The mastering issue, nerve grinding

[John Dyson]

Dear Hornet's nest:

 

The left over DolbyA imprint just struck again yesterday:  I just ran into the 6th or more (not 100% sure) premium download album that didn't sound quite right.  These are apparently still 'noise reduction' encoded.

 

Note before reading further:  I am not criticising anyone in particular, and definitely not codemning the people doing the mastering/remastering/etc, a lot of the time there are documentation problems and other complications that make their jobs a little tricky -- I am only commenting on the state of affairs, and the problem of not-so-good material reaching the consumer. I generally do not include proper NR decoding as a part of mastering, because it *should* be a part of the normal process to prepare for the consumer -- mastering should be considered the extra steps done to make the recording better/cleaner/etc for the consumer. (Mastering might include massaging the material to fit better on vinyl, for example.)

 

This observation has happened on numerous pieces premium 'material', 192/24 or 96/24...   I have 3/Simon&Garfunkel 192/24, Roberta Flack 192/24, Carpenters 96/24, PaulMc 96/24 and maybe a few others -- this is enough of a set of examples.  It seems likely that the 'quality' problem is still a left-over DolbyA-like imprint on these recordings.  The most ugly thing isn't the HF compression per-se, but the thin (compressed depth) stereo space left on these recordings.   *This quality experience/difference is difficult to describe, and is best experienced.  Suffice to say, I am NO golden ears, but can clearly hear the difference/improvement of proper handling of recordings.

 

These more egregious examples are mostly from well-known pay-for-download sites ( the download sites are not especially at fault -- they are selling what is available to them.  The distribution
is the problem.)  The problem doesn't come from any one seller -- I have some old CDs from eons ago also with the left-over imprint.  Of course, we don't expect the old CDs to be perfect, but it would be nice.  However, with proper re-EQ, and decoding, many CDs can be incredibly improved.

 

This last Roberta Flack example was a bit of a 'oh, my' kind of moment.  Luckily, I did borrow a song from a friend (I already had the normal CD) so I could determine if I wanted the 'pristine' copy.  I guess I can say -- no bother about getting the 'premium' album, that is unless I need to hear the above 20kHz noise/other artifacts on the album.  Unfortunately there appears also a DolbyA encoding imprint on the 'premium' album, but seems to have been further damaged by some compression.  So, this 'premium' album is even less useful to me than the old CD -- at least, I can clean-up the CD.

 

Why do I notice this FRUSTRATING problem?  I am in a almost unique position that I can actually diagnose the problem, and can resolve it - not so much at the consumer level, however.  The software to 'detect'/'correct' the problem is not a simple thing to do, and is definitely not a weekend project.  Using the decoding software is NOT for the consumer either -- we need to advocate for more complete digital preparation before we are sent the digital copies.  So, after these 30yrs of digital distributions, I am finding that many of them from over the years are 'just not right'.  Very often, the material appears to be DolbyA encoded.  Recordings are STILL being distributed not 'quite right' even nowadays.

 

With all of the discussion over the years about mp3, opus, 16bit PCM vs 24bit PCM, 192/96 vs 44.1k, etc... to me, knowing what I have learned in the last 3-4yrs, it appears that they are all diversions, because only the slower lossy compression are more important than the 'damage' from non-NR-decoded material reaching the consumer.  (choice between properly mastered 192kmp3 vs. undecoded 192/24, I'll take the mp3 quality anyday.)  I am NOT advocating for mp3 though -- just that the un decoded material sounds worse.

 

I almost blew a 'gasket' with this last album (Roberta Flack, 2012 remaster.)  When purchasing/downloading a 'premium' album, I'd expect that the material wasproperly prepared for the consumer -- but it seems like a lot of material is just not properly handled.  I do have some premium albums (Nat King Cole, for example) which also APPEAR to also have the residual DolbyA imprint, but cannot prove it, even for myself.

 

Given the fact about MQA causing such an uproar (In my opinion ONLY, I still believe that the worst thing about MQA is the obfuscation/complication and possible DRM), I wonder why this missing-mastering-step travesty hasn't caused a riot!?!?!? :-).  Is the thin stereo space, compressed high end material actually good enough?  If that is good enough, then why worry about needing more quality than mp3 at 128k? (rhetorical.)

 

PS: I am only pushing for the distribution chain to do the right thing and properly master the material -- no need to get attention by hypercompressing conventional releases, just do the basic
preparation correctly!!!   Every user cannot practically do their own mastering (or re-mastering), it is the job of the distributors with the actual master tape copies to finish the job!!!

 

John

 

[StephenJK]

I feel for you brother - and for all of us.  

 

Your comments on Dolby encoded artifacts being prevalent with any recording from the (late?) 70’s through the 80’s is something that your average bear wouldn’t know much about in this day and age, but as you have pointed out it just won’t go away.

 

Sure, I saw that Dolby logo on my Onkyo TA-2080, but they all had that.  I wasn’t aware that Dolby had made such inroads into the studio world, nor that it would still be around.

 

I don’t think anyone in the 70’s could have even imagined that music still being played today.  As Randy Bachman says on his radio show (Vinyl Tap), it was a lot of work to get a hit record and then you had weeks or a couple of months to play it at high school gyms and local arenas trying to make a buck from it.

 

Who could see a world of classic rock radio stations with the same songs played over and over every day all over the world.  

 

They don’t like Rush, they like The Spirit Of Radio.  

 

They don’t like Led Zeppelin, they recognize that wedding song.  Play Fool In The Rain and you get “Who the hell is that”?

 

I think you need to repurpose your development for correcting Dolby encoding not as a separate program, but perhaps as a filter for media server programs. 

 

It would seem that the commercial side of things doesn’t care much about whether things sound as they could or should, but the folks who hang out here sure do.

 

Maybe you need to have your program as a filter for people like JRiver.  

 

Hang in there.  Years from now we’ll all laugh about this.

[John Dyson]

You know, the filter idea is pretty good...  Actually, much of the time, I run the decoder realtime (kind of a filter) for my own listening, so it CAN run realtime in some cases.   With a different set of tradeoffs, it could work reasonably as a filter on the larger scale.   The really big problem is that it would overload a service, tieing up a big part of a CPU for the highest quality.  Actually, though, I haven't done a massive architectural enhancement yet -- could probably create a very reasonable quality result (front-end anti-IMD mitigation only, perhaps) to run on perhaps 1/4 of a core2 CPU core.   The highest quality at 96k or 192k currently fills up my 4 core Haswell and then some -- very very CPU intensive, and the code is 99% 8 wide floating point SIMD!!!   In that advanced mode, the decoder does a lot to gouge out as much MD (modulation distortion) as makes practical sense -- probably more than actually practical.

