Jump to content
IGNORED

'FeralA' decoder -- free-to-use


Recommended Posts

In doing some demos for someone on a private list, a new anti-distortion update came to mind.

 

I promise, as long as something doesn't tie me up tomorrow, there WILL be a release.   The decoder is doing VERY WELL.

 

Every time I start considering a release, during a code review, I figure out a new trick -- really!!!

 

Gotta do a release some time, and if all goes well, will be a semi-final type thing.

 


 

Link to comment

Updated demos, pretty much exactly as the planned release tonight.

The best way to hear the improvement, if you like the group -- listen to the ABBA examples.   There is a profound amount of clarity, but the material is the minimally compressed available.   Sometimes strong compression will brighten the sound, but the recordings in these examples get all of their clarity by using true and honest methods without any compression-boost or multi-band brightening.  ( Low level dynamics measurements show the ABBA recordings to have a fairly strong dynamic range, unlike most other releases -- the examples unfortunately, don't show the full improvement.)

 

All of the improved clarity comes from the suppression of several kinds of distortion, including the strong gain-control fog (sidebands) that exist in most consumer recordings.  These WILL sound different, and SOMETIMES I choose the wrong bass settings (yes, sadly -- we are back to three bass modes.)

 

No longer two versions of the snippets demos -- the new EQ works for most recordings, but some need a change in EQ shape (not just bass boost/cut.)   This EQ is much more compliant with the recordings, and I got rid of my over-engineering (which is a bad thing) in the EQ.

 

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

Link to comment

V2.1.1A is available.

A few last minute LF EQ changes were added (newer than the demos.)   The EQ is now more full, but still follows the rules.  (There are rules to fit correctly, or there will be bumps in the response.)

 

The anti-distortion is further improved from before.  It can be difficult to hear the benefits, but is obvious on vocal chorus type material.   Perhaps the most obvious are ABBA chorus or perhaps '05 Dreamer' on 'Crime of the Century' amonst other POP.   A lot of classical is also more detailed.    I just came-up with another kind of anti-distortion/phase stabilization that just might help further.  Might appear in the next release if it works.   All of these timing stabilization methods appear to come from some kind of phase scrambling done to FA recordings -- helps to smoothout peak.

 

LF EQ has been very difficult for me, and still might have some troubles.   There are three modes:

--fa (normal, hopefully.)   Perhaps a little too full for classical, but matches most pop quite reasonably.

--fa=L (more thin sounding.)  Useful for mostly ancient pop and some classical.

--fa=M for recordings that are too thin sounding when using the default.  Love Over Gold SACD really needed this.

 

Both --fa and --fa=M can sometimes appear to be similar sounding, but if it does sound tolerable, best to use '--fa' alone.   When used improperly, the '--fa=M' mode will sound oddly bass heavy.

 

Since the LF/lower MF EQ has been troublesome -- constructive feedback is desperately welcome.   Frustratingly, LF has been troublesome, but HF and upper MF (1k-3k) EQ is almost provably correct.  My hearing is not a good judge of LF, and even though I have derived some EQ rules for the LF, there are still a lot of plausible choices.   Help is welcome.   After resolving any corrections, I can do a simple release in about 1-2Hrs.   So, corrections are easy once I know what to do.

 

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

 

Link to comment
3 hours ago, John Dyson said:

V2.1.1A is available.

A few last minute LF EQ changes were added (newer than the demos.)   The EQ is now more full, but still follows the rules.  (There are rules to fit correctly, or there will be bumps in the response.)

 

The anti-distortion is further improved from before.  It can be difficult to hear the benefits, but is obvious on vocal chorus type material.   Perhaps the most obvious are ABBA chorus or perhaps '05 Dreamer' on 'Crime of the Century' amonst other POP.   A lot of classical is also more detailed.    I just came-up with another kind of anti-distortion/phase stabilization that just might help further.  Might appear in the next release if it works.   All of these timing stabilization methods appear to come from some kind of phase scrambling done to FA recordings -- helps to smoothout peak.

 

LF EQ has been very difficult for me, and still might have some troubles.   There are three modes:

--fa (normal, hopefully.)   Perhaps a little too full for classical, but matches most pop quite reasonably.

