'FeralA' decoder -- free-to-use

[John Dyson]

53 minutes ago, MarcelNL said:

the difference in how the material presents itself is very interesting, rather intriguing.....I'll have to listen more, as on my current setup (which is mono, errr one channel, due to space restrictions in the house we currently live in) there appears to be lots of merit (the middle and higher frequencies sound so much more natural) yet there is something that might indicate a phase issue (which could mean that phase was messed up tp begin with or got mangled in the process) 

There is DEFINITELY a change in behavior of phase.

The original/encoded version is severely phase and amplitude modulated, but mostly amplitude modulated.

There COULD be a phase problem in the result, and my mind is open about some ideas.

 

My whole reason for talking online about this is to get ideas about what people think.  NOT for ego boost, but instead for helping with the

project.   Some parts might really suck -- so lets see what is wrong!!!

 

Unlike other projects, like at work or FreeBSD, I have NO ONE to help or bounce ideas off of, other than the AS people.

Help is really needed, and so if you have any good ideas, let me know.  I will REALLY think  about them!!!

 

I think that the decoder is either 1) as good as it can be, or 2) so close that just a few minor nits need to be fixed...

 

 

 

[MarcelNL]

thanks for that open minded attitude! refreshing in an environment seemingly predominantly occupied by males with hyperinflated ego's!

 

I'm not sure where the 'issue' is if it is one at all, I just sense there is a difference, I have sent some samples to a friend of me who designs audio gear that sounds incredibly transparent and organic, he has great ears and the most openminded attitude of all the people I know. I'll listen to the samples many times over the next days, yet I think it'll be hard to come to a more clear opinion without playing stereo at my end. 

I really like the DEC files for their organic quality, so much I dare say!

 

[John Dyson]

14 minutes ago, MarcelNL said:

thanks for that open minded attitude! refreshing in an environment seemingly predominantly occupied by males with hyperinflated ego's!

 

I'm not sure where the 'issue' is if it is one at all, I just sense there is a difference, I have sent some samples to a friend of me who designs audio gear that sounds incredibly transparent and organic, he has great ears and the most openminded attitude of all the people I know. I'll listen to the samples many times over the next days, yet I think it'll be hard to come to a more clear opinion without playing stereo at my end. 

I really like the DEC files for their organic quality, so much I dare say!

 

Most important to me is the comment about the egos.   I was probably the major contributor to the kernel on the FreeBSD project, but nasty egos kept hurting me so much that i had to quit.  There is nothing worse than frustrated man-children with too much testosterone and not enough time with their girlfriends....  Usually, the problem with egos isn't usually about being 'too big' (I have a big ego),  but it is that the egos are weak, and with weak egos, people can become defensive. 

 

My only weakness is that I want to help/serve, and sometimes the offer is not understood.

 

I don't see that problem here at AS as much as frustration and let-down about the project.   I wouldn't be participating here at AS if the egos were out-of-control.  Maybe I didn't communicate the need for participation strongly enough, or make enough progress that people would be truly interested.   I know of at least one person who really tried hard to contribute, but I wasn't ready for the part that was contributed.  It might be a lack of leadership thing about me, or perhaps simply the project being so darned challenging that even I was over my head.

 

This thing is now mostly behind us, and if not totally ready to use, so very close that it is within a few weeks of comments and feedback IF there are any major nits.   I cannot tell if the sound is good enough,  and I really did have to depend perhaps 90% on my engineering background, because my own hearing is extremely glitchy (just like my brain.)  I still had somewhat reliable hearing when I first started about 8yrs ago.  I couldn't stomach the sound of CDs again, that is after giving up back in about 1989-1990 because they sounded bad to me back then.  (I have a curious background, one of them is doing orchestral recordings myself -- so I had high standards.)

 

Bottom line -- try to enjoy the decoder...   It is a gift. with the intention of being kind.  I have had a good life, and want to do something good for others.

 

 

 

[John Dyson]

Since there are some new individuals looking at the decoder, I have a 'usage note' that might help improve the enjoyment.

 

When the decoder cleans-up the recording, the goal is to revert the recording to the state just before it was encoded/damaged.   When doing this reversion, the result might not be in the ideal listening state.   I believe that the bass should be relatively accurate, and if it isn't, then we still have minor mods needed on the decoder.  This bass fix will happen over the next week or two -- pretty much drop dead for finishing this phase.

