hard center or full center preference?

[raquete]

i was thinking to do a poll but as i will read all answers...

if your (ficticious music) stereo source have:
sax------voice------trumpet
where sax and the voice are in the left,the trumpet and the voice in the right.

i can separate:
left: sax and voice
right: trumpet and voice
surround left: sax only
surround right: trumpet only
and
center (hard) : only voice
or
center (full) : mono with sax,voice and trumpet

what center channel you choose and the reason?
(hard or full)

[setarip_old]

what center channel you choose and the reason?Hi!

The only reason for either choice, that I'm aware of, would simply be personal preference...

[raquete]

Hi!

The only reason for either choice, that I'm aware of, would simply be personal preference...
After long time and observing the first post my preference remains in

center (hard) : only voice

Do you know how i can do that, please?

[tebasuna51]

You can try the old http://www.moitah.net/download/latest/Center_Cut_GUI.zip.

Like input you need a stereo wav with a simple header.

[raquete]

Sincerely, I helped just a little (as beta tester) to Moitah with CCGui with sincronism between channels extracteds. (more than 20 tears ago)
I use a long time but CCGui let some undesirable artifacts even if the source is from LPs(rippeds in VHR, 192/32 for exemple)
Now i sum C from CCGui together with Audacity - effects - Robert Haenggi - vocal reduction and isolation – isolating the center or isolating vocals.
The result is very good but not perfect.
Moitah dissapear a long time ago but CCGui remains the best tool to extract center and surrounds that i know.
CCGui extract the center but not only the voice or what is really only in the Center.
I'm searching "pure center only" without 'extremes' and don't know what tool can do that.

Thank you tebasuna51. :-)

[Emulgator]

Usual processing of vocals may add non-mono FX.
A: any ambience like Delay, Reverb etc.
B: any modulated delays like Chorus, Flanger etc.
C: any distortion like Exciter, etc.
D: artificial voicing like Melodyne etc.
One can not expect to remove these with basic algos.

[raquete]

Don't know if is permissive post sample here in the forum than i ask:

Can i post a short example in Mega, mp3 format with only voice for you download and hear, Emulgator?

[tebasuna51]

You can upload a link to a short example, but if filters (such as A-B-C-D or others) have been applied to it, recovering the original voice is difficult.
And even if you succeed, there is no guarantee that the method used will work in every case.

[raquete]

The method I use works in every case, but it is extremely difficult and tedious to carry out.
How do I do it? I don't know; I just mix the ingredients, put it in the oven, and wait for the result. If the cake tastes bad, I start all over again.

Sometimes it takes a whole day to achieve the result, and I lack the ability to explain the step-by-step process due to my limitations in writing in English;
➡️ that is why I asked for a way to achieve the result using ffmpeg:

https://mega.nz/file/Ym5TmJ5B#xLa0w65nbsqwYS_DvaPM6klckp7H2fM95gz0preZkXg

Thank you for permission to post the sample, tebasuna51. :thanks:

[tebasuna51]

Although the audio contains no musical instruments—only the human voice—an analysis of its frequency spectrum reveals significant content in both the low-frequency range (below 40 Hz) and the high-frequency range (up to 15 kHz).

This suggests that filters were applied to modify the original voice, rendering the original signal unrecoverable.
Clearly, your example requires no separation, as it consists solely of the human voice; the work is already done.

Try with other sample.

[raquete]

Exactly, the work is already done(i removed the voice from the music, i can remove any instrument too but is very hard!
If you sum the voice, the guitar, the bass and the drums you'll have the original music.
...or i don't understand what you mean tebasuna51, sorry.
I can post 4 parts of one music with voice(drums, bass, guitar) removed from the music.
Then you sum the parts extracteds to get the 'original' source again without loses
My target is get this results with ffmped but i don't know how to do.

Give me a little time to post the examples, please.

[raquete]

tebasuna51,
please download and open in your ➡️wave editor to hear: track1 drums; track2 guitar; track 3 bass; track 4 voice
you can select individual track, 2 or 3 together or all together

https://mega.nz/file/x2xljRSC#hkwIk_LaSNv5hWa0YvzdFWQUK2uQKrSZ2ItlKvWSlfM

[tebasuna51]

...
Then you sum the parts extracteds to get the 'original' source again without loses
I see; extracting that probably isn't easy.
Although I doubt the result sounds the same as the original.
Could you upload it so I can check?

My target is get this results with ffmped but i don't know how to do.
I have no idea how to achieve that with FFmpeg.
I even doubt there is a fixed combination of filters that could be applied to any source.