 

You really do make a good point about doing a filter...   I think that I am going to see what I could do -- right now, the code isn't doing much interpolation/decimation, so there IS opportunity to speed up the code.

 

John

 

[StephenJK]

That's great.  Keep us informed of any developments - I think you might have a lot of interest in this.  

[fas42]

The long term solution is unmixing, which is being dealt with currently from different angles; one of them is https://en.wikipedia.org/wiki/Signal_separation. Ultimately, all recordings will be able to be decomposed fully into every sound element that participated, and full mastering control will be with the consumer to completely reassemble the musical parts into whatever they wish.

[Teresa]

On 7/30/2019 at 6:56 AM, John Dyson said:

...The left over DolbyA imprint just struck again yesterday:  I just ran into the 6th or more (not 100% sure) premium download album that didn't sound quite right.  These are apparently still 'noise reduction' encoded...

 

The box of the analog master tape should say “Dolby A“ and at the start of the master tape are Dolby A test tones to set the Dolby levels. All studios that used Dolby A should maintain active Dolby A decoding units since Dolby noise reduction is a dynamic process and cannot be adequately fixed with EQ. The fact that recording companies are releasing undecoded Dolby A recordings is a crime IMHO.

 

I've read many of your posts and am unsure what your correction is? Is it Dolby A decoding software? The patents should be expired by now?

[StephenJK]

22 hours ago, fas42 said:

The long term solution is unmixing, which is being dealt with currently from different angles; one of them is https://en.wikipedia.org/wiki/Signal_separation. Ultimately, all recordings will be able to be decomposed fully into every sound element that participated, and full mastering control will be with the consumer to completely reassemble the musical parts into whatever they wish.

 

Sure, we can all get a copy of Melodyne and tweak a track so that it’s unrecognizable from where it started. That’s not hard to do, you can do that today!

 

But what’s your point with this long term solution?  We all turn into latter day Miles Davis artists and explore the deep and inner meaning of discordant skreeks and skronks, but without the talent and skill to take it somewhere?

 

 

 

 

[fas42]

26 minutes ago, SJK said:

 

Sure, we can all get a copy of Melodyne and tweak a track so that it’s unrecognizable from where it started. That’s not hard to do, you can do that today!

 

But what’s your point with this long term solution?  We all turn into latter day Miles Davis artists and explore the deep and inner meaning of discordant skreeks and skronks, but without the talent and skill to take it somewhere?

 

 

 

 

 

No, you misunderstand. Take a jazz trio track, and split into the three tracks: the piano with its associated acoustic imprint in the recording space, and again for double bass, and drummer. You now have what the mastering engineer has on three tracks on his work console; and now you get to play with the balances between those instruments, and placements; also, remove the acoustic from an individual sound producer, extracting a "perfect" recording booth version, and then invoke and add a new acoustic which responds to that 'raw' sound.

 

The artistry remains the same; you merely move the chess pieces so that it's more pleasing, etc, to you ...

[John Dyson]

22 hours ago, Teresa said:

 

The box of the analog master tape should say “Dolby A“ and at the start of the master tape are Dolby A test tones to set the Dolby levels. All studios that used Dolby A should maintain active Dolby A decoding units since Dolby noise reduction is a dynamic process and cannot be adequately fixed with EQ. The fact that recording companies are releasing undecoded Dolby A recordings is a crime IMHO.

 

I've read many of your posts and am unsure what your correction is? Is it Dolby A decoding software? The patents should be expired by now?

You are correct that the labels should be true, but I have direct evidence from people actually doing 'greatest hits' type collections that the state of the metadata on the tapes (the labels) are sometimes (for some record labels -- OFTEN) wrong.  Also, have numerous examples of indirect evidence that shows the leakage of DolbyA encoded material into the consumer realm.

 

*  So, the built-in DolbyA units aren't always being used, and the DHNRDS can do an infinitely better job (you gotta listen!!!)

 

Also, there is the issue of the Library of Congress standards for backing up existing analog archives.  They totally demure on whether or not to actually pass the material through a decoder.  I have some evidence that decoding doesn't always happen.

 

Next, there is a HUGE amount of evidence that DolbyA material is DEFINITELY getting into the consumer realm.  This means that DolbyA is NOT being decoded in MANY MANY MANY cases.

 

Finally, DolbyA HW *decoding*  (esp by those who were looking into being early adopters a long time ago) produces a kind of stealthy IMD that softens the sound.  It almost sounds like a profound kind of 'tape compression' all by itself.  I can show NUMEROUS examples of undecoded material in the CONSUMER realm, equivalent decoded versions in the CONSUMER realm, and the DHNRDS DA decoder producing very improved results.

 

So, IN THE IDEAL WORLD, the tapes have proper labels, and the instructions/information on the labels is followed.  There is a huge amount of existance proof that not to be followed much of the time.  Even one of the earliest CDS (Nena 99Luftballons) has both the CD preemphasis AND DolbyA encoding compression (that is, not decoded.)  I have existance proof of that and literally 100's of other examples, and I did NOT specially select my archives.

 

Consumers are simply accepting the DolbyA compression with mastering EQ as a decoding substitution.

 

Again, the DHNRDS produces amazingly less decoding distortion (most of DolbyA distortion is in DECODING, not in ENCODING.)

 

I can provide literally 100's of satisfying decoding results from consumer materials, and willing to decode short selections of masters that you might have.  I can even provide a timed-out version of the decoder for free (totally legal.)   My project partner is Richard Hess -- a farily well known restoration expert -- and the lack-of-decoding problem IS REAL.   IT IS NO JOKE.

 

John

[gmgraves]

On 7/30/2019 at 6:56 AM, John Dyson said:

Dear Hornet's nest:

 

The left over DolbyA imprint just struck again yesterday:  I just ran into the 6th or more (not 100% sure) premium download album that didn't sound quite right.  These are apparently still 'noise reduction' encoded.