 

However, for the HF, I fully expect there NORMALLY will be an upward tilt at approx 9kHz.   This has always appeared on decoding results, and actually makes some sense wrt the history of making recordings.   GIven this, for the best enjoyment on some recordings try the following EQ.  I'll also show how to implement it using the decoder, and also SoX.   On a DAW, there'll be more obvious ways to do it.

 

Often, a recording might need -0.75 or as much as -1.5dB at 9kHz.   Ideally, the EQ should be done as 1st order (single pole.)   However, it can be approximated, when necessary, but 2nd order with Q=0.50.   Sometimes, one additional EQ might be needed at 12kHz.   Also, even as small as -0.375dB might help.

 

Add one of the following command options to the decoder:

 --pvdh=9k,-0.75 or --pvdh=9k,-0.375 or --pvdh=9k,-1.5

(you can also add on the same thing, except 12kHz.)

 

For SoX, the following will work pretty well:

 

treble -0.75 9k 0.50q

or

treble -0.375 9k 0.50q

or

treble -1.5 9k 0.50q

 

Sometimes, when using 2nd order, a Q value of 0.8409 instead of 0.50 might be helpful.   0.8409 is an approximation to a Chebyshev filter, and 0.50 is an approx to Bessel.   I have found that 'Butterworth' (Q=0.707) to be not as helpful.

 

This is a lot of tech, and perhaps the best thing to do is to simply use the following by default, until getting used to the decoder.  You might even want to use -1.5dB.  I do suggest that -0.75dB is a good tradeoff.

 

--pvdh=9k,-0.75

 

The addiitonal EQ isnt default in the decoder as a matter of integrity, the decoder does 'clean' decodes.

 

 

 

 

[Jud]

This is very, very good (subjectively speaking of course) at this point, good enough that I would trust measurements (if there are informative ones to be done) over my own hearing.  So take this comment in that vein: I feel there may be either a bit missing at the top end of the human vocal range, or else something near that frequency is emphasized and so the very top end of the vocals feels a bit hidden, relatively speaking.  I don't actually know whether that subjective feeling is accurate, whether I'm accustomed to some inaccurate overemphasized "air" in that range, or even whether I'm hearing things that aren't there and measurements will demonstrate to be nonexistent.  Of course it could also be my desktop speakers - I haven't had the time to download snippets and run them through my main system.  And I'm not talking about the decoded snippets versus the raw - there *is* no comparison there as far as I'm concerned.

 

The comment is only because I'm trying to be hyper-alert to come up with some possibly useful remarks - I'm really enjoying listening to these snippets.

[John Dyson]

7 hours ago, Jud said:

This is very, very good (subjectively speaking of course) at this point, good enough that I would trust measurements (if there are informative ones to be done) over my own hearing.  So take this comment in that vein: I feel there may be either a bit missing at the top end of the human vocal range, or else something near that frequency is emphasized and so the very top end of the vocals feels a bit hidden, relatively speaking.  I don't actually know whether that subjective feeling is accurate, whether I'm accustomed to some inaccurate overemphasized "air" in that range, or even whether I'm hearing things that aren't there and measurements will demonstrate to be nonexistent.  Of course it could also be my desktop speakers - I haven't had the time to download snippets and run them through my main system.  And I'm not talking about the decoded snippets versus the raw - there *is* no comparison there as far as I'm concerned.

 

The comment is only because I'm trying to be hyper-alert to come up with some possibly useful remarks - I'm really enjoying listening to these snippets.

Good description

Yes -- there is a high end problem -- I slipped in an new 2.2.4F version to fix it -- I am no longer announcing the releases, because the changes are so small now.   Most changes are close-in adjusments are minor error corrections.

