'FeralA' decoder -- free-to-use

[jabbr]

I'm going for broke right now with Joni Mitchell, whose vocal recordings have some notorious difficultues:

 

So far from Studio Albums(2012): Clouds 

settings: --floatout --cddemphasis --fcsequence="6,-44.5,fcx=G" --dh=off --fz

[John Dyson]

30 minutes ago, jabbr said:

I'm going for broke right now with Joni Mitchell, whose vocal recordings have some notorious difficultues:

 

So far from Studio Albums(2012): Clouds 

settings: --floatout --cddemphasis --fcsequence="6,-44.5,fcx=G" --dh=off --fz

I don't have a copy of that recording, but at first -- maybe try starting with 4 layers instead of 6.   Also, not using --fz will speed things up while you are tweaking the settings.   Then, add --fz when you are happy (--fz doesnt' change anything, for FA, just does the gain control incrementally while filtering out the lower frequency IMD.)   If you enable the full 'monty' by doing --dh=off & --fz, then it does some sideband demodulation games to clean up the sound -- but the sidebands are splattered all over the place with FA anyway, so there is sometimes only slight benefit of turning on all of the MD removal by --dh=off (disable Hilbert is disabled.  It only disables the hilbert processing on the anti-MD, the detectors keep the Hilbert detectors all of the time.)   The 'full monty' strongly helps decoding raw DolbyA material though.   I have sometimes seen big improvement by turning on the --dh=off switch on FA, but sometimes where it is really needed (ABBA SuperTrouper), its results are mixed.

 

Vocals can be a real pain in the b*tt!!!   After all of this time, I have just barely figured out 'The Carpenters', and 'ABBA' -- problems because of vocal sibilance and vocal chorus.

 

ADD-ON:  Since I don't know if you are having troubles or not -- assumed that you were.  However, if you are getting good results, then let us all know!!!

 

Later on today, I'll check with my industry sources for something to try.

 

John

 

[jabbr]

48 minutes ago, John Dyson said:

Vocals can be a real pain in the b*tt!!!   After all of this time, I have just barely figured out 'The Carpenters', and 'ABBA' -- problems because of vocal sibilance and vocal chorus.

 

ADD-ON:  Since I don't know if you are having troubles or not -- assumed that you were.  However, if you are getting good results, then let us all know!!!

 

I'm just starting on this one, also court and spark. The initial settings aren't bad!

[John Dyson]

4 hours ago, jabbr said:

 

I'm just starting on this one, also court and spark. The initial settings aren't bad!

Here is a caveat -- the g, vs Gg vs G vs no-g at all, this is a slippery slope.   I sugget being careful about the tradeoff between the 'body' of the vocal vs brightness vs. sibilance.   Once you get the sound 'just right', but perhaps a little bit of sibilance leaks through, then consider clamping down a little on the "resonant 's'" sound.   Depending on signal level, but if it is full scale, then maybe try --as=8, and see if the sibilance is SLIGHTLY suppressed, then push down a little (down as low as 4 or so) to see if the 's' is better controlled.

Some recordings are done VERY WELL, and have just the right 'S' sound, but esp older recordings, they 'sweeten' the sibilance too much.

 

The anti-sibilance switch '--as' does things that cannot be done by using a notch filter.  However, if you want to manually control sibilance, a fixed equalizer notch of approx Q=1,2,3,4 at 6kHz might help.

 

The DHNRDS has the '--as' switch which automatically finds where the notch should be and pushes down a little on an overly strong 'S'.   i designed the anti-sibilance to sound relatively natural, but too much (too low --as value) can sound strange.   It can also dynamic range compress the 5kHz to 10kHz frequency range if misused.   The anti-sibilance is best used sparingly, but can really help.

 

John

 

[jabbr]

Yeah I just ran --floatout --cddemphasis --fcsequence="6,-44.5,fcx" --fz and its a little brighter and interestingly DR when from 7 to 8

[John Dyson]

14 hours ago, Confused said:

 

I must admit it has surprised me also how little interest this thread has gained, considering some of the other crazy stuff being tried elsewhere.  (no offence to anyone, I like a bit of crazy stuff myself)   

 

Do you think it might help if someone could collate a few examples of successful decodes?  At the moments there are dozens of versions and examples of ABBA and Carpenters tracks, but not much else.  Personally I would be interested in a small selection of tracks or short samples covering other artists and genres.

 

OK - I could try it myself, but being perfectly honest I am as busy as hell at the moment, and I think it would take me too much time for me to try to get it running and usable, mindful that it has been about 30 years since I have needed to do much with a command line software or similar.  (maybe a simple starters guide "for dummies" might help too, just to get the software up and running?)

