Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Possible reason why many old music releases sound harsh... (Read 12218 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Possible reason why many old music releases sound harsh...

I have had reason to try to improve/reclaim some music from some old master tapes...  Well, I am not the typical audio professional, but I have found something interesting.  My my aggressive, desperate attempts to reclaim an old tape, I had to end up writing a DolbyA decoder.   I noticed that the decoder was the solution to the harshness of the source material.  I also remembered that several of my recently acquired music collections also had the harshness.   I applied the DolbyA decoder to several music sources, including music from the Carpenters, Carly Simon, etc.   It is pretty clear that either sidechain expansion really makes a bigger improvement than I had thought, or the material was DolbyA encoded.   Note that using the decoder isn't just using it in one L+R pass, but often benefits from an M+S and an inbetween pass.   I found that the 3axis decoding worked especially well on some Carpenters stuff, but a straightforward deocde worked well on Carly Simon's digital copy.  So -- I might have a line on some of the harshness that had been mentioned over the years.   Please forget all of the historical comments that DolbyA encoded stuff is unlistenable -- it is DBX that is very aggressively compressed.  DolbyA only does about 10dB of compression over a 40dB range (but it isn't a db linear compression like typical DBX.)  So, what do you think about ths?  I plan to release a freely copyable (but limited due to a lack of time) DolbyA decoder if this might be helpful.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #1 –
Quote from: jsdyson on
I have had reason to try to improve/reclaim some music from some old master tapes...  Well, I am not the typical audio professional, but I have found something interesting.  My my aggressive, desperate attempts to reclaim an old tape, I had to end up writing a DolbyA decoder.  I noticed that the decoder was the solution to the harshness of the source material.  I also remembered that several of my recently acquired music collections also had the harshness.  I applied the DolbyA decoder to several music sources, including music from the Carpenters, Carly Simon, etc.  It is pretty clear that either sidechain expansion really makes a bigger improvement than I had thought, or the material was DolbyA encoded.  Note that using the decoder isn't just using it in one L+R pass, but often benefits from an M+S and an inbetween pass.  I found that the 3axis decoding worked especially well on some Carpenters stuff, but a straightforward deocde worked well on Carly Simon's digital copy.  So -- I might have a line on some of the harshness that had been mentioned over the years.  Please forget all of the historical comments that DolbyA encoded stuff is unlistenable -- it is DBX that is very aggressively compressed.  DolbyA only does about 10dB of compression over a 40dB range (but it isn't a db linear compression like typical DBX.)  So, what do you think about ths?  I plan to release a freely copyable (but limited due to a lack of time) DolbyA decoder if this might be helpful.

John Dyson

AFAIK only studio master tapes were Dolby A encoded.

Regular commercial tapes were generally  encoded with Dolby B.

There was such a thing as Dolby C, but while recorders with it were made for the mainstream home market, I never saw any prerecorded tapes that were any kind of Dolby NR but Dolby B.

There is a freeware product with Dolby B decoding, but I can't remember its name.

Re: Possible reason why many old music releases sound harsh...

Reply #2 –
You are right that only studios did DolbyA encoding/decoding to maintain SNR (signal-to-noise ratio.)  However, I also just found out that some recordists did use DolbyA for high-frequency enhancement for voices.  One example, is that there is a rumor that they sometimes used DolbyA for the purpose of brightening Karen Carpenter's voice.   I do have some audible proof that either DolbyA or a sidechain expander like a DolbyA decoder makes AMAZING improvement to some of the overly bright releases.
One comment that I read from a while ago was that the overly bright edge couldn't be EQed out...  That likely means alot of clipping or some kind of HF compression.   
I AM NOT claiming that DolbyA is being actively used in consumer releases, but on the contrary, I am claiming that it is possible that some of the 'anthologies' or 'high quality' music releases could be LEFT DolbyA encoded instead of being properly decoded.  If so, it is a benefit to us people who like good sound, because a DolbyA decoder can be much less expensive than one used to be.
Again, I am claiming the possibility that some releases are now being done by running off the master tapes, or worse yet -- some might be 'finalized' (dynamic range modified) without doing an explicit DolbyA decode.
I have one simple example from the Carpenters (The 'We'll be right back after we go the bathroom") directly from the anthology copy and then the example after being DolbyA decoded.
If enough people are interested, I can slightly repackage my DolbyA processor -- wont call it that, but instead my sidechain expander) and distribute it for free.   It currently only works at 96k, with .wav files and with 16bit signed integer or 32bit floating point.
If this is true, then there is a minor breakthrough here for those of us who hav wondered why some CDs and digital releases don't sound right.

John

Re: Possible reason why many old music releases sound harsh...