 

Note before reading further:  I am not criticising anyone in particular, and definitely not codemning the people doing the mastering/remastering/etc, a lot of the time there are documentation problems and other complications that make their jobs a little tricky -- I am only commenting on the state of affairs, and the problem of not-so-good material reaching the consumer. I generally do not include proper NR decoding as a part of mastering, because it *should* be a part of the normal process to prepare for the consumer -- mastering should be considered the extra steps done to make the recording better/cleaner/etc for the consumer. (Mastering might include massaging the material to fit better on vinyl, for example.)

 

This observation has happened on numerous pieces premium 'material', 192/24 or 96/24...   I have 3/Simon&Garfunkel 192/24, Roberta Flack 192/24, Carpenters 96/24, PaulMc 96/24 and maybe a few others -- this is enough of a set of examples.  It seems likely that the 'quality' problem is still a left-over DolbyA-like imprint on these recordings.  The most ugly thing isn't the HF compression per-se, but the thin (compressed depth) stereo space left on these recordings.   *This quality experience/difference is difficult to describe, and is best experienced.  Suffice to say, I am NO golden ears, but can clearly hear the difference/improvement of proper handling of recordings.

 

These more egregious examples are mostly from well-known pay-for-download sites ( the download sites are not especially at fault -- they are selling what is available to them.  The distribution
is the problem.)  The problem doesn't come from any one seller -- I have some old CDs from eons ago also with the left-over imprint.  Of course, we don't expect the old CDs to be perfect, but it would be nice.  However, with proper re-EQ, and decoding, many CDs can be incredibly improved.

 

This last Roberta Flack example was a bit of a 'oh, my' kind of moment.  Luckily, I did borrow a song from a friend (I already had the normal CD) so I could determine if I wanted the 'pristine' copy.  I guess I can say -- no bother about getting the 'premium' album, that is unless I need to hear the above 20kHz noise/other artifacts on the album.  Unfortunately there appears also a DolbyA encoding imprint on the 'premium' album, but seems to have been further damaged by some compression.  So, this 'premium' album is even less useful to me than the old CD -- at least, I can clean-up the CD.

 

Why do I notice this FRUSTRATING problem?  I am in a almost unique position that I can actually diagnose the problem, and can resolve it - not so much at the consumer level, however.  The software to 'detect'/'correct' the problem is not a simple thing to do, and is definitely not a weekend project.  Using the decoding software is NOT for the consumer either -- we need to advocate for more complete digital preparation before we are sent the digital copies.  So, after these 30yrs of digital distributions, I am finding that many of them from over the years are 'just not right'.  Very often, the material appears to be DolbyA encoded.  Recordings are STILL being distributed not 'quite right' even nowadays.

 

With all of the discussion over the years about mp3, opus, 16bit PCM vs 24bit PCM, 192/96 vs 44.1k, etc... to me, knowing what I have learned in the last 3-4yrs, it appears that they are all diversions, because only the slower lossy compression are more important than the 'damage' from non-NR-decoded material reaching the consumer.  (choice between properly mastered 192kmp3 vs. undecoded 192/24, I'll take the mp3 quality anyday.)  I am NOT advocating for mp3 though -- just that the un decoded material sounds worse.

 

I almost blew a 'gasket' with this last album (Roberta Flack, 2012 remaster.)  When purchasing/downloading a 'premium' album, I'd expect that the material wasproperly prepared for the consumer -- but it seems like a lot of material is just not properly handled.  I do have some premium albums (Nat King Cole, for example) which also APPEAR to also have the residual DolbyA imprint, but cannot prove it, even for myself.

 

Given the fact about MQA causing such an uproar (In my opinion ONLY, I still believe that the worst thing about MQA is the obfuscation/complication and possible DRM), I wonder why this missing-mastering-step travesty hasn't caused a riot!?!?!? :-).  Is the thin stereo space, compressed high end material actually good enough?  If that is good enough, then why worry about needing more quality than mp3 at 128k? (rhetorical.)

 

PS: I am only pushing for the distribution chain to do the right thing and properly master the material -- no need to get attention by hypercompressing conventional releases, just do the basic
preparation correctly!!!   Every user cannot practically do their own mastering (or re-mastering), it is the job of the distributors with the actual master tape copies to finish the job!!!

 

John

 

I have used both Dolby-A and DBX  noise “companding” systems and found that as long as one used a 400 Hz “calibration” tone at the head and tail of every recording, and made sure that the tones match “exactly” on playback, that aside from a tiny bit of insertion distortion (which without a direct A/B to the source through headphones, I would defy anyone to notice) that Dolby-A was pretty benign. DBX, OTOH, could be heard to “pump” and noise modulate the signal no matter how carefully levels were matched.

[John Dyson]

8 hours ago, gmgraves said:

I have used both Dolby-A and DBX  noise “companding” systems and found that as long as one used a 400 Hz “calibration” tone at the head and tail of every recording, and made sure that the tones match “exactly” on playback, that aside from a tiny bit of insertion distortion (which without a direct A/B to the source through headphones, I would defy anyone to notice) that Dolby-A was pretty benign. DBX, OTOH, could be heard to “pump” and noise modulate the signal no matter how carefully levels were matched.

As I have written -- the DolbyA decoding distortion is stealthy.  Unless one does a direct A/B comparison on intense material -- most people won't hear it.  DBX does NOT have the same characteristics of DolbyA when it comes to the distortion.  After learning what to listen for -- the problem becomes apparent.  I have found that the 363 decoding is better than an old cat22 in a 360 chassis for some reason -- but the modulation distortion (again -- it is a lack of clarity) exists on both versions.

 

* EDIT -- most of the DolbyA distortion is on the DECODE side, because of the delays and structure of the feedback loop for decoding.  The feedback loop for the decoding on the surface seems correct, and in an ideal world with zero delays -- it could work.  However, the DolbyA units work in the real world, and the delays in the additional layer of feedback create lots of opportunity for imprecision.  In this case, the imprecision is audible.  The DHRNDS does NOT have this flaw.

 

I can show multiple POP examples of DolbyA 'distortion (that is, material which has been publically available), unfortunately, the best controlled cases where there is only 1 pass of DolbyA are on material that is non-public (master tape material.)  I can provide NUMEROUS examples as desired, where I can decode leaked DolbyA vs. a decoded DolbyA copy -- the DHNRDS blows away the commercial DolbyA decoded copy for clarity in places like vocal chorus, and other complex mixes of material.  (Too complex, there is an improvement -- like on 'The Cars' stuff, but not as obvious as vocal chorus.)   On the Cars stuff, you can actually hear the intermod from synthesized material on DolbyA deocded versions, while pristine on the DHNRDS version -- again, ALL PUBLIC MATERIAL (of course, still under copyright, but I can do snippets.)