--fa=L (more thin sounding.)  Useful for mostly ancient pop and some classical.

--fa=M for recordings that are too thin sounding when using the default.  Love Over Gold SACD really needed this.

 

Both --fa and --fa=M can sometimes appear to be similar sounding, but if it does sound tolerable, best to use '--fa' alone.   When used improperly, the '--fa=M' mode will sound oddly bass heavy.

 

Since the LF/lower MF EQ has been troublesome -- constructive feedback is desperately welcome.   Frustratingly, LF has been troublesome, but HF and upper MF (1k-3k) EQ is almost provably correct.  My hearing is not a good judge of LF, and even though I have derived some EQ rules for the LF, there are still a lot of plausible choices.   Help is welcome.   After resolving any corrections, I can do a simple release in about 1-2Hrs.   So, corrections are easy once I know what to do.

 

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

 

Just uploaded the snippets to correctly match the released version of the decoder.

Link to comment

Got some of the feedback that I asked for.

With the suggested changes, we are back to mostly the same EQ except the old Carpenters stuff, which also needed a change.  (my hearing doesn't detect the problems, but also can hear the advantage when I accept the suggestion.)

 

The decoder is DARNED CLOSE to perfect.

The LF EQ is very very close.

 

New demos (in the same place) are coming after the decoding session tonight -- great improvement, clarity and more consistency.


Thank goodness for the help.  A lot more recordings can be decoded with just --fa again.   Only old Carpenters (and just a few others) need --fa=L.

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

 

da-release2.1.1D-win-2021-03-05.zip StartUsage-V2.1.1A.pdf da-V2.1.1D-Linux.txz

Link to comment
32 minutes ago, John Dyson said:

Got some of the feedback that I asked for.

With the suggested changes, we are back to mostly the same EQ except the old Carpenters stuff, which also needed a change.  (my hearing doesn't detect the problems, but also can hear the advantage when I accept the suggestion.)

 

The decoder is DARNED CLOSE to perfect.

The LF EQ is very very close.

 

New demos (in the same place) are coming after the decoding session tonight -- great improvement, clarity and more consistency.


Thank goodness for the help.  A lot more recordings can be decoded with just --fa again.   Only old Carpenters (and just a few others) need --fa=L.

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

 

da-release2.1.1D-win-2021-03-05.zip 5.96 MB · 0 downloads StartUsage-V2.1.1A.pdf 138.31 kB · 0 downloads da-V2.1.1D-Linux.txz 828.52 kB · 0 downloads

Error in the files -- here the correct ones are -- multiple computers -- nice, but confusing:

 

 

da-release2.1.1D-win-2021-03-05.zip da-V2.1.1D-Linux.txz

Link to comment
1 hour ago, John Dyson said:

Error in the files -- here the correct ones are -- multiple computers -- nice, but confusing:

 

 

da-release2.1.1D-win-2021-03-05.zip 5.96 MB · 2 downloads da-V2.1.1D-Linux.txz 828.49 kB · 2 downloads

We caught a bug -- sorry...

Doing releases like this (without system test) is sometimes a problem.  It is all my fault, trying to get tihs finished!!!

 

da-release2.1.1E-win-2021-03-06.zip da-V2.1.1E-Linux.txz

Link to comment

Reminder -- use the 2.1.1E version -- I made a transcription error on the earlier versions.   There was an error in

transcription (after getting some help with test code) -- and only noticed it when running the demos, then

double checked the code.

 

Think that I am done with bugfixes.   This is a VERY complex piece of software -- there are simpler OSes.

 

My next project is to internally document the code, and put together (slowly) a document for easier use.

 

Specifcially requested bugfixes are welcome.   EQ troubles are also welcome to talk about!!!

There might be material that needs an odd EQ (e.g. Love Over Gold SACD) or the old Carpenters stuff.   We have plenty of command

space to add other kinds of EQ!!!

 

 

 

 

Link to comment

I just realized something -- just out of respect for AS and the wonderful forum, I'll be posting minor updates on the Dropbox site.

It is best to just use the download site below.