[raquete]

Stereo Source from SACD ISO: https://mega.nz/file/ti5BiRJZ#BeUa9bSEcXDPD90CnwwIM__aO8cDFJqc1STG-aalUTM (22.4MB)

Seventh attempt to get an acceptable 5.1 upmix by mixing the levels of the previously separated channels.
https://mega.nz/file/U6ZkBD7R#BHcborLPXq13bquI-GD8hnq91Es5r9o0ymSUdhLgBBA (57.2MB)

If you have no idea how to do this with FFmpeg, imagine what I could achieve using my "homo-frog science."

ps: of course the levels change using wave editors or foobar to hear but the sum of the 5.1 channels give me the exact source adjusting the levels of each channel.
The downside: I don't know if I'm being clear in my explanations.

Thank you tebasuna51.

[hello_hello]

Have you listened to the upmixed version after it's been downmixed to stereo again?
I only ask because I can only listen to it on my stereo PC speakers, but it sounds a bit flat compared to the original, plus the upmixed version still has the vocal predominantly on the right side (front and surround). It's in the center channel too but if the object of the exercise was to move it to the center it's not quite right. Or are you after the same balance of left and right, but spread over more speakers?

As I said I'm only using stereo PC speakers and I do have roaring tinnitus in my right ear today, so maybe I'm wrong. Although if anything I'd probably go the opposite way and mix it down to mono, assuming you're mainly doing this because you don't like the way the vocal is panned mostly right. Or is it left? I should check my speakers are connected the right way around.

Chances are this is nothing like the result you're after, but I was bored so I played around a little, and after loading the stereo track into Audacity twice, I swapped the channels of one track then reduced it's volume by 4dB (compared to the other) before downmixing into a single stereo track again. It centers things a bit more without it being completely mono.

06 - I Don't Live Today.flac (https://files.videohelp.com/u/210984/06%20-%20I%20Don_t%20Live%20Today.flac)

[tebasuna51]

To be honest, the stereo remix—after separating the components—strikes me as inferior to the original, which comes as no surprise.

Regarding your speaker setup, if the vocals are appearing in the right channel in addition to the center channel, then your speakers are connected correctly.

If you have the time, feel free to experiment however you like; however, in my opinion, you shouldn't draw broad conclusions based on a single specific example.
A technique that works well for one track might fail completely for another.
5.1 AVRs themselves feature various filters (applying echo, reverb, etc.) that convert stereo audio into 5.1 for different listening environments; the results are interesting, and—since taste is subjective—some people might enjoy using them while others might not.

[raquete]

@hello_hello and @tebasuna51.

you're right, the sound is a bit(too much) flat cos i did a fast upmix for test the separation between channels to show what can be done,
not for impressive quality, was only a 'sketch'.
Distributing the audio elements (vocals, guitars, etc.) in desired channles requires multiple tests to achieve the final result; that is why the vocals (and other parts) are assigned to the specific channels I chose. Rest assured, however, that perfect separation is possible—though it involves many stages and numerous temporary files before the final outcome is, at the very least, desirable!

But wait, i'll do another upmix and will post in mega together with the source and with all possibles separeteds channels.
--> i can post the separeted channels extracteds too than you'll see and will join the channels as you desire and taste.
The very hard part is separete the instruments(bass, drums guitars,voices,etc, join to have the desired resul is the hard part.
i'll send a channel only with pur voice, another only with pure drum, another only with pure bass, etc.
I never 'dope' the separeteds channels with reverbers, limiters, phasers, normalizations, etc , i only mix the separeteds channels as i desire.
The LFE i extracted from the stereo source using equalizers(not a special programs,)cos one my systems need the separeted LFE channel.
The LFE contain only the 'bumbo' from drum and Basses below 150Hz !

Give me time please and i'll do my best. you are the judges with the hammers that will judge the separeteds audio channels files(this is the best part).
The result will let me free to go ahead or i'll go to jail.

@hello_hello
I never did upmix to later downmix to stereo. My target is upmix stereo to 5.1 !
After the 5.1 i don't mind with little differences comparanding with stereo, i never did it and never will do.
Compare for example Santana Abraxas quad with stereo..are completely differences, the same with TYA - A space in time comparing 5.1 with quad and with stereo. 'everything change.
Beatles stereo with the news 5.1 upmixes inside Abbey Road...'everything change'
If you dowmix this multichannels sources you never will have the original stereo again IMHO as simple musician, electronis technician retired and with memory forgething what i done . I did tube amplifiers a long time, equalizers and lots more...nothins 'spectacular' but very good.
i do my upmix for fun and pleasure and i'll do some that you will really like. :) My time is short, i'm old (72) and want to hear in the best way that i can upmix. Perfetion target in what i do? Never!

