenters stuff has some explict ambiance added.)

I realize that this mp3 defect is off topic, but still, it is interesting that for precision purposes, even at 320K, mp3 doesn't quite cut it.  I didn't upload an opus example, but opus seems to keep both of Karen's vocals (both the normal and delayed one.)


SimplePortal 1.0.0 RC1 © 2008-2019