 

No new development is expected, but things like table entries or bona fide error-fixes are still on the horizon.   Maybe someday, the hooks for a GUI, but that is IN THE FUTURE, after the internals are documented.

 

 I hate to use forum space when until settle down, there might be a few updates == more opinons help, but also might force updates.   One of my first mistakes just as I started the release -- everything worked fine until I tried an alternative EQ.  Different EQs cause different EQ array contents -- but the array went out of bounds.   Sometimes it worked, sometimes not.   Yes. I still make mistakes like that, but the array def is perhaps 100 lines away from using the array, and for max perf, I do everything barebones.   This mistake was fixed.   The next was a transcription error.  Now, there just might be some feedback about EQ.   The updates will be stabilized at +4Hrs and every 24Hrs after that.  (That is: 9:00AM each morning.)   I am organizing like this simply to keep REALLY interested people sane.   The first bugfix might be this morning.

 

* For an update, ONLY the da-avx, da-win, da-core (one) needs to be copied to the original location.  Everything else is being kept stable.  

 

The EQ is a matter of opinion, and I am getting different opinions.  Currently, there is feedback about the bass being a bit too strong -- for me, even reasonable bass is irritating.   Even pros vary in their opinons.  Each of us are different, but I'll have the matter resolved before the update time at +4Hrs.   If there are any changes, please refer to the dist site below.  I have to check with 2-3 people before a good answer is gotten.

 

The version numbers is currently: V2.1.1E and will increase letter by letter.   There will be NO usage changes when the decoder stays as V2.1.X.   A change in the 2nd number field means that there could be a usage change..   The letters are bugfixes and minor audio improvements.   The third number field is for profound improvements/changes, or if I run out of letters.

 

NO NEW FEATURES ARE CURRENTLY PLANNED.   No new algorithms or capabilites are planned, but there are some already built-in things that I haven't discussed much and will document better as time goes on.

 

The same is also true for the DolbyA stuff.   DolbyA decoding is infinitely more mature, and previously had growth pains also.  FA is 4-10X more complex, depending on how you measure it.

 

Updates will be at a standard time (when needed):  daily at or before 9:00AM USA EST time, and will be available at the locations below.   No need to check other times.  There will be weeks when there are no changes, but might be daily for 2-3days after a release.   You CAN check the site early, possibly get a one-up on the changes.

 

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

I know that this is different than my claim of being finished, and in a relative sense, the code is finished.   Maybe table entry type things, or bugfixes migiht happen.

I'll make a post just as I do an update -- the contents will be obvious for those who quickly scan things.

 

Unless problems need to be public, and if you have troubles -- PM is probably best to keep traffic down.   Some problems are best discussed publically, like 'need more bass' or 'need less bass'.

 

John

 

Link to comment
4 hours ago, John Dyson said:

I just realized something -- just out of respect for AS and the wonderful forum, I'll be posting minor updates on the Dropbox site.

It is best to just use the download site below.

 

No new development is expected, but things like table entries or bona fide error-fixes are still on the horizon.   Maybe someday, the hooks for a GUI, but that is IN THE FUTURE, after the internals are documented.

 

 I hate to use forum space when until settle down, there might be a few updates == more opinons help, but also might force updates.   One of my first mistakes just as I started the release -- everything worked fine until I tried an alternative EQ.  Different EQs cause different EQ array contents -- but the array went out of bounds.   Sometimes it worked, sometimes not.   Yes. I still make mistakes like that, but the array def is perhaps 100 lines away from using the array, and for max perf, I do everything barebones.   This mistake was fixed.   The next was a transcription error.  Now, there just might be some feedback about EQ.   The updates will be stabilized at +4Hrs and every 24Hrs after that.  (That is: 9:00AM each morning.)   I am organizing like this simply to keep REALLY interested people sane.   The first bugfix might be this morning.

 

* For an update, ONLY the da-avx, da-win, da-core (one) needs to be copied to the original location.  Everything else is being kept stable.  

 

The EQ is a matter of opinion, and I am getting different opinions.  Currently, there is feedback about the bass being a bit too strong -- for me, even reasonable bass is irritating.   Even pros vary in their opinons.  Each of us are different, but I'll have the matter resolved before the update time at +4Hrs.   If there are any changes, please refer to the dist site below.  I have to check with 2-3 people before a good answer is gotten.