 

Here is the trouble (it might be what you are seeing):

There is a necessary 3k to 9k 1st order pre-emphasis, along with a 9k to 18k decrease  total:(3k +9dB, 9k -6dB).   On the 'flip-side' for the de-emphasis, of course, there needs to be a 3k to 9k decrease (-9dB.)   The problem side is 9k to 18k.    The complicating factor is the 9k to 20+kHz band, which effectively gives a gain of +6dB, but it is dynamic.   What should the 9kHz to 18kHz de-emphasis be?   When looking at this from an engineering point of view (and understanding filters), it seemed like a de-emphasis starting at 9kHz * sqrt(6dB) would be the correct de-emphasis frequency.   Well, that was wrong.   Basically, there should be NO de-emphasis for the 9kHz+ band.   It is all taken care of by the DolbyA '9kHz +5dB boost'.   *The reason for the question/confusion is that the choice of the de-emphasis AND pre-emphasis is based on what the original designer chose*

 

The effect of my erroneous EQ (yes -- I take ownership for my mistakes) was to enhance the 9kHz maybe up to 16kHz frequency region, because there was also a 16k decrease.  I know that this is goblety gook -- but basically it all adds up to a bad 9kHz to 12kHz (maybe a little higher) approx 1.5dB boost.   Even the smallest error like that sounds REALLY REALLY bad.   I  went through a few days of bad HF hearing, so I didn't catch it.  I just got my ear-full on some orchestral stuff -- the violins were almost dissonant -- my hearing came back.

 

* We are at the level of very small changes now.   This might even be the last change, maybe not.   The reason why I am pulling back is that there is more to do, and SOME work needs to be done privately.

 

This thing is a mix of engineering knowledge (thank goodness I do deeply understand DolbyA), and also double checking by listening/hearing.   Sometimes my hearing fails -- and yes, sometimes my engineering guesses fail me.  My hard engineering is really good -- but my guesses could probably be better.

 

See if this 'V2.2.4F' is more tolerable.  I guess that I probably should do snippets again -- look for the new snippets with the correct "V2.2.4F" version (probably in +3Hrs.)

 

BTW -- I'll be checking for questions in the morning about this time of the day - maybe a little later.   I'll be focusing on some new anti-MD code along with code-sanitation so that I wont be embarassed for others to see it...   Been lots of changes -- it was SOOOOO nice to remove that LF EQ trash though 🙂

 

Thanks -- chat later.   Let me know what you think, and I'll try again.

 

 

 

[John Dyson]

Thought that I'd make something available here -- an ALMOST proof of the decoder and a bit of background.   I do NOT expect you to read the whole thing -- offerring it as a casual infomration source...  When the text file mentions the diagrams, the pointers are here:

.odt of the text is attached.   The formatting sucks!!!

 

Waveform:

https://www.dropbox.com/s/or0nb665k52ff8f/Screenshot from 2021-04-09 09-00-03.png?dl=0

 

Spectrogram:

https://www.dropbox.com/s/xzbcj6uiri6073w/Screenshot from 2021-04-09 09-05-33.png?dl=0

 

 

Going back into my 'hole' trying to do good stuff!!!

      
	


tdescription.odt

 

[Jud]

11 hours ago, John Dyson said:

Good description

Yes -- there is a high end problem -- I slipped in an new 2.2.4F version to fix it -- I am no longer announcing the releases, because the changes are so small now.   Most changes are close-in adjusments are minor error corrections.

 

Here is the trouble (it might be what you are seeing):

There is a necessary 3k to 9k 1st order pre-emphasis, along with a 9k to 18k decrease  total:(3k +9dB, 9k -6dB).   On the 'flip-side' for the de-emphasis, of course, there needs to be a 3k to 9k decrease (-9dB.)   The problem side is 9k to 18k.    The complicating factor is the 9k to 20+kHz band, which effectively gives a gain of +6dB, but it is dynamic.   What should the 9kHz to 18kHz de-emphasis be?   When looking at this from an engineering point of view (and understanding filters), it seemed like a de-emphasis starting at 9kHz * sqrt(6dB) would be the correct de-emphasis frequency.   Well, that was wrong.   Basically, there should be NO de-emphasis for the 9kHz+ band.   It is all taken care of by the DolbyA '9kHz +5dB boost'.   *The reason for the question/confusion is that the choice of the de-emphasis AND pre-emphasis is based on what the original designer chose*

 

The effect of my erroneous EQ (yes -- I take ownership for my mistakes) was to enhance the 9kHz maybe up to 16kHz frequency region, because there was also a 16k decrease.  I know that this is goblety gook -- but basically it all adds up to a bad 9kHz to 12kHz (maybe a little higher) approx 1.5dB boost.   Even the smallest error like that sounds REALLY REALLY bad.   I  went through a few days of bad HF hearing, so I didn't catch it.  I just got my ear-full on some orchestral stuff -- the violins were almost dissonant -- my hearing came back.

