Aften 0.0.8 is out [Archive] - Page 3

Kurtnoise

28th July 2006, 05:49

i'm still searching answer for this question(for more than i read i can't find:
for audio(audio-dvds) i have to adjust Aften to -31, -27 or -15 in dialog normalization?

thanks.
From the sticky (http://forum.doom9.org/showthread.php?t=56020)...

Referencing Volume to a Known Level - Dialogue Normalization

To meet the Dolby Digital requirement that different programs should have approximately the same listening level (thus the consumer does not have to adjust volume level between programs), Dolby Digital incorporates a parameter called dialogue Normalization. This metadata parameter tells the decoder how far away from the reference level the average sound pressure level of the material's dialogue is.

The movie industry masters their soundtracks in a specific way. The maximum rated sound level (where all amplifiers are putting out their rated power) is 0 dB. Sounds below that level are rated in terms of how many decibels (dB) they are down from that maximum level. As such, these values are negative. The movie industry typically masters the "normal" listening level of dialogue (where people are speaking in a normal voice) at -31 dBFS. In other words, a speaking voice is at an average of -31 dB when referenced to the 0 dB maximum sound level, hence the term decibels of full scale (dBFS).

Since movie content is the largest class of programs to go on DVD, Dolby chose -31 dBFS as the reference level for audio on DVD, where 0 dB represents the maximum encodable digital sound level (full scale).

The dialogue normalization parameter needs to be set to the LAeq level of your program material's dialogue. LAeq stands for the long-term A-weighted sound pressure level. Loosely, this is the average volume level of your source material's dialogue. Us lowly consumers really don't have a tool that can measure this parameter, but we can get close. Sonic Foundry's Sound Forge has a "Normalization" feature that can measure the RMS level of a .wav file (or the portion thereof containing dialogue). CoolEdit may also have a feature like this. To use it in Sound Forge, open your .wav file containing the movie audio. Select a section containing dialogue (no sound effects or music). Go to "Process"/"Normalize". Select the "Average RMS Power (Loudness)" radio button. Then click the "Scan Levels" button. The displayed "RMS" level is very close (within 1-2 dB) to the LAeq level.

That RMS level is the number that the dialogue normalization parameter should be set to. In other words, if the RMS level in Sound Forge shows as -17.6 dB, set the dialogue normalization parameter in your Dolby Digital encoder to -18 dBFS.

The decoder will perform an attenuation of (31 + dialnorm) dB to the program material when played back. So, in this case, the decoder will attenuate by (31 + -18) = 13 dB. This will bring the average sound level of the material to (-17.6 - 13) = -30.6 dBFS. The program is now played back at approximately -31 dBFS, the reference level.

-31 dBFS is a lower average volume level than what is typical from other sources. It will be noticeable that you will have to turn the volume up on your system when playing a DVD versus playing broadcast, tape, or other non-Dolby Digital program material.

Rockaria

28th July 2006, 07:29

trying to help the thread after got great gratifications:

Pro Logic II ... Left .. Right Center Rear Left Rear Right
Left Total .... 1.000 0.000 0.707 .. j0.8165 .. j0.5774
Right Total ... 0.000 1.000 0.707 . k0.5774 .. k0.8165

j = + 90º phase-shift , k = - 90º phase-shift
AFAIK, this coefficient(matrix) issue is handled in other thread where yourself also participated in. I might be wrong here...
Many have confirmed the 1:3dB sound pressure ratio model(0.866, 0.5) seperates the rears better than the quoted 1:2 ratio model.( I also TESTED and confirmed it)
The downmix itself is currently beyond the scope of several AC3 encoders(DP569, sodtEncode...) while they only provide the option for the rears to be 90deg phase shifted to be ready for the decoder-downmix(where one rear coef channel will be inverted) for the 2ch speaker set compatibility.

Also this functionality is mentioned in the manuals of current Dolby's related devices, which is why I wanted to mention it in this thread.
Anyway, It would be great to know what functionalities from the revised specs and standards you want to integrate in this s/w....

BTW, I thought just reflecting the memory would be enough than quoting the whole contents in this case... I might be wrong here also.;)

raquete

28th July 2006, 15:52

@ Kurtnoise13
From the sticky...right,i knew that stick.:)
there,SomeJoe was treating "Films Light"(http://pages.sbcglobal.net/wilsondr/ddexacid4.gif)soundtracks,i was unclear when asking(and maybe again here,sorry :eek: )
i mean "audio dvds" as dvs with musics only (audio-dvd with single menu with pictures),and reading 18_metadata.guide.pdf(2005 Dolby laboratories) from Dolby.inc
in
6 Metadata Combinations(bottom of the .pdf)
note: these parameter settings are provided as examples to demonstrate that different settings can be saved,named,and brought up as needed for quick use in different situations.
the settings are not recommendations,but could be used as a starting point from which to create your own metadata values.
examples of possibles metadata settings
...
Dialogue Level:
Action Film -27 dB, Drama -27 dB, Local News -20 dB, Music -15 dB, Live Sporting Events -18 dB.