What does the distortion sound like?   Don't listen for the typical grinding kind of distortion -- even though in the extreme, DolbyA can produce it.   Listen for something that sounds like tape compression.  It is NOT a left over compression though -- even though it does sound like compression IN A WAY.   The best way to explain:  listen to a clean vocal chorus before encoding.   Then, go through an encode/decode cycle -- notice that the 'clean' vocal chorus is no longer clean, but the voices loose their individual nature.  The result of vocal chorus in the higher registers is a 'blob'.   One reason why I use ABBA a lot in testing is that they trigger the problem in most extreme ways.  However, I do have 1st generation encoded DolbyA where an A/B comparison even shows it in a natural recording in a large area setting.

 

What does DolbyA do on other, less intense material?  Generally it isn't intrusive, but there is a loss in detail -- much much much worse than anything that might happen by rate conversion down to 44.1k/16 and frankly, worse than mp3 at 192/256k.  (The loss in time resolution from mp3 doesn't have nearly the negative effect of the conversion of a vocal chorus into a blob.)

 

Where is DolbyA somewhat transparent?  Single instruments, single vocals.  Any time that there aren't multiple complex sources, then DolbyA isn't soo bad.   It is really great for first pass of mixing down material (simple components.)

 

The DHNRDS reduces this modulation distortion very significantly (I cannot say how much, but I'd suspect it is close to 1/10th as much), and makes the results more clean/clear than the alternative.  DHNRDS can make ABBA (as the extreme example) sound MUCH more clear/clean, less 'blaring' in the sound because of more distinct mixed vocals.  However, Olivia is as clean/clear as she has been before DolbyA encoding, and the IMD from 'The Cars' synthesizers is much diminished.

 

Examples available upon request.

 

John

 

 

 

[StephenJK]

10 hours ago, fas42 said:

 

No, you misunderstand. Take a jazz trio track, and split into the three tracks: the piano with its associated acoustic imprint in the recording space, and again for double bass, and drummer. You now have what the mastering engineer has on three tracks on his work console; and now you get to play with the balances between those instruments, and placements; also, remove the acoustic from an individual sound producer, extracting a "perfect" recording booth version, and then invoke and add a new acoustic which responds to that 'raw' sound.

 

The artistry remains the same; you merely move the chess pieces so that it's more pleasing, etc, to you ...

OK, that makes more sense.  I’m not sure how many would take the time and trouble to go through that process, to say nothing of where and how I find that new acoustic I want to use.  

 

Are you talking about a sound file or a specific configuration?

 

Like others, I have all types of recordings with very poor sound but even if the process you describe could be done at home I wonder how many of them would be salvageable.  You know, how do you insert something that was never there? 

 

It still sounds like you're describing a DAW setup but with new and improved plugins. 

 

 

[jabbr]

You highlight the problem of so-called Master tapes being stored in an encoded format — unfortunately what is done is done in the case of the old DolbyA but hopefully this problem can be prevented in the future if the issue is brought attention (as you are doing). MQA is unfortunately similar for the modern day (worse actually). The labels seem to want to reissue with the least amount of work. I hope you can nudge the labels into at least decoding before remastering!

[John Dyson]

1 minute ago, jabbr said:

You highlight the problem of so-called Master tapes being stored in an encoded format — unfortunately what is done is done in the case of the old DolbyA but hopefully this problem can be prevented in the future if the issue is brought attention (as you are doing). MQA is unfortunately similar for the modern day (worse actually). The labels seem to want to reissue with the least amount of work. I hope you can nudge the labels into at least decoding before remastering!

Yes -- (avoiding the evil MQA word after this sentence :-)), a lot of material IS in DolbyA form.  Even when properly decoded with DolbyA hardware, the resulting quality is *less* than what is actually on tape.  The DolbyA defect is MOSTLY on the decoding side, and there is an alternative solution that produces almost the maximum quality residing on the tape.

 

I do know that the old DolbyA HW does have a decoding problem -- because I have narrowed it down, but have had some problems finding the actual 'flaw'  -- I think that I have it:  *for decoding, the DolbyA places the encoding side into a feedback loop*.  For *encoding*, the variable delays through the compressors aren't much of a problem -- we do not hear the associated delays.   However for decoding, *in the feedback loop*, the differences in delays produce enough imprecision that the NORMAL hardware decoding process produces more flawed audio than what we expect for new recordings today.

 

 

The DHNRDS DA decoder was written to eliminate the effects of the hardware  decoding precision, but also to go further to remove modulation products (which are forms of distortion) that aren't needed for the fast gain control to happen. *modulation distortion from DolbyA doesn't usually have a 'rough' sound, but rather a softening -- sounds similar to tape compression itself.

 

So, the DolbyA HW has some defects that produce a 'fuzzier'/almost compressed sound.  These defects take NOTHING away from R Dolby's genius, because the resulting quality was probably 'good enough' for the 1960's.   However, the DHNRDS doesn't have the 'feedback imprecision' mechnaism of the DolbyA HW, plus DHNRDS plays complex/nonlinear math games (similar to Orban patent US 6,205,225 -- but taking more advantage of computer math -- also using different operations on the signal components) that weren't available for the HW design.  The DHNRDS actually minimizes the modulation distortions, unlike the Orban patent which mostly just stashes the components in a less obtrusive spectral location.  (Being specific -- the DHRNDS DA attack/release curves are quickly & dynamically bent to minimize production of modulation products, while still doing the  attack correctly in the larger, important time scale.)  The distortion mitigation method is implicit -- and operates on the signal at exactly the same time as the gain control multiplication operation -- tricky and very effective.

 

So, we 'all'  have lots of material in DolbyA encoded form, on master tapes, on digital back-ups, and yes, on consumer digitial distributions.  Let's ignore what the consumer can do -- that is a messy thing, but lets look at the two non-consumer sources...   1) the master tapes can be decoded by the almost built-in DolbyA units -- but why isn't it done more often?   2) the digital back-ups can be decoded by what?