 

* We are at the level of very small changes now.   This might even be the last change, maybe not.   The reason why I am pulling back is that there is more to do, and SOME work needs to be done privately.

 

This thing is a mix of engineering knowledge (thank goodness I do deeply understand DolbyA), and also double checking by listening/hearing.   Sometimes my hearing fails -- and yes, sometimes my engineering guesses fail me.  My hard engineering is really good -- but my guesses could probably be better.

 

See if this 'V2.2.4F' is more tolerable.  I guess that I probably should do snippets again -- look for the new snippets with the correct "V2.2.4F" version (probably in +3Hrs.)

 

BTW -- I'll be checking for questions in the morning about this time of the day - maybe a little later.   I'll be focusing on some new anti-MD code along with code-sanitation so that I wont be embarassed for others to see it...   Been lots of changes -- it was SOOOOO nice to remove that LF EQ trash though 🙂

 

Thanks -- chat later.   Let me know what you think, and I'll try again.

 

 

 

 

I like the G version a lot.

 

As @PeterSt likes to say, if you listen and say something like "The vocals really come through well," or "Great detail from the strings" - really, if any separate component calls attention to itself - then something is wrong.  I didn't come away with an impression that a particular piece of the rendition was outstanding or diminished, just that the entire thing was very natural and musical, and compared to the raw versions, clearer.

 

All quite subjective, of course, but that's what it sounds like to me.

[John Dyson]

21 minutes ago, Jud said:

 

I like the G version a lot.

 

As @PeterSt likes to say, if you listen and say something like "The vocals really come through well," or "Great detail from the strings" - really, if any separate component calls attention to itself - then something is wrong.  I didn't come away with an impression that a particular piece of the rendition was outstanding or diminished, just that the entire thing was very natural and musical, and compared to the raw versions, clearer.

 

All quite subjective, of course, but that's what it sounds like to me.

Thanks!!!

I knew that the decoder was close -- I REALLY needed encouragement and/or some help where the audio problems are.  I started with good hearing about 8yrs ago, but because of meds, my hearing varies EXTREMLY over the day, and never know how I am going to perceive things -- SO I GREATLY appreciate kind, non-judgemental help.

 

I need more help in making a decision -- which project to do next...   Here is my list (in no particular order):

 

1) Try to make the higher quality modes faster.

2) Source code.

3) Better docs.

4) More ideas?

 

Source code will take me about 1month (I'll still lurk), Better docs about 1-2wks, higher quality modes faster, maybe 1wk or maybe 4wks.  I am HORRID when estimating time to program.  In the olden days, I'd say 1wk or less, but THAT WOULD BE TRUE :-).   My old boss used to say that I could create the world in one week, but of course, I was working with my favorite co worker, and we worked together with checks and balances.   I don't think that I am nearly as good as I was some 20yrs ago.  I really wish Bill was still available -- I had more respect for him than he'll ever know.  I believe that he might have been on the 'spectrum' a little, but he was wonderful to me.

 

Anyway -- got any ideas for the next project.   I want EVERYONE to benefit.

 

John

 

[John Dyson]

Got hyper good news, even though it probably doesn't seem important.

 

It took until a few months ago (probably about 6-9) to get rid of the 'buzzsaw' sound from the Carpenters' recordings.  Of course, all recordings had problems, but were mostly hidden.

 

Over the last few days, the decoder is now able to do an *accurate* decode, to the limits of the recording, of SuperTrouper (the song) and 'Dreamworld'.  This is especially good because all known copies of 'Dreamworld' are damaged.   SuperTrouper has a kind of ambient chorus at the beginning that always seemed to have random glitches in the sound.

Supertrouper and 'Dreamworld' ARE relatively rough sounding, but appear to be correct decodes, given the limitations of what is left in the recordings.  Decoding 'SuperTrouper' is a real challenge.

 

This is really a nice day given my *last* technical goals are met.   This probably means that ALL recordings are now possible to be correctly decoded.  Normally, the only major variation appears to be whether or not to use '--fw=classical'.   '--fw=classical' specifies the kind of M/S conversion to do (it has to do with stereo image.)

 

*I don't intend to imply that decoding is always desirable, just that decoding appears to be correct now*

 

Sure hope I never lose the source code (at least, before I distribute it so that someone else might be able to retrieve it for me.)