Reply #3 –
Quote from: jsdyson on
You are right that only studios did DolbyA encoding/decoding to maintain SNR (signal-to-noise ratio.)  However, I also just found out that some recordists did use DolbyA for high-frequency enhancement for voices.  One example, is that there is a rumor that they sometimes used DolbyA for the purpose of brightening Karen Carpenter's voice.  I do have some audible proof that either DolbyA or a sidechain expander like a DolbyA decoder makes AMAZING improvement to some of the overly bright releases.
One comment that I read from a while ago was that the overly bright edge couldn't be EQed out...  That likely means alot of clipping or some kind of HF compression.   
I AM NOT claiming that DolbyA is being actively used in consumer releases, but on the contrary, I am claiming that it is possible that some of the 'anthologies' or 'high quality' music releases could be LEFT DolbyA encoded instead of being properly decoded.  If so, it is a benefit to us people who like good sound, because a DolbyA decoder can be much less expensive than one used to be.
Again, I am claiming the possibility that some releases are now being done by running off the master tapes, or worse yet -- some might be 'finalized' (dynamic range modified) without doing an explicit DolbyA decode.
I have one simple example from the Carpenters (The 'We'll be right back after we go the bathroom") directly from the anthology copy and then the example after being DolbyA decoded.
If enough people are interested, I can slightly repackage my DolbyA processor -- wont call it that, but instead my sidechain expander) and distribute it for free.  It currently only works at 96k, with .wav files and with 16bit signed integer or 32bit floating point.
If this is true, then there is a minor breakthrough here for those of us who hav wondered why some CDs and digital releases don't sound right.

Well what sounds right is always subjective.

I try to avoid subjectivity as much as possible when I describe things, preferring to comment on certain audible effects.

The idea of using Dolby A record processing as a channel EFX is one I've heard before, but not so  much about. Confirmation that this has been done at times might be found here: https://www.soundonsound.com/techniques/bright-idea . The article also talks about simulating the effect using a modern DAW (Cubase).  

Having a Dolby A encode channel as software might be an attractive alternaive.

I'm a bit of a Carpenter's fan and went back and looked at my archives, as well as Youtube versions of various song tracks from TV shows of the days and Karoke videos. I think that an example of the sound effect you may be describing may be here: https://www.youtube.com/watch?v=rzbZaVPrz0I.  It may have been applied to Karen's vocal track, as something cute and unnatural seems to be going on. 

There are a number of examples of this vocalist singing this song in various contexts. I suspect that the BBC version was probably engineered to sound as natural as possible.  Sometimes there are contract riders that attempt to provide technical consistency but what happens in every instance may or may not go as planned.

Long ago I engineered the live sound for some large performances of some very famous acts of the day. As I said, what happened is what happened. ;-)

Re: Possible reason why many old music releases sound harsh...

Reply #6 –
The origional Dolby A patent reveals most relevant details:

https://www.google.com/patents/US3846719

Figure  7 summarizes it pretty well. It split the audio into 4 bands,

30-150 Hz
150-1500 Hz
1.5 KGz-5 KHz
5 KHz-20 KHz.
And performed dynamics compression on each band. This amounted to gain reduction that started above about -40 dB, and increased with increasing level up to about 20 dB, at which time the limiting was constant up to 0 dB.  In short, all audio above the approximate noise floor was limited at least a little.  The nature of the limiting was more like compression than clipping.

Also: https://en.wikipedia.org/wiki/Dolby_noise-reduction_system

At playback, the process was approximately reversed.

Dolby B differed in that there were only 2 bands, and the frequency where the bands were divided varied with input level.

When executed in the analog domain, thie Dolby A system required frequent manual adjustments to be as sonically transparent as it was. It was sensitive to variations in overall levels, and since the sensitivity of analog tape was not rigidly fixed, it may have also required manual adjustment every time you used a different roll of tape. The sensitivity of the tape may also vary along its length.

In time obtaining this kind if EFX  was separated from the Dolby hardware, and devices called Multiband compressors were evolved and used. for making recordings. 

The frequency bands could be used to process fundamentals, harmonics, and sibilance separately.

Re: Possible reason why many old music releases sound harsh...

Reply #7 –
Regarding DolbyA's need for adjustment -- I looked at an old schematic, and it required alot of selected resistors/etc, for it to work correctly.  It used discrete components and depended upon the threshold of a JFET (notoriously different from part to part.)  Once it was set-up, then even playback was best done after checking the same kind of  thing as a DolbyB 'dolby level.'  Frankly, after playing with consumer DBX back in the day, I had originally thought that the actually better undecoded DolbyA would sound as bad as DBX, but it isn't.   It just pretty much sounds like an HF emphasized compression scheme when not decoded.  For an extreme case of DolbyA, it appears that ABBA might have used multiple passes of DolbyA (or something similar) to partially achieve their characteristic sound.
Myself, I have developed a very complete 'finalizer' that does expansion, compression, and limiting in about 6-7 steps, but doesn't really cause any significant intermodulation (and no cascading of compression effects.)  However, when DolbyA is used in succession the best way to describe it is kind of an extreme ABBA sound or something that sounds like 'Shake it Off".