 

Answer, a superior AUDIO QUALITY way to decode the old DolbyA material is to use the DHNRDS DA decoder.   One might ask:  why not use the almost built-in DolbyA units (they are usually cards in an associated rack)?   Well, I'll answer and say -- *why don't they use the built-in decoders more often -- there are too many consumer leaks of DolbyA*?

 

The final, best answer -- just use the DHNRDS DA decoder before any manipulations -- esp in lieu of the very common use of simple EQ instead of proper decoding.  You would not believe (really) the cases where I can effectively produce a DolbyA encoded recording by starting with a consumer release, add 3dB @ 3kHz with Q=0.707 or Q=1.0, and produce pristine material (really pristine.)  I can produce demos ad-infinitum -- happily decode a few of interested parties personal collection as examples (will keep material private/secret & destroy on completion.)  I treat peoples' private materials MUCH MUCH more securely than normal consumer released materials -- my master tape material is stashed in a very different place to avoid mistakes!!!

 

Using the DHNRDS DA over simple EQed material or even DolbyA HW material produces superior results (in some cases, VERY superior.)

 

The distributors should BEST use the DHNRDS DA decoder (it is almost free of charge -- much less than $500), produce beautiful results, much closer to the quality that we actually expect nowdays, instead of a (sadly) "sloppy" decode from DolbyA HW or simple, ugly sounding EQ.  I only expect to sell, maybe, 20 copies of the software -- with irritating support issues.  I am happy to give timed-out versions to consumers for free -- when they are really motivated.  This is NOT a profit siphon that MQA is (whoops -- used the bad word again :-)).  A single copy of DHNRDS DA decoder can produce approx 100 decoded masters per day -- that amortizes to an infinitesimally small cost for a MAJOR quality improvement.

 

John

 

* PLEASE never consider my criticism of DolbyA as a disparagement of R. Dolby or the Dolby Inc people.  The DolbyA was designed with amazing economy -- very few components for a lot of functionality.  Also, the quality was within acceptable limits for the 1960s pop music (and doesn't sound horrible on non-complex recordings of almost any type.)   R Dolby was a genius (esp by todays standards), and produce an amazing, ground breaking product.  Just so happens that yr2000+ computers and software can finally do better for the decoding operation -- that was no mean feat by R Dolby!!!!  I have great respect for him.

 

 

[gmgraves]

5 hours ago, John Dyson said:

As I have written -- the DolbyA decoding distortion is stealthy.  Unless one does a direct A/B comparison on intense material -- most people won't hear it.  DBX does NOT have the same characteristics of DolbyA when it comes to the distortion.  After learning what to listen for -- the problem becomes apparent.  I have found that the 363 decoding is better than an old cat22 in a 360 chassis for some reason -- but the modulation distortion (again -- it is a lack of clarity) exists on both versions.

 

* EDIT -- most of the DolbyA distortion is on the DECODE side, because of the delays and structure of the feedback loop for decoding.  The feedback loop for the decoding on the surface seems correct, and in an ideal world with zero delays -- it could work.  However, the DolbyA units work in the real world, and the delays in the additional layer of feedback create lots of opportunity for imprecision.  In this case, the imprecision is audible.  The DHRNDS does NOT have this flaw.

 

I can show multiple POP examples of DolbyA 'distortion (that is, material which has been publically available), unfortunately, the best controlled cases where there is only 1 pass of DolbyA are on material that is non-public (master tape material.)  I can provide NUMEROUS examples as desired, where I can decode leaked DolbyA vs. a decoded DolbyA copy -- the DHNRDS blows away the commercial DolbyA decoded copy for clarity in places like vocal chorus, and other complex mixes of material.  (Too complex, there is an improvement -- like on 'The Cars' stuff, but not as obvious as vocal chorus.)   On the Cars stuff, you can actually hear the intermod from synthesized material on DolbyA deocded versions, while pristine on the DHNRDS version -- again, ALL PUBLIC MATERIAL (of course, still under copyright, but I can do snippets.)

What does the distortion sound like?   Don't listen for the typical grinding kind of distortion -- even though in the extreme, DolbyA can produce it.   Listen for something that sounds like tape compression.  It is NOT a left over compression though -- even though it does sound like compression IN A WAY.   The best way to explain:  listen to a clean vocal chorus before encoding.   Then, go through an encode/decode cycle -- notice that the 'clean' vocal chorus is no longer clean, but the voices loose their individual nature.  The result of vocal chorus in the higher registers is a 'blob'.   One reason why I use ABBA a lot in testing is that they trigger the problem in most extreme ways.  However, I do have 1st generation encoded DolbyA where an A/B comparison even shows it in a natural recording in a large area setting.

 

What does DolbyA do on other, less intense material?  Generally it isn't intrusive, but there is a loss in detail -- much much much worse than anything that might happen by rate conversion down to 44.1k/16 and frankly, worse than mp3 at 192/256k.  (The loss in time resolution from mp3 doesn't have nearly the negative effect of the conversion of a vocal chorus into a blob.)

 

Where is DolbyA somewhat transparent?  Single instruments, single vocals.  Any time that there aren't multiple complex sources, then DolbyA isn't soo bad.   It is really great for first pass of mixing down material (simple components.)

 

The DHNRDS reduces this modulation distortion very significantly (I cannot say how much, but I'd suspect it is close to 1/10th as much), and makes the results more clean/clear than the alternative.  DHNRDS can make ABBA (as the extreme example) sound MUCH more clear/clean, less 'blaring' in the sound because of more distinct mixed vocals.  However, Olivia is as clean/clear as she has been before DolbyA encoding, and the IMD from 'The Cars' synthesizers is much diminished.

 

Examples available upon request.

 

John

 

 

 

Agreed.

[Josh Mound]

On 7/30/2019 at 9:56 AM, John Dyson said:

Dear Hornet's nest:

 

The left over DolbyA imprint just struck again yesterday:  I just ran into the 6th or more (not 100% sure) premium download album that didn't sound quite right.  These are apparently still 'noise reduction' encoded.

 

Note before reading further:  I am not criticising anyone in particular, and definitely not codemning the people doing the mastering/remastering/etc, a lot of the time there are documentation problems and other complications that make their jobs a little tricky -- I am only commenting on the state of affairs, and the problem of not-so-good material reaching the consumer. I generally do not include proper NR decoding as a part of mastering, because it *should* be a part of the normal process to prepare for the consumer -- mastering should be considered the extra steps done to make the recording better/cleaner/etc for the consumer. (Mastering might include massaging the material to fit better on vinyl, for example.)