What you done using Audacity was a crime with the music.
Don't let the Audacity developers knows about you done.(lol, just kidding) :) :p

@tebasuna51

I do tests all day long every day and don't know why muy wife still don't kill me. (lol)
again you're completely right: "A technique that works well for one track might fail completely for another."
This is true in major cases. Each music have to be treated with new forms to have 5.1 results.
The same with receivers and speakers quality and size and lots more details.

I thank you both for your comments, guidance, and sincerity; I deeply appreciate what you have done—your help is invaluable. :)


...and sorry about the Google translations, because it's dumber than I am, and together we're creating a new language called 'angrysh' where nobody understand nobody and this post was edited dozen times to have a little sense. :p

[hello_hello]

@hello_hello
I never did upmix to later downmix to stereo. My target is upmix stereo to 5.1

Fair enough. I only mentioned it because if the upmix is likely to be downmixed on playback when using a stereo system, it'd pay to make sure there's no phase issues you can't hear while it's 5.1ch, but if that'll never happen it doesn't matter.

[raquete]

Fair enough. I only mentioned it because if the upmix is likely to be downmixed on playback when using a stereo system, it'd pay to make sure there's no phase issues you can't hear while it's 5.1ch, but if that'll never happen it doesn't matter.

Your reasoning is valid, logical, and spot-on!
After a good upmix, we can downmix it to stereo and check the result by comparing it to the stereo source used at the start of the process! :)
(It seems I was clear about what we can do.)

[raquete]

@hello_hello and @tebasuna51.

Please read everything carefully since English isn't my native language,
I learned on my own—figuring everything out through trial and error—
and you can carry out the process yourself if you like.

I can guarantee good results because I have been using this method for approximately 17 years.

Starting to upmix-downmix sample. first wiil be upmix the stereo source and later downmix the proceesed file
to do comparisons.stereo and, depending on interest, it can be upmixed to 5.1 and that same 5.1 downmixed to stereo.

In this example was used as source Uriah Heeo - Tears in My Eyes
from Look at Yourself album ripped from CD(44.1k 16 bit) this music
have good transitions between channels, variable levels and is a good option
for the test of upmix the stereo source and downmix to stereo resul for comparidons!
-->The result will be STEREO for test of the process. (later to 5.1 IF someone have interest)


-->You can use ANY STEREO source and follow the EXACT STEPS to get the right result!

VERY IMPORTANT:
Load the original Tears in My Eyes 48k 24 bit PCM in Audacity
click in 'effects - volume and compression - amplify
READ the Amplification (DB) level. COPY AND SAVE THIS LEVEL but DON'T CHANGE ANYTHING,
only COPY AND SAVE THE LEVEL FOUND in any place (in notepad for example)
--> CLOSE the original Tears in My Eyes 48k 24 bit PCM WITHOUT do changes.

Using CenterCutGui (default adjusts):

Loading Tears in My Eyes 48k 24 bit PCM in CenterCutGui to extract
Tears in My Eyes 48k 24 bit PCM-C (Mono channel only with central Channel)
and
Tears in My Eyes 48k 24 bit PCM-S (Stereo Surround file without Center sounds)


* HOW to have the Source again after separt the audio using CenterCutGui

If you load the Tears in My Eyes 48k 24 bit PCM-S file in Audacity,
separate this stereo file in 2 mono files.
--> CREATE 2 EMPTY MONO FILES in the BOTTOM of the Tears in My Eyes 48k 24 bit PCM-S separated files.
Now you have 2 tracks on the top with sound and 2 EMPTY mono tacks in the BOTTOM.

Open a new archive in Audacity and load the Tears in My Eyes 48k 24 bit PCM-C

Copy the Tears in My Eyes 48k 24 bit PCM-C file and paste in one empty mono file
of the separated track created in the bottom of the separated file.
MOVE this file to the TOP of the separated Tears in My Eyes 48k 24 bit PCM-S file

Copy again the Tears in My Eyes 48k 24 bit PCM-C and paste in the
EMPTY track in the BOTTOM of the separated Tears in My Eyes 48k 24 bit PCM-S file

IMPORTANT:
Amplify to -6 dBs each one of the Tears in My Eyes 48k 24 bit PCM-C
(One is in the TOP, another is in the BOTTOM)

JOIN the 2 mono tracks in the TOP as Stereo Track.
MIX this stereo Tacks on the top to MONO.
Now you have one mono file in the TOP.