ONE VERY IMPORTANT NOTE -- if someone wants to use a real DolbyA unit to decode some of the music that would benefit -- it is important to realize that just one pass of L+R isn't enough.  I have found that 2 or 3 (maybe 4) passes are often needed:  L+R, halfway between L+R/M+S, and then M+S.  Sometimes an additional L+R is beneficial.  When just running one L+R pass, you might see a benefit, but not as much as can be obtained.
When trying to use non-purpose-specific expander software to try to emulate DolbyA, please note that you'll never get the 9K+ range correct, DolbyA is a sidechain based expander -- not inline, DolbyA is not dB linear (like DBX), thresholds are important -- but not extremely critical, it might be best to try to expand the 80Hz on up and ignore the 80Hz on down, and the timeconstants are RC in a linear domain -- not squared -- alot of compressors/expanders work in the squared/dB linear domain, and finally the DolbyA attack timeconstant is NOT CONSTANT, but has a formula (after much experimentation) where the attack time starts the same as the decay time, but then speeds up with a second order factor of 70, and the 3rd order factor of 0.25 (which means that the attack timeconstant decreases kind of like DECAYTC/(1+70*(in-out)^2 + 0.25*(in-out)^3), where the attack/decay TCs start off as being equal, but then the attack speeds up like the factor that I described (that is directly from my code -- after 100s of iterations to provide the best decoding.)
The DECAY factor for the digital filter is about 0.0004 (at 44.1k) for LF/MF -- and 0.00075 for both HF bands.   The first order attack time factors are 0.01 for LF/MF, and 0.02 for the HF bands.   The second order for the attack is that complicated calculation above -- and note AGAIN that the attack/decay times start out as being the same for slow level transitions, but increase fairly rapidly as the attack speeds up.

Lots of techie here - but suffice to say that DolbyA is gentle enough that some mild, careful expansion by a general purpose expander can help a little bit.

I intend to distribute a copy of the working Linux version today much later on (might be late, but that is my intention), and a command line windows version in a few days.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #8 –
Quote
It just pretty much sounds like an HF emphasized compression scheme when not decoded.
Are you sure it's not CD pre-emphasis?  (A CD with pre-emphasis should play OK on a CD player, but it won't he de-emphasized when ripped.)

Or, were these recordings made when Aphex was popular?

As you probably know, many effects are not reversible (including any effects applied before mixing) and some effects that might be theoretically reversible (some kinds of compression for example) are not reversible in practice because of the unknown settings/parameters.   I'm not saying that you can't "improve" it, or make it more acceptable to your tastes, but there are limits to what can be done.

Re: Possible reason why many old music releases sound harsh...

Reply #9 –
I am very sure that it isnt' just CD pre-emphasis.  Other people have noted that the harshness couldn't be eq'ed out.  There is definitely a compressor in there somewhere.   Also, cleaning up the sound quality is too well correlated with the parameters of DolbyA.   There might certainly be the results of a primitive (but effective) compressor in some of the test results, but it is pretty clear that they are using a relatively mild compression which matches DolbyA pretty closely.
I have attached two examples, not totally definitive, but definitely show that the undecoded.mp3 benefitted very significantly from the decoded.mp3.  In this case, the improvement was less related to sibilence rather than ambience.

The second cases (the tst-files) show a general improvement -- where the processed version reminds me very much of the old carpenter's sound.  The source was from an anthology set.

I have found similar results from a high quality Carly Simon download.

UPDATE:
WRT my promise of a decoder today -- it will probably have to be tomorrow -- but I did make good progress...  There were lots more hooks from the entire finalizer than I had remembered, so it will take a bit more time to resolve them and to make the command line interface (however primitive) more complete for the application.   The version that I plan to distribute tomorrow will only be .wav file/96konly/32bitfp or 16bit signed integer, and should have the ability to do L+R,M+S and inbetween in any order, and the ability to specify thresholds and gains for each step.  That command interface isn't difficult, but is tedious to put together.  SO it should be someting like this:  da <infile.wav >outfile.wav (command params) or da --inf=infiile.wav --outf=outfile.wav --cmd="(l,-6.0,0.0)(m,-6,0),(b,-6,0)"
the second command above, where there are command params will mean;  L+R, -6dB threshold, 0dBgain ... M+S, -6dB threshold, 0dB gain, between, -6dB threshold, 0dBgain.
I used the name 'b' above for the 'between' matrix.   I did alot of testing, and found the best choice for that between matrix, and it does seem to be commonly used.