 

This observation has happened on numerous pieces premium 'material', 192/24 or 96/24...   I have 3/Simon&Garfunkel 192/24, Roberta Flack 192/24, Carpenters 96/24, PaulMc 96/24 and maybe a few others -- this is enough of a set of examples.  It seems likely that the 'quality' problem is still a left-over DolbyA-like imprint on these recordings.  The most ugly thing isn't the HF compression per-se, but the thin (compressed depth) stereo space left on these recordings.   *This quality experience/difference is difficult to describe, and is best experienced.  Suffice to say, I am NO golden ears, but can clearly hear the difference/improvement of proper handling of recordings.

 

These more egregious examples are mostly from well-known pay-for-download sites ( the download sites are not especially at fault -- they are selling what is available to them.  The distribution
is the problem.)  The problem doesn't come from any one seller -- I have some old CDs from eons ago also with the left-over imprint.  Of course, we don't expect the old CDs to be perfect, but it would be nice.  However, with proper re-EQ, and decoding, many CDs can be incredibly improved.

 

This last Roberta Flack example was a bit of a 'oh, my' kind of moment.  Luckily, I did borrow a song from a friend (I already had the normal CD) so I could determine if I wanted the 'pristine' copy.  I guess I can say -- no bother about getting the 'premium' album, that is unless I need to hear the above 20kHz noise/other artifacts on the album.  Unfortunately there appears also a DolbyA encoding imprint on the 'premium' album, but seems to have been further damaged by some compression.  So, this 'premium' album is even less useful to me than the old CD -- at least, I can clean-up the CD.

 

Why do I notice this FRUSTRATING problem?  I am in a almost unique position that I can actually diagnose the problem, and can resolve it - not so much at the consumer level, however.  The software to 'detect'/'correct' the problem is not a simple thing to do, and is definitely not a weekend project.  Using the decoding software is NOT for the consumer either -- we need to advocate for more complete digital preparation before we are sent the digital copies.  So, after these 30yrs of digital distributions, I am finding that many of them from over the years are 'just not right'.  Very often, the material appears to be DolbyA encoded.  Recordings are STILL being distributed not 'quite right' even nowadays.

 

With all of the discussion over the years about mp3, opus, 16bit PCM vs 24bit PCM, 192/96 vs 44.1k, etc... to me, knowing what I have learned in the last 3-4yrs, it appears that they are all diversions, because only the slower lossy compression are more important than the 'damage' from non-NR-decoded material reaching the consumer.  (choice between properly mastered 192kmp3 vs. undecoded 192/24, I'll take the mp3 quality anyday.)  I am NOT advocating for mp3 though -- just that the un decoded material sounds worse.

 

I almost blew a 'gasket' with this last album (Roberta Flack, 2012 remaster.)  When purchasing/downloading a 'premium' album, I'd expect that the material wasproperly prepared for the consumer -- but it seems like a lot of material is just not properly handled.  I do have some premium albums (Nat King Cole, for example) which also APPEAR to also have the residual DolbyA imprint, but cannot prove it, even for myself.

 

Given the fact about MQA causing such an uproar (In my opinion ONLY, I still believe that the worst thing about MQA is the obfuscation/complication and possible DRM), I wonder why this missing-mastering-step travesty hasn't caused a riot!?!?!? :-).  Is the thin stereo space, compressed high end material actually good enough?  If that is good enough, then why worry about needing more quality than mp3 at 128k? (rhetorical.)

 

PS: I am only pushing for the distribution chain to do the right thing and properly master the material -- no need to get attention by hypercompressing conventional releases, just do the basic
preparation correctly!!!   Every user cannot practically do their own mastering (or re-mastering), it is the job of the distributors with the actual master tape copies to finish the job!!!

 

John

 

 

Hi John, can you list some of the specific albums (the S&G and Flack albums, etc.) as well as any others you’ve noticed? It would be great if this thread could try to compile a list to vet. 

 

It definitely does seem that the noise reduction info on many albums gets lost or ignored over time. Famously, there was/is a huge debate about whether Stevie Wonder’s Songs in the Key of Life is Dolby encoded or not.

[John Dyson]

26 minutes ago, JoshM said:

 

Hi John, can you list some of the specific albums (the S&G and Flack albums, etc.) as well as any others you’ve noticed? It would be great if this thread could try to compile a list to vet. 

 

It definitely does seem that the noise reduction info on many albums gets lost or ignored over time. Famously, there was/is a huge debate about whether Stevie Wonder’s Songs in the Key of Life is Dolby encoded or not.

 

(Most of the time, the per song EQ is identical, and un-normalized material all work with one single calibration value.  In limited cases, like when noted about a few pieces from ABBA, need minor mods to EQ.)  Most material is happy with calibration of -14.40dB or -15.15dB, sometimes -13.70dB -- tweaking in to within 0.10dB can sometimes improve sibilance or mitigate some minor ducking.

 

I'll compile a much larger list later on -- a lot of CDs are in storage, but trying to gather metadata that I still have online:

 

before decoding, most require EQ +3dB @ 3kHz/Q=0.707 or Q=1.0.  ABBA mostly wants Q=1.0.

some ABBA songs (like SuperTrouper on SuperTrouper, or SOS on ABBA need 3dB additional boost, but with additional Q of 0.707) AFTER decoding to match the sound of released material.

 

Of course, the CDs weren't made in '1971', but that was the date of the base album -- the CD was made as a release of the album.

 

Olivia Newton John's 48 Original Tracks 1971-1975, EMI

ABBA Gold, 1992, Japanese release

More ABBA Gold, 1993, Japanese release

ABBA, ABBA 1986, Polydor K.K, P33P 20056

ABBA, Voulez Vous, 1984, Discomate, CDP-106

 

(lots more ABBA!!! I have various copies of each album -- I have chose the least molested versions as my examples.)  Sometimes there is an attempt to 'master' material that is still DolbyA encoded -- not a pretty sight or listen!!!