i still have to follow the SomeJoe's guide for audio-dvds(music only) ? :confused:

@ Rockaria

from back to front:
BTW, I thought just reflecting the memory would be enough than quoting the whole contents in this case... I might be wrong here also.
...is handled in other thread where yourself also participated in. I might be wrong here...
of course,i remember.
i found this informations from wikipedia "table" yesterday,was tasting as "fresh news" for me.

.( I also TESTED and confirmed it)ok,it's the end of my doubts! :cool:

:thanks: you all.
:)

Rockaria

29th July 2006, 13:01

ok,it's the end of my doubts! :cool: Cool! I am relieved now you are attended. :thanks:
And one more :thanks: for the links probably mostly useful for my future resarches(bases, references)...
@Justin,
I think the DPL(II) downmix implementation won't have the higher priority in the Aften development. It requires the 90deg phase shift on the rears(might be a hard work to correctly implement) to be ready for the simple downmix in the decoders(without the expensive all-pass filter equipped)
And I guess the annex->dmixmod is ideal to be set automatically(internally) by the 'rear phase shift' option selection when the source actually has the rear channels(4.0~).

Thanks.

Kurtnoise

30th July 2006, 20:28

A new dev build (http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften-0.03-dev.zip) for testing...Check Justin's blog to have some extra informations.

@Justin: about wavrms. To avoid some errors during compilation I added

#include "../wav.h"
#include "../wav.c"

in wavrms.c. Not tested on my linux box though. Only on windows.

chickenmonger

1st August 2006, 06:36

Disclaimer: I'm not a programmer, just a user. I apologize if this seems demanding.

If I recall correctly, both Aften and Besweet's ac3enc.dll are taken from the same FFMPEG sources? If that's the case, does the current version of Aften suffer from the same three ( 1 (http://forum.doom9.org/showthread.php?s=&threadid=43466) 2 (http://forum.doom9.org/showthread.php?s=&threadid=34061) 3 (http://forum.doom9.org/showthread.php?s=&threadid=52263) ) bugs currently unfixed?

Can these be investigated?

Also, could Aften be folded into BeSweet, so it could be used as a drop-in replacement? Thanks.

tebasuna51

1st August 2006, 10:34

If I recall correctly, both Aften and Besweet's ac3enc.dll are taken from the same FFMPEG sources? If that's the case, does the current version of Aften suffer from the same three ( 1 (http://forum.doom9.org/showthread.php?s=&threadid=43466) 2 (http://forum.doom9.org/showthread.php?s=&threadid=34061) 3 (http://forum.doom9.org/showthread.php?s=&threadid=52263) ) bugs currently unfixed?
At least 3 Low volume ac3 (http://forum.doom9.org/showthread.php?s=&threadid=52263) bug is solved since ffmpeg CVS 2006-04-28. Of course also in Aften.

Others know bugs are solved also in ffmpeg (at March 2004, maybe not in ac3enc.dll for BeSweet). If still remain any problem with Pioneer DVD hardware players must be detected with new test, with my hardware/software players works ok.

jruggle

1st August 2006, 15:24

At least 3 Low volume ac3 (http://forum.doom9.org/showthread.php?s=&threadid=52263) bug is solved since ffmpeg CVS 2006-04-28. Of course also in Aften.
:) Yep...I think that was the hardest sell I've ever had to make for a patch.

Others know bugs are solved also in ffmpeg (at March 2004, maybe not in ac3enc.dll for BeSweet). If still remain any problem with Pioneer DVD hardware players must be detected with new test, with my hardware/software players works ok.
I don't know anything about the Pioneer problem...I'm not sure it was ever addressed on FFmpeg-devel. There was a bug fix quite a while ago that dealt with an error in how the header was written. That might have been to fix this bug, but I'm not sure.

I can't really tell what bug #2 was...the thread goes all over the place. It looks like it might have been a BeSweet or Azid issue. Part of the problem seems to be CRC errors. I do know Aften's CRC code works just fine, as I just reorganized it a bit and added a double-check in the code for safe measure.

-Justin

jruggle

2nd August 2006, 03:20

Hi everyone,

I am getting pretty close to a new release for Aften. I thought I would post here because I have made what I think are some pretty good changes. The best one has to be the bandwidth filter. I was previously skeptical, but I tried it anyway since the Dolby guidelines recommend it. It makes a substantial difference in the high frequencies, and I don't even have a good ear for those kind of things. To hear for yourself...try encoding this sample (http://www.xiph.org/vorbis/listen/41_30sec.wav) at 192kbps. I have been using it quite a bit for testing because it has lots of dynamic range and prominant cymbals. I suspect many of you will have to wait for a Windows binary release...sorry. I will get a cross-compile environment setup one of these days.