John

John

Re: Possible reason why many old music releases sound harsh...

Reply #10 –
Quote from: Arnold B. Krueger on
At playback, the process was approximately reversed.
Approximately, that's exactly the reason why many engineers of my generation had avoided to use it at all costs.

Re: Possible reason why many old music releases sound harsh...

Reply #11 –
Quote from: jsdyson on
I am very sure that it isnt' just CD pre-emphasis.  Other people have noted that the harshness couldn't be eq'ed out.  There is definitely a compressor in there somewhere.  Also, cleaning up the sound quality is too well correlated with the parameters of DolbyA.  There might certainly be the results of a primitive (but effective) compressor in some of the test results, but it is pretty clear that they are using a relatively mild compression which matches DolbyA pretty closely.
I have attached two examples, not totally definitive, but definitely show that the undecoded.mp3 benefitted very significantly from the decoded.mp3.  In this case, the improvement was less related to sibilence rather than ambience.

I'm confused.   Two IMO separate and distinct issues have been discussed:

(1) Dolby A processing via new software

(2)  How Dolby A encoding only can be used as a special effect

There logically might be one or more pairs of files to illustrate each.

There would be set(s) of file pairs would show how good of a of a job the Dolby-A software works as a noise reduction tool.  IOW, how sonically transparently a round trip is both for a perfect recording medium, and also for recording mediums that have common sorts of spurious responses and noises.  It is probably critical that these files be 44/16  or 44/48 .wav files.

Ideally, some test tones such as sine sweeps at various levels could be used to determine how well the new software conforms to the documented (patent and other) the standards for Dolby A processing.

There would be another set(s)  of file pairs that illustrate how Dolby A encode works as an EFX processor.

The most ideal situation would be if the software Dolby A encoder be available in a form that could be easily run in Windows and/or Android, as they are far and away the most popular OS's.   Then people could invent tests, challenges and applications to their heart's content and find out what they want to know, limited only by their general audio technical skills.




Re: Possible reason why many old music releases sound harsh...

Reply #12 –
My software currently works only as a decoder, however due to the necessary structure of DolbyA decoding, an encoder shouldn't be too difficult to make.   I am currently focused on restoring old music, but the matter of using DolbyA as an effect has popped up elsewhere when discussing this subject.

My focus is mostly on recovering old recordings so that they sound like they do when they were originally released.   However, I must say that given some of my experiments with some ABBA material (the stuff that I have hasn't been finalized or compressed beyond that of DolbyA), that they seem to have used several passes of DolbyA for the sound effect.  'Shake It Off' truly appears to have had several passes of DolbyA also (to give that crunchy sound during the entire song -- not the short 'carbon microphone' sounding dialogue).

So, I promise to keep my focus on the restoration aspect of this matter, and another thread might best be used for the 'sound effect' aspect.  I am TOO BUSY getting a workable/usable and fully separated version of the decoder working today to have the time/energy to work out all of the details of an encoder also.   Frankly, I find very little use for a DolbyA encoder nowadays -- digital recording is generally so very accurate that adding DolbyA would be a liability from a quality standpoint.   I don't live on the 'creative' side of the world, so using DolbyA as an effect isn't my 'thing.'

I apologize for confusing the matter -- again, I am focused on recovery, not using the DolbyA for effect...

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #13 –
I have attached an amazingly short Linux x86-64 program that does the dolbyA decoding that we have been talking about.  This program is somewhat limited due to the time that I have spent on this specific version.  It does appear to work,and it supports all kinds of decoder configurations (e.g. just normal stereo l+r, or the whole shebang witih many steps.)  This most useful would be doing l+r, in-between and then m+s.  The likely most useful command line might be this:

da <infile.wav >outfile.wav --info --cmd="(l,-6,0)(b,-6,2)(m,-6,2)"

which does the full 3axis decode and provides a realtime progress output.  This is a primitive user interface, so be careful.  I'll produce a Win32 and Win64 version soon -- no GUI, just nasty raw command line like this.  Let me know of any problems, and I'll fix ASAP until this is stable.  I have included a rather wordy help file (dainst.txt) along with the binary.

THE PROGRAM NEEDS TO BE RENAMED FROM "da.program" to "da", and execute permissions need to be added for it to work.
This is only a Linux x86_64 program -- that is where I do all of my research/programming, but I intend to produce a Win32/Win64 version -- exact replica -- within a week.

