'FeralA' decoder -- free-to-use

[PeterSt]

31 minutes ago, The Computer Audiophile said:

Thank you for the information. I will put his posts back.

 

Thank you, Chris.

[PeterSt]

I wrote this yesterday. I'll let remain the introduction of a much larger post.

 

3 hours ago, KSTR said:

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.

 

It has become quite difficult to contribute to this thread - and John's work in general - in a fashion which is appreciated as constructive. The last time I tried was since attempts to really make something of this, last July/August. I held back for countless months in order to have some diplomacy in my post, a phenomenon I usually averse to.

I think I was as neutral as I could be, but it did not work out. I appear to not understand, can not see the objectives, do not relate to the complexity ... and if not that I'd have to be deaf (the latter is my own conclusion in order to avoid the accusation).

 

But Dear John,

I'll try it again.

 

I am not deaf at all.

Does that help ? :-)

 

@KSTR - in my view - deserves a little backup, although I am sure he can do without just the same. However, like me in the beginning, he spent a lot of time understanding what you are doing, John. If then, all what comes back is an unfair dose of shouting, telling us that we clearly don't understand a hoot what this is about, then all what remains is let you be.

 

[...]

 

I' will leave that post at that, as since then some PMs were exchanged. Still, John needs to know that nobody is attacking him. But apparently it feels like that, when anything negative is put forward. So John, that is part of the game. Without that, no improvement possible !

Heads up !

[John Dyson]

I have no more energy to discuss this --

First -- people who will not listen to evidence are not worth listening to.

Next -- I have lots of evidence that the decoder is structurally correct -- might make mistakes.

 

Since I am not listened to about those who ALREADY  KNOW EVERYTHING? -- geesh, judging the MULTI-BAND project on a static frequency response -- is that competent?

Using already multi-band compressed materials as a spectral reference -- is that competent?

I like to think that people who write NEW programming or design NEW circuits without just copying tend have a little more REAL experience  than pontification and mosty just measuring things.

 

Here is the location of the decoder and examples.

The expansion is now close to what I expect, and much easier to achieve once the bass was fixed.

However, some recordings appear to produce too much bass, even though the bass gain control is active.  I don't know what is going on there -- it might be something that I cannot see yet.  (I know my limitations, and figure that things could be better.)

There is a problem with 'Year of the Cat' -- I am thinking that they used one less layer, or perhaps another kind of compressor (a bit of gating.)

 

When comparing -- listen for the 'telephone' vocals on FA,and the usually distorted bass (there is a higher frequency distortion envelope accompanying

the actual low bass frequencies.  There are other things, much better mitigated now -- cymbals stronger -- not cymbals SMUSHED like FA recordings.

 

I cannot explain all of the technical attributes here, but those who will not listen, making proclamations without asking questions as if interested will never learn anything new, and their expertise is suspect, even though a few people might have seen some facts from them.

 

Here is the location of the decoder -- take it or leave it.   There is a lot of research in it that competent people in the field might even be interested,but I don't understand the lack of curiosity in those who claim expertise, and have done so for a long time..  My own technical expertise comes from 40+ years of real engineering work -- that doesn't say that I am always right, but also shows a lot of experience in judging whether or not something is possible.

 

Here we go - I dont' want to take away options like pontificators do:

 

Location of decoder, examples and comments...   I wont know about anyone who downloads these, but when possible, progress will be made.  There will be an improvement at least 1 timeper week -- will try to be stable on 9:00PM on Friday nights.  Already, in the last few days, it has significantly more expansion in the -30dB range, which makes a profound difference in the frequency ranges that lurk around that level (e.g. the higher freqs.)  I had mistakenly and wastefully (and sometimes encouraging gating) expanded in the -60 to -70dB range (remember thiis is a density, so there really is active signal in that range from time to time.)  The levels that I speak of arent like total signal level.  The #of layers is still 7, because anything different (like different compensatory EQ for DolbyA), will sound bad (sometimes even worse than FA.)

 

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

[John Dyson]

2 hours ago, opus101 said:

 

You could try asking 'What exactly is this alleged 'poison' he's dispensing?'. And why use such emotive language as 'stalking' and poisoning? Isn't that generally indicative of loss of impartiality?

He (@kstr) lost his open mind on my first post elsewhere.

Am I not allowed to be peaceful?

[John Dyson]

1 hour ago, PeterSt said:

I wrote this yesterday. I'll let remain the introduction of a much larger post.

 

 

It has become quite difficult to contribute to this thread - and John's work in general - in a fashion which is appreciated as constructive. The last time I tried was since attempts to really make something of this, last July/August. I held back for countless months in order to have some diplomacy in my post, a phenomenon I usually averse to.

