DolbyA or Pre-emphasis, which is it or both?

Your comment comes close to what I am thinking. First, I do NOT think that DolbyA encoding was intentional. I *think* that the syndrome was based upon the raw Analog-to-Digital conversion to digital archives. After that, it had been a real pain to DolbyA decode the digital copies AND they pretty much forgot about the DolbyA encoding. Then, the digital archives get utilized to produce the digital copies (e.g. CDs or whatever.) THAT IS MY GUESS.

Also, *paraphrasing* your comment about essentially NOT BEING RELIGIOUS ABOUT THE ENCODING/PREMPHASIS DETAILS... Don't decide before hand that screwed up pre-emphasis does/does not happen, and don't decide before hand that probable unintentional distribution of DolbyA encoded material does/does not happen. Just do what makes it sound best/most clean-accurate.

During some experiments, I found MORE than I had expected that DolbyA decoding was the answer. I thought that I happened to have an errant ABBA copy, knowing that it was DolbyA encoded -- but then with a fully working decoder, I found more material. My DolbyA decoder project was not meant for consumer use other than myself (professional use for others.) When I realized that OTHERS MIGHT BENEFIT, then I offer it for free.

It just seems like there is so much rejection without really knowing the situation. I don't mean to make claims contrary to others' beliefs, but only talking about what I have clearly seen. I even equivocate when I am not sure, and quickly deny that something is DolbyA encoded if I find that it isn't... I am quite sane and very honest...

John

 

If it sounds better, then it sounds better.

 

Indeed. But that is a matter of preference, not a matter of fact.

John's flawed thinking seems to be that if something sounds like it is Dolby A encoded and sounds subjectively better Dolby A decoded, then it factually is Dolby A encoded, intentionally or unintentionally.

However, that is not good inductive nor deductive logic. It is just educated conjecture that overlooks other possible explanations.

And that is an issue. Please do salt to taste all you want -- I may offer a cooking analogy later -- but do not assert you factually know based on your taste that the chef failed to salt the broth.

AJ

 

You are being rude -- i have no flawed thinking on this matter. You lose credibility by making ill-mannered descriptions about my competency.
If you try to use my decoder on non encoded material, it usually blows chunks and makes things not-pretty. Note that I didn't say 100% of the time. I have also NEVER claimed that this will work for 75% of the material in EVERYONES library. This is not a panacea, but can solve a big quality mystery. Frankly, almost anything commercial tthat you listen to has already been heavily processed so, even if I am wrong and the decoder does wonders on non-encoded material, then what is wrong with that? I have gotten responses from people who have had strong emotional manifestations (crying from beauty) because of the original music, and how the decoder cleans it up from being overly harsh, and it is more accurate sounding.

For some reason, this is attacking your belief system. Bottom line, I doubt you have even looked into it.

I have a wonderful general purpose expander that almost always works on ANYTHING, but that is because it adapts. The DolbyA decoder DOESN"T work on everything. There no adaptive capability in the decoder. If the attack/decay on the source material doesn't match FAIRLY CLOSELY, then it really works poorly. Now, you might claim -- well, they just happen to be using that attack/decay time, right? That is crazy -- who uses 1ms atk/ 30ms decay and 2ms atk/60ms decay very often for a compressor? How many compressors even have settings like that? These settings even cause distortion because a real DolbyA unit has no real defense from disrtorion in the design (the encoder/decoder does/undoes distortion.) The intermod mitigation in my code cleans up some of the mess that doesn't get undone. (Actually, there is a really smart attack trick in the HW, that helps a bit -- but my SW does his design justice by expanding the concept to be near perfect -- easier in software!!!) It still distorts, but symmtrically (encode/decode) does help by some percentage. AND IF THAT 'ATTACK TRICK' is not used, it is difficult to get correct results because the threshold never seems to be set correctly.

Here is a simple proof that the decoder works very poorly on non-encoded material -- the expansion curve has to sit right on the compression curve because the expansion is so aggressive that things start sounding bad if they don't match well. It ducks, it distorts (a little), the spatial relaitions get messed up, just sounds bad IF IT DOESN"T MATCH.