 

But if there were a few convincing samples to try, it might be enough to convince me and a few others to give it the time it maybe deserves.

 

Just a thought.

 

As it happens, I have just been listening to the Cure's "Staring at the Sea" compilation Album from a late 80's CD rip.  Sound quality seems pretty good to me, but there are few tracks with very noticeable tape hiss.  (In particular on the second track, 10:15 Saturday Night, much hiss at the end of A Forest too)  From what I have read here, audible tape hiss is one of the more obvious "FerralA" tells.  As I said, the overall sound quality of the album seems pretty good to me, certainly in comparison to similar recordings of the era, if this one could be improved it would certainly help to convince me.

 

 

I finally got a copy of a few selections from the Cure disk.   Did some reviews...

 

Thank you so much for finding something that I have not found in a few years of searching -- YOU WIN THE PRIZE.

The Cure disk is NOT FA.   The hiss on the disk is natural tape hiss, not enhanced hiss from the FA compression.

 

I tried decoding it, even though my prejudiced opinion would be that it would fail.   Indeed, all of the manifestations of expander

surging.   This is a good *commercial* example that I can use now -- I  only have 'master tape' type material to demo

non-FA that I can cannot publically release.

 

Here is a failed decode of the 2nd cut, Saturday Night.  (50 seconds of unlistenable recording.)

The only good news -- no hiss :-).

 

John

 

 

TheCure-SaturdayNight.mp3

[John Dyson]

Right now, the decoder is VERY VERY deficient in documentation.   I am spending much of my time this week producing some useful documentation, esp for beginning users.  Most likely, on Fri late afternoon, there should be improved documentation, with an updated/complete exhaustive man page (converted to pdf for reading), and a better beginners guide.   Also, I hope to have a more general 'user guide for success' type thing.   Using the decoder now is actually reasonably easy for people who are comfortable while using Linux/Windows command lines, and a simplified version new set of options might fit into the limiting scope of a GUI.   The spectre of a GUI appears to be more and more practical for both DA and FA decoding.

 

Also, on Friday (or Saturday), there will be a minor quality improvement release, targeted to improving one of the auxiliary features.  There are no more features to be added or substantively changed.   The expected change is a slight retuning of the anti-sibilance.   Even though the current version works pretty well, it is a little too likely for the anti-sibilance to add a 'hitch' to a vocal.   I have made the slight change to better avoid creating the 'hitch'.   (The change is not really 'tuning' per se, but is slightly changing the behavior of the dynamic notches.)   Associated with the changes, there is a performance improvement in the anti-sibilance code, so that more fine control can be provided without penalizing the speed.    This improvement already exists and is part of my ongoing decoding testing.

 

John

 

[Night Rain]

Thanks for the effort.

[Confused]

18 hours ago, John Dyson said:

I finally got a copy of a few selections from the Cure disk.   Did some reviews...

 

Thank you so much for finding something that I have not found in a few years of searching -- YOU WIN THE PRIZE.

The Cure disk is NOT FA.   The hiss on the disk is natural tape hiss, not enhanced hiss from the FA compression.

 

I tried decoding it, even though my prejudiced opinion would be that it would fail.   Indeed, all of the manifestations of expander

surging.   This is a good *commercial* example that I can use now -- I  only have 'master tape' type material to demo

non-FA that I can cannot publically release.

 

Here is a failed decode of the 2nd cut, Saturday Night.  (50 seconds of unlistenable recording.)

The only good news -- no hiss :-).

 

John

 

 

TheCure-SaturdayNight.mp3 1.15 MB · 12 downloads

So I win the prize for being useless at detecting FeralA.  The kudos is overwhelming!  I presume I win a FeralA decoder?🙂

 

Joking apart, I think there is an interesting observation here.  I had started out thinking that audible hiss is a FerralA "tell".  The Cure album has much hiss, but I also note that it has a very respectable DR score (DR 10 to 16, depending on track, 13 average, for the 80's CD)  So now I am thinking that a more accurate FeralA "tell" would be audible hiss, combined with a low DR?

 

Does this make sense?  I shall keep hunting!

[John Dyson]

2 hours ago, Confused said:

So I win the prize for being useless at detecting FeralA.  The kudos is overwhelming!  I presume I win a FeralA decoder?🙂

 

Joking apart, I think there is an interesting observation here.  I had started out thinking that audible hiss is a FerralA "tell".  The Cure album has much hiss, but I also note that it has a very respectable DR score (DR 10 to 16, depending on track, 13 average, for the 80's CD)  So now I am thinking that a more accurate FeralA "tell" would be audible hiss, combined with a low DR?