 

John

 

[John Dyson]

Sorry for posting over and over again today (I do this once in a while), but this is important usage information.

I just was privately suggesting some usage info to another, and realized this is IMPORTANT general information:

 

*  The decoder tries to faithfully represent what was originally FA encoded.

*  The FA encoded material sometimes has a rising frequency response in the 9kHz region.

 

Given this, I suggest adding this to the command line when expecting to directly listen to the decoding results:

 

--pvdh=9k,-0.75

 

This is a special EQ, almost impossible for a user to do (it is very intricate), but the effect is an ultimate rolloff about -0.75dB in the 20kHz

range.   The actual rolloff is  about -0.4dB at 9kHz, but can be important for listening.

 

The EQ above seems to counter the rising response in recordings.  If this EQ isn't done, sometimes there is a rough sound (e.g.

violin strings can sometimes sound hideous.)   Also, POP vocals can sometimes be a little harsh sounding.

 

ALSO -- just caught it -- I found a recording that needs a variation on --coff.  Instead of working correctly with --coff=-2 (the

default), it seems to need --coff=0.   This is the first recording that I encountered with this characteristic.  Therefore,

I need to update the online docs with this fact.  Since my sample set is only about 50-100 albums, and most are POP,

and I encountered this on a CLASSICAL recording, my guess is that this happens more often than I thought.   THIS

is infinitely more important than the EQ.

 

I probably should build-in the EQ by default, then allow removing it.   Since the decoder was originally designed to TRY

to be faithful, brain-lock kept the EQ from being the default setting.

 

If I get any significant number of complaints about slightly rough sound, I'll do something to add the EQ by default.   The

counter to that -- sometimes the EQ isn't just '--pvdh=9k,-0.75', but also add '--pvdh=12k,-0.75', but that is best

avoided unless really needed.

 

THANKS -- will try to disappear for another day!!!

 

[John Dyson]

I wouldn't break my promise, except a major sound improvement.

 

Back when the LF EQ was wrong, I had REAL troubles with the MF EQ that I thought was correct.  Nothing worked right.  So, I reverted the MF EQ to something simple, but I didn't like it.

 

On the current decodes, I always felt that the midrange was too metallic sounding.   I was right.

 

Re-instated the correct MF EQ (in my experimental, upcoming V2.2.4H), and it just might be more correct.

 

HOW EMBARASSING!!!   This only amounts to about a 10% error in frequency -- considering everything going on, that error is small.

 

After a few hours of testing, there will probably be a V2.2.4G.  I wish this damned thing would come to an end!!!

 

 

 

[John Dyson]

I know i wasn't going to announce releases, but I didn't expect audio problems anymore.

V2.2.4H fixes about 0.35dB to about 0.50dB at 3kHz.  (Yes, I am worried about that much precision.)

Demos are also up -- not everyone will notice much difference, but I did (probably because I listen all of the time.)

The 'grain' really bothered me, but I thought that it was as good as it gets.   You know that some of the

recording is lost in the encoding/decoding -- and the DHNRDS does a lot to retrieve whatever is possible

 

The biggest difference is that the highs are a little less grainy.

 

[Skip Pack]

To me, the improvements from G to H are distinct. What you describe as 'graininess' I would call 'congestion' that gets in the way of the upper midrange. 

[Jud]

I know for darn sure I can't hear .35-.50dB in the middle of other music, because I've tried test signals and distinguishing a .4dB change was extremely difficult for me, right on the edge of what I could (at least consciously) perceive.  So take the following with a large degree of skepticism: The background (for example, bass) in the current H versions, as of 10 pm and a little later US Mountain time, seems elevated in respect to the vocals and the rest of the midrange, which now seem to be relatively a little reticent.  No idea whether this subjective impression is accurate at all.  (I note the snippets page says "Correction Coming," but for all I know it's already occurred.)

[KSTR]

Hi,

I did not follow the whole thread and do not know how many people have noticed what I have noticed on the demo tracks provided as well as on all decodings I ran myself on various music. I saw at least some have while skimming through the thread, though.

 

Besides the general core function (downward expansion) which works well, there is a very strong overall EQ curve on the resulting output data which in almost all cases spoiled the tonal balance for me big time. Tubby lower mids and harsh treble.

I have not evaluated the various EQ command-line switches in my trials, I just the defaults.