Good luck, and tell me if problems -- I am happy to fix.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #14 –
I tried to update my previous post -- sorry if this is a repeat -- I forgot that there is a major CPU dependency in the released version of the decoder.   it only works on processors with AVX2 (maybe AVX.)  So, I happen to use a Haswell.   I feel guilty, but tied up tonight.  I'll get a non-AVX2 dependent version working on Linux tomorrow (it is just a rebuild, but will run significantly more slowly.)   I'll do a 32bit version also, which will prepare me for the Windows versions in a few days.
SORRY AGAIN!

Re: Possible reason why many old music releases sound harsh...

Reply #15 –
Quote
My focus is mostly on recovering old recordings so that they sound like they do when they were originally released.
Do you mean the original vinyl?

If the CD was remastered, that's usually compression/limiting and maybe some EQ.

Re: Possible reason why many old music releases sound harsh...

Reply #16 –
Quote from: DVDdoug on
Quote
My focus is mostly on recovering old recordings so that they sound like they do when they were originally released.
Do you mean the original vinyl?

If the CD was remastered, that's usually compression/limiting and maybe some EQ.

I believe that there  the results of decoding sound CLOSER to the original vinyl type sound, but there is a certain warmness to vinyl (they had to use lots of eq/feedback cutting amplifiers  to make vinyl work correctly.)   i don't think that an exact emulation of vinyl sound would be possible (or in some cases, desirable -- click/pop :-)).   However, when using the decoder on appropriate source material, the resulting general balance of the sound is MUCH closer to the original 'Carpenters' sound.     DolbyA encoding isn't flat, and will tend to boost the frequency regions which are of lower level, and also will give a double boost to the frequencies above 9kHz.  (Not 3 or 6dB louder, but a double boost of compression which will tend to make the HF sound more intense than just the simple rebalance caused by the multi-band sidechain compressor.)

Regarding 'remastering'...  I have found that SOME 'remasters' or 'anthologies' APPEAR to be runoffs of the master tapes (or slight variations thereof.)   When/if they do add compression/limiting to a tape that should be, but hasn't been DolbyA decoded, then it is much more difficult to do the DolbyA decoding.   Just EQ is very likely possible to be mostly undone, but dynamic processing can cause the modification or loss of the 'hints' in the audio used by the DolbyA or other NR processes.   The DolbyA decoding is mostly just a carefully crafted expander, so 'DolbyA decoding' a damaged DolbyA encoded recording might result in an improvement, but the natural decoding information would be damaged also.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #17 –
Quote
I believe that there  the results of decoding sound CLOSER to the original vinyl type sound, but there is a certain warmness to vinyl...
 However, when using the decoder on appropriate source material, the resulting general balance of the sound is MUCH closer to the original 'Carpenters'...

DolbyA encoding isn't flat...
My point is, if Dolby A was used as an effect  it would be on the "original release" too.    If you like the results you're getting with the Dolby A decoder, that's fine.   But, nobody applied Dolby A when re-mastering for digital release.  


BTW -  Do you still have the vinyl?     If you are comparing the CD to 40-year old memories you probably don't remember what the original sounded like.  ;)    And if you do have the vinyl and you prefer the original sound, it might be easier to de-click a digitized version of the vinyl.    

Re: Possible reason why many old music releases sound harsh...

Reply #18 –
Honestly -- I still don't have the vinyl anymore, but I certainly would have remembered the (mostly apparent) HF compression artifacts of DolbyA even back then.   I was working on audio AGC back in the '70s, and have always been very aware of compression effects and negative artifacts.   I am even aware of artifacts that most people dont know about or understand (one example: the intermodulation effects that are so common with fast or poorly designed expansion or AGC systems.  harmonic-type distortions aren't the only troubles caused by AGC systems.)

So, I shouldn't have said 'sounded more like the vinyl', but instead 'removed the compression/noise artifacts in newer releases so that the sound appears closer to what I remember from the vinyl.'   I used shorthand when I wrote the original claim.

For example, there is an obvious artifact in an 'anthology-type' release from Carly Simon.  The apparent power buzz from one of her songs is much worse than it used to be in the original release.  That buzz is fairly well suppressed by using the DolbyA decoder -- and I remember that buzz from doing design of audio AGC when the music was new -- as you know, simple compressors can make background noises worse.  Once listening to the decoded version of the song (mitigating a big part of the buzz), then the compression artifacts occuring when not using the decoder become very obvious.