I really think that you state opinion about things, including me, without any knowledge...

Gain knowlege before making claims. I don't think that you can find anything anywhere where I have helped anyone with anything fake, all the way back to the beginning of the real internet, when YAHOO adopted some of my technology.

John

 

I just pulled-up an example that shows how closely the decoder matches real DolbyA. I have a piece that I decoded, which if the parameters don't match closely, the bass beat has a window around it -- kind of sounds odd and incorrect. This merges the music together seamlessly. Expanders can easily do damage to this material. Even my general purpose expander can't do such an accurate reproduction - because it doesn't precisely match the needed attack/decay and the left over freq-response skew. (My GP expander will adjust the various kinds of expansion to come close, but won't have the seamless sound that this does.) The example is 'Under Attack', and the chorus comes through fairly clearly. (ABBA chorus sometimes sounds like a blob, but this still has some seperation of the voices -- not perfect, but some -- even with mp3 encoding.) The location is Hightail Spaces

I can probably break the decoder or use a general purpose expander to show the difference. In a way, a GP expander might have a little less trouble in some ways -- where it will sound plausible because of the overriding compression that is meant to be in the music will mask the left over mistakes from the GP expander.

BTW -- I am not claiming that UnderAttack is the perfect example of beautiful/accurate music, but I am only talking about the challenge of making sure that the music components match up so that there are no substantial expander artifacats.

John

 

I started a thread about that topic, without any technical knowledge though, 13 years ago, but it didn't go anywhere:
The Dolby Thread

My observation was and is that for albums recorded around 1970 in the UK, we have different CD masterings where one is far brighter and the other one has far more pronounced bass, and none sounds right. Some examples:

Led Zeppelin III: The Marino remaster is much brighter and also a bit hissier than the Diament CD
Elton John (s/t): The WG DJM CD sound is dark and a bit booming with very little hiss. On the other hand, the UK DJM and the UK Polydor CD sounds very different
Genesis Trespass: The MCA CD is bright, the Virgin/Charisma CD is dark
Joni Mitchell Ladies Of The Canyon: The original Reprise CD is rather muffled, the Gastwirt remaster is very bright
Black Sabbath (s/t): The Castle CD is very dark (with de-emphasis correctly applied), other incarnations are far brighter
David Bowie Station To Station: The WG RCA CD is dark, the Japanese for US RCA CD is rather bright
Yes The Yes album: The original Atlantic CD is a bit thin, the MFSL CD --- reportedly the only one with correct Dolby decoding --- is much more bass-heavy and comparably compressed

 

If you are interested (and have the facility), take a few examples that you are especially interested in -- no more than about 30seconds -- mp3 encode so that it isn't so big (30seconds and mp3 encoding should be pretty short), then email to me (send me a private message, and I'll send you my personal email for sending the music files), and I'll happily look at them and produce decoded versions IF THEY DO DECODE. In fact, I'll send you the results anyway -- and tell you which ones worked and which ones didnt.

Again -- the only reason why I suggest the short/mp3 copies is for space reasons, not trying to make it complicated.

The copies most likely to be DolbyA encoded would be slightly overly bright with the highs compressed sounding and/or voices having a suspciously hard edge. Another hint, even though I haven't run into it very often is that the LF is suspeciously distorted (mostly for live-type recordings, not so much with big mixes.) The moderately highly compressed pieces with high average level tend to have less apparent HF boost when DolbyA encoding. (DolbyA mostly boosts the highs at lower levels -- kind of weird until one recognizes that the intent is to push the various bands above the noise level, but not to overdrive the electronics.)