 

Does this make sense?  I shall keep hunting!

 

===

I really hope this helps -- learning to reliably detect/determine FA has taken me a fairly long time, and I still do make mistakes!!!

===

 

Probably a little lower DR is an indicator.  FA compression is done 'inside' of a recording, so DR sometimes doesnt' change much, and then other times the difference can be profound.

 

The confusing factor about The Cure -- FA tends to have louder 'hiss' than normal tape hiss levels.   The hiss on 'The Cure' was a little stronger than usual for normal tapes.

 

The 'woody' sound of FA is the biggest tell for me other than the compressed sound itself  -- but many people might prefer the satisfaction of a stronger lower midrange, sometimes (not always) 'nasal' sound.   The sound of the compression is another tell - cymbals and other percussion become 'squishy' or ''swish' instead of 'tick'.

 

I really suggest listening to the decoded and undecoded copy below -- it will be very educational, and also show why I have problems using words describing the difference.

 

 

My own history was that AT FIRST, I couldn't reliably tell if something was FA or not.   Even though I hadn't listened to music all that much for 30yrs, my brain was still somewhat programmed to accept the FA sound as normal.   Ths 'accommodation' is probably the reason why industry has gotten by with the deception for so very many years.  I actually gave up on the HiFi hobby in about 1990 because all of my CDs sounded sucky, and I didn't know why.  I figured that I would only be disappointed, no matter what equipment I would buy -- so my hobby dwindled.

 

====================

 

Maybe the best way to describe the difference is to demonstrate an undecoded and decoded copy of the same thing.

in fact, this decoded copy was decoded directly  from the undecoded copy.  YOU WILL NOTICE THAT THE DECODED IS MUCH MORE QUIET.   The bad thing is that there is MUCH MUCH less 'sparkle' in a decoded recording, but this is especially true of old stuff like the Bread example.  The undecoded 'sparkle' gives reassurance about the highs -- but the indication is over-emphasized, distorting the impression.   The material might not have all that much HF to begin with, but the FA would boost the highs by 10-20dB at least, when doing the FA encoding, IT APPEARS THAT THERE IS A LOSS OF HIGHS.

 

Bottom line, if the sparkly highs are close to the same level as the tape hiss -- so, if there are highs in a CD that appear close to the hiss level, then they shouldn't be very audible.    However, use of a tone control is VERY effective, because the resulting quality is MUCH higher.   You have the material to do some of your own mastering,  and adding careful compression to a decoded result isn't always bad.  When doing  a decode, the material might not have much compression beyond the mix!!!   Sometimes the result might be too 'dry' sounding.

 

* Also, it is a little frustrating that vocals on old recordings fairly often have uncontrolled sibilance...

 

It is SO difficult and tricky to verbally describe the difference.   Here are a decoded  recording and the undecoded version distributed on CD.
You will probably be able to describe the difference yourself in your own terms, than I can.

The real problem is that none of the 'tells' are absolute...

 

When I do a casual comparison -- the very notable facts is that the decoded version has stronger dynamics.  The bass on FA is distorted, so is more obvious that it almost says -- I am going to boom, and here it is.   The decoded version just makes fairly clean  bass.

 

This is example was done quickly -- I can hear a minor defect in the sibilance, but otherwise the comparison should be valid.

Using mp3 here is valid -- the difference is so profound that the difference is communicated pretty well.

I simply do not want to waste a lot of the very kind provider's disk space for this ephemeral comparison.

 

John

 

 

 

 

 

 

 

 

Bread-decoded.mp3 Bread-undecoded.mp3

[Wilderness]

Will this be offered in a version for Mac at some point?

 

Thanks.

[John Dyson]

23 minutes ago, Wilderness said:

Will this be offered in a version for Mac at some point?

 

Thanks.

Mac does have an emulator for Windows or Linux?

 

A port to Mac would be a significant amount of work, and I am the only programmer on the project.   In fact, other than my pro project partner for DA decoding,  and people who help me around here, almost everything is done by me.  I dont have the resources to do it -- and I do feel regretful that I cannot do it.

 

Some day (year or two), I will free the source, and also willing to help plug-in developers to do a DolbyA decoder or FeralA decoder (FeralA is a complex combination of DolbyA encoding and EQ.)   The technology is challenging to say the least, the program is very detailed, subtle and is a lot of software to support!!!