Frankly, I haven't been intending to discuss my ability to determine the quality of playback. We'd best do individual comparisons with the decoder that I have offered.   I don't have the decoder working on Windows or earlier Intel Chips yet, but I do promise to have the decoder using less capable Intel CPUs ready within another day (and most likely tonight unless there is an emergency of some kind.)  It will take another day or two of work (most likely a few hours of work) to get a Windows version working.  However, the calendar time might be  on the order of 1wk for the Windows version...  (For a couple of reasons, Windows is a lot less convenient for me to do development on.)

No matter what, I intend to make raw version of my decoder (like what I have distributed) available for free use and distribution.  it will end up being a little nicer/cleaner in a week or so.  For now, people now have a primitive version of the program that seems to help mitigate some about of the compression artifacts and re balances the audio spectrum.  There might become a commercial/easier to use/plugiin version made available in the future -- and I will probably offer appropriate use of the source code for commercial purposes, but that is NOT my focus.  I might even give away the software source code once I remove all of the technology that I actually care about continuing to control.

I don't have golden ears, so I am certainly willing to let other people listen.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #19 –
okay -- I have produced the programs that I promised -- each one appears to work well, and I even have a windows version working.  Because I am trying to avoid confusion, I am not distributing all of the versions that are optimized for every CPU -- there are too many choices (I havent added the automatic CPU selection stuff into the code yet.)   So -- here is what I am giving you (I'll send you more optmized versions if you ask):

LINUX versions:
da32-sse3 -- works on most machines made since the middle 200X timeframe.   Might not work on the earliest Atoms. 32bit or 64bit
da32-p4 -- works on practically all machines, but is also the slowest version. 32bit or 64bit
da-avx -- works on most recent CPUS, 64bit only -- the fastest version so far -- often works in realtime.
(when running the 32bit versions on 64bit systems, you have to already have the 32bit shared libs installed also.)
Also, before running the file that you downloaded, you'll need to rename the program (remove the .prog extension), and then add execute permissions, e.g.:  'mv da32-p4.prog da32-p4", then "chmod +x da32-p4".

Windows version (zipfile containing program & needed cygwin libraries -- unzip & run in same directory as the binary):
dawin64.zip

I already downloaded the short instruction file on an earlier posting.  As long as there aren't any unforseen incompatibilities, after downloading these, you should probably have a useful DolbyA decoder that helps to clean up alot of old music that has been recently distributed!!!

Tell me if there are problems or if you need/want help!

John

Re: Possible reason why many old music releases sound harsh...

Reply #20 –
I'm not sure if this "background" will help much, it's sort of a brain dump.  But within the past month I had a pair of Dolby 361 units on the bench and tested them for tracking encode>decode over a wide dynamic range.  Tracking is easy and spot-on from -20 and up, it gets more challenging at lower levels, so I wanted to see if this vintage gear was working up to spec.

To start with, everyone that used Dolby A for noise reduction also used the built-in "Dolby Tone" generator to head each tape with a reference tone.  The specific tone level was to be the reference fluxivity the studio used, which ranged from 185nWb/m to 250nWb/m for those using Vu meters.  On playback that tone would be adjusted to hit a little black dot in the center of a horizontal level meter on a Dolby unit, but it also corresponded with 0VU, so not much adjustment was usually required.  Once that was done, and assuming the recorders were reasonably flat, Dolby tracked itself excellently, and far better than dbx.  But reasonably flat response was key to tracking for any NR system, and non-flat response was the un-doing for any single-band system like dbx or even Dolby B or C.

So my test units were first checked for noise and distortion, a bad power supply filter cap replaced, then connected to each other, one set to "record" (encode), the other to "play" (decode) and calibrated to each other using the record unit's Dolby tone. and FR measurements made with sweeps from 20Hz to 20kHz at levels from reference to ref-40dB.  Tracking was quite remarkable, with the worst being at -40 where variances of 1dB were seen.  Now, I couldn't really test for dynamic tracking, but that was something Dolby was pretty good about with careful component selection resulting in tight time-constant tolerances. 

Now about practical use of Dolby A.  I worked at a studio for 15 years in the late 1970s, to early 1990s, and we used Dolby A (and eventually SR) for most if not all analog tape recordings, eventually having 361 units on every recorder in the house.  We never used it as an effect, though, being focussed on classical music.  But I did hear quite a bit of encoded, but not decoded Dolby A, and it was really not pleasant or listenable. Worse was decoded but not encoded Dolby A, something which used to happen quite often in TV as VTRs eventually had on-board Dolby A cards in them, but the use of Dolby A wasn't consistent. Stuff was broadcast with decoding all the time, and sounded terrible.  To use encode as an effect I would expect it to be used in a side-chain where the effect could be controlled and moderated, because it was pretty wild otherwise.  Also, it would never have been applied to the entire recording, but may be on the track level, making the reversal of it in a stereo mix impossible. 