I think I was as neutral as I could be, but it did not work out. I appear to not understand, can not see the objectives, do not relate to the complexity ... and if not that I'd have to be deaf (the latter is my own conclusion in order to avoid the accusation).

 

But Dear John,

I'll try it again.

 

I am not deaf at all.

Does that help ? :-)

 

@KSTR - in my view - deserves a little backup, although I am sure he can do without just the same. However, like me in the beginning, he spent a lot of time understanding what you are doing, John. If then, all what comes back is an unfair dose of shouting, telling us that we clearly don't understand a hoot what this is about, then all what remains is let you be.

 

[...]

 

I' will leave that post at that, as since then some PMs were exchanged. Still, John needs to know that nobody is attacking him. But apparently it feels like that, when anything negative is put forward. So John, that is part of the game. Without that, no improvement possible !

Heads up !

That is not true --- read my comments, carefully.

Did @kstr ask for me to explain -- or did he explain how it wasn't flat?

 

At least, it isn't 8 yrs down the drain - Have you even tried to listen to the decoder recently?

I had someone on the other group talk about the single layer expander working better -- sure, it might have expanded, but I found that it didn't really do what I wanted.

The decoder really IS perceptually flat now.

 

How many people do you know who could design something THIS COMPLEX from scratch, without ANY substantive help other than criticism, and have to infer every specification?

This is hard - and sure, I am a little (truly a little) adverse to NON SPECIFIC CRITICISM -- please be specific -- that REALLY helps.   Blasting me or the project with pontification, that does NOT help.   The project WILL be completed, but I guess with very little help.

 

A REAL pro in audio asked why I was treated poorly...   Reflect on this -- I am doing something VERY SPECIAL in the eyes of people who know this stuff -- some people are even angry about it.

 

[Jud]

I know everyone likes to gawk at fights, but perhaps this once we could sit back, let John do his work, help him in the way that is most helpful to him; and meanwhile, see what can be gathered from the critiques that may eventually in some fashion prove helpful to John, since I think all agree that the goal is worthwhile.

 

I don't know if KSTR is posting any more, but is the basis of the critique that all these files cannot have been screwed up so much and in roughly the same way, and that therefore an attempt to "unscrew" them to such an extent and in roughly the same way cannot be correct?

[John Dyson]

27 minutes ago, Jud said:

I know everyone likes to gawk at fights, but perhaps this once we could sit back, let John do his work, help him in the way that is most helpful to him; and meanwhile, see what can be gathered from the critiques that may eventually in some fashion prove helpful to John, since I think all agree that the goal is worthwhile.

 

I don't know if KSTR is posting any more, but is the basis of the critique that all these files cannot have been screwed up so much and in roughly the same way, and that therefore an attempt to "unscrew" them to such an extent and in roughly the same way cannot be correct?

Thank you for an open mind.

I REALLY WAS MISTREATED -- I was offering something FREE, and was almost 100% rejected immediately, with NOT asking for what was going on. 

(removed a comment like 'woe is me' -- it was just stupid -- keeping the facts here.)

 

For a perfect, worst case example of what the decoder does -- listen to 'Take a chance on me'.   I am not sure that it is perfectly correct, perhaps needs some EQ -- but you can hear the garble on the FA version.   Admittedly a worst case, but that is the kind of thing when I listen to FA -- but I had done recordings myself when younger -- and kind of trained to hear audio problems.  This is even true though parts of my hearing do not work well.  Maybe even I can hear distortions MORE nowadays -- not sure about that though.

 

 

[John Dyson]

Removed incorrect comment about levels -- my snippet demos were NOT even level.   I screwed that up also.   I was wondeering about the level difference -- NOW I KNOW :-).

 

 

 

[fas42]

6 hours ago, Jud said:

 

I don't know if KSTR is posting any more, but is the basis of the critique that all these files cannot have been screwed up so much and in roughly the same way, and that therefore an attempt to "unscrew" them to such an extent and in roughly the same way cannot be correct?

 

Yes. In part. John is of the opinion that a high percentage of recordings are 'damaged', on their release. IME, and for others as well this is not so - flaws of the playback chain are what is revealed when these recordings are played, and by trial and error one can adjust the recording in many cases to make them more pleasant to listen to - it's an alternative to remedying the root cause, which is shortcomings of the replay mechanism. When John's algorithm gets it right, it certainly does a nice job of presenting the recording in a good light on an average playback quality setup, as I recently heard on my laptop speakers.

 

So his tool has great usefulness for many, and is therefore worthy of consideration by those who want to use such a method to improve the listenability of their music. But, just because the program can do this does not mean that "the files have been screwed up" - IMO, a far more effective solution is to resolve the lackings of the reproduction system, which then means that all recordings benefit ...

[Racerxnet]

1 hour ago, fas42 said:

Yes. In part. John is of the opinion that a high percentage of recordings are 'damaged', on their release. IME, and for others as well this is not so

Really Frank,

 

How many people do you really think are following your lead? Inquiring minds want to know...

[fas42]

18 minutes ago, Racerxnet said:

Really Frank,

 

How many people do you really think are following your lead? Inquiring minds want to know...

 

So, you believe "a high percentage of recordings are 'damaged', on their release" - what makes you think this?

[Racerxnet]

39 minutes ago, fas42 said:

So, you believe "a high percentage of recordings are 'damaged', on their release" - what makes you think this?

 

You have a knack for pulling things from your rear end. Where did I say I thought any recording was damaged?? Guesss you can't! 

 

It would be nice if you answered the question asked. How many people do you really think are following your lead? Inquiring minds want to know...

[opus101]

8 hours ago, John Dyson said:

He (@kstr) lost his open mind on my first post elsewhere.

 

Link please? If its the thread on ASR I've already looked and saw no evidence of what you're claiming.

[fas42]

34 minutes ago, Racerxnet said:

 

You have a knack for pulling things from your rear end. Where did I say I thought any recording was damaged?? Guesss you can't! 

 

It would be nice if you answered the question asked. How many people do you really think are following your lead? Inquiring minds want to know...

 

Dear me ... 🙄

 

Go back and study what I said,

 

Quote

Yes. In part. John is of the opinion that a high percentage of recordings are 'damaged', on their release. IME, and for others as well this is not so

 

Now, think really carefully about what I'm saying here ... a clue: I'm disagreeing with what John is saying ...

[Don Hills]

John,

It does appear from Klaus' measurements that the decoder frequency response is not flat, at least at high levels where the decoder isn't applying dynamic  gain or EQ changes. Assuming that the non-flat response is deliberate, what might be your reasoning behind it? For example, might the response be to correct perceived response changes applied by the original "FeralA" processing?

[PeterSt]

9 hours ago, John Dyson said:

Have you even tried to listen to the decoder recently?

 

Hi John - Yes I did. This was 10 hours ago, so before you asked. To a small quarter of the ~80 demo tracks, but two times. :-) First the Decoded, then the RAW. I have two pages with remarks, per track one line.

Mind you please, the 50 seconds is already too short to have a sufficient judgement and write something about it as well.

 

 

This is how they (accidentally) appeared in the player (selection per DEC and later per RAW in the file name).

Here you see me one way or an other writing down the title names :

 

 

So this is DEC(oded) first. Btw, with the "short" I mean that all the reverb has been taken out. Cut piano notes and such (like Y.o.t.C.).

 

There is no general consensus of what to do for improvement (this is no different from the various times I tried before over the years), which makes it the most difficult to help out. However, with my experience on these matters, the verdict would be that the application is wrong in the basis, that allowed for the non-consensus and all completely different remarks. One think I could say is that for almost everything counts that it sounds hollow and inside out (this latter may be more difficult to explain, but it comes along with the hollow and definitely implies a phase problem).

I said "-Stop-" there in a KSTR way of ending the session, because at some stage it becomes clear that continuing is useless. However, it would be very easy to do the remainder tonight. N.b.: I could prepare better to print the playlist in advance and avoid the witing of the titles.

With "Highs above mid" I mean: emphasized highs (like in 4K and up) but no mid under it.

"beep beep beep": hardly power and LF.

 

...

 

I don't even know how legit it is to listen to the RAW, but my very personal opinion tells me that there has to be some merit because it *is* the base for further processing. Next it will depend on that further processing obviously, but without that decent base, I'd say there's nothing to go for further. It now completely depends on the specific remarks.

The track sequence is the same (first one is Year of the Cat), so you can compare the remarks, if needed.

 

 

So mind you, if I with e.g. Y.o.t.C. now don't mention "hollow" it for you means explicitly that where the DEC version is hollow, this one is not. I am looking myself over the remarks of RAW and see no hollow anywhere. Thus, the phase problem emerges in the decoding, right ? ... so this is how your thinking judging is allowed to be. Partly relative but in the end explicit (give us 200 seconds per track and more comment is possible ;-).

 

My wife did not know what I was doing, but I noticed her singing along with the DEC snippets. Into the third of the RAW snippets she asked whether these were cover bands contrary to the DEC's. Mind you, this was into Mrs. Robinson.

This should be telling, namely that the process of making the RAW is completely killing everything, which you so hard try to restore by EQ which I too regard subjective.

I expected my wife to have remarks during the DEC set, but she didn't have any (I asked nothing either, but if things go bad she will announce herself, usually). It is a measure, over here.

 

For myself I noticed that I could listen to this, contrary to the DEC's.

 

I can't read what I wrote at Mrs. Robinson, but the "Cool" refers to stone cold with an arrow to ""opposite to the DEc".

 

The consensus here is quite clear for me: way too much emphasis on the mid. In aftermath you (I) can see this on "tonality", "Voice too profound".

 

The 4 but last says "no high mid" and refers to "bell". This is super crucial because when rendered wrongly, bells disappear (or become profound, which would be good/nice but would still be wrong-ish). Thus, it tells (me) that the top of (human) voices is lacking here.

 

What should be telling (if I had to solve things) is that the one comes along with "power" (which will be relative and as a surprise, to the DEC version), while the other just the same does not show any power at all. This will be the low mid, FWIW.

 

The "unrecognizable" at Mama Mia (one but last) was already so in the DEC version (there noted as "super bad"). So again, something which is wrong in the base, won't allow restoration.

 

You may investigate 15 - Walking my Baby back home, as I judged that "OK". It is a rarity because the only one.

 

...

 

A last remark, because this is crucial for someone like me:

As someone who is not continuously trying to improve his playback means, but who is very explicitly working on making all real-life representation (this takes 40+ years of taking hurdles), it already goes wrong in my mind (starts to hurt) if whatever Agneta does not sound like her. Olivia Newton John no different. And remember, I am that guy with grand piano's, full drum kits, a bunch of guitars and what not, especially for the purpose. I try to reproduce reality. Now:

When a piano does not sound like one, I am not able to judge the remainder any more (of the track of concern). This is how Year of the Cat fails at the very first tones. Btw, YotC is a most difficult one to begin with, because it is a first which sounds thin in a digital system and will take years to get right (with a full body etc.).

 

Peter

[PeterSt]

 

10 hours ago, John Dyson said:

I REALLY WAS MISTREATED -- I was offering something FREE, and was almost 100% rejected immediately, with NOT asking for what was going on. 

 

Hey John - the fact that something is for free, will not justify its being right or wrong. So when we're a bit into the real judgment of matters, you may now be able to try to digest more real life realities. So below is still not the last post I could not post day before yesterday because the post of KSTR (and mine) disappeared, but this one is I think relevant to the comments people have, like I just invested some time in collecting mine.

Don't be angry about the below, but instead please try to reset a little and see the good intentions we all have. This should not derail because you don't like the truthful comments. And btw, what's distortion and what not, is not only for you to decide. I think this is crucial. We all may be as old as you are with possibly similar experiences. And just saying: when my ears do not behave, all starts to sound like distortion. A day to skip.

 

So from two days back, unaltered:

 

 

7 minutes ago, John Dyson said:

I am tired of @KSTR's possibly misguided statements (I won't call them lies, but probably are.)

 

 

I understand. But first it has to be true. And I thought I was going to work that out nicely.

Btw, I hope you don't have deletion etc. rights to this topic, or else I may have blamed Chris unjustified.

 

Btw, personally I hate it when my time is spoiled by deleting my posts (and that just happened).

 

In that thread over there, all was nice and cozy. The last thing prior to "debate" was this from your hand:

 

Quote

Maybe the EQ problem is what you heard?

 

Mind you, no faul words anywhere. Not from you, not from him. But then in response he showed stuff like this:

 

plus two more and ended with 

 

Quote

I think we can stop here... the exact same pattern on all four files.

 

... and ended with a clear exhibit of being disappointed and to invest no further time. But also with your current explanation (no earlier) which went radically different in there. Example:

 

Quote

I understand your plight and limitations, but it is really best to be kind to people, even the disabled. So, I will do so.

 

So John, instead of such nastiness - which from there on grew for obvious reasons - you could have explained those "huge" EQ changes. But since you did not, while it was repeatedly asked, the result went as it did.

*My* original response to it wasn't even posted yet (finding the stuff I respond to deleted in the first place). But I sure think I still will.

Meanwhile, your explanations regarding this look to be orchestrated because they are without reference now, so you can say what you want and be right always. This does not work for me.**

 

**): Especially not because I had similar remarks and you did not respond to those either and instead suggested that I don't understand (in itself correct, but ...).

 

All 'n all:

 

23 minutes ago, John Dyson said:

If he is allowed to lurk and ruin my attempt to do something good

 

Nobody is doing that, that I can see. Not even me. :-) However, if you post your results like they are on an other forum, you can expect people to be as deaf as I am and next measure. You should be able to cope with that other than thinking that people are ruining you. 

 

On the danger of me getting too confused about not yet sent posts because the posts they refer to being deleted, what about this 10dB of difference which would blow the windows or would be a stone cold result otherwise ?

Them 10dBs really are there. So ... that was my question, since someone took the effort of measuring it (PKane did too but showed the LF only, most probably because that was the subject (somewhere last week)).

 

So John, or you have a real life explanation other than scoffing me to observe the difference, or you may say "oops".

It just can't be.

Last week I was hinting that your RAW file (I only listened to the Fleetwood Mac examples) should not be your reference because it was terrible not the best to my ears. Maybe you are correct that you are applying the very best theoretical EQ or whatever it is exactly that I and others can't comprehend ... but that your RAW version is off all the way while you think it has to be good.

 

For now, and NOT trying to make it worse ... listen to those who try to help you listening. Bear critique (my unsent post is about that). Don't see that as criticism. Can you do that ?

[John Dyson]

This, with a few personal comments removed, was sent to someone REALLY helping instead of 'highs suck' or something like that.  Also, some other people were way quick to judge -- assuming I am incompetent -- but also wasn't  really open to ask questions.   This could have been resolved before problems if questions were asked.   Also, it is true that the decoder IS NOT FLAT, and that fact is immaterial to a multi-band device.   Such judgement makes no sense, and without query -- that was injustice.

 

 

 

 

These new iterations is MUCH less complex - I really missed the mark, but also I really needed help.  As soon as I got any substantive help -- I said thank you.  But if something 'sucks' or equvalent -- I don't know what to do with that.  I really thank you for the useful hints given -- they REALLY

helped me and accelerated the breakthrough that my hearing was REALLY bad, not just a little.

 

ALL of the fixes were trivial for me -- just knowing what was needed, that was the problem.

 

The V2.2.5F results are ready (using my working hearing), but still not 100% sure of details -- it generally sounds 100% better than the previous results -- relative to RAW.  There are some regressions against RAW -- mostly because of EQ before encoding, so a slight amount of EQ (e.g.+3dB at 3kHz maybe) might REALLY open up the recordings.  This EQ would be CORRECTIVE and not 'tweak'.  I NEVER intentionally  'tweak' -- it breaks the rules.  l WILL  NOT EQ THE RESULTS, EVEN IF CORRECT, I depend on the foresight and competency of those doing the evaluations.

 

Sometimes on the snippets I measure slight decrease in SoX measured dynamics -- but a lot of processing isn't seen by the SoX measures.   Also, the snippets aren't showing the full recordings, just the snippets -- I will definitely measure against the full recordings and see what is going on.   (The Beatles recordings -- showed the most profound increase in dynamics.  Some of Nat KingColes recordings were disappointing.)

 

https://www.dropbox.com/sh/tepjnd01xawzscv/AAB08KiAo8IRtYiUXSHRwLMla?dl=0

 

I do have some warnings about interpretation:

1)  The bass is sometimes a LOT more tight, but I believe more accurate about the instruments being played (the sound and percussion of a drum.)

2)  The vocals are less telephone-like...   On A/B comparisons, that can sometimes sound like the voices are a little more buried.   They might actually be because of item 3 below.

3)  I am not sure that I have still got the lower midrange/bass correct.  This is one thing that I stripped out a LOT of complexity.  I had all kinds of crazy EQ that I thought was necessary (some EQ shaped like the DolbyA curves -- WRONG.)  I think that we went down from 6EQ down to 2.   Some of the EQ previously were 2nd order, but now there is NO 2nd order anywhere.  (That bodes wll for my normal 1st order 'rule'.)

4) I can further increase the 20Hz bass by about 1.5dB without a lot of work, but I tend to be bass adverse (esp low bass), so I did a slight amount of trimming, but it just might be correct.  The lowest bass is the only place where I have had to use ad-hoc techniques -- but the 3dB and even freq rules are still followed.

5) The HF eq also had a profound decreases in complexity, and made the biggest improvement.  I think that I dropped from about 5EQ  (2 was pre, 3 was de empahsis).   Now, it is 1 and 2 (the reason for two has to do with the shape of what the DolbyA units do -- super low Q filters, and they had to be emulated with 1st order EQ.)

 

* The between-layer EQ ended up being 100% correct -- any changes are disaster.

 

In the last day, I have stripped out 7-8 equalizers and getting closer to origiinal sound (sans noise.)  Sometimes the result is a little more muddy (probably legit needs EQ), sometimes a little better (usually removing noise), and sometimes PROFOUNDLY better.

 

The other reason (other than over-engineering EQ)  was my hearing -- I kept chasing which choice to make on the EQ.   Out of frustration, I quit my meds about 2days ago so that I could hear more reliably.   Of course, I am tearing (dropping water from my eyes.) with super-high BP right now, so a little shakey.   Will take a day or so to come back to normal, but once I took the medicine, a few hours later -- back to crazy hearing.

 

So -- I BELIEVE (guess) some various reasonsfor FA is to slightly increase the density, make vocals a little more intense, and

hide the originals.   (Vocals being more intense is the 'telephone sound' that I complain about. I believe that the main reason was to hide the IP so that the record companies keep the originals.  I got STRONG pushback on this project, well before the local naysers.   The pushback was WORSE than the DolbyA project, WHICH I SUCCEDED AT.   Everyone says IMPOSSIBLE, but I have pretty good judgement with a perfect record.

 

When the decoder runs, it can only recover what was put into tje decoder.  I sometimes believe that there was EQ before the encoder so that the sound will be better for the listeners.   Maybe some of the ONJ stuff and a few others were EQed a little.   I didn't delete any of my tests -- integrity is important.

 

if there was more specific help (no 'sounds bad'), maybe with a spec, I could have dealt with the hearing earlier.   Believe me -- I got lots of sucks when moving to multi-layer, but single layer WILL ALWAYS SUCK.

 

Just give me some chances now I know what is wrong.   THIS IS REALLY COMPLEX, and probably NO ONE ELSE COULD DO THIS!!!

 

[John Dyson]

8 hours ago, PeterSt said:

 

Hi John - Yes I did. This was 10 hours ago, so before you asked. To a small quarter of the ~80 demo tracks, but two times. :-) First the Decoded, then the RAW. I have two pages with remarks, per track one line.

Mind you please, the 50 seconds is already too short to have a sufficient judgement and write something about it as well.

 

 

This is how they (accidentally) appeared in the player (selection per DEC and later per RAW in the file name).

Here you see me one way or an other writing down the title names :

 

 

So this is DEC(oded) first. Btw, with the "short" I mean that all the reverb has been taken out. Cut piano notes and such (like Y.o.t.C.).

 

There is no general consensus of what to do for improvement (this is no different from the various times I tried before over the years), which makes it the most difficult to help out. However, with my experience on these matters, the verdict would be that the application is wrong in the basis, that allowed for the non-consensus and all completely different remarks. One think I could say is that for almost everything counts that it sounds hollow and inside out (this latter may be more difficult to explain, but it comes along with the hollow and definitely implies a phase problem).

I said "-Stop-" there in a KSTR way of ending the session, because at some stage it becomes clear that continuing is useless. However, it would be very easy to do the remainder tonight. N.b.: I could prepare better to print the playlist in advance and avoid the witing of the titles.

With "Highs above mid" I mean: emphasized highs (like in 4K and up) but no mid under it.

"beep beep beep": hardly power and LF.

 

...

 

I don't even know how legit it is to listen to the RAW, but my very personal opinion tells me that there has to be some merit because it *is* the base for further processing. Next it will depend on that further processing obviously, but without that decent base, I'd say there's nothing to go for further. It now completely depends on the specific remarks.

The track sequence is the same (first one is Year of the Cat), so you can compare the remarks, if needed.

 

 

So mind you, if I with e.g. Y.o.t.C. now don't mention "hollow" it for you means explicitly that where the DEC version is hollow, this one is not. I am looking myself over the remarks of RAW and see no hollow anywhere. Thus, the phase problem emerges in the decoding, right ? ... so this is how your thinking judging is allowed to be. Partly relative but in the end explicit (give us 200 seconds per track and more comment is possible ;-).

 

My wife did not know what I was doing, but I noticed her singing along with the DEC snippets. Into the third of the RAW snippets she asked whether these were cover bands contrary to the DEC's. Mind you, this was into Mrs. Robinson.

This should be telling, namely that the process of making the RAW is completely killing everything, which you so hard try to restore by EQ which I too regard subjective.

I expected my wife to have remarks during the DEC set, but she didn't have any (I asked nothing either, but if things go bad she will announce herself, usually). It is a measure, over here.

 

For myself I noticed that I could listen to this, contrary to the DEC's.

 

I can't read what I wrote at Mrs. Robinson, but the "Cool" refers to stone cold with an arrow to ""opposite to the DEc".

 

The consensus here is quite clear for me: way too much emphasis on the mid. In aftermath you (I) can see this on "tonality", "Voice too profound".

 

The 4 but last says "no high mid" and refers to "bell". This is super crucial because when rendered wrongly, bells disappear (or become profound, which would be good/nice but would still be wrong-ish). Thus, it tells (me) that the top of (human) voices is lacking here.

 

What should be telling (if I had to solve things) is that the one comes along with "power" (which will be relative and as a surprise, to the DEC version), while the other just the same does not show any power at all. This will be the low mid, FWIW.

 

The "unrecognizable" at Mama Mia (one but last) was already so in the DEC version (there noted as "super bad"). So again, something which is wrong in the base, won't allow restoration.

 

You may investigate 15 - Walking my Baby back home, as I judged that "OK". It is a rarity because the only one.