 

Sergio Mendes & Brasil 66 (the Very Best), 1986, A&M Label 396 907-2

The Very Best of Herb Alpert, 1991, A&M, Label 397 165-2

 

Close To You, Carpenters, A&M, 1970

The Carpenters, Carpenters, A&M, 1971

A Song For You, Carpenters, A&M, 1972

(the rest of the standard Carpenters albums)

 

Complete Greatest Hits, The Cars, 2002

 

(I have many many others, some from premium sources -- one example, recently purchased -- so it has better provenance):

Singles 1969-1981, Carpenters, 2011, from HDtracks.

 

I know that there will be disagreement -- and if so, give me a chance to decode some cuts that might cause valid skeptcism & present the results -- the improvement is usually eye-opening (including better stereo depth.)  Undecoded DolbyA material tends to be a bit 'thin' in the stereo sense.

 

Also, on an individual basis, I can provide all decoding information, and willing to provide a timed-out version of the decoder for verification.  (Hell, not just for verification -- but material in your archives -- with a few hints, you can decode them yourselves.)  Doing DolbyA-like decoding without tones is NOT for the faint of heart, though.  The decoder runs on Windows, and much prefers a CPU of at least Haswell class (needs AVX2), or runs slowly on an SSE2 type machine (recent ATOM or i3000 class.)  The development is on Linux, and is frankly easier to use on Linux, I don't normally produce a product version of the Linux code -- but would do so, if desired.

 

* disclaimer, even though the DHNRDS DA can decode DolbyA material probably just as accurately as a true DolbyA, and with less modulation distortion, the DHNRDS DA is NOT a Dolby device and not sanctioned by DolbyA at all!!!

 

PS: will look up the S&G and Flack albums poste haste.  Give me an hour or so!!!!

 

John

 

 

 

 

 

[John Dyson]

Additional albums:

Bridge Over Troubled Water, 1970, HDtracks 2014, 24-192

Softly With These Songs/Best of Roberta Flack, 1993

(have several other S&G albums --- some are HDtracks type things, others are CDs.)

 

The reason for the delay -- I always double check when I lose immediate memory.

 

Truly -- and this is my guess, that my random selection of pop (well, you can probably guess

my taste in pop), through all of my collection, perhaps 2/3 or more are DolbyA or molested DolbyA (undecoded, but further 'mastered'.)

 

The worst thing that seems to happen recently:   undecoded DolbyA being further hyper compressed (e.g. The Complete Studio Recordings for ABBA.)

 

Also, I could probably SERIOUSLY disappoint many individuals who thought that they got 'remastered' or 'better masterered' material that still has a DolbyA imprint on it!!!   I am taking care not to divulge some of those -- I don't want conflicts or to hurt anyones feelings.  If more than 10-20 audiophiles had copies of the DHNRDS decoder, it COULD cause lots of trouble...  That really worries me.

 

The saving grace is that some of the incompletely mastered 'premium' releases don't really sound too awfully bad, but they do have more hiss than they should (a hint for detecting the 'forgotten' DolbyA decode operation.)

 

John

 

[John Dyson]

For fun, I have included snippets of a Carpenters decode from an HDtracks download, Simon & Garfunkel from HDtracks download, 'Killing me softly' from a premium 192/64 download and numerous others.  I am NOT a mastering engineer, but I can detect the need for a bit of EQ to soften the sound a little -- that is VERY typical for vinyl targeted material.  The ABBA stuff is probably as clean as anyone has ever heard, and the Olivia stuff is pristine (if downloaded -- I detected serious botches in the mp3 playback on her material.)

 

These are mp3, but the online player sometimes distorts a little.  Best to download, but can play online for quick checks.

 

https://www.dropbox.com/sh/ab9nhtqjforacd8/AABvt7IYgoob7VXxpN0ekK6ra?dl=0

 

[Doak]

53 minutes ago, John Dyson said:

For fun, I have included snippets of a Carpenters decode from an HDtracks download, Simon & Garfunkel from HDtracks download, 'Killing me softly' from a premium 192/64 download and numerous others.  I am NOT a mastering engineer, but I can detect the need for a bit of EQ to soften the sound a little -- that is VERY typical for vinyl targeted material.  The ABBA stuff is probably as clean as anyone has ever heard, and the Olivia stuff is pristine (if downloaded -- I detected serious botches in the mp3 playback on her material.)

 

These are mp3, but the online player sometimes distorts a little.  Best to download, but can play online for quick checks.

 

https://www.dropbox.com/sh/ab9nhtqjforacd8/AABvt7IYgoob7VXxpN0ekK6ra?dl=0

 

GREAT stuff, John.

Thank you for this very important work. 

[gmgraves]

On 8/4/2019 at 3:55 AM, John Dyson said:

As I have written -- the DolbyA decoding distortion is stealthy.  Unless one does a direct A/B comparison on intense material -- most people won't hear it.  DBX does NOT have the same characteristics of DolbyA when it comes to the distortion.  After learning what to listen for -- the problem becomes apparent.  I have found that the 363 decoding is better than an old cat22 in a 360 chassis for some reason -- but the modulation distortion (again -- it is a lack of clarity) exists on both versions.

 

* EDIT -- most of the DolbyA distortion is on the DECODE side, because of the delays and structure of the feedback loop for decoding.  The feedback loop for the decoding on the surface seems correct, and in an ideal world with zero delays -- it could work.  However, the DolbyA units work in the real world, and the delays in the additional layer of feedback create lots of opportunity for imprecision.  In this case, the imprecision is audible.  The DHRNDS does NOT have this flaw.

 

I can show multiple POP examples of DolbyA 'distortion (that is, material which has been publically available), unfortunately, the best controlled cases where there is only 1 pass of DolbyA are on material that is non-public (master tape material.)  I can provide NUMEROUS examples as desired, where I can decode leaked DolbyA vs. a decoded DolbyA copy -- the DHNRDS blows away the commercial DolbyA decoded copy for clarity in places like vocal chorus, and other complex mixes of material.  (Too complex, there is an improvement -- like on 'The Cars' stuff, but not as obvious as vocal chorus.)   On the Cars stuff, you can actually hear the intermod from synthesized material on DolbyA deocded versions, while pristine on the DHNRDS version -- again, ALL PUBLIC MATERIAL (of course, still under copyright, but I can do snippets.)

What does the distortion sound like?   Don't listen for the typical grinding kind of distortion -- even though in the extreme, DolbyA can produce it.   Listen for something that sounds like tape compression.  It is NOT a left over compression though -- even though it does sound like compression IN A WAY.   The best way to explain:  listen to a clean vocal chorus before encoding.   Then, go through an encode/decode cycle -- notice that the 'clean' vocal chorus is no longer clean, but the voices loose their individual nature.  The result of vocal chorus in the higher registers is a 'blob'.   One reason why I use ABBA a lot in testing is that they trigger the problem in most extreme ways.  However, I do have 1st generation encoded DolbyA where an A/B comparison even shows it in a natural recording in a large area setting.

 

What does DolbyA do on other, less intense material?  Generally it isn't intrusive, but there is a loss in detail -- much much much worse than anything that might happen by rate conversion down to 44.1k/16 and frankly, worse than mp3 at 192/256k.  (The loss in time resolution from mp3 doesn't have nearly the negative effect of the conversion of a vocal chorus into a blob.)

 

Where is DolbyA somewhat transparent?  Single instruments, single vocals.  Any time that there aren't multiple complex sources, then DolbyA isn't soo bad.   It is really great for first pass of mixing down material (simple components.)

 

The DHNRDS reduces this modulation distortion very significantly (I cannot say how much, but I'd suspect it is close to 1/10th as much), and makes the results more clean/clear than the alternative.  DHNRDS can make ABBA (as the extreme example) sound MUCH more clear/clean, less 'blaring' in the sound because of more distinct mixed vocals.  However, Olivia is as clean/clear as she has been before DolbyA encoding, and the IMD from 'The Cars' synthesizers is much diminished.

 

Examples available upon request.

 

John

 

 

 

Well John, I'll certainly take your word for it. I don't listen to pop recordings and certainly have never recorded any pop material. I have recorded various symphony orchestras, classical wind ensembles, and jazz performances by such luminaries as Dizzy Gillespie, Aaron Copland (conducting), Stepan Grappelli, Phillipe Entremont, Nathan Milstein, etc. I have used Dolby A and tried DBX (didn't like the sound at all). In my early forays into recording the San Jose (CA) Symphony, I used Dolby B, which was pretty benign (the big two-metered, round-trip TASCAM unit - forget the model number, it's been 40 years). But once I bought my (used) Dolby-A setup, I never looked back until I went DAT in the early 90's.

[John Dyson]

2 hours ago, gmgraves said:

Well John, I'll certainly take your word for it. I don't listen to pop recordings and certainly have never recorded any pop material. I have recorded various symphony orchestras, classical wind ensembles, and jazz performances by such luminaries as Dizzy Gillespie, Aaron Copland (conducting), Stepan Grappelli, Phillipe Entremont, Nathan Milstein, etc. I have used Dolby A and tried DBX (didn't like the sound at all). In my early forays into recording the San Jose (CA) Symphony, I used Dolby B, which was pretty benign (the big two-metered, round-trip TASCAM unit - forget the model number, it's been 40 years). But once I bought my (used) Dolby-A setup, I never looked back until I went DAT in the early 90's.

 

I have to admit that back  in the day (in the 1980s'), I played with the consumer dbxII, and thought it was nonsense (too much noise-pumping/noise modulation.)  That was back when I was not really into audio, not so much a quality crazy like I am now, and still thought dbx to be useless.  DolbyB was more benign and barely adequate (and what I accepted for my own personal use at the time.)  Boringly, my upcoming projects do include dbxI (and II) (yech!!!),  DolbyB/C (boring, HW already does a good job).  Currently working on Telcom C4 -- producing test tape material right now.  SR might be coming in the future, but not sure if I'll live that long (complex.)

 

* Want to clarify -- my project is meant to UNDO DolbyA -- no further encoding.  To me, there is no good benefit for encoding anymore.  The DHNRDS DA is meant as a one-time-only almost 'forensic' recovery of the material, so that it can be  placed into the 'perfect' digital domain.

 

On the DolbyA side of things -- it is pretty good on orchestral, less dense, stuff.*   It is very plausible that most would feel that it is okay -- it isn't HORRIBLE on many kinds of material.    It does have audible flaws (at least one recording engineer from the time that I know didn't like it, him recording soft Jazz type stuff) -- but it isn't terrible for such material though -- but pretty bad  for processed pop.  I wish I could show you a 1st generation copy of a master tape that I have (I mean, direct recorded DolbyA, only one encode operation on the material), and comparing a DHNRDS DA decoded version, a 360/cat22 decode, and two different 363 decodes -- even though all sound 'okay' -- the DHNRDS version is eye opening with detail.  (This is the -- don't know if I can say -- but a recording of a fairly famous music persons memorial service.)  I have other materials (safety copies/master tapes) where the benefit is also noticeable.  Describing two -- one is a Tchaikovsky full two sided tape, and the other is a to-be-released ancient master tape type thing -- soft/pop-like jazz.  There are significant improvements to cymbals type material, and violins (you know -- they slightly-bigger-than-violin type instruments) are improved -- not 'enhanced' though.

 

* As there is more and more chance for IMD (and just MD), the worse that DolbyA becomes.  Simpler material is not as badly damaged.

 

Now, comparing with an 'improvement' vs. being 'a lot better' -- this 'a lot better' decode is much more noticeable on 'pop' and on the straight 1st generation encode only cycle.  When there have been multiple encode/decode cycles,then the improvement is less.  Oddly, I am seeing greater improvement on the multi-pass stuff than I had thought -- but definitely more loss than with pure, 1 encode-only material.

 

DolbyA is not cr*p when recognizing the hiss from the past (esp multi-track, mixing down with simple -- single sources) -- but the other starting-in-1980s 'digital' problem isn't about the DolbyA  quality loss.   The big problem in pop is the too-often TOTAL LACK OF DECODING -- and it leaks into the consumer realm.  I would not have ONE pop recording without the 'feral' DolbyA being leaked all over the place -- 2/3 of my collection, not specially selected.  None of my contacts/project contributors have had any true, studio-with-tones pop DolbyA material to test with, it all has been more of the orchestral/jazz type stuff.

 

So, I happily accept and agree with your position - you have a very practical and rational position -- sounds like pretty much my own view.   We (consumers) do have a bit of a problem -- frustrates the h*ll out of me when people whine about 44.1k/16 bits when they are listening to DolbyA compressed material!!!   Even at 192/24 -- it won't undo the DolbyA imprint...  It needs to be decoded.   It is crazy-making...

 

John