JOIN the 2 mono tarcks in the BOTTOM as stereo track
MIX this stereo Tacks in the bottom to MONO.
Now you have a mono track in th BOTTOM

--> JOIN THIS 2 MONO TRACKS AS STEREO TRACK.

--> Click in effectts - volume and compression - amplify
AND APPLY THE LEVEL FOUND IN THE BEGINNING OF THE PROCESS THAT YOU TAKE NOTE AND SAVE.
SAVE WITH DIFFERENT NAME OF THE SOURCE USED!

NOW OPEN THE SOURCE TO HEAR
OPEN THE RESULT OF THE PROCESS USING AUDITION AND CENTERCUTGUI TO HEAR.
COMPARE ANYTHING YOU WANT BETWEEN THE ORIFINAL FILE USED AS ORIGINAL SOURCE WITH THE FILE PROCESSED
WITH AUDITION AND CENTERCUTGUI QIRTH AND TELL IF YOU CAN HEAR DIFFERENCES.

THE BEST to use this for simple but very good 5.1 file
using a similar process summed with the original source,
you will have a simple but VERY GOOD 5.1 File

--> IF THE MUSIC USED TO DO THE 5.1 THIS WAY AND THE MUSIC HAVE TRANSITIONS THE IN
THE AUDIO FROM LEFT TO RIGHT ND FROM RIGHT TO LEFT, THE TRANSITIONS WILL BE PERFECTS.
USING Jimi Hendrix - May this be Love, you will have PERFECTS transitions
between ALL channels without change the levels in the transitions.

If you're interested, i do one TUTORIAL, is very simple and fast.

Regards and excuse the 'caps lock', It is pure enthusiasm. :)

(later i'll remove unecessary caps locks, leting only the necessary for highlight)

Thanks

[tebasuna51]

If I understand ok you do:

1) Volume to normalize the source after:
click in 'effects - volume and compression - amplify
READ the Amplification (DB) level. COPY AND SAVE THIS LEVEL

2) Obtain Source_S (L-R) and Source_C (mono) using CenterCutGui

3) Mix the channels with Audacity to obtain a new:
L' = L + 0.5*C
R' = R + 0.5*C
And amplify the output with the value obtained in 1)

For me, it is easier to create an .avs file to use with BeHappy than to use Audacity.
At the same time, we ensure there are no clips during the process.
We cannot guarantee that the 1 + 0.5 mixture—subsequently amplified—will not produce values ​​exceeding 1 that get clipped.
Working with floating-point values ​​and then normalizing ensures that no clipping occurs.

s=WavSource("Source_S.wav").ConvertAudioToFloat()
c=WavSource("Source_C.wav").ConvertAudioToFloat()

fl = mixaudio(s.GetLeftChannel(),c,1.0,0.5)
fr = mixaudio(s.GetRightChannel(),c,1.0,0.5)
MergeChannels(fl,fr)

SSRC(48000)
Normalize(1.0)
ConvertAudioTo24bit()

[raquete]

EDITED:
"If I understand ok you do:

1) Volume to normalize the source after:"
No, you don't normalize the audio,
you only take note of the level found to be used in the future, don't change anything in the source to be used!

Yes tebasuna51,
your 'manner' to do this steps really seems more easy and more precise but i never used
'Be Happy' cos don't understand 'a thing' of this program.
-->I really can't deal with 'command lines', is one of my limitations.

The mannear that i do using Audacity don't give clips, differences in basses or trebles
(Comparing the source with the result, both are exacts in Audacity and in Foobar 2000 (using 'Peek Meter' together with 'VU Meter').
Don't have 'clips', don't have differents levels and the sound seems equals to my ears and show the exact amplitudes
in waveforms in any audio editor(i use more Audacity but have others audio editor) and foobar 2000 too to compare!

The more important:
I never did: "Obtain Source_S (L-R) and Source_C (mono)."

I don't know how to do in Audacity L' = L + 0.5*C R' = R + 0.5*C
And amplify the output with the value obtained in 1)
Please tell me how to do that in Audacity(or in any audio editor!

(I'm thinking loud): A doubt: L-R don't will eliminate what is common in both channels?
This method don't will give me one kind of 'oops'?

Using CenterCutGui give me the Center Channel as i want using like i described cos in 'my' method the Center is Mono but
--> without any sound of the extremes, is 'pure' Center Channel.

I will try 'your method' to obtain the Center Channel. (using Audacity)
After i compare my result with 'your' method and post the result.

Thanks for your help and good will! :-)
:thanks:

[raquete]

@hello_hello and @tebasuna51.