 

I do regret my need to answer the way that I have.

John

 

[Skip Pack]

In my most recent effort using V1.6.2E, working on Albert Collins -- Live in Japan, I used the following command line to get a very significant improvement:

 

da-win --input=$outfile --overwrite --output=$newout --info=1 --fcs="2,-33,fcf=G" --as=5 --fz

 

I believe the '2' setting for the number of passes results in 3 passes being executed. I tried seven variants, and the larger number of passes result in a super-cleaned result that was bland an uninvolving. I reduced the number of passes and used what the document suggests is a milder initiator to reach this set of command parameters. 

 

What I missed on the over-cooked results was the bounce and vitality in the bass and lower midrange. What I wanted to diminish in the original was mid-to-upper-mid congestion that included some lower HF unpleasentness as well. Would snippets and corresponding command lines be of use to you?

 

[John Dyson]

1 hour ago, Skip Pack said:

In my most recent effort using V1.6.2E, working on Albert Collins -- Live in Japan, I used the following command line to get a very significant improvement:

 

da-win --input=$outfile --overwrite --output=$newout --info=1 --fcs="2,-33,fcf=G" --as=5 --fz

 

I believe the '2' setting for the number of passes results in 3 passes being executed. I tried seven variants, and the larger number of passes result in a super-cleaned result that was bland an uninvolving. I reduced the number of passes and used what the document suggests is a milder initiator to reach this set of command parameters. 

 

What I missed on the over-cooked results was the bounce and vitality in the bass and lower midrange. What I wanted to diminish in the original was mid-to-upper-mid congestion that included some lower HF unpleasentness as well. Would snippets and corresponding command lines be of use to you?

 

First -- the V1.6.4 series is absolutely  better.   We found that the EQ is most definitely based on the CD emphaiss frequencies, and not the more DolbyA natural EQ of 3kHz/9kHz. (3.18kHz/10.5kHz, 3.18kHz/18kHz, 3.18kHz/21.5kHz, 3.18kHz/forever, instead of 3kHz/9kHz, 3kHz/18kHz, 3kHz/21.5kHz and 3kHz/forever.)   The very *sllght* difference of 180 Hz is like the differnece of night and day.

 

But, if you try to run more layers, it is better to decode with a lower calibration starting point.   After a lot of testing and design updates (and infinitely better accuracy), a starting point of approx -4X dB is usually the best.   (e.g. -44.5, -46 or -49 usually, but -43 is still possible.)   Then, the equivalent of your two layers, perhaps three when starting at -4XdB instead of -3XdB.

 

There is a variation in the way that FA is used.   Usually, it seems to start with like -44.5dB, then each layer is always 10dB higher up to the -12 to -22dB range.  So, the typical 4 layer decode, which produces something close to a high quality commercial recording, would be 4 layers at -44.5dB.  This means that the virtual DA units are set at calibration levels of -44.5, -34.5, -24.5 and -14.5dB.   Some recordings don't stop at the -14.5dB level, but instead work best if you stop at -24.5dB level insteadd.

 

It appears that you might have that kind of recording whose highest calibration might be in the -20dB range instead of -10dB range, where the highest calibration should be -24.5 instead of -14.5dB.   This isn't very common, but it has happend in about 5% of my digital recordings.

 

So, maybe for a little better result, with less congestion in the low level highs, maybe something like --fcs="3, -43, fcf=G" might be better.  Recently,  I have found that most recordings prefer the 3.18kHz starting point instead of the 3kHz starting point.   This means, instead of using the --fcd series (which is --fcd, --fce, --fcf,  --fcc, and then with the G modifiers), use the new --fcx series (which is 3.18kHz instead of 3kHz).   When you use the --fcx series, also use the -cdd switch.  A possible (cleaner) equivlanet  using the V1.6.4series, try this:  --cdd --fcx="3, -44.5, fcx=g", or maybe even try 4 layers.

 

Let me know if you want some information on the examples -- send me a private message.   The decoder settings reside in the metadata of the demo recordings.

 

As of the V1.6.4 series, the decoder is essentially complete and fully functioning (and sounds REALLY GOOD.)   I recently did a marathon effort, disconnecting for a few days, and workd out practically all of the bugs.   I can discuss certain things in more detail in PM.   Questions about quality of decoding are now totally and 100% gone.  

 

John

 

[John Dyson]

I think that I might have forgotten to announce V1.6.4A.   I am working on improved docs right now, but the Using document will get you started wiith the new commands.