 

The version numbers is currently: V2.1.1E and will increase letter by letter.   There will be NO usage changes when the decoder stays as V2.1.X.   A change in the 2nd number field means that there could be a usage change..   The letters are bugfixes and minor audio improvements.   The third number field is for profound improvements/changes, or if I run out of letters.

 

NO NEW FEATURES ARE CURRENTLY PLANNED.   No new algorithms or capabilites are planned, but there are some already built-in things that I haven't discussed much and will document better as time goes on.

 

The same is also true for the DolbyA stuff.   DolbyA decoding is infinitely more mature, and previously had growth pains also.  FA is 4-10X more complex, depending on how you measure it.

 

Updates will be at a standard time (when needed):  daily at or before 9:00AM USA EST time, and will be available at the locations below.   No need to check other times.  There will be weeks when there are no changes, but might be daily for 2-3days after a release.   You CAN check the site early, possibly get a one-up on the changes.

 

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

I know that this is different than my claim of being finished, and in a relative sense, the code is finished.   Maybe table entry type things, or bugfixes migiht happen.

I'll make a post just as I do an update -- the contents will be obvious for those who quickly scan things.

 

Unless problems need to be public, and if you have troubles -- PM is probably best to keep traffic down.   Some problems are best discussed publically, like 'need more bass' or 'need less bass'.

 

John

 

Minor update:

V2.1.1F

https://www.dropbox.com/sh/1srzzih0qoi1k4l/AAAMNIQ47AzBe1TubxJutJADa?dl=0

 

Got complaints about too much bass.   If you need more bass with the new table entires, then use --fa=M.   Both --fa and --fa=M should cover everything except older recordings, then use --fa=L.   If you need more bass, also use the decoder option --pe75=2,1.5 and/or --pe90=2,1.5, etc.  The resulting decodes witihout USER EQ are not necessarily air-worthy, but there are plenty of EQ options in the decoder, specifically the --pe commands!!!

 

If I try to do too much EQ with the 1st order standard EQ for decoding, the results can be a real mess.   The 2nd order as consumers typically use work best for 'tone control'!!!

 

The next update, if helpful, will be 24Hrs from now.

 

* About the LF issues.   Apparently, on some recordings, because of the FA compression bass boost, they are doing a bass cut.   It appears to be 2nd order because I cannot accurately compensate for it with 1st order.   Here is an example EQ -- works well on some recordings:

 

 --pe120=2,1.5 --pe90=2,1.5 --pe75=2,1.5 --pe45=2,1.5

 

THIS IS NOT A HACK -- but likely undoing something that was done before FA encoding.   There simply is NOT a match on some recordings of proper, time correct EQ vs response balance.

 

 

 

 

 

 

 

Link to comment

Remember -- use the 'V2.1.1F' version now.

 

About the EQ issue.  Once we get a good balance, then there appears to be another problem somewhere else.   When I have seen this behavior, it usually means that there is something fundamentally missing in the EQ.

 

I am now convinced that there is often a -dB 2nd order EQ at approx 80Hz to partially compensate for the DolbyA compression bass boost.  That compression is the fundamental reason for the 'boosted' bass on FA materials.   Of course, they dont' want an extreme boost, so sometimes add a small 2nd order EQ before the encoding process.

 

This is most likely the reason for the EQ challenge/problems that I have had.   I know what to listen for when the responses are correct, correct timing has a magic 'sound'.   The apparent timing isn't so screwed up by 2nd order (consumer) EQ.

 

There is a consistent bass cut (or bass boost after deocding) at about 80Hz and a Q of 1.17.  (Geesh, how do I know that?)   A new subsubmode will be added for tomorrows release (yep, my promise of no new development is put aside, but this is REALLY a bug.)

 

I don't know whether to make the EQ a default setting -- default on?  default how many dB?   how many dB when enabled?

 

If you want to do the EQ yourself after decoding, because the sound is too 'thin' -- my previous suggestion was approximate, but WRONG.   Here is the correct thing to do:

 

--pe90=1,0.75,1.17 --pe90=1,0.75,1.17

 (LF 2nd order shelf, 0.75dB at 90Hz/Q=1.17...   Need two of them)

 

Other EQ doesn't work as well -- seems to be the magic number.