John

 

New version of the decoder is now available. One of the leaders of another board ran some tests and gave me some critques about the code, and I complied and -I think- fixed all but one of the bugs that he noticed. Most of the bugs were things like a misunderstanding about using the input/output redirection, so I added an --input=filename and --output=filename switch. Also, the switch characters should be a 'double dash' and not just a 'dash'. The office program that I use merged the two dashes into one big one, which caused some confusion -- so that was mostly a documentation bugfix. Also, now the program gives an error on single channel files instead of being confused. Maybe some day I'll implement mono operation if there is demand (would be quite simple, actually.) The biggest bugaboo is that he found LOTS of distortion, but then he and I both realized indepdendently that the output was clipping. He was testing with a 0dB signal (rightfully so), and the output of the decoder showed a 0.33dB gain. This is technically correct, and exactly the expected behavior of an exact DolbyA emulation. However, I do need to improve the documentation to note that it is important to use the --outgain=-3 option to decrease the output level when creating an integer file. I always use --ingain=3 (because I always attenuate the input by 3dB, which compensates), and --outgain=-3 to keep from clipping. Need to document a bit better. Otherwise the distortion even at high and low levels (below clipping) doesn't appear on the Audacity spectrum analysis at all. A real DolbyA shows -66dB at 3kHz with a 1kHz signal. So the raw harmonic distortion at high and low levels had been found to be adequately low (EVEN WITH EXPANSION ENABLED.)
Most of the changes are superficial, and the program still has an approx 3000 sample loss (which I'll start fixing soon -- it should have zero loss.)
There is a new DecoderA.pdf, that includes the necessary documentation updates.

It sounds as good as ever (for encoded materials), and the filename is da-win-17may2018A.zip, in the repo: Hightail Spaces

John

 

It's a known fact that some releases contain pre-emphasis mastering, but don't have flags. Obvious examples: Culture Club's 1st album (CDV...), as mentioned in an interview in serious site by producer himself (not their fault), Black Sabbath Japanese CD issues (apparently same masters as in UK PE releases, but no flags). Errors do happen. Same even more easily could happen with Dolby A. Back in 80s they were in rush to put out as much reissues on CDs as fastly as possible, since there was a lot of money to be made. Considering circumstances, it was easy for mistakes to happen.

 

Yea -- I agree -- my REAL GUESS -- just a guess is that a lot of material was transferred (probably by contract help) to digital media in the late 1980s, and the idea of DolbyA encoding being marked on the tapes was ignored. Just the complication of turning off/on DolbyA decoding was probably ignored -- possibly requiring tone set-up when being played on different decks. So, the majority made it onto digital media, without indication of encoding/compression/phase of moon or anything like that. Once the material is on digital media, the cost of doing a Digital-to-Analog conversion, then DolbyA decode, then Analog-to-Digital conversion, then end user media production was short cut (either due to ignorance of the encooding, or "it is just a reissue, and sounds good enough') to just be a playout.
About the ongoing complaints about the high end being harsh or 'something different about digital', or raspy, or crunchy or something like that (I remember those complaints from the 1980s), I suspect that the media companies were pretty much immune to it. They are $$$ making companies, and people keep buying their stuff, why do they have to go to the expense of a few $1000 of DolbyA decoding for each CD that is likely a re-issue? The mgmt probably didn't even think about pre-emphasis or DolbyA decoding -- and it costs money, negatively affecting profits (in their minds.) They probably didn't even think about the small volume outfits like MFSL making money from trying to do it more correctly...

That is truly just a guess, but once the material is on digital, they certainly don't want to pay for an additional transfer/wear & tear on the original masters/etc. Many of the masters are probably sitting in a vault, under a mountain somewhere.

I am not making ANY money on my project, and my reputation in my field is set in stone, so I am doing the DolbyA project because I really think (know) that it is fixing a problem once in a while at least.

John

 

Right. I have no idea what this thread is about. Nobody could listen to a Dolby A encoded tape and not know something is very wrong. Nobody would master a Dolby A tape without decoding it first.

I have Dolby A 360 units around her someplace. It is not at all like Dolby B. With Dolby B you can listen to an encoded tape and it won't sound horrible. It will not sound right but it will be listenable. A Dolby A tape will not be listenable and was never intended to be anything a consumer would deal with. Dolby A is a closed system that requires proper setup, encoding and decoding and was only used by professional engineers.

This thread makes absolutely no sense to me.

 

If so, the mastering engineer had no clue what he was doing. None whatsoever.