The details of what I have done and what I have left before releasing version 0.04 are on my Aften development blog (http://aftenblog.blogspot.com/).

Thanks,
Justin

raquete

2nd August 2006, 08:24

jruggle,
thanks for Aftenblog news (short-term plans,Bandwidth stuff and everything)

i download the sample that you posted and after hear i need to do (again(and again) the same answer that i posted in some threads in the audio forum and in this thread:
http://forum.doom9.org/showpost.php?p=856483&postcount=103

...
Dialogue Level:
Action Film -27 dB, Drama -27 dB, Local News -20 dB, Music -15 dB, Live Sporting Events -18 dB.

i still have to follow the SomeJoe's guide for audio-dvds(music only) ? :confused:

why i'm asking the same again?
...because your sample is too loud and is clipping :p
and if later is used to encode AC3 ...

regards.
;)

BigDid

2nd August 2006, 20:01

...Dialogue Level:
Action Film -27 dB, Drama -27 dB, Local News -20 dB, Music -15 dB, Live Sporting Events -18 dB...

Hi,

So if I use -27db instead of default -31db, I increase the dialogue level of 4db?

Why do I ask? because
- I'm noob/dumb/difficult learner in audio ;)
- I reencode Ac3 audio cause in actual movies audio is often with too much expansion (for me) and dialogues are too low.
ATM I use Behappy input NicAudio with DRC/Normalize/ output to ac3-ffmpeg 448-> 384/320; 384-> 320.
I sure would like to use aften instead cause it is very promising.
I cannot use aften in Behappy unless Dimzon fixes the aften interface (I have asked for it but he has no time atm) so I will try converting to wav with behappy than back to ac3 with aften unless...

@jruggle, thanks for the good work.
Would it be possible/doable to make a special aften release for BeHappy (until fixed by Dimzon) that convert the aften bp/s to Behappy kbp/s? ie aften 320.000(bp/s), being converted to 320(kbp/s) for Behappy... It seems the only bug given by Tebasuna...
I may be wrong on that matter so maybe wait for reactions from authorised persons: Kurtnoise, Tebasuna, others?

Did

jruggle

3rd August 2006, 01:19

So if I use -27db instead of default -31db, I increase the dialogue level of 4db?
If I understand your question right...no. The dialnorm setting does not affect the audio at all, only how it is decoded. You are just telling the decoder what the dialog level is. If you set the dialog level at, say -27, the decoder will actually turn down the volume by 4dB to "normalize" it to -31. This way all dialog is presented at around the same volume level when switching programs (i.e. movie to commercial).

@raquete
As far as dialnorm for music. I think it depends on what you define as "dialog". If there is singing, maybe find a section where the singing is prominant and measure the RMS of that. Really, unless it is being broadcast side-by-side with other AC-3 streams or you plan to add dynamic range control, it isn't much use to change the dialnorm. The listener will adjust the volume to a comfortable level.

@jruggle, thanks for the good work.
Would it be possible/doable to make a special aften release for BeHappy (until fixed by Dimzon) that convert the aften bp/s to Behappy kbp/s? ie aften 320.000(bp/s), being converted to 320(kbp/s) for Behappy... It seems the only bug given by Tebasuna...
I may be wrong on that matter so maybe wait for reactions from authorised persons: Kurtnoise, Tebasuna, others?
I guess if a majority of people agree, I can change Aften to use kbps instead of bps in release 0.04. I'm not particularly attached to doing it one way over the other. If it will be simpler to use kbps I have no problem changing it.

-Justin

tebasuna51

3rd August 2006, 01:28

I cannot use aften in Behappy unless Dimzon fixes the aften interface (I have asked for it but he has no time atm) so I will try converting to wav with behappy than back to ac3 with aften unless...
You can use previous BeHappy version and aften.extension (http://forum.doom9.org/showthread.php?p=851058#post851058) in BeHappy folder. You can select only the bitrate, but if you want any other parameter, edit aften.extension lines before run BeHappy:
<Value>-b 320000 - "{0}"</Value>
and put your required parameters, for instance:
<Value>-b 320000 -dnorm 27 -v 0 - "{0}"</Value>

BigDid

3rd August 2006, 02:02

@jruggle, thanks for the answer and the proposal for the 0.04. If done I apology to Kurtnoise who will have to change his GUI :o

@Tebasuna, I'll give it a try at home with everything at hand but I am a kind of dumb with commands and command line (had a hard time with avs/avsi at first) :( Thanks anyway for the proposal.
Now i understand why you didn't push Dimzon, you could do it that way :)

Did

tebasuna51

3rd August 2006, 02:08

why i'm asking the same again?
...because your sample is too loud and is clipping
and if later is used to encode AC3 ...
There are a problem with jruggle sample but it isn't related with Dialog Norm.