Note that it takes TWO EQ at the same frequency.   It might work better with more, and that is what I need to experiment with.   The new submode will be a letter, either use the letter to disable it, or use the letter to enable it.   What should the default be?   Gotta figure that out, and will be fixed tomorrow!!!

 

John

 

Link to comment
4 hours ago, John Dyson said:

Remember -- use the 'V2.1.1F' version now.

 

About the EQ issue.  Once we get a good balance, then there appears to be another problem somewhere else.   When I have seen this behavior, it usually means that there is something fundamentally missing in the EQ.

 

I am now convinced that there is often a -dB 2nd order EQ at approx 80Hz to partially compensate for the DolbyA compression bass boost.  That compression is the fundamental reason for the 'boosted' bass on FA materials.   Of course, they dont' want an extreme boost, so sometimes add a small 2nd order EQ before the encoding process.

 

This is most likely the reason for the EQ challenge/problems that I have had.   I know what to listen for when the responses are correct, correct timing has a magic 'sound'.   The apparent timing isn't so screwed up by 2nd order (consumer) EQ.

 

There is a consistent bass cut (or bass boost after deocding) at about 80Hz and a Q of 1.17.  (Geesh, how do I know that?)   A new subsubmode will be added for tomorrows release (yep, my promise of no new development is put aside, but this is REALLY a bug.)

 

I don't know whether to make the EQ a default setting -- default on?  default how many dB?   how many dB when enabled?

 

If you want to do the EQ yourself after decoding, because the sound is too 'thin' -- my previous suggestion was approximate, but WRONG.   Here is the correct thing to do:

 

--pe90=1,0.75,1.17 --pe90=1,0.75,1.17

 (LF 2nd order shelf, 0.75dB at 90Hz/Q=1.17...   Need two of them)

 

Other EQ doesn't work as well -- seems to be the magic number.

Note that it takes TWO EQ at the same frequency.   It might work better with more, and that is what I need to experiment with.   The new submode will be a letter, either use the letter to disable it, or use the letter to enable it.   What should the default be?   Gotta figure that out, and will be fixed tomorrow!!!

 

John

 

i got to hand it to you, John.  Seems like you will have finally solved the problem, all through brute force. You are a persistent man.  My hats off to you.

mQa is dead!

Link to comment

More good news -- fixed the LF!!!   This fixes a LOT of things -- this reverse engineering is VERY tedious.

Even though the previous version was perfect in most ways -- this version gets rid of the final major issue.

 

YOU NEED NO SUBMODES NOW!!!   Just FA.

 

(I'll be answering some of the questions in another post.   Part of the problems are misunderstanding what a spectogram measures, and the other is that there really has been a problem with the LF -- actually between 1k -> 10Hz that the 1st order EQ couldn't deal with. LF/MF problems can be misdiagnosed as problems in other bands.)

 

Here are the released versions and usage help information:

 

https://www.dropbox.com/s/kbdyz4hv088f95v/da-release2.1.2E-win-2021-03-07.zip?dl=0

https://www.dropbox.com/s/2uf1vrlrot0v5ih/da-V2.1.2E-Linux.txz?dl=0

https://www.dropbox.com/s/7bgzpfc70dt4dlm/StartUsage-V2.1.2E.pdf?dl=0

 

Shortend group of snippets -- the decoder is really worth using now (absolutely no need for --fa submodes, just --fa itself.)

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

 

Working on Dire Straits and Supertramp again -- been busy with the rework and LF correction.

 

Even though most EQ is 1st order, there needs to be a dash of 2nd order EQ at 80Hz to compensate for the approx 3.0dB threshold difference on the DolbyA LF compression.   By allowing myself to use the 2nd order 80Hz EQ, then I could revert to the totally standard, and hoped for EQ between 1kHz and 10Hz.   Even the 15Hz and 10Hz EQ are totally critical for proper sound.  (I knew this all along but was frustrated that I couldn't 'get there' by 1st order only.)