I listened to some Carpenters to try to see what the fuss was about.  Karen's voice certainly has some boosted top end, but it doesn't sound anything like Dolby A to me.  I would suspect an Aphex Aural Exciter[ may have been used, but there is a lot of Carpenters material that would pre-date that device.  What I hear sounds more like some simple EQ on the vocal.

Dolby did make a unit that used the Cat. 22 Dolby A card as a single-ended noise reduction device, called a Cat. 43, which gave users access to the band thresholds making it a useful multi-band, single-ended (no encode) noise reducer.  It was used a lot in film production.

I would be highly skeptical that applying Dolby A decoding to any commercial release would improve anything.  The OP stated these were "old master tapes", but of course they can't be masters, as those would be in studio vaults.  We don't have any info on the tapes themselves, format, speed or where they came from, so it's hard to even theorize.  But they can't be actual "masters".  They could be a dub of a Dolby A master somebody absconded with, and thus may not have the requisite yellow Dolby sticker on them.  And those would indeed require full Dolby A decoding to sound right, but there also should have been a Dolby Tone at the head or tail for calibration.

Since I still own a Dolby 365 card frame and Cat. 22 "A" cards for it (also have Cat. 280 "SR" cards), we could do some testing with actual Dolby processing.  Just a thought.

Re: Possible reason why many old music releases sound harsh...

Reply #21 –
Regarding the 'listen ability' of undecoded material, and the real improvement that I am hearing.   I am certainly willing to email you the before and after when using the decoder over very obvious examples (I am carefully following the schematic and specs WRT my decoder.)   As far as I can tell, DolbyA only a limited gain change over a 40dB input range, but concentrates its gain change over a narrower range -- it is basically 4 parallel fet compressors with carefully controlled attack/release times (but not using normal RC time constant characteristics in the attack.)   The expanders that I wrote tries to mirror the behavior of those old FET compressors.  (Actually, it is NOT 4 disjoint compressors, but they work in a brilliant combination for the 1960s timeframe.)  Maybe by emulating the effect of the DolbyA decoder, I am taking advantage of the work of a genius -- but my decoder does choke (sound bad) on material that doesn't sound like it is encoded or heavily compressed.

So, I am seeing HUGE improvement and very good control of background noise (where results in constant level -- no subjective noise pumping with residual noise -- but a slight/mellow rush at very low level that would be expected.)  Maybe I am getting the benefits of the sidechain expansion, but i get NO surging, and the ambiance in some recordings are very significantly improved by doing a three axis decode.
I have had NON-AUDIOPHILES mention the incredible improvement in some Carpenters stuff.
I havent' directly claimed that the material is DolbyA encoded, but I am definitely using DolbyA parameters to do the decoding.  If you wish, send me a short copy of some music (any common format) for reasonable quality so that it can be decoded, and an idea of the reference level -- and I can send you the results.   Also, I have distributed a Windows64bit version of the program which I believe will work (I had to add some shared libraries from the cygwin environment -- hope that I sent all of them)   The program does NO mischief or networking -- it is doing something VERY good.

Frnakly, it did wonders with Simon&Garfunkel music that I have just tested it with.   The original was much more hissy/noisy than the other music that I tested it with, and it works very well.   I had felt that much of the recently ancient-recorded & downloaded music was messed up somehow, and my project has basically undone most of the damage that I hear.

I truly do not know what is going on here -- other than an expander built on the approx parameters and character of DolbyA undoes most of the damage.   Also,  I just found a patent online for a DolbyA decoder of very different design, but from what I can tell -- it would do similar to my decoder.   Their (patented) design appears to miss an aspect of decoding that might make it sound differently.  My decoder only approximates the behavior of the 9K+range, but the differences don't' appear to be audible (my hearing is gone at 15k anyway.)

If the decoder works well as a general purpose expander, I can do much/much better than the 'DolbyA' decoder approximation that I wrote!!!   I mean, as  a GP expander, there is A LOT more that I can do to better track normal music material.   I have dynamic attack/release algorithms that I have designed (and use very well) that have literally ideal attack/release times and cannot practically be done in hardware... (e.g. the lowest level filter has an algorithm that if you have a 1msec attack/dwell time, you'll get a 1msec decay, likewise it mirrors the dwell & decay times all the way to seconds timeframe -- but that is not the only filter used,  but it is a good tool.)

So, I'd REALLY APPRECIATE ANY HELP THAT YOU GIVE ME.  Most of this work product is being distributed for free.

John Dyson

Re: Possible reason why many old music releases sound harsh...