 

When chatting on this forum, I had assumed that I had done the announcement...   It has been in testing by a small group of people, and I simply forgot to do the announcement -- SORRY ABOUT FORGETTING!

 

Here it is:

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

[John Dyson]

Heads up -- there is a new, much more clean sounding decoder ready.  V1.6.5A (I left the old version on the repo also -- get the right one!!!)

 

There is a short usage guide, and a Linux version also.

I had to reschedule the more complete manual sets because of personal matters, but the new manuals are coming next week instead.

 

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

 

No real time to describe things, but the Usage document explains the differences.   Basically, the EQ is made more useful and number of EQ per frequency has been increased.

 

The decoder itself now has much more precise calculation for the HF0/HF1 split (people who know much about DolbyA recognize that there are two overlapping compressors for the high frequencies, and it is a MAJOR TRICK to do the split correctly.)

 

Other, somewhat profound quality improvements also.

 

John

 

 

[John Dyson]

New, bugfix release -- V1.6.5B.

 

Everyone -- I made a 0.5dB mistake, and it really makes a big difference!!!   There is a mistaken 0.5dB boost between about 10kHz to 20kHz, and it sucks.   My hearing started missing the HF detail, and I had been adjusting the decoder, forgetting that I had the wrong setting.

 

The 0.5dB mistake doesn't have any effect till about -10dB, then it starts burning your hearing!!!

 

The new, replacement version is in the usual spot:

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

 

Sorry about this.!!!

 

John

[jabbr]

I am having good initial results with 1.6.5B using

 

--cdd --fcs="4,-44.5,fcx=g" --floatout

 

[Night Rain]

10 hours ago, jabbr said:

I am having good initial results with 1.6.5B using

 

--cdd --fcs="4,-44.5,fcx=g" --floatout

 

Tried those and it sounds muffled.

[sandyk]

18 minutes ago, Night Rain said:

Tried those and it sounds muffled.

Did you first make sure that the album actually had Feral-A encoding, which is sometimes done to make the recording brighter sounding ? 

One of the giveaways is often excessive sibilance, but some artists do often have a little too much sibilance in their recordings. Kenny Rogers for example

 

[Confused]

4 minutes ago, sandyk said:

Did you first make 100% sure that the album actually had Feral-A encoding, which is sometimes done to make the recording brighter sounding ? 

Is there a way to be 100% sure a recording has "Feral-A" encoding?  I have asked this before and the answer seems to be no.

[Night Rain]

13 minutes ago, sandyk said:

Did you first make sure that the album actually had Feral-A encoding, which is sometimes done to make the recording brighter sounding ? 

One of the giveaways is often excessive sibilance, but some artists do often have a little too much sibilance in their recordings. Kenny Rogers for example

 

  Yes, tried ABBA and Supertramp. It mostly affects the mids.

[jabbr]

Genesis, Lamb Lies Down on Broadway -- the upper details are removed e.g. the shimmering in the intro which comes in at 22 seconds

 

 

[sandyk]

9 hours ago, Confused said:

Is there a way to be 100% sure a recording has "Feral-A" encoding?  I have asked this before and the answer seems to be no.

I would still say no, as even John gets it wrong occasionally sometimes. You tend to get a feel for most of this after a while though.

 Look out also for harshness ,with sometimes female voices tending to sound a little higher in pitch overall, perhaps even "screechy" (upper harmonics accentuated) where other recordings from them suggest that they should sound sweeter, with perhaps more individuality to their voices.

Sometimes, YouTube videos of the same recording may help for comparison due to the rolled off HF detail .

 Some male voices may also sound a little harsh with loss of natural low end warmth.

Look out for loss of natural sounding low end warmth in general.

YMMV.

 

Alex

[Skip Pack]

2 minutes ago, sandyk said:

Look out also for harshness ,with sometimes female voices tending to sound a little higher in pitch overall, perhaps even "screechy" (upper harmonics accentuated) where other recordings from them suggest that they should sound sweeter, with perhaps more individuality to their voices.

Sometimes, YouTube videos of the same recording may help for comparison due to the rolled off HF detail .

 Some male voices may also sound a little harsh with loss of natural low end warmth.

YMMV.

 

Alex

Look also for congestion from around 600hz up to say 4Khz when multiple voices and instuments are playing. I now view the process here (goal) as removing the screechiness/congestion while not reducing the vitality/richness. John has cited excess 'boominess' as a feralA issue, and he is probably right, but I do want bass (and bass harmonics) to still be nicely audible. Might just be a personal preference of my old ears.

 

Skip