 

Other minor improvements:

Better choice of Hilbert windows.

Better locality of reference for some of the FIR filter coefficients.   Will be further improving in the future.  (slight performance improvements.)

 

Link to comment
17 hours ago, Night Rain said:

The left is "decoded" and the right is the original. As you can imagine it sounds completely different and not in a good way.

koskokdopkspok.jpg

 

First, I have known about the LF problem and mentioned it OVER AND OVER AGAIN, asking for ideas and help.   Happily, I figured it out myself.   The rest of the decoder has truly been perfect (beyond perfect in the sense of what a DolbyA can do) for months.

 

A spectrum measurement needs to be taken with a HUGE grain of salt.  Long average interval, and it will totally miss the resuls of the super-high speed expansion.   If anything, a peak reading spectrum measurement that sees information in the 5msec and faster region would be most helpful.   Even then, interpreting the results would require a bit of a learningcurve.

 

 

Link to comment
18 hours ago, lucretius said:

i got to hand it to you, John.  Seems like you will have finally solved the problem, all through brute force. You are a persistent man.  My hats off to you.

Thanks -- they used to give me the most difficult projects at Bell Labs because I didn't give up, and use every resource available to me.  Of course, at the Labs, I did have access to some of the biggest experts in the country (perhaps world.)   Standing mostly alone (thankfully with encouraging help from time to time) has been challenging.   When I have gotten encouraging help and comments -- it showed kindness and faith.   Also, it showed there are people who realize that real progress had been made, even though there was more needed.   I had to take a positive outlook, and would NEVER have stopped -- and HAPPILY, THIS VERSION *IS* a showing of success, whether or not totally perfect.

 

This last LF problem is one of my long, random walks, finally finding my way.   I kept trying to morph what should have been a simple MF->LF st order EQ into something that sounded right at the low bass.   So, I had unknowingly been damaging the curve to get a reasonable lower bass.  2nd order was the answer -- but is the ONLY place in the decoder where 2nd order was needed!!!

 

By recognzing that a 0.75+0.75dB 2nd order LF shelving boost at 80Hz/Q=1.0 allowed using the  rest of the curve that was correct.   I am pretty thorough -- I ran all kinds of test, trying to figure out if a single 1.5dB boost, a staggered boost, or using Q=0.50, 0.7071, 0.8409, 1.0 or even 1.17 (the 80Hz DolbyA band splitting filter).   The best result was Q=1.0.

 

Also, the lowest freq EQ (with a bunch of -6dB 1st order EQ) had to include 25Hz, 20Hz, 12.5Hz and 10Hz.   Also, the 25Hz isn't really 25Hz, but has an anti-distortion pattern added to it (it is -6dB at 35Hz, +6dB at 20Hz, -6dB at 5Hz.)   I doubt that most people would be focused enough to work out these details (it is a blessing and curse for me.)   The HF EQ is even more intricate, but was also less dependent on my sense response balance in my hearing.   The HF band was more about distortion cancellation and response errors are more profound to me at >3kHz.)

 

The MF range between 1kHz and 3kHz was another tricky game.   It requires three pairs of shelving filters.

 

If all of these aren't done correctly, the sound will have all kinds of weird distortions.   I am still listening for troubles, and WILL hear them if a cogent explanation is provided (I mean real help instead of 'ugly sound' or some other condemnation!!!)

 

Thanks again -- and I sure hope that this new release:  V2.1.2E really fills the bill!!!

 

 

Link to comment
2 minutes ago, John Dyson said:

Thanks -- they used to give me the most difficult projects at Bell Labs because I didn't give up, and use every resource available to me.  Of course, at the Labs, I did have access to some of the biggest experts in the country (perhaps world.)   Standing alone (thankfully with encouraging help from time to time) has been challenging.

 

This last LF problem is one of my long, random walks, finally finding my way.   I kept trying to morph what should have been a simple MF->LF EQ into something that sounded right at the low bass.   So, I had unknowingly been damaging the curve to get a reasonable lower bass.