Reply #22 –
Here is a short portion of the beginning of Simon&Garfunkel 'Scarborough Fair' (45seconds, probably too long -- but it needs to show what it does.)   i wanted to make it clear that both channels are working.  This expander has two independent channels, so you can hear the decrease in noise (hiss) on the left channel -- and also MUCH more correct ambiance.  If this is just working as an expander, it is doing really good things without significant artifacts.  IF the music had not been compressed in a controlled way, then the two expanders would not track so very well.

The original: sgundecode.mp3  too much ambiance -- too much background noise -- noticeable, but not horrible hiss.
The expanded (or decoded): sgdecode.mp3 sounds more normal -- less hiss -- less extreme ambiance.

I have provided this example because it has a simpler sound -- so some of the components of the damage can be easily heard.  I was wondering why the recording seemed awfully compressed for the style of music.   I set the threshold for approx 0dB of expansion loss at about -7.5dB full scale of the input.  (That is the only way that I can set the threshold since I don't have a reference level.)

PLEASE FEEL FREE TO HELP -- if you think of any ideas -- let me know.   If this is just an expander, I'll continue the project to improve further and make the result available. I have a lot of unique gain control (attack/release -- linear/squared/dB timed) algorithms that work really well as an expander, but most of my previous work was inline expansion/compression unlike the necessary DolbyA behavior.

Re: Possible reason why many old music releases sound harsh...

Reply #23 –
Doesn't quite sound like encoded Dolby A, though.  What kind of tape, track configuration and speed is this?

Re: Possible reason why many old music releases sound harsh...

Reply #24 –
The music is part of an anthology collection.
I have lots of collectons that I have purchased over time, but often noticed excess compression and feeliing like a boost of the HF also.  There is a very consistent kind of compression being done in alot of recent releases of old stuff.  I cannot describe it better other than by listening to various releases.

My most recent purchased Carpenters singles collection from HDtracks also has same problems (I am very, very aware of compression and such artifacts), which are very well undone by the expander.   The Carpenters' music sounds pretty clean, but still has too much compression to be of the utmost quality -- and the amount of compression (approx 10dB over 40dB) seems to sound about right to what I have been hearing.  It is nowhere near 2:1 over the entire range, but could be 2:1 over a very narrow range (which comes close to the curve that DolbyA apparently has.)   I found someone else who has done a more 'graphical' showing of the encode/decode curve in a patent -- I just found it a day or so ago:  US 6,807,278 B1.  My design isn't based upon the same (or similar in any substantial way) approach, but works around the same problems of a digital implementation. They didn't even handle the variable attack time like I did (I cannot see where they handled it at all.)

DolbyA cannot be super-aggressive (fast) because the intermodulation would otherwise be impossible to undo because of the realizable phase issues with almost any analog recorder.  So, a release time of approx 50msec (one set of channels is 2x faster than the other), and a dynamic attack time that varies between a bit slower than 2msec and as fast as 50msec doesnt' really sound horrid at the amount of compression that DolbyA can provide.  It certainly wouldn't be pretty, but has to be slow enough that there isn't too much intermod so that the processing can be undone.  A fixed release time is not meant to sound 'good' but is doing that for a reason.  Even the most primitive compressors mostly have a simple pseudo-dual time constant scheme (to help with transients and avoid pumping.)  The DolbyA is (apparently) meant to be as fast as it can be without having lots of unrecoverable intermod -- and to make sure there is minimal noise pumping.

(Of course, the way that the DolbyA compression works makes tracking at the lower levels much more difficult.)   At higher levels, the feedback from the compressor because less critical (because the feedback gain is less), but at lower levels the feedback into the summing node results from higher gain and therefore more sensitive.  YOUR MENTION IMPLYING MORE TEDIOUS TRACKING AT LOWER LEVELS MAKES SENSE FROM THE ARCHITECTURE.

If the material wasn't compatibly compressed(especially considering the two independent channels), there would be alot more 'surging' and pumping than what I have heard.  Here is an example:  if I had deviated from the DolbyA specs (including just removing the dynamic attack time portion of the timeconstant code), it doesn't work nearly as well.   I know compressors/expanders very well (but admittedly not an expert or hands-on knowledgeable about DolbyA), and the 'hand in glove' compatibility between the expansion parameters and the music material seems to be an odd coincidence.

I'd REALLY like to have a short clip of something that is PROVEN DolbyA compressed.   With that, I might be able to figure out what is going on.  If the expander just happens to be working over its less sensitive range, and it could be improved as an expander -- eventually it would be an amazing device given the other technologies that I haven't even thought about adding to it yet!!!

Again, I am NOT claiming that the music is definitely DolbyA compressed, but it is definitely being compressed in a way that appears to be undone by the expander designed from the known DolbyA information.

John Dyson