Encoded with Aften (last binary from Kurtnoise), ffmpeg or SoftEncode (and with dnorm 17 or 31), when is decoded to wav with NicAc3Source (Bepipe/BeHappy) crash, and decoded with BeSweet-Azid there are more than 20 errors like:
[00:00:13.235] W7: Downmix overflow (0: +1.7dB)

The players work fine and I can't listen differences or clipping.

I wait for a new windows binary (maybe Kurtnoise is in holidays) to test the new bandwidth methods.

jruggle

3rd August 2006, 03:49

There are a problem with jruggle sample but it isn't related with Dialog Norm.

Encoded with Aften (last binary from Kurtnoise), ffmpeg or SoftEncode (and with dnorm 17 or 31), when is decoded to wav with NicAc3Source (Bepipe/BeHappy) crash, and decoded with BeSweet-Azid there are more than 20 errors like:
[00:00:13.235] W7: Downmix overflow (0: +1.7dB)

The players work fine and I can't listen differences or clipping.

Now that is strange! That sample is just a regular wav file... It is one of the clips used in the vorbis listening tests. Maybe the problem is a decoder thing. What other settings are you using?

On another note. I have redone the configure/build system and made a separate libaften. It does not produce a .dll right now, only a static lib, but it will eventually. I will leave aften-current.tar.bz2 alone for a few days to allow for some testing before releasing version 0.04.

-Justin

raquete

3rd August 2006, 04:36

@ jruggle and tebasuna51
Now that is strange! That sample is just a regular wav file...
There are a problem with jruggle sample but it isn't related with Dialog Norm.i was talking about the "sample .wav".
i mean "clipping" as "more than 100%(0db),i got big red bars in audition using the sample .wav,then,i don't encode this source for test.

As far as dialnorm for music. I think it depends on what you define as "dialog". i don't define,this is the hard doubt,i'm lost.
if i use -31 sounds too loud,using -16(value found using my source following SomeJoe sticky is -16.5) sounds poor,without "life"...
do you have recommendations to encode AC3 musics?
edit i don't use decoder/receiver but 6 channels amplifiers with dvdplayer Dolby Digital 5.1 decoder built in.

ps: adjusting in AftenGui dialogue normalization 0(zero) means "none" ?

thanks.
:)

Rockaria

3rd August 2006, 09:35

Disclaimer : I am here just to exchange the facts and opinions audio specific, not for any boaring artificial attempts. Also any corrections will be appreciated.

That 2ch wav file is just max-gained, the peaks with no sign of clipped or hard-limited when expanded in Audacity, also played all-green(no red) with no attenuations on the channels. Some misinterpretations in the utils might have caused the clippings.

My understanding on the dialnorm is that it is used to normalize the listening environment(not to be deaf or frigntened) when switching between the clips or channels(sources) on Dolby decoders(receivers) based on the average volume level of (relative) signals that we call or define as 'dialogue' in Dolby clips.

The dialogue measured signal will be finally attenuated(adjusted) to the traditional (quiet) -31dB in the Dolby players/decoders/receivers based on the given relative meta value.

As used in most dd live solutions, the -31dB dialnorm won't adjust the decoding time dialogue level, and I guess the focus on the DRC and the volume knob(or album normalization) will be more useful if we consider the overall clarity and even volume level within/between(especially dolby/non-dolby preprocessed) the source(s).

In general there can be no default setting for dialnorm; the value depends on the nature of the program, and in the context of mixed programming it is essential for the setting to change from item to item. For a channel with uniform material, a fixed (but appropriate) setting may be acceptable. Generally a setting for dialnorm of -31 is unusual, required only for a few unprocessed wide-range movie soundtracks. For typical broadcast material (speech and popular music), the setting lies more often in the range of -15 to -20.
http://web.archive.org/web/20040716131627/http://www.dolby.com/tech/L.mn.0002.DDPEG1.pdf

tebasuna51

3rd August 2006, 13:41

i was talking about the "sample .wav".
i mean "clipping" as "more than 100%(0db),i got big red bars in audition using the sample .wav,then,i don't encode this source for test.
A integer wav like the sample can't have more than 100%(0db). Is a maximized wav (or normalized at 0 dB) with peaks at 0 dB.

if i use -31 sounds too loud,using -16(value found using my source following SomeJoe sticky is -16.5) sounