'FeralA' decoder -- free-to-use

[John Dyson]

10 minutes ago, PeterSt said:

 

John, despite your hints about your perception of "bass", you might want to know that in the Crime album there plainly is no bass in the first 50 seconds. It all starts right after that (by heart, and knowing the tracks. A Supertramp feature, I suppose. Only just over half way on Crime of The Century (track) itself, there's supposed to be bass. But there is still virtually nothing; what is there is heavily underwhelmed.

 

You should never take Crime of the Century as a test case because the digital album is flowed for output level to begin with (and way too dynamic because of that - lacking compression if you will).

There is one hell of a lot of approx 80Hz on down.  If you are talking about super-low bass, there is an intentional rolloff at 18.75Hz.   The decoder  has a *CORRECTIVE* rising response below 80Hz to about 30Hz, then starts trailing off.  I carefully rolled off at/below 18Hz.   To me, the approx 60-80Hz is repressive.  But, I have headphones that are close to flat down to about 10Hz.  The headphones with certain kinds of music  can reallly be repressive when there is super low bass - my ears almost pop from air pressure :-).

 

I don't like strong bass very much -- it is repressive to me.

 

But. most importantly -- thanks for telling me that this is normal.   Probably one reason why some people didn't like some of the LF characteristics of the older decoder versios.   I just cannot stand the sound that they like (and what they like is probably more accurate.)   No matter -- it is yet another bias in my hearing, driving me crazy.

 

If only I had specs, or even before/after examples of unencoded and encoded FA together,  this thing would have been correct, literally within weeks or months.

 

 

 

[PeterSt]

21 minutes ago, John Dyson said:

https://www.dropbox.com/sh/lgccx2jdrj1dgtf/AACCQwLxlAyEZPE_1AHX8QWOa?dl=0

 

Same thing, except for two tracks. But it is so wrong in combination with the somewhat higher regions that it is impossible to listen to (it hurts).

 

I understand that you don't want to hear people talk about technicalities (I have no clue what you do want as you don't answer that repeatedly), so skip ...

But if you ask me things are digitally clipping (+32768 becoming -32767 instead of +32767). Audible on track 03 and 07 of Crime, and track 07 and 08 on Crisis.

 

I will now play the latter from CD and report in comparison about that re the bass.

[PeterSt]

13 minutes ago, John Dyson said:

There is one hell of a lot of approx 80Hz on down.

 

Nope, there is NOTHING. Intentionally (except for the last track). Mind you, to be certain I felt my woofers (they drop in at 230Hz). They don't move a bit. Remember, the first 50 seconds. And don't compare with your RAW versions, because there it may (or will) have gong wrong to begin with.

[John Dyson]

Just now, PeterSt said:

 

Same thing, except for two tracks. But it is so wrong in combination with the somewhat higher regions that it is impossible to listen to (it hurts).

 

I understand that you don't want to hear people talk about technicalities (I have no clue what you do want as you don't answer that repeatedly), so skip ...

But if you ask me things are digitally clipping (+32768 becoming -32767 instead of +32767). Audible on track 03 and 07 of Crime, and track 07 and 08 on Crisis.

 

I will now play the latter from CD and report in comparison about that re the bass.

Off topic -- after this, no further discussion about it -- we gotta focus on the project issues:

 

Honestly -- I am having severe comprehension problems right now.  It has been a bad time all around for me  I can talk, I can write (excepth with organization problems)I can think, but the other  interfaces to the world are scrambled, and I get imperceptable glitches all fo the time -- I dont' feel them, no-one sees them, but disrupts some kind of activities and are diagosed..  Sometimes, if my finger is on a key to skip to a next song, the key gets struck without having control over it. This is strange and is a realtively organized activity without consious control.   It has gotten infinitely worse recently  THATS THE SIMPLE TRUTH.  Along with all of that, my hearing is telling lies.  The tech part is EASY -- trying to collect the informatoinand kindly interface with people  is difficult.  (I am on meds for it, and they are effective, but along wiht everything else, I have always paniced very easily).

 

I have been the best touch typist that anyone can imagine, now - if I depend on touch typing, both hands get far out of sync.  Same thing for sentence structure/paragraph organizatoin, etc.   Maybe it is fatigue, but I need to simplify soon.  It will be easy to leisurly clean up the code, maybe tweak here or there, write my S-Z converter class, etc -- that is nice, easy stuff.  I can do it without much external interface.  Maybe I am beyond fatigued, I just dont' know.  A minor encumberance when I was 40-50 has become a hell nowadays.

 

You don't need to hear about my problems, but maybe try to understand why I am a jerk sometimes -- because, in my brain, without anyone (including me) seeing, I am a jerk inside. (literally)  I want this damned thing to get done, and my ability to adjust is very poor.  It is not about control -- it is about being afraid of losing balance, and someimtes distrusting reality, even though I have no problems in that regard.

 

Whether or not the decoder is 'Gods gift to the world', or a 'piece of cr*p', there is a responsibility to finish it.

 

 

 

 

[PeterSt]

28 minutes ago, PeterSt said:

Same thing, except for two tracks. But it is so wrong in combination with the somewhat higher regions

 

At listening to Crisis CD, I can now see that you seem to try to get something of bass by emphasizing the higher regions of it. So this runs into the lower mid (or upper bass if you want) which

a. makes voices to profound (I am not the only one any more on this one)

b. makes all as nasal as can be (though a bit Supertramp prone) though my horn speakers.

 

Otherwise I think I can see how you (or others) start working on this album. It would not be my favorite for best sound.

 

Last week I coincidentally played their first album once again. That is a whole different story ...

 

 

[John Dyson]

27 minutes ago, PeterSt said:

 

Nope, there is NOTHING. Intentionally (except for the last track). Mind you, to be certain I felt my woofers (they drop in at 230Hz). They don't move a bit. Remember, the first 50 seconds. And don't compare with your RAW versions, because there it may (or will) have gong wrong to begin with.

Okay, you are right -- I agree.  I got confused on snippet vs. full, then somehow got it all jumbled....  Sorry about that.  There are portions where there is A LOT of 80Hz, but I got it scrambled where  it is.   That big bass sound is too much for my taste -- for some reason, I prefer lesser bass, more about the transient/structure.   Perhaps eneded up with a bias after workiing on listening for transiets/signal structure/tells/glitches.   Yea -- I know a song by its glithces.  I like those glitches, they sound good to me :-).

 

 

[John Dyson]

43 minutes ago, PeterSt said:

 

Same thing, except for two tracks. But it is so wrong in combination with the somewhat higher regions that it is impossible to listen to (it hurts).

 

I understand that you don't want to hear people talk about technicalities (I have no clue what you do want as you don't answer that repeatedly), so skip ...

But if you ask me things are digitally clipping (+32768 becoming -32767 instead of +32767). Audible on track 03 and 07 of Crime, and track 07 and 08 on Crisis.

 

I will now play the latter from CD and report in comparison about that re the bass.

I'll look at that -- I don't use a DAW, even using something like audacity very seldom, even though it is very useful at times.

Everything is done in FP (or should be done in FP) until the last step.   My scripts are writtten by me, so are not perfect.   I could certainly have done a conversion, using SoX, which stupidly cannot handle floating point inputs greater than +-1.   Sometimes, I mistakenly depend on something that will cliip.  When I am aware, careful andthink  that I might want to do EQ, I use the decoder in --equalizer mode, then simply do --outgain=-10.   Another alternative si to use --outgain=-10 or whateve during the acutal decoding operatoinr, that will generaly make sure that there isn't clipping.

 

One of the attributes of the FA decoder design, is to keep the calbirations in sync from DA to DA, there is a fixed sqrt(2) -- 3dB gain between each layer.  If that is done, generally the levels don't go nuts.  However, with all of the EQ that goes on, the output can certainly go over 0dB, but it does so very seldom given most material.  Back in the bad old days when the decoder wasn't so precise, these levels would drift.  That was one of the original attributes that informed me that i wasan't doing something correctly.  it took a while, but eventually the correct architecture was inutited, hard core reverse engineer, and maybe a few seances with dead engineers.  (Humor implied.)

 

I'll correct those scripts, thanks for the warning. I do NOT like it when clipping occurs.

 

[PeterSt]

John, last bother before I am off for dinner etc. ...

 

After Supertramp and being quite annoyed of its sound already from normal CD, I put on this one:

 

 

So much more stereo. So much more open. So much more warm bass. So much more good sound from those days.

 

IOW, don't pull on dead horses too much.

(okay, I understand if you deem this cause by DolbyA then that's your purpose in the first place ... but the difference is too huge - try it)

 

Again, heads up.

[John Dyson]

1 hour ago, PeterSt said:

 

At listening to Crisis CD, I can now see that you seem to try to get something of bass by emphasizing the higher regions of it. So this runs into the lower mid (or upper bass if you want) which

a. makes voices to profound (I am not the only one any more on this one)

b. makes all as nasal as can be (though a bit Supertramp prone) though my horn speakers.

 

Otherwise I think I can see how you (or others) start working on this album. It would not be my favorite for best sound.

 

Last week I coincidentally played their first album once again. That is a whole different story ...

 

 

The thing is -- I am not enhancing anything.  Everything is standard.

The onlything that I see as questionable at all -- the Q value of 1.414, which I think is high, but also works perfectly for the Anne Murray albums.  I check a lot of material, not just my 'favorites'.  If I just checked my favorites, then nothing would work correctly.

 

Methinks they are doing EQ before FA encoding.

Or, it might not be FA, but another kind of compression.

On the full FA decoder, Supertramp has always sucked -- and I should have probably stopped it before trying.  I am too fast to accomodate unfavorable requests,

but I also have a very strong sense of integrity.  One should accomodate favorable and unfavorable tests.  Some WILL fail.

Supertramp, however worked easily on the cr*p single layer simple modified  DolbyA expander.

 

It is almost as if Supertramp is a benchmark programmed to fail -- really.  Audiophiles seem to like it, yet is real trouble to deal with.  Maybe they favor it

because the variant of FA (there are variants) or they didn't use FA.

 

I did think that 'Crisis' sucked.   There is a possibility that they used a different configuration (that is quite possible)

My opinion about this kind of problem is not reliable so I trust people to suggest that it isn't FA.  Or at least, ask  some questions

about the matter because this whole thing is a lot more complex than the simple black box that the decoder artificially provides.

I do know that another configuration is sometimes used, and I am still thinking about the possibility of it being 4+2 or even just '4'.

 

I will make sure that the Supertramp dead horse on FA wont be used as a benchmark again.

 

There are too many recordings that work super well.  Something is different about Supertramp,and the sound of 'Crisis' is ALMOST

as bad as trying to decode material that wasn't encoded.  However the tells all say 'FA'.  This must be a different config than 5+2.

 

When you hear 'Crisis', that does approach the kind of sound when I directly try to FA decode a DA decoded master tape.

 

That is, FA decoding of non-encoded materials  sounds similar to the Crisis decoding attempt, but is even more distorted.

 

So, giving a more complete review/analysis -- Crisis is probably a variant of FA.  I dont' have the energy to mess with it, but people who are interested,

and have a decoder might try:  --fa=4, --fa=6, perhaps even --fa=+5.   The normal mode is --fa=+7, and appears to be the most common

configuration.  Since Crisis didn't appear to work right away, and so few recordings have that problem, I just decided that in the future, proclaim

it isn't compatible.  I haven''t adequately documented the possible variants, but even in its current state, it is too complicated for the non-committed

person to use.  Even I forget about the variants, because they are so seldom needed.  If, someday, I get the time, and that is NOT right now, I'll search though the variants.

 

The general FA scheme is NOT simple to deal with.  I am lucky that most recordings are +7.

 

No-one else is going to want to search all of the possible modes, are they?  It is so much easier to avoid trying to fix it, unless

someone who really likes supertramp is willing to search the variious calibration levels and layer configurations.

 

I know that this seems selfish in a way, but I have a LOT of test material that doesn't require a special configuration.  I want to comply with a special request, but anything that I do will be 'desperate' rather than planned.  Supertramp might as well be called 'John-bane' after the troubles it has given me.

 

 

 

 

 

 

[John Dyson]

1 hour ago, PeterSt said:

John, last bother before I am off for dinner etc. ...

 

After Supertramp and being quite annoyed of its sound already from normal CD, I put on this one:

 

 

So much more stereo. So much more open. So much more warm bass. So much more good sound from those days.

 

IOW, don't pull on dead horses too much.

(okay, I understand if you deem this cause by DolbyA then that's your purpose in the first place ... but the difference is too huge - try it)

 

Again, heads up.

I did check and I gave a long response that should be ignored once I did the check about the recording (remmeber my unreliable hearing.)

IGNORE THE PREVIOUS, RAMBLING response -- more concrete information is here:

 

The Supertramp 'Crisis' appears to be an --fa=6 instead of tbe default --fa=+7.  I might still have something wrong as I am keeping the focus narrow right now.

 

The nasal effect that you mention is very similar to the 'decoding the unencoded' syndrome.  The only difference is that 'nasal' isn't as severe because the material IS FA, just not as much FA.   6 is a less processed recording than a +7.

 

 

[John Dyson]

This is about 'Crisis'.  First, I was checking around to find a correct setting, if there is one.

I thought that '6' would work, and it almost did.  It required --fa=7, but not the default --fa=+7.

 

Tried to run quick decodes without --fz or -fx, but just like a real DolbyA (but better).  The result was bad.  When doing certain recordings, at least --fz needs to be used.

When I ran some tests in that mode, it actually sounded pretty good.  This is another indicator that they didn't expect FA to be decodeable, because the available technology at the time (DolbyA HW) wouldn't have worked

 

Cannot tie-up my main machine any longer today, but in the next few days, I'll redo 'Crime' and 'Crisis' in --fa=7 mode.  I think that most people

would hear the improvement as being very substantial.  Here is why +7 was worse than 7 on this recording:

 

+7 is a decoding sequence with calibration levels like this:

-63, -53, -43, -33, -23, -53, -43

 7 is a decoding sequence like this:

-63, -53, -43, -33, -53, 43, -33

 

So, what is the big difference, and why did the default '+7' sound almost like trying to decode material that wasn't FA?   Here is the reason, the expansion is more intense as the calibration number is a bigger value.  So, -53 is more intense than -63, etc.   If you look at the +7 sequence, notice that the highest calibration number is -23?   If you look at the sequence that was successful, on '7' the highest calibration number is '-33'.   When trying to decode with too high a calibration level in the sequence, the result becomes similar to decoding the wrong material.  It really is just a combination of expansion mismatch, and too much expansion at the wrong time.  When you do too much expansion, it is almost like a profound dead zone in linear amplifier terms.  We all know that dead zones (e.g. under biased push pull amplifiers) can be ugly to listen to.  Of course, the expansion dead zone doesn't produce the high order components of a hard analog dead zone, but is still creates distortion products.

 

As mentioned above, I'll redo the decodes near the end of the weekend, when I am away and the computer is doing nothing else anyway.

 

 

[PeterSt]

8 hours ago, John Dyson said:

It is almost as if Supertramp is a benchmark programmed to fail -- really.  Audiophiles seem to like it, yet is real trouble to deal with.  Maybe they favor it

because the variant of FA (there are variants) or they didn't use FA.

 

FYI: I know Supertramp from 2 year prior to anyone knowing them and their first single. Back at the time (say 15 yo) I loved it. Today, however, nothing wants to be n my "demo" Galleries except these:

 

 

Normally "Demo" is used for the best sounding tracks hence it is ready for auditioners, but it may also contain albu,s which I want to revisit regularly because they fail but I feel that it is my system and they can be improved upon. Crime is in there for that reason (I still have hopes) already because it is a best album ever. The right-hand is their first I just talked about in my post from yesterday (sounds superb even though from end 60's begin 70's), and the middle one apparently sounds good enough to have a couple of tracks in my Demo.

 

 

If you want full harmonic bass from Supertramp, go for it, there.

I said "apparently" because tracks end up in my Demo in real time by the press of a button during playback at normal daily listening sessions for pleasure.

The fact that apart from their first for the good reason, nothing but this one shows up in there for good sound, is definitely telling. This already is so because back in the days they were among my favorites. Now, including yesterday's forced playback of Crisis, they don't "want to go" in there. They sound too poor.

 

Best of Times could be differently recorded / processed because it is a live album. So John, you may incorporate that in your analysis.

 

Btw funny, that It's a Hard World track, there's really singing through a telephone line. "Operator ...".

[PeterSt]

7 hours ago, John Dyson said:

+7 is a decoding sequence with calibration levels like this:

-63, -53, -43, -33, -23, -53, -43

 

My remark about Year of The Cat was about all being too "short". This is about the decays being shorter than reality. So focus on that first piano notes again in order to learn what I mean by this. Now:

 

Throughout for Crime I observed the same, if only the tracks are soft (and about all are the first 50 seconds). And because this is about "electric piano" as they were used in the days, the effect of sheer "cutting" becomes profound (those pianos flanger). So not short, but cutting. How this technically may work is this:

 

If there's certain "voices" (this includes instruments) which are at say +63dBFS only, they will be at zero or less when attenuated with 63dB. This is of course not the full voice, but the lowest frequencies in there first. 

... And what becomes zero (or less) will never expand to x any more. It's gone forever.

 

So John, that is what I hear now. But it is biased because of your explanation in that post from the quote above.

From Crime, from yesterday I recall to perceive the sheer hard cutting during the decay. I did not mention anything about it or other stuff, because you asked for "bass response" (pun).

I don't want to interfere with something I don't know anything about anyway, but at attenuating, the lower (volume) parts have to be treated less aggressive. As in : -96dB should not be attenuated at all or else you lose it (when processing in 16 bits (you will not, but it is about the gist)).

 

This is also how you lose the ambience out of everything. And how it thus becomes "dead birds" (my book).

[dkskl]

1 hour ago, PeterSt said:

 

My remark about Year of The Cat was about all being too "short". This is about the decays being shorter than reality. So focus on that first piano notes again in order to learn what I mean by this. Now:

 

Throughout for Crime I observed the same, if only the tracks are soft (and about all are the first 50 seconds). And because this is about "electric piano" as they were used in the days, the effect of sheer "cutting" becomes profound (those pianos flanger). So not short, but cutting. How this technically may work is this:

 

If there's certain "voices" (this includes instruments) which are at say +63dBFS only, they will be at zero or less when attenuated with 63dB. This is of course not the full voice, but the lowest frequencies in there first. 

... And what becomes zero (or less) will never expand to x any more. It's gone forever.

 

So John, that is what I hear now. But it is biased because of your explanation in that post from the quote above.

From Crime, from yesterday I recall to perceive the sheer hard cutting during the decay. I did not mention anything about it or other stuff, because you asked for "bass response" (pun).

I don't want to interfere with something I don't know anything about anyway, but at attenuating, the lower (volume) parts have to be treated less aggressive. As in : -96dB should not be attenuated at all or else you lose it (when processing in 16 bits (you will not, but it is about the gist)).

 

This is also how you lose the ambience out of everything. And how it thus becomes "dead birds" (my book).

@John Dyson

For a very obvious example of what PeterSt means: Please listen to the decoded snippet of "Finlandia". Especially the decays at 5, 10 and 30 sec. Total silence where there should be a little reverb/ ambience.

 

Apart from that. You ARE getting closer and closer to hitting bull's eye imo 👍

[fas42]

Crisis, what crisis? 😉 ... Just put on my standard copy of the Supertramp CD ... what's the problem?

 

 

is so mellow ... what an intro!

 

And,

 

 

is just downright amazing!!

 

Fabulous stuff ... 🙂

[fas42]

... 'cuse me - I was just boppin' and bouncing around the living room, listening to the other tracks on the Crisis album ... bloody good stuff!! 😁

 

At times, you forget how good these albums are ...

[KSTR]

V2.2.6C, Anne Murray "Danny's Song" (a critical piece as the levels are quite low, not to many sources playing at once, and the general quality is excellent), compare RAW vs. DEC-V2.2.6C vs. DEC-2.2.6C de-EQ'd with DeltaWave.

 

RAW vs DEC-V2.2.6C : EQ dominates (though the dynamic effects can be heard if one tries to ignore the EQ mentally as much as possible). It is a bit better balanced than earlier versions, less 300Hz honk and less 7kHz screech. But still completely unbearable. Again, I don't believe the original (whatever that was) sounded that way.

 

RAW vs DEC-V2.2.6C de-EQ'd : Failed for me, the decoder turned the steelstring guitar into a nylon-string, very muffled. Plate reverb on vocals... the same, killed decay and killed HF. Once levels get louder when the whole band starts playing, things get better as the expander is working less.

 

 

Comparison of large-signal frequency responses for V2.2.4E vs 2.2.6C:

 

It's gotten better with 2.2.6C in relative terms. In absolute terms, still completely off.

And no attempt seen to fix those strange jumps at 3kHz and 9kHz, where I'm still waiting for an explanation from John (also for the less severe roughness around 1.3kHz). Those jumps are linear phase and by this must produce pre-ringing when hit by a transient.

 

And, surprise, surprise, that is exactly what is seen in the large-signal impulse response:

The snippet highlighted is 30samples long and a part of the pre-ringing building up before the main pulse (it also produces post-ringing but that is audibly much more benign), at 88.2kHz sample rate that equates to 3kHz. The 9kHz overlay can readily be seen thanks to the integer ratio of the frequencies.

 

It is not clear if that pre-ringing is any audible, though. The effect is almost impossible to be judged in isolation with so much else going on, to get there a more thorough investigation and analysis would be needed.

 

 

[PeterSt]

3 hours ago, fas42 said:

is so mellow ... what an intro!

 

What's wrong with that is that all is in the background, especially the (singer) voice. Mind the horn of the car. It is there but nothing more. Try John's version and you'll jump to the sealing. So all exactly the other way around. Also, around 1:05, what do you actually hear for instrument ? OK, I know it is an electric guitar chord-played (famous Supertramp sound). But here ? it's mush. Mellow much if you want ...

 

3 hours ago, fas42 said:

And,

 

 

is just downright amazing!!

 

Not so much so. Again the singer is way underwhelmed. Listen at 2:05. I am used to distinguish instruments. Here you can wrongly blend something like a ride cymbal into other noises. The cymbal changes (or disappears) because of those noises. So this is "nothing" in my book. I guess this is emphasized even more when I'd play this on the main rig (where all the highs pop out). 

 

Quote

bloody good stuff!! 😁

 

That is the point. It is ...

[fas42]

11 hours ago, PeterSt said:

 

What's wrong with that is that all is in the background, especially the (singer) voice. Mind the horn of the car. It is there but nothing more. Try John's version and you'll jump to the sealing. So all exactly the other way around. Also, around 1:05, what do you actually hear for instrument ? OK, I know it is an electric guitar chord-played (famous Supertramp sound). But here ? it's mush. Mellow much if you want ...

 

Being in the background is the intent of the song, to my ears - people complain that there are no dynamics, or light and shade, in modern music ... then when they get a song with it, they still complain ... 😄.

 

Yes, the harmonically rich guitar sound - but I don't hear mush; I hear it clearly defined 🙂.

 

11 hours ago, PeterSt said:

 

 

Not so much so. Again the singer is way underwhelmed. Listen at 2:05. I am used to distinguish instruments. Here you can wrongly blend something like a ride cymbal into other noises. The cymbal changes (or disappears) because of those noises. So this is "nothing" in my book. I guess this is emphasized even more when I'd play this on the main rig (where all the highs pop out). 

 

 

That bit at 2:05 is the magic - it's rich, dense, subjectively so powerful - the impact of this section contrasting to the lead up just makes the track so satisfying ... okay, the mastering of it makes this very much a 'difficult' recording; but the rewards of getting it spot on are what it's about. There is little 'space' between the instruments, and the system has to be on its best behaviour for it all to make sense to the ears - in fact, in spite of the interference isolation tweaks I've been using, I did detect a slight edge to the sound - which turned out to be the TV, which was on, and is very close by - switching off fixed that. In conductive terms, the set is way down the end of the street; but there appears to be just enough RF, perhaps from the screen circuitry, to impact. For a 'normal' recording this causal link is close to being inaudible - but when the playback gets to the "everything matters!" zone, then I can hear it ... I will need to think more about some strategies to get this under control.

[John Dyson]

There is a new release V2.2.6G, along with fully phase-descrambled test snipets/results.  The sound will be qualitatively different.

These results are NOT from dynamics decoding alone. (Might be too much bass, might not -- but does have MORE.)

 

https://www.dropbox.com/sh/i6jccfopoi93s05/AAAZYvdR5co3-d1OM7v0BxWja?dl=0

 

Please PM me for now -- I might read the public messages tomorrow, but trying to see if there will be some 'self-organization' happening.

Obviously, with yours-truly's shortcomings, solving this LF problem has stagnated.

 

On 'more interesting' matters:

Instead of focusing on FA decoding EQ, I have been  playing with more interesting new-tech/secondary portions of the project.

There will seem to be less progress, or maybe a lot more -- it all depends on how quickly the remediation can happen.   Tell me

what you think of the sound with phase descrambling fully enabled?  Phase descrambling is 99% independent of the bass,

so the higher frequency clarity should be considered seperately from the character of the bass.

 

Also, a secondary matter because of the other things going on, I believe that the LF matter is improved yet again -- maybe gone a little too far?.   I tried

a different methodology to try to help mitigate some of the variability of my hearing. It IS possible that the LF goal has been overshot,  All I really

can do is to use 'tells', since myself, I don't trust measuring frequency response alone (all kinds of ffactors, even though it might help, then

I will be VERY VERY happy to say that I was very wrong about this.)

 

Will possibly be doing some work with a few others, based on the stabilized version V2.2.6G as I just uploaded.

 

 

 

 

 

 

[fas42]

I will put this here, John ... just to let you know, as I monitor how it's going with the Mamma Mia track, that this latest version sounds good! Can't speak for how the bass sounds, as I have already explained - but elsewhere it all hangs together well; I can't pick anything obvious to complain about, 😉.

 

[PeterSt]

7 hours ago, John Dyson said:

Please PM me for now

 

This is harder for me because of less required responsibility for what I'm spouting. But it depends on the subject. Here's something that is nicely vague, as I am guessing it largely:

 

John, it could well be that your processing of the downsampling fails (massively). It could also well be that you don't even listen to the snippets yourself (for a longer time by now, because why would you once it worked).

 

You could investigate the 24 to 16 bits process, as it seems that you could do something wrong with the thought of how to do this. For example: Ever very long back I had the idea that adding 8 bits should lead to louder volume "on top". This, while it appeared to be the other way around: it adds lower level at the bottom.

Now why do I think that you do similar in wrong fashion ?

 

Your 16/44.1 (or 16/48 - I did not check them all) is not really 48dB too loud, but if anything it is 24dB. I myself took this for granted for "years" (really so) as it would be some odd way of doing things for you. But what I keep on noticing is that you don't go into that, when the remark about it is made. I did a couple of times (in not urging fashion) and @KSTR did too. Also, it without further thinking, for me it seems the result of expansion.

But how can you ever: the one time imply my volume to be at -40dBFS and the other time have it at -16dBFS. Man, people's windows go out if this goes unexpectedly. One thing why others may not be bothered: I have my volume normalized by standard, and thus any lowest level "congestion" works out to an enormous boost in level. You talked about this yourself regularly, so you should know what I mean without further explanation.

I never talked about it, but to me this is also an indication of "the wrong", in the sense of: if that much expansion is possible from a (new) native signal, a lot has to have vanished from the original signal. Too much compression undone. ... I may refer to my "too short" and the "cut" sound, I talked about yesterday.

 

If I am right, this public posting is justified, because you may have plagued yourself for a very very long time without reason; people might learn how a project almost terribly failed without real reason.

For now I will leave it at this as a teaser. ... Things to work out ... (and more by PM).

[John Dyson]

12 hours ago, PeterSt said:

 

This is harder for me because of less required responsibility for what I'm spouting. But it depends on the subject. Here's something that is nicely vague, as I am guessing it largely:

 

John, it could well be that your processing of the downsampling fails (massively). It could also well be that you don't even listen to the snippets yourself (for a longer time by now, because why would you once it worked).

 

You could investigate the 24 to 16 bits process, as it seems that you could do something wrong with the thought of how to do this. For example: Ever very long back I had the idea that adding 8 bits should lead to louder volume "on top". This, while it appeared to be the other way around: it adds lower level at the bottom.

Now why do I think that you do similar in wrong fashion ?

 

Your 16/44.1 (or 16/48 - I did not check them all) is not really 48dB too loud, but if anything it is 24dB. I myself took this for granted for "years" (really so) as it would be some odd way of doing things for you. But what I keep on noticing is that you don't go into that, when the remark about it is made. I did a couple of times (in not urging fashion) and @KSTR did too. Also, it without further thinking, for me it seems the result of expansion.

But how can you ever: the one time imply my volume to be at -40dBFS and the other time have it at -16dBFS. Man, people's windows go out if this goes unexpectedly. One thing why others may not be bothered: I have my volume normalized by standard, and thus any lowest level "congestion" works out to an enormous boost in level. You talked about this yourself regularly, so you should know what I mean without further explanation.

I never talked about it, but to me this is also an indication of "the wrong", in the sense of: if that much expansion is possible from a (new) native signal, a lot has to have vanished from the original signal. Too much compression undone. ... I may refer to my "too short" and the "cut" sound, I talked about yesterday.

 

If I am right, this public posting is justified, because you may have plagued yourself for a very very long time without reason; people might learn how a project almost terribly failed without real reason.

For now I will leave it at this as a teaser. ... Things to work out ... (and more by PM).

This is interesting -- everything on the decoder  is +-1 (24bits, 16bits, both full scale in normal fashion), except FP is whatever it is.

Whatever is happening, lets try to keep this discussion going until it is resolved -- I'll try to explain what I see, and the context of how the decoding process creates the output files..

I almost NEVER share direct decoder output, everything (almost absolutely everything) has been processed by standard command line utilities after the decoder has created the file.  I need to figure out if the command line utilities are causing problems?!?!?!?  Is there a way that the decoder created a bogus output?!?!!?   All decoder output is .wav, so if someone has a non-compliant .flac file created by me, then I need to find what utility is doing the evil.  Could this be something like a bogus REPLAY GAIN  or a bogus metadata flag being set?

 

Below is blather about how files are created -- again, trying to explain, in my jumbled way, the kinds of files supported and a little bit about how they are treated.

 

 

The decoder itself never creates 16bits, only supports 16bits on input.  In that case, the conversion is 'full scale' is +-1 (-32768 is -1, and 32767 is +1).

As I normally use the decoder, everything is FP on output, then converted by sndfile utils or sox to 16bits, if appropriate.   Very specifically, I removed 16bit output support from the decoder because I didn't want to have to bother with dither issues at the time (really -- that is the simple truth.)   I really don't think that 16bits should be used any where that there might be additional editing/processing/etc because itis very easy to damage the dither.   On FP, the normal level is +-1 is 0dB.   If there was a format with scaled values, then the other utilities should do the scaling  because FP and any normal 24bit that I have seen is a +-1 based 0dB.     So, the decoder outputs FP and 24bits as +-1 full scale (where 24bit +1 full scale is (1<<23) - 1.   FP full scale is just FP value +1.   Of course, on FP you can go beyond full scale and usually works okay.  I don't like to use direct 24bit output on the decoder, is that what you are talking about?  Is it that I didn't adequately test the 24bit output code for the decoder itself, or is the problem in the subsequent utilities?  (Decoder 24bit output IS dithered, but I am sloppy doing it -- I don't want filtering to cause the dither to disappear.)

 

There is no conversion that I know about whose result is hidden from SoX or other tools like that.   The direct decoder output is just not normally directly seen in my demos, and any 'bad value' events would create an indication during the .flac' file creation process (overflow.)  Also, the mechanism for normalizaton on an album basis is manual, so I add/subtract the dBs manually except for the snippets demos, but those are normalized with an attempt to keep the RAW vs DECODED versions at the apprx the same level. I don't think that I have ever added more than 12dB, and that would be an extreme event.  Usually, the decoder output is -9dB to -5dB, so adding values like 5-9db are typical.

 

I don't want to confuse the matter with my blather (I know that I blather...)   At the bottom of this post, there SoX measuremens for the 'raw' 01-School.wav file, and the 'flac' file that we used as a test demo. I edited the 'boring' information from the SoX output..

The decoder doesn't care at all about the signal output level because the decoder output  is all FP.    The flac file shows the results after all of the conversions, the local processing doesn't go any further than creating the flac file.   The .wav file is raw from the converter -- no utiliies that i have show any extreme values during my process.    The .wav vs. .flac  numbers technically match, but they look different because the level sare different by about 8-9dB.   The .flac version has been normalized on an album basis, while the .wav file is raw raw raw.  (For example, look at the first Pk lev dB vs RMS lev dB.)   The difference on the .wav file is (31.41 - 8.24) is pretty much the same as (23.66 - 0.49).

 

I don't see anything wrong with these numbers.  Also, SoX does say that there are 31 bits of 'stuff changing' in one of the removed lines.

The only thing that I know can do odd things that you are talking about is 'replay gain', but the problems that you are talking about are a lot more profound than

'replay gain'.   I alaways try to avoid 'replay gain' n the examples because tit can obfuscate comparisons.

 

I'll try to think about how this incompatility is happening.   Can we find more data on this?

As far as I know, 24bit signed values are usually the same +-1 as 16bit numbers...   Is this where we seem to have an incompatibility?  Is there a scale factor scheme that

is somehow being triggered, something that I don't know about?   Is there a bit that is getting set somewhere that I don' tknow about?  I am seeing good numbers

from the utilities that I have.  MAYBE AN UNSUPPORTED EXTENSION that SoX, sndfile, flac, etc don't support, but are supported by DAW or whatever tools that you

are using?

 

Direct from decoder:

[jdyson@i10900X V2.2.6D-2]$ sox 01-School.wav -n stats (FP file)
             Overall     Left      Right
DC offset   0.000000  0.000000 -0.000000
Min level  -0.345295 -0.345295 -0.336206
Max level   0.387468  0.387468  0.316939
Pk lev dB      -8.24     -8.24     -9.47
RMS lev dB    -31.41    -31.73    -31.12
RMS Pk dB     -23.03    -23.69    -23.03
RMS Tr dB    -137.91   -135.98   -137.91
Crest factor       -     14.95     12.10
 

[jdyson@i4770 crime]$ sox 01-School.flac -n stats (24 bit file)
             Overall     Left      Right
DC offset   0.000000  0.000000 -0.000000
Min level  -0.842734 -0.842734 -0.820550
Max level   0.945661  0.945661  0.773528
Pk lev dB      -0.49     -0.49     -1.72
RMS lev dB    -23.66    -23.98    -23.37
RMS Pk dB     -15.28    -15.94    -15.28
RMS Tr dB    -130.09   -128.18   -130.09
Crest factor       -     14.95     12.10

 

 

 

 

 

[John Dyson]

18 hours ago, fas42 said:

I will put this here, John ... just to let you know, as I monitor how it's going with the Mamma Mia track, that this latest version sounds good! Can't speak for how the bass sounds, as I have already explained - but elsewhere it all hangs together well; I can't pick anything obvious to complain about, 😉.

 

There is a few of us trying to beat the decoder into submission.  A process of cooperation seems to be developing, and there has been more constructive process going on. I am in the midst of decodes and documentation as I am writing this now.

Just forewarning, as there is currently positive progress, it seems to be most productive for me to hold back on the blather, but will definitely let you (and everyone) know as there is REAL progress.  I might be sending some private comments to those corresponding with me, but the focus really is on getting this thing finished before I cannot hear at all :-).

Hoping to have an interesting PM for you sometime tomorrow, and as time goes on, we can make more stuff public. 

 

John

 

[KSTR]

2 hours ago, John Dyson said:

The decoder doesn't care at all about the signal output level because the decoder output  is all FP

Yes, that's one of really good things about it. I've hit the decoder with +40dBFS Diracs (to keep it from expanding) etc with no apparent issues.

 

SoX is limited as it's 32bit integer internally so while it can take FP input and output it still clips internally when "overdriven".