 

Being an accomplished engineer I tried to find out what going on and this is what I found on basically all examples I've tried (about 20 so far) and with the two decoder versions I've tested myself (V2.2.3E and V2.2.4E):

https://www.audiosciencereview.com/forum/index.php?attachments/dysondecoder-fr-mag-phase-png.122626/

A dip/peak valley (blue curve) spanning more than 12dB(!) in the range from 50Hz to 20kHz (there might be small differences between versions but you'll get the picture). Ignore the absolute levels, just look at the span.

Note those kinks at exactly 3kHz and 9kHz and the irregularity at ~1.5kHz  which don't have corresponding wiggles in the phase (red) which means this overall EQ is not fully minimum phase (not a big deal, but still an interesting detail).

 

I've had an encounter with Mr.Dyson before and therefore will not engage again, everything has been said.

I just wanted to share that bit of information, backed up with data (which was created properly and competently) which is IMHO elementary to understand why the decoder sounds the way it sounds, at least this is the dominating factor.

 

I've also created a compensating EQ that restores the original tonal balance which I could share if anyone is interested (or just follow the link and find it there).

Once this is applied, it's much easier to hear (and judge) the dynamic processing that takes place, being the only variable left (a key requirement for fair comparisons in any field of engineering/science). One can still apply any needed (or preferred) EQ after (or before) decoding to get the best results.

 

On 4/10/2021 at 1:01 AM, John Dyson said:

2) Source code.

It would be a pity if this valuable project were lost.

[John Dyson]

10 hours ago, Jud said:

I know for darn sure I can't hear .35-.50dB in the middle of other music, because I've tried test signals and distinguishing a .4dB change was extremely difficult for me, right on the edge of what I could (at least consciously) perceive.  So take the following with a large degree of skepticism: The background (for example, bass) in the current H versions, as of 10 pm and a little later US Mountain time, seems elevated in respect to the vocals and the rest of the midrange, which now seem to be relatively a little reticent.  No idea whether this subjective impression is accurate at all.  (I note the snippets page says "Correction Coming," but for all I know it's already occurred.)

Well -- I found another mistake -- expansion in the wrong place...  Fixed it, a lot more expansion, sounds signifcantly more dynamic than previous.

Try V2.2.5C...

Doing demos right now.

This is a progressive effort -- once I fixed one thing, then another thing pops up.

 

About the EQ - the result IS flat relative to the original recording.

I think it was an old Ronstadt non-FA recording that I demoed -- the sound was essentially the same on output of the decoder.

We are step by step improving.   I believe that the non-FA Ronstadt was a little normally-compressed.

 

This new version is  even less compressed.   When the transition to V2.2.4 was released, something really got fixed.  I forgot about the fact that

the calibration (expansion steps) needed to be revisited -- I still felt that there was too much compression in the sound of cymbals.  I was correct.

(The original sequence was suspect all along, but it worked.)  We DO know that 7 layers is correct, or the sound is all wrong.  I did check that

also.

I sure hope this settles things down, but as @jud said, the decoder was already pretty close -- but I am a perfectionist.

The decoder IS getting to the point that the result is tolerable to me (that is, sounds like a natural mix.)

 

 

[John Dyson]

Watch out for malicious people -- there have been ugly trolls, just because someone cannot hear the improvement, doesn't mean that it isn't there.

Honestly, I truly believe that if someone cannot hear the improvement, they should leave things alone.

 

[PeterSt]

6 minutes ago, John Dyson said:

This guy has been trolling me all over the place -- a real jerk.  Makes up competency -- read, but he has hate.

 

John, you may go too far with this. I hope you can show it and justify it.

As far as I can see "this guy" tried his best at understanding your work, then you bashed at him (like you lately do in here no different) and next he worked out what your decoder does for net result (quite the same as I told you and what should not be). Lastly your thread was closed because of your "shouting" as it was called over there, but we can't see that because the posts have been removed. This leaves what we can see at this moment.

[John Dyson]

8 minutes ago, PeterSt said:

 

John, you may go too far with this. I hope you can show it and justify it.

As far as I can see "this guy" tried his best at understanding your work, then you bashed at him (like you lately do in here no different) and next he worked out what your decoder does for net result (quite the same as I told you and what should not be). Lastly your thread was closed because of your "shouting" as it was called over there, but we can't see that because the posts have been removed. This leaves what we can see at this moment.