Please, download this test to hear: https://mega.nz/file/h24gjYaY#1wattc5bj6Q9frfE3pU7znLCkrGFd4S9Qs2qGenMdPk

--> And don't forget to read my last post : https://forum.doom9.org/showthread.php?p=2049937#post2049937

Edit:
Simple 5.1 upmixes: (download if you really want to hear, in this moment i did this 2 samples only as 'preview'< they are not 'perfects''

With LFE channel: https://mega.nz/file/93Zi1Q5L#vJBipKvz4qkUNP3KvoVu3iliPtgb8Fha_Nv0Io8aSYE

without LFE: https://mega.nz/file/56YjwQpC#ERatJxr9XTRWknXxwiGi_wFNqdUFqO8LvLJmJCskNGE

[tebasuna51]

1) Volume to normalize the source after:"
No, you don't normalize the audio,
you only take note of the level found to be used in the future, don't change anything in the source to be used!

I say "after", not at this moment.
For what that value? You can't know the apropiate value to amplify after the mixes.

The mannear that i do using Audacity don't give clips
This example might work, but the process followed cannot guarantee that clipping will not occur—either when mixing the L and R channels with the center channel, or during subsequent amplification using a value calculated prior to the mixing stage.

The more important:
I never did: "Obtain Source_S (L-R) and Source_C (mono)."
You say:
Using CenterCutGui (default adjusts):

Loading Tears in My Eyes 48k 24 bit PCM in CenterCutGui to extract
Tears in My Eyes 48k 24 bit PCM-C (Mono channel only with central Channel)
and
Tears in My Eyes 48k 24 bit PCM-S (Stereo Surround file without Center sounds)


* HOW to have the Source again after separt the audio using CenterCutGui

If you load the Tears in My Eyes 48k 24 bit PCM-S file in Audacity,
separate this stereo file in 2 mono files...

That Tears in My Eyes 48k 24 bit PCM-S I named Source_S (L-R), it's not the Surround channels, but the left and right Side channels.

I don't know how to do in Audacity L' = L + 0.5*C R' = R + 0.5*C
And amplify the output with the value obtained in 1)
Please tell me how to do that in Audacity(or in any audio editor!
Your method is correct.
1) When you load "Tears in My Eyes 48k 24-bit PCM-S" (a stereo Side file without center-channel audio)—which I’ve abbreviated as "Tears-S"—into Audacity, it automatically converts the file to 32-bit float (see image, right click over Tears-S) to prevent issues.
This is equivalent to my loading command in AviSynth:
s=WavSource("Tears-S.wav").ConvertAudioToFloat()

The same for Tears-C.

2) To do L' = L + 0.5*C R' = R + 0.5*C is exactly your process:
Copy the Tears in My Eyes 48k 24 bit PCM-C file and paste in one empty mono file
...
--> JOIN THIS 2 MONO TRACKS AS STEREO TRACK.

3) Now you can't do:
--> Click in effectts - volume and compression - amplify
AND APPLY THE LEVEL FOUND IN THE BEGINNING OF THE PROCESS
Because you can clip the sounds.
You must apply the value level obtained now, it can be positive (no problem) but maybe negative (clips detected).
Thats is the avs: Normalize(1.0)
And now can be saved as integer samples (24 bits) without clips.

(I'm thinking loud): A doubt: L-R don't will eliminate what is common in both channels?
This method don't will give me one kind of 'oops'?

Using CenterCutGui give me the Center Channel as i want using like i described cos in 'my' method the Center is Mono but
--> without any sound of the extremes, is 'pure' Center Channel.
I don't understand your doubts; that's what CenterCut is supposed to do, isn't it?
I will try 'your method' to obtain the Center Channel. (using Audacity)
After i compare my result with 'your' method and post the result.
I don't understand you; I don't have any other method to get the center channel.

[raquete]

@tebasuna51

if you had download the https://mega.nz/file/h24gjYaY#1wattc5bj6Q9frfE3pU7znLCkrGFd4S9Qs2qGenMdPk
you will get the original source and the result that i redid for comparisons but...

Seems that you don't download the 'Simple 5.1 upmixes too: one with LFE and another without LFE'
This samples can remove lots of doubts! I did one sample without LFE cos i know that you don't like or is useless for you.
But in my case, i need the LFE cos one of my receivers have separateds inputs.(6 inputs for RCA cables)
With LFE channel: https://mega.nz/file/93Zi1Q5L#vJBipK...Fha_Nv0Io8aSYE
without LFE: https://mega.nz/file/56YjwQpC#ERatJx...O8LvLJmJCskNGE

Trying to answer each doubt or question:

A- "You said after, not at this moment" but don't said when and '"WHEN" is vital in my case, in the end of explanations you will understand. (i wish)

B- "For what that value? You can't know the apropiate value to amplify after the mixes."
I know when the value have to be used, this is why in the start i tell to take note of the level found in the source!

C- "but the process followed cannot guarantee that clipping will not occur"
I can garantee!

D- "it's not the Surround channels, but the left and right Side channels." I said: "(Stereo Surround file without Center sounds)
- the result is a stereo channel where the Left channel have only what is in the Left channel and the Right Channel will have only is in the Right channel. I call as 'surrounds' cos Moitah called the C channel as C and the S as Surrounds when i was his beta tester.

E- "...which I’ve abbreviated as "Tears-S"—into Audacity, it automatically converts the file to 32-bit float (see image, right click over Tears-S) to prevent issues."
Oh yes, i know that! :) ...but you can save as 24 bit and this is what i do!

F- "This is equivalent to my loading command in AviSynth: "
Sorry, as i wrote before somewhere here, i can't deal with command lines and don't use AVISynth cos i dont know ' a thing' about AVISynth. For sure must be very cool !

G- "Now you can't do: "... "Because you can clip the sounds."
No, the way i do don't clip the sound i and was recommended to read this post with some explanations:, i wrote:
- And don't forget to read my last post : https://forum.doom9.org/showthread.p...37#post2049937
You have to use the same level that you take note !
I don't use 'normalize', i use 'volume and compression - amplify...amplify to' and use the level that you take note!

H- "I don't understand your doubts; that's what CenterCut is supposed to do, isn't it?" I don't know because i dont know how to use AVISynth.

I- "I don't understand you; I don't have any other method to get the center channel." your method is with AVISynth that i don't know how to use. l:p

Thanks for the picture, for extreme good will and for ask deep questions and show doubts about what i'm doing, this help a lot cos in step by step, i learn more esch day.

Thank you very much tebasuna 51 and please download my samples in mega to check, to put doubts, to find erros or if you agree with anything. :thanks:

[tebasuna51]

C- "but the process followed cannot guarantee that clipping will not occur"
I can garantee!

I'm sorry, but that isn't correct.

- I load Tears-S (the Side channels obtained via CenterCutGUI) and calculate the maximum amplification value as 1.175 dB (Figure 1).
- Then I load Tears-C (the center channel obtained via CenterCutGUI; I never calculate the center channel using AviSynth, I always use the one generated by CenterCut).
- I amplify it to -6 dB (equivalent to 50%, or 0.5*C).
- Following your instructions, I mix it with the left channel of Tears-S: L' = L + 0.5*C
- And also with the right channel: R' = R + 0.5*C
- This yields a new stereo audio track that I have named 'stereo'.
- I request amplification, and the program offers the option to amplify it by up to 1.16 dB (Figure 2).
- If I enter the 1.175 value calculated earlier, the 'Apply' option becomes disabled because the requested amplification exceeds the allowed limit (Figure 3).
- ...unless we explicitly check 'Allow Clipping' (Figure 4).

On the other hand, you are performing a 3.0 to 2.0 downmix using:
L' = L + 0.5*C
R' = R + 0.5*C
However, to maintain sound power (which is proportional to the square of the amplitude), the recommended method is:
L' = L + 0.707*C
R' = R + 0.707*C
That mix would allow for an amplification of only 1.052 dB without clipping.

I have downloaded your upmix tests, but I am not going to pass judgment for the simple reason that I do not trust my 75-year-old ears.
I will limit myself to technical, non-subjective matters.

[raquete]

Thanks for the reply, tebasuna51; you gave an excellent description—albeit a complex one for my limited English—but...
PLEASE,
download what I created using the method I explained and listen to it (please).
It is very difficult for me to be more precise given my limited English:

With LFE channel:
https://mega.nz/file/93Zi1Q5L#vJBipKvz4qkUNP3KvoVu3iliPtgb8Fha_Nv0Io8aSYE
Without LFE:
https://mega.nz/file/56YjwQpC#ERatJxr9XTRWknXxwiGi_wFNqdUFqO8LvLJmJCskNGE

If you don't listen to what I made—and if you want to follow the method I described—you will never understand me.

Once you download the files from the links, unzip them, listen to the results,
and follow the steps I posted using any stereo source, you will get exactly the result I described!

https://forum.doom9.org/showthread.php?p=2049919#post2049919

P.S.: No "calculation" is necessary; just follow the steps!
The target is not to have 0(zero) Db,
The result don't have to have different level found in the source used, never!
--> is to have the same level that was found and take note when start the process!

:thanks:

edited:
"I have downloaded your upmix tests, but I am not going to pass judgment for the simple reason that I do not trust my 75-year-old ears.
I will limit myself to technical, non-subjective matters."

...But music is to hear tebasuna51, if we can't hear, what we doing? lol :)

[tebasuna51]

The target is not to have 0(zero) Db,
The result don't have to have different level found in the source used, never!
--> is to have the same level that was found and take note when start the process!

Of course, the output cannot exceed 0 dB if you intend to convert it to integer samples (16-bit or 24-bit) or encode it using any codec.

That is the purpose of normalization: to amplify the signal to the maximum permissible level.
If you exceed that maximum, the samples are clipped, resulting in distortion.

In this case, the difference is imperceptible, yet it exists. Therefore, it cannot be guaranteed that distortion will be absent in every instance.

On the other hand, changing the input level does not alter the quality of the result.
Especially when it involves more than just a slight variation in the peaks.

Audible volume is proportional to the root-mean-square (RMS) value, and—as I mentioned—to maintain it, the mix should be created by attenuating the center channel by only -3 dB (0.707*C).

[tebasuna51]

I’ve tried listening tests with your samples, and honestly, I didn't notice any improvement when listening to the REDONE version compared to the SOURCE.

I also can't tell the difference between the 5.1 versions with and without LFE (given my AVR's configuration, they ought to sound the same). They clearly engage all my speakers, just like the upmix I used to create a while back using foobar and the VI plugin.

In any case, listening tests are highly subjective; besides my aging ears, the fact that the music genre isn't really my cup of tea doesn't help matters.
The high frequencies (electric guitar, I think) were quite bothersome to my left ear, and I felt like I was hearing noise rather than music.

[raquete]

Ok but:
I don't wrote to change the level of the source.
Don't have the purpose of normalization!
I don't tell in post that is to have 0 dB, is to have the SAME level of the source in the result,
this is the reason of that you have to take note of the level of the source before start the steps.

The difference is imperceptible but can exists. Who care if is imperceptible? :D

If the lack of distortion is virtually imperceptible in any situation—
or if you can't tell the difference even when comparing it to the original source—then it's fine.

It’s like a blind person asking you to turn on the light! You don’t actually turn it on, but you say you did!
To that person, it makes absolutely no difference, since they don’t perceive any change at all. :p

Honestly, you didn't notice any improvement when listening to the "REDONE" version compared to the original (SOURCE)?
That's great, because that is precisely the whole point of the "redone" version.

Remember what hello_hello asked?
"Did you listen to the *upmixed* version after it had been converted back to stereo (*downmixed*)?"
So, if you don't "sense" any differences between the original and the "redone" version when listening to my samples,
then hello_hello's question has been answered by my "redone" version!
Then if you don't 'feel' differences between source and the redone hearing my 'samples', the hello_hello answer was done in my redone way!

You don't notice the difference between the 5.1 versions with and without LFE because your system You don't need 5.1; 5.0 is sufficient for your system. Your system "decodes"(don't know the right word) or separates the low frequencies and redirects them to the LFE channel (a unit to amplify the LFE, where you can adjust the desired level for that channel).
-->That is correct for your setup!
In my case, I need LFE because I send individual channels to my receiver via RCA cables, and without the LFE channel, I lose the bass, etc. I need LFE for only one of my systems; the other two don't, 5.0 works perfectly for those other two systems but i do only 5.1 upmixes cos i can you anywhere!

Aging years?
We’re respectable older gentlemen now tebasuna51(lol),
don’t let audio editors and waveforms impress you—let what you hear impress you!
Audio engineers create remixes that degrade the original quality; dozens of remixed albums sound terrible, and the "loudness war" is here to stay. Have you heard the latest Deep Purple album? I can't listen to three songs in a row even with cotton in my ears—*everything* is saturated! (seems that have DR 3 or 4) :eek:

"Redone" should actually be called "REMADE," but 'my' English confuses even me.
If I can barely understand what I write, who is actually going to understand me? :D

--> Thanks for your usual kindness, tebasuna51! :)

[tebasuna51]

Ok but:
I don't wrote to change the level of the source.
Don't have the purpose of normalization!
I don't tell in post that is to have 0 dB, is to have the SAME level of the source in the result,
this is the reason of that you have to take note of the level of the source before start the steps.

The way you are using the Amplify function does not seem to yield the desired result—namely, having the output match the original level—because:

1) Volume level is proportional to RMS and is unrelated to the highest peak detected during the Amplify process.

2) When you apply the value calculated from the source, you do not get the original peak level but rather a higher value. Instead, you should enter that value—with a negative sign—into the 'New peak amplitude' field and let the system calculate the required amplification value.

3) I don't know why you object to Normalization; it is a method to ensure there is no distortion while achieving maximum output volume.
It is recommended at the end of the process whenever working with floating-point audio samples.

The difference is imperceptible but can exists. Who care if is imperceptible? :D

If the lack of distortion is virtually imperceptible in any situation—

However, there is no guarantee that this will always happen.

I repeat: what I strongly recommend is forgetting about the source's peak level and normalizing at the end of the process to avoid clipping or to maximize output volume.

[raquete]

��

1- i never had thinking about the RMS value, i'll to pay attention about it!

2- I apply with a negative sign—into the 'New peak amplitude' field and let the wave editor(Audacity) apply the required amplification value.
That’s exactly how I do it.

3- 'Normalize to' will give me the result but as i read the value using 'amplify to'....now i will use 'normalize to', you're right . :)
(It’s down to my "partial knowledge of the parameters" :
not knowing what makes it easier to get the result using another tool in the audio editor—my mistake).

Regarding the 'imperceptibility of the difference between the source and the result'
There’s no guarantee this will always happen, but the peaks are always the same, and the overall volume of the entire track,
the waveforms seems equals, "sounds the same" to me. I’m going to compare the RMS levels of the source and the result. Good tip! :)

you strongly recommend is forgetting about the source's peak level and normalizing at the end of the process to avoid clipping or to maximize output volume.
-->However, I need to know the source level to apply it to the result;
so, I check the peak, "note down the value in a text file," and apply it later in the result.

I’ve never had clipping issues doing this;
using the level found in the source before 'the job' and apply the same level in the result, I’ve always achieved exactly the same level!

-->very very important: i never had to 'decrease the level' in the result, never is clipped,
i always have to use what i 'take note' in the result and the level always encrease just a few to get the same level of the source.

Using the "normalize to" function in the result using the value I "noted in the .txt file"
will give me the exact level in the final result just like the "Amplify to" function.

I will follow your tips to compare the levels, waveforms and RMS results between the source and the output.

:thanks: for more tips and explanations tebasuna51! :)

ps: today soon i was answering you and the forum go down (out :confused: ) but i could copy what i wrote in .txt file and i'm posting only now!

[tebasuna51]

Ok, make your test.
If you obtain less RMS in the REDONE than the original maybe you can test the downmix with 0.707*C (amplify C by -3 dB instead -6 in your process).

[raquete]

All Right! :)

I will do like you recommend but remove my doubts cos i forgot lots of details, i'm retired for more than 24 years and going (more) insane

-6 dBs = 50% ( half or 1/2of the level, right? )

how much is -1dB?
how much is -2dB?
how much is -3dB?
how much is -10 dBs?

:confused:
Be sure, i forgot lots of things and electronics details.
Remedy to sleep don't make me sleep, only remove my memory and each week i don't sleep for 2 or 3 days.
I'm going to turn in one kind of 'Zumbi' :(

[tebasuna51]

-6 dBs = 50% ( half or 1/2of the level, right? ) = 0.5

how much is -1dB? = 0.89
how much is -2dB? = 0.79
how much is -3dB? = 0.71
how much is -10 dBs? = 0.32


If you have UsEac3to GUI, you can use a small built-in calculator.

Simply open it and—without loading any file—click on:

1) Add (next to F-FFMPEG)
In the window that opens:
2) Enter the value in dB
3) Click the <- arrow
You will get the equivalent Gain value.

Or do the reverse: enter the Gain and click ->

There are also other utilities available, such as time formatting, frame equivalencies based on fps, etc.

[raquete]

I don't use UsEac3to GUI (I need to know this GUI)
but you answerd me when 'quoted' my questions:

how much is -1dB? = 0.89
how much is -2dB? = 0.79
how much is -3dB? = 0.71
hoe much is -6 dB? = 0.5
how much is -10 dBs? = 0.32

I found complicateds answers too (was hard to find):

dB..........Power ratio.........Amplitude ratio
...0.........1........................1
−1.........0.794..................0.891
−3.........0.501 ≈ 1⁄2...........0.708 ≈ √1⁄2
−6.........0.251 ≈ 1⁄4...........0.501 ≈ 1⁄2
−10.......0.1......................0.3162
.
As always, thank for your good will. :)