Have any of the people commenting here about Dolby A, ever heard a Dolby A encoded tape? My guess is no.

 

I have heard DolbyA material. You can also hear some for your own reference on my repository - so you can say that you heard DolbyA also. I am listening to real-time DolbyA decoded ABBA right now!!! I also have access to material with tones -- but I am effectively under an NDA. (Dreamworld from ABBA actually sounds pretty darned good once it has been properly decoded with a VERY LOW INTERMOD decoder. Still some left-over fuzz, but not nearly as bad as using a normal decoder.)

 

I've heard Dolby A plenty of times in the studios back in the 70s.

 

May I ask what type of Dolby A decoder you are using?

 

I am using my own -- it is being picked up by pros because it works better than anything else out there. It isn't just an analog of a 360/cat22 or even the earlier DolbyA design. I took advantage of some technology that Ray Dolby didn't have available. Also, the decoder has been carefully compared dB for dB with a real DolbyA -- it works nearly perfectly. The dynamic tracking is superior also -- the intermod avoidance that my code does can match the necessary dynamics, yet avoid the intermod products and aliasing.

So, if you disable the aliasing processing, it comes close to the slight fuzz of a real DolbyA, then with everything enabled -- you can hear (e.g. individual voices in a chorus) much more clearly.

The tonal balance is between a real DolbyA and a Satin, but MUCH CLOSER to the real DolbyA. Apparently, they tried to emulate the 360's filters -- not really what you want. My decoder sounded like that until I had a 'Eureka' after the recording engineer who is working with me implied that we could do better. I worked it through, and WE DID!!! (I am speaking of the freq resp balance.)

The intermod handling is just really good.

John

 

Interesting. I was never a big fan of Dolby A. I can often tell when Dolby A or DBX was used on a project. Once you get to know their sounds it's not so hard to detect.

 

i agree -- I don't really 'like' either, but they are a necessary evil to have to decode when you have material encoded with those 'things.'

I have done a COMPLETE analysis of DolbyA and what it does (not a pure theoretical analysis, but more of an observation.) There are some artifacts that are really evil to deal with -- it is really bad in software, because when you do gain control with fast gain changes, then it is a nonlinear event, then the idea of intermod causing aliasing splatting all over the place starts to become real. This is one reason why some SW compressors (I mean FAST ones) sometimes have a harsh sound -- it is inappropriate handling of the intermod and aliasing.

But, DolbyA itself has a really bad anomaly -- which I am disclosing for the first time (don't want to tell the competition too much if they are listening), but a true DolbyA encoder appears to splat nonsense in the 9-10k range. It appears to result from the mix of fast attack and slow attack compressors in the specific DolbyA configuration. Otherwise, I don't know any other reason -- I have modeled the electronics -- which doesn't really directly help the software design because of arcane DSP issues -- but I don't see a source for oscillation or instability problems like that. So, it must be a mix of the attack/decays of the three associated compressors 80-3k, 3k-20k and 9k-20k, thereby producing enough hash in the 9-10k range to be noticeable under certain circumstances. My decoder does deal with that hash -- but would have been nearly impossible in the old HW technology. (I am an EE, software & OS designer, and DSP developer), so I am multi-disciplinary enough to figure out that something bad is happening (superficially at least!!!) . That problem, along with a few others, makes me unhappy with DolbyA. DBX has its own problems even though it is much simpler. RMS detection is almost necessary because it is wide-band, yet RMS is so slow that along with the normal pumping effects that cause a kind of modulation noise, the RMS being slower in some ways (but faster in others), lets more of the modulation noise through.

I rather like the concept of multi-band, non-rms schemes (or if rms, it must be fairly quick.) However, the trick of having different attack/decay for each band is evil also -- causes problems similar to the 9k-10k hash being worse than it should be.

Complicated stuff -- I love this world of 90+dB SNR essentially for free .

John

 

This demo will disappear in a few days, but I really got floored by this song -- DolbyA decoded with my decoder. The natural sound and really good spatial stability shows how reasonably good the decoder can do with well made material:

repository: Hightail Spaces

file name: 01-L-O-V-E (Remastered).mp3

Knocked my socks off. Absolutely NO eq or anything at all except normalization. Don't look for extreme highs or anything surprising -- it is just nice. I didnt upload a before and after -- but will do so if someone wants proof.

John

 

The 33PD is actually a hair darker than the '86 Castle (but incomplete, alas). But by the same token, I'm now convinced after listening to so many versions of this album, that the master tape is on the brighter side.

I'll have to check out your link though. 13 years ago, and my attention was elsewhere.

 

My guess about the varying descriptions about DolbyA in the past (being very compressed) and today is that we are really used to a lot of compression nowadays. I had to retrain my hearing to detect the HF emphasis and/or compression sound with DolbyA. Also, an uninitiated ear might think that DolbyA can sound similar to DolbyB (without doing a direct comparison), because at normal signal levels, the DolbyA compression is most active in the HF region. If you watch the gain logs of my decoder/DolbyA expander (very accurate DolbyA emulation), a lot of music almost saturates the 80-3k channel with between -1.0db and 0dB gain during most pop musice pieces (esp if fairly well compressed.) The HF channels (both) tend to bounce around, where the 3k-20k channel (actually, I do a remapping/remixing, and on my design it is the 3k-9k channel) tends to be the next to saturate towards -3dB to 0dB when the music becomes generally louder with significant upper MF intensity (the upper part of 80-3k and 3k-9k do intermix a lot, where the 80-3k channel actually passes significant signal up to 12k or so.) The highest channel is the last to approach 0dB, and almost never does. During a strong HF signal or fairly sibiliant voice, the 9k-20k channel will tend to approach -6dB to -3dB, and ONCE IN A BLUE MOON hit 0dB.

What this means for the actual encoded signal is that the MF (80-3k) gain stays generally much more constant than the other bands*, and there are often compression artifacts due to both the increased and changing gain of the two upper channels. Those two upper channels are in the same frequency region as most of the DolbyB activity also (but DolbyB does a sliding frequency band rather than gain change of fixed bands.) The sound would be superficially similar.
*When I do my 'acid' decoder testing, I look for material which causes significant activity in the MF frequency band -- because that is the band where the ear is more sensitive to correct tracking, and also the decoder is much
more active when the levels are lower in the MF band. It is usually most active at the beginnings and endings of a recording, and atk/dcy is also much more critical because of the way that a feedback compressor attack/decay works --
that is one of the big challenges of the very accurate emulation.

The other 'often' dynamic DolbyA effect is the LF range. The LF frequency range is officially 20-80Hz, but since the filters are fairly low Q (approx 0.6), the LF range includes perhaps 150Hz, esp if the MF range has a depressed gain (that is, for the decoding side.) The gains all interact in a really complex way -- but there is some audible LF compression which sometimes manfests as LF distortion. The 'hand in glove' effect of the DolbyA decoder will undo the distortion. The effect of DolbyA on the LF range can appear to either be somewhat neutral or a boost of low levels (which the ear doesn't really hear well.) So the LF compression isn't really very audible other than the 'distortion' and perhaps a slight change in level.

The sum total of those effects would result in a DolbyA sound similar to DolbyB in cases where the recording is is generally in the -15dB or higher level. At lower levels, the HF emphasis of DolbyA is more pronounced.

John

 

Yes, I don’t understand why this happens. Last year I posted about an LP from 1966 that had an 18 dB top shelf boost on about half the tracks, earbleeding stuff, and got told to adjust my ears.

Mastering stuff ups happen, and sometimes material gets released into the wild where it’s obvious that nobody was listening to what was going to disc, or had no idea. Sometimes these releases wind up in our hands, and we want to correct them. This should be encouraged, not derisively snorted at.

 

I am moving my discussions about the DolbyA stuff to the HW forum. Even though my decoder isn't hardware per-se, it is a very partial replacement in some cases.

 

Thanks for that. I would be interested especially in trying John's software on the Elton John discs. Also the Carpenters disc or discs.

 

Interesting thread.