As I posted before, I truly believe that he cannot hear the improvement.  Even though he holds his hearing in high regard, doesn't mean that it doesn't exist.   When someone deeply insults me, I do believe that response in kind is important.

You know, I believe in being passive, but some people do not -- yet represent themselves as authority.   WHAT TO DO?

What about the 'fake facts?'   Refer to the Ronstadt example - it really IS very similar.

 

Trolls really should prove their facts, but instead pontificate...   if the decoder wasn't DEAD ON accurate, you WOULD hear gating.   I have tried HARD to avoid gating.

 

Again, what to do when the insults and deeply incorrect pseudo-facts start.   When he talks about 'major'EQ -- I have been very careful to do ANYTHING to bring the EQ as close to natural as possible.   I am worred about 0.5dB, gosh sake.

 

Yes, there IS EQ -- have you ever heard a DolbyA unit?   I am ONLY compensating for a DolbyA unit...   Sure, there is anti-distortion (cancellation) in the code, but it is flat.   Why?  Is it only because he cannot hear what is so clear to me?  I used  to do CLEAN recordings -- what we get from the distributors ARE NOT CLEAN.

 

 

 

[John Dyson]

 

 

Please read this -- then judge when there are trolls.

FIRST, IMPORTANT:  Errors in expansion are infinitely more audible than errors in compression.   If there were errors, the sound would have STRONG artifacts.

 

I want to explain why a static measurement of freq response is non-flat.

 

This is because ANY multi-band compressor/expander is non-flat, well -- unless adjusted to beflat.   There will be people, who have no idea, will measure something like a multi-band device and proclaim not flat.  In production, non-flat multi-band compression is used -- yet, the general sound isn't as profound as you might think...   Also, they are doing compression, which naturally 'homogenizes' the signal.  With expansion (the decoder uses 7 DolbyA expanders), then errors are magnified. When I make minor errors before one of the expansion steps, the results can be painful (literally.) -- remember, expansion MAGNIFIES errors.

 

Howeverr, one has to look at the shape of the curve relative to what happened to the recording before hand, and also the fact that the dynamic frequency response matches

what was already done to the recording.   Also, a measured frequency response is total nonsense for the decoder, just as it probably is (depending on settings) for a multi-band compressor.

 

There are some reasons why the result is flat, but there are also reasons why the compensation is very tricky.   One reason is because the natural threshold of each band on a DolbyA unit is different.  Even the HF1 band has a different compression ratio!!!   Errors in expansion, especially when a big part of the step-by-step DolbyA EQ is sitting BEFORE an expander, would be audbile -- usually as extreme treble or extreme bass.  All (I mean ALL) of the EQ is based on standard 1.5dB, 3dB, 6dB, 9dB, and maybe 10dB increments.  There are NO in-betweens anywhere except the pure DolbyA which has to emulate selected semiconductor devices.   All of the EQ frequencies are even.  The difference between an 80Hz and 75Hz EQ is VERY significant, in fact some of the bass problems can be attributed to me choosing thewrong one, and I cannot hear LF very well (I had to select between the two, because one was a DolbyA characteristic, the other is a standard frequency -- but which one did the designer use?  If I had to tweak, then questioning the design would be good -- I have ONLY been selecting standard choices based on my idea of what the original designer did.   When a change is made, the ONLY change is to change the standard dB values or move a frequency (moving frequencies is done very seldom.)

* sometimes, instead of using 3dB, I use 1.414, and instead of -3dB, I use 0.7071...   That is because I don't want to do the conversion internally.  Sometimes I do use dB values, but those are on set-up.  Doing log operations (or power operations) are expensive, and I don't want to do them on every sample.

 

Given all of that, if the curves  of the encoding didn't match the decoding, because expansion DOES show errors MUCH more than compression, then the sound would not just be defective, but be very defective.   The small errors that I am working right now are mostly sub-dB.   I heard the sub-dB error in the upper midrange, and it always irritated me, but I though that was normal EQ for pop.   I didn't really hear it on classical/orchestral, but the strings were alittle strong.   The sub-dB error was a manifestation of a mistake that I made by removing a subtle taper in the EQ for midrange (which REALLY is needed.)   I re-instituted the taper, and the bump went away.  The correction is approx what I perceived.   These sub-dB errors were due to using 3ea 3kHz instead of 2.75kHz, 3kHz, 3.25kHz -- that kind of thing EASILY gives subdB errors.

 

So -- if you read a skeptic, don't let his representation of golden hearing or engineering prowess be considered without reading this.  I can give a LOT of further technical backing onthis design to those with open minds.

 

 

 

[PeterSt]

Aha, I see that someone is clearly tweaking the posts. This should not have happened.

Why did I expect this ?

 

Nice collaboration (not).

 

And without the possibility of @KSTR'post being worked out, well, he is right. Sorry.

John, you may ask Chris to put all back instead of asking to get rid of the things you don't like.

 

This really did not help.

 

Maybe I fell between tweaking jobs and with a wink of the eye all is back ... and I dreamt it.

[John Dyson]

29 minutes ago, PeterSt said:

Aha, I see that someone is clearly tweaking the posts. This should not have happened.

Why did I expect this ?

 

Nice collaboration (not).

 

And without the possibility of @KSTR'post being worked out, well, he is right. Sorry.

John, you may ask Chris to put all back instead of asking to get rid of the things you don't like.

 

This really did not help.

 

Maybe I fell between tweaking jobs and with a wink of the eye all is back ... and I dreamt it.

I am tired of @KSTR's possibly misguided statements (I won't call them lies, but probably are.)   If he is allowed to lurk and ruin my attempt to do something good, then it becomes non-fun for me, and basically will dissude anything that I am trying to do for the audio community.

This is hard enough for me -- I thought that it would be done over 1yr ago -- I am on the edge of quiting.

Do you want a meddler to destroy something?

 

He is explcitly following me to destroy something that is most probably good, and he knows it -- Iyesterday - to protect me and my project -- just turned off all of the forum that he lurks at -- I cannot trust proven destructive people to have open minds.   It is tiresome .

My information above is correct and justifies the design concepts.

One mistake that I kept making -- I over designed -- the resulting design is MUCH simpler (except for my add-ons.) than my original misunderstanding.  I have learned A LOT.   I mean, REALLY simple like a 500Hz, 75Hz EQ bettween each layer (plus a 3dB or 1.5dB -- forget) level change.   THAT IS SIMPLE -- if it was wrong, it would be a LOT more complex.

 

 

 

[John Dyson]

I guess I have been a jerk -- but I need people to COMMUNICATE DETAILS to me -- my hearing has been tricking me for a long time.

I really tried to give a gift.  Sometimes the 'gift horse' syndrome has manifest, but I have given EVERYTHING (literally.)

 

The project is suspended.

 

[PeterSt]

1 hour ago, John Dyson said:

I guess I have been a jerk

 

No, not true at all.

On a side note: Things got heavily out of context by now, with deleted posts and such and not posted posts which refer (and quote) deleted posts (including my own).

 

I guess I can't keep on responding similar in an always changing slight context difference (me knowing what I read and wrote), so I will try to be brief now (if I ever can):

 

Communicating as such includes plain listening to what people have to say. Or, what they deem impossible to happen (for measurements). Communicating is - I guess, also not scoffing even the slightest too much. This is not about being a jerk, but about you being the controller collecting all the data, being happy that ears are around to help out, you being responsible for the intellect of bringing all together into something sensible, which includes the iteration(s) of listening and for example the judging of something going down hill. This obviously (!) in combination with your math now having an other bug less. --> Usually in such situations something very else is just amiss. 

 

A new one (I thus did not type in vain today) would be:

We must realize that all we have is ears and not any of your math or further knowledge. So all we can really do is have one or another complaint of something - unless all sounds as good as we are used to (yeah, what to do with that eh ?). Thus sadly, all what ever can be there is complaints.**

 

**): This is born by the sheer fact that we don't have complaints otherwise. So for example, not only I - but also my customers with some more full-fletched setup, don't even want anything better. We are done with improvements (although I will continue and people will keep on buying the presented novelty).

John, please, how much chance does that give you ? ... it can only happen when things sound enormously better all the way. Meanwhile you'd have to give the people the fair headstart with their own systems they almost exclusively will find OK (in whatever stage of their ($) life). So if they deem that the decoder makes it worse, if just is so. It would be your task to bring all the comments together and make "chocolate of that" as we say over here. ... Things are only out of the way if nobody complains about the same thing any more and with sufficient contenders in the first place.

So indeed, people should not be scared away.