 

By recognzing that a 0.75+0.75dB 2nd order LF shelving boost at 80Hz/Q=1.0 allowed using the  rest of the curve that seemed correct.   I am pretty thorough -- I ran all kinds of test, trying to figure out if a single 1.5dB boost, a staggered boost, or using Q=0.50, 0.7071, 0.8409, 1.0 or even 1.17 (the 80Hz DolbyA band splitting filter).   The best result was Q=1.0.

 

Also, the lowest freq EQ (with a bunch of -6dB 1st order EQ) had to include 25Hz, 20Hz, 12.5Hz and 10Hz.   Also, the 25Hz isn't really 25Hz, but has an anti-distortion pattern added to it (it is -6dB at 35Hz, +6dB at 20Hz, -6dB at 5Hz.)   I doubt that most people would be focused enough to work out these details (it is a blessing and curse for me.)   The HF EQ is even more intricate, but was also less dependent on my sense response balance in my hearing.   The HF band was more about distortion cancellation and response errors are more profound to me at >3kHz.)

 

The MF range between 1kHz and 3kHz was another tricky game.   It requires three pairs of shelving filters.

 

If all of these aren't done correctly, the sound will have all kinds of weird distortions.   I am still listening for troubles, and WILL hear them if a cogent explanation is provided (I mean real help instead of 'ugly sound' or some other condemnation!!!)

 

Thanks again -- and I sure hope that this new release:  V2.1.2E really fills the bill!!!

 

 

 

Is the truth a condemnation? Okay here goes, the sound has been turned into what one might expect from an AM radio. There's no depth, the midrange is muffld and the highs are like white noise instead of a cymbal tap. I cannot find one area that is improved, indeed it's like putting a blanket over the speakers. There's also a terrible tightening of the bass like the drums were misstuned.

 Tracks used were Jennifer Warnes - The Hunter, Holst/Walter Susskind - Jupiter and Gordon Lightfoot - Sit Down Young Stranger.

Link to comment
17 hours ago, Night Rain said:

The top is the original and the bottom is decoded.

guggyguy.jpg

Decoded should have more dynamics, as it does.   This is one reason why an averaged spectrum isn't as useful in all of the ways that one might initially think.   A kind of 3d spectrum might be more helpful.   A spectogram is 'interesting', but also requires intepretation.

 

An important item of note:   when one might see signal varations above 20kHz on the output of the decoder, that is NOT distortion.   The decoder produces so little distortion, I dobut that it could be easily measured without a gain up.   The 'signal' above 20kHz is mostly noise modulation, and the noise can appear even though not obvious to begin with.   (This is one reason for the false idea that there is signal above 20kHz in a lot of pop material.)

 

 

Link to comment
1 minute ago, Night Rain said:

 

Is the truth a condemnation? Okay here goes, the sound has been turned into what one might expect from an AM radio. There's no depth, the midrange is muffld and the highs are like white noise instead of a cymbal tap. I cannot find one area that is improved, indeed it's like putting a blanket over the speakers. There's also a terrible tightening of the bass like the drums were misstuned.

 Tracks used were Jennifer Warnes - The Hunter, Holst/Walter Susskind - Jupiter and Gordon Lightfoot - Sit Down Young Stranger.

I don't hear the 'white noise' on the decoded versions -- I hear a correlated signal instead of a time smeared FA signal.   You must compare the current version anyway to be able to make cogent comments.   Peoples brains who are accomodated to FA might actually be resmearing a clean signal.   (Keep an open mind about this -- I used to think that my hearing was 'absolute' also.)

 

I cannot show you the upcoming SACD Love Over Gold publically, but that is more of my kind of test material.   I also use 'Crime', 'Breakfast', 'Quiet' and 'Crisis' on Supetramp.  ABBA is interesting also -- because of the extreme time smearing, and the re-correlation of the signal by the decoder.

 

Those who are accomdated to FA might have some extra processing going on that will actually distort the clean signal that they are listening to.   This 'accomdation' happens just like adapting to smells and colors.

 

Whenver listening to the decoder, stay away from the FA listening for at least a few hours, if not a few days.   I know what to expect now, but also have real experience on these matters.