 

...

 

A last remark, because this is crucial for someone like me:

As someone who is not continuously trying to improve his playback means, but who is very explicitly working on making all real-life representation (this takes 40+ years of taking hurdles), it already goes wrong in my mind (starts to hurt) if whatever Agneta does not sound like her. Olivia Newton John no different. And remember, I am that guy with grand piano's, full drum kits, a bunch of guitars and what not, especially for the purpose. I try to reproduce reality. Now:

When a piano does not sound like one, I am not able to judge the remainder any more (of the track of concern). This is how Year of the Cat fails at the very first tones. Btw, YotC is a most difficult one to begin with, because it is a first which sounds thin in a digital system and will take years to get right (with a full body etc.).

 

Peter

I would have LOVED to see whatever notes made -- would probably have helped.  There isn't a lot to say as I have begged for help, but it seems like there

were too many filters.   I do have track record for saying thank you when helped.  I probably would have benefitted from your records with some additioinal explanation,

and then you might have gotten the biggest thank you ever.

Of course, you might have done so -- but you have to realize about the complexity of this thing and everything on my mind.  No one else could do this -- no one else could do a GOOD DolbyA decoder either.   Had lots of 'impossibles' there also.   Of course, I had an ancient schematic and test files given to me by some major players ( the few who would gamble a little.)

Since I could hear yesterday (with major health risk -- maybe could you try a listen to the demos.)   I think that they are close enough that some SIMPLE comments might be helpful.

 

PS:  IMPORTANT NOTE ABOUT EVALUATION -- some of the DEC demos are lower level.  All demos were normalized, but COMPRESSED material will tend (not always) be louder.   PLEASE CONSIDER THIS FACT.   When the signal is softer, it is usually because of stronger dynamics.  It can also be because of greater density.   DECODED material CAN be louder -- but that is usally because compressed is more dense.

 

 

 

[Jud]

10 minutes ago, John Dyson said:

I think that they are close enough that some SIMPLE comments might be helpful

 

John, I'm not sure Peter does "simple." 😀

[John Dyson]

29 minutes ago, Jud said:

 

John, I'm not sure Peter does "simple." 😀

Yea -- you are probably right -- and with that complexity without a kind word, I could have taken some of his stuff as a reason why it cannot work.

Evidence is difficult to deal with, but also there was a general loss of faith.  I was fighting a lot of obsticles, a LOT more than I had thought.   DolbyA, which was deemed impossible, was infinitely easier than FA!!!   I am sure that the decoder is not fully finished, but it is within earshot that even the biggest skeptic and those turned against it should be able to estimate the possibility of sucess.  (Man, I hate my hearing sucking -- it is my meds!!!)

 

However -- PLEASE HEED THE COMMENT ABOUT LOUDNESS (sorry for what is sometimes considered yelling -- but it is very very important.  Even I get confused when there are different levels.)

 

PS:  thanks for the suggestions!!!

Earlier, I made a comment about 'ad-hoc' techniques on the bass.   What I meant is that normally, if one can hear, the sound 'locks-in'.   On the bass, I just used judgement and not hearing 'lock-in'.   Man, the inter-layer EQ has to be dead-on!!!

 

 

 

[John Dyson]

PS: later on today, I'll be uploading V2.2.5F (the version for the demos.)   They'll be in the demo area.   I am still really shakey and mistake prone.  I might hold off untili tomorrow.  The build process for Linux is basically uploading the files.  However, for Windows there is about a 5step process where I usually make at least one mistake.

 

[John Dyson]

11 hours ago, Don Hills said:

John,

It does appear from Klaus' measurements that the decoder frequency response is not flat, at least at high levels where the decoder isn't applying dynamic  gain or EQ changes. Assuming that the non-flat response is deliberate, what might be your reasoning behind it? For example, might the response be to correct perceived response changes applied by the original "FeralA" processing?

Here is what happened -- FA processing isn't really flat like DolbyA is at high levels.   It is weird.  I'll explain FA -- and also recognizing that a low level multi band process will never be flat.   (Esp with the varying thresholds on different bands like on DolbyA, and the units being cascaded at different calibration levels.)

 

* I call each DolbyA unit a 'layer'.

 

Some genius figured out a way to cascade a bunch of DolbyA units to get a very wide dynamic range compressor of some kind.   I am not sure why  -- perhaps IP protection?   Perhaps to make vocals louder?  It is a workable concept for some reason, and i'd guess that it was formulated back when the subtle nature of DolbyA design wasn't well understood, and a fast compressor ended up working pretty sickly without the DolbyA attack/release design concepts.   The design concepts ARE VERY subtle, and caused me to design a lot of 'superior' versions missing the nonlinear charactertsic of the attack/release & sucking.  DolbyA units have NO attack/release as we normally understand it -- Dolby was a total, absolute genius,   The damned design is almost perfect at following the waveforms/envelope except frequencies below about 500Hz, and then depends on tracking the shape of the waveform to cancel the distortion on playback

 

Anyway -- with that DolbyA genius, I guess they designed the FA fiasco to do what they wanted.   I don't really hear much expansion myself,when listening to 'decodes',  but noise magically disappears, and the transients are a little stronger. and less 'garble'   FA somehow compresses the insides of the signal.   I know WHAT it does, just don't understand the goal.

 

So -- how I came onto the FA scheme was some experiments in 2012, where I was trying to design a GP expander.  When experimenting, I found this weird super-fast attack release.  Also, eventually I found a periodic nature (somewhere between 9dB and 10dB) between compression curves -- totally strange.   After a while, I designed an approximate DolbyA decoder (I mean, really approximate), and it seemed to work.  So, even though originally I coudn't imagine that they were using DolbyA -- somehow they WERE.

 

--

 

At this point, the serious DolbyA project started , with the help of Richard -- even though other parties were interested in just DolbyA, I was interested in both DolbyA and solving the 'garbled signal' FA problem.  Basically, I think that the FA signal is MUSH (assuming the term FA is the generic last stage compression used on most consumer recordings.)   My hearing is somewhat trained when I was fairly young --  so I know what a good signal sounds like (my 64 yr old hearing notwithstanding.)

 

Once I had a moderately good DolbyA decoder, then I started experimenting with FA decoding.  I started with the standard -13dB or so threshold, but found that FA used at least -1.5dB lower.   After a lot of work, I got a single layer decoder working, but something was wrong...  There was still the pseudo-DolbyA compression -- EQed, and some dB lower, of course.  Eventually, I tried a multi-layer design (the DHNRDS can emulate up to about 7-10 DolbyA units all in one.)   Over time, the DolbyA side ended up being damned good (of course not perfect), and a lot more clean than a true  DolbyA.   Once the DolbyA decoding started working super well, I started playing with multli-level  FA decoding.  BTW, the only thing keeping the DHNRDS from doing 16 tracks at a time is my I/O code and a little 'rewiring' of how the layers are put together and splitting how it works.  (It is like a high level set of block diagrams and can be re-adjusted in design.)

 

This is where the trouble is -- the multi-layer.   Of course, one must consider dealing with the screechy DolbyA sound -- what kind of EQ?   This is where I have been having troubles, esp my hearing.  Here is what I have found out, and I think that it is correct:

----

 

At the input, exact 9khz, -3dB.   At the output, along with a couple of 9kHz to 21kHz approx -7dB filters (the -7dB being 9k/21k, which is  something like -7.xx , I still consider it an 'even' number because it is a frequency ratio)  and the 9kHz -6dB and 12kHz -3dB.  (the -dB eq at 9kHz is to counter the 9kHz boost.)

 

ADD-ON:  also at the output (just checked)  75Hz+3dB, 37.5Hz+3dB, plus significant amount of 20Hz rolloff.

 

The real complication comes in between.   They use this crazy tapered set of -3dB EQ  each between 750Hz to 2750Hz, 1kHz to 3kHz and 1.25kHz to 3.25kHz -- totalling about -9dB.   The design of the upper MF EQ  is a dip betwen those frequencies.   IT isn't really an upwards/downwards filter, it is a DIP in the upper midrange.

 

Next, we gotta handle the bass between the layers -- 500Hz -3dB, 75Hz -3dB.

 

The above is the current state of practically ALL of the EQ and the in-between -- the result is now close-to audibly flat (much better than even 1 day ago.)  Now, there are just a few deviant defects in my demos (tests) which I partially attribute to EQ before encoding the FA recording.

 

After all of this -- ALL of the EQ is even numbers, and I end up with essentially flat audible response +-.   Note that the 500Hz -3dB and 75Hz -3dB is hyper crtical.  ANY change ends up either blasting your hearing or cannot hear any bass!!!

What is going on other than FA existing?

 

John

[PeterSt]

3 hours ago, John Dyson said:

I would have LOVED to see whatever notes made -- would probably have helped.  There isn't a lot to say as I have begged for help, but it seems like there

were too many filters.   I do have track record for saying thank you when helped.  I probably would have benefitted from your records with some additioinal explanation,

and then you might have gotten the biggest thank you ever.

 

I am totally clueless as to what you want to tell me there.

Maybe someone can translate it ?

 

[Allan F]

24 minutes ago, PeterSt said:

 

I am totally clueless as to what you want to tell me there.

Maybe someone can translate it ?

 

Perhaps this might help: https://mybigmonkey.com/gibberish/