 

 

Link to comment
2 minutes ago, John Dyson said:

I don't hear the 'white noise' on the decoded versions -- I hear a correlated signal instead of a time smeared FA signal.   You must compare the current version anyway to be able to make cogent comments.   Peoples brains who are accomodated to FA might actually be resmearing a clean signal.   (Keep an open mind about this -- I used to think that my hearing was 'absolute' also.)

 

I cannot show you the upcoming SACD Love Over Gold publically, but that is more of my kind of test material.   I also use 'Crime', 'Breakfast', 'Quiet' and 'Crisis' on Supetramp.  ABBA is interesting also -- because of the extreme time smearing, and the re-correlation of the signal by the decoder.

 

Those who are accomdated to FA might have some extra processing going on that will actually distort the clean signal that they are listening to.   This 'accomdation' happens just like adapting to smells and colors.

 

Whenver listening to the decoder, stay away from the FA listening for at least a few hours, if not a few days.   I know what to expect now, but also have real experience on these matters.

 

 

Just sharing my opinion. It's like the life has been sucked out of the recordings and instead of a 3 dimensional reproduction it has been completely flattened. Have been following this project for about a year.

Link to comment
10 minutes ago, Night Rain said:

Just sharing my opinion. It's like the life has been sucked out of the recordings and instead of a 3 dimensional reproduction it has been completely flattened. Have been following this project for about a year.

Yea, that is how I think of the highly compressed consumer FA recordings.   The dynamics are wierdly compressed -- sure bringing up otherwise unaudible details, but the compression is evil and doesn't sound like the mixdown.   Can you even hear a coherent stereo image on a consumer recording?   Didn't you notice that there was a weak 45degrees on them.   You get left/right, forward, but the sides are all messedup.   The scrambled timing on FA is attrocious, all of the minimium phase filters with varying gains -- create varying time.   Consumer recordings are just ugly stuff.

 

If you like the strange, scrambled, segmented compression FA, that is cool.   I walked away from HiFi back in 1990 because consumer recordings sounded so bad to me.    Since CDs weren't going to solve the problem -- I was finished.   Of course, I used to do REAL recordings with REAL stereo -- nothing fake.

 

John

Link to comment

About fuzz or lack of definintion -- oh, come on now -- listen to this.   This is very typical of the results.  Sure aint fuzzy, and you can hear the details in the percussion, not just a compressed FA , time smeared blob...  Note even the ambience is more natural,  isn't amplified/compressed like in FA, and essentially inaudible hiss.   Many FA recordings will tend to be hissy, especially done on analog tape, but even those done on digital tape also have some hiss.

 

With the FA process, digital isn't really digital, but has been passed through an analog processor.

 

FA is just YUCK!!!

 

03 - DIRE STRAITS - INDUSTRIAL DISEASE-snippet.flac

Link to comment
Just now, John Dyson said:

About fuzz or lack of definintion -- oh, come on now -- listen to this.   This is very typical of the results.  Sure aint fuzzy, and you can hear the details in the percussion, not just a compressed FA , time smeared blob...  Note even the ambience is more natural,  isn't amplified/compressed like in FA, and essentially inaudible hiss.   Many FA recordings will tend to be hissy, especially done on analog tape, but even those done on digital tape also have some hiss.

 

With the FA process, digital isn't really digital, but has been passed through an analog processor.

 

FA is just YUCK!!!

 

03 - DIRE STRAITS - INDUSTRIAL DISEASE-snippet.flac 4.46 MB · 0 downloads

You asked for opinions and you got one. I didn't kick your dog.

Link to comment
Just now, Night Rain said:

You asked for opinions and you got one. I didn't kick your dog.

General comments of the general terms like 'muddiness' , not 3d or other things that aren't specific and oriented towards fixing the problem, just not kind.   BTW -- compressed hash does give a sense of detail, and I do suspect that is what people sometimes are missing.

 

It is obvious that FA is a problem, and the FA compression is  a deficit -- and the FA decoder is really working  produces results much closer to a raw recording.

 

I tell people who prefer FA -- just talk into a good quality mic, through a board, no processing.   You will NOT get FA sound.   Even if you use a good quality compressor, the sound is NOT like most consumer recordings.

 

Sometimes people REALLY like the music.   When people REALLY like the music, they'll even listen to AM radio.

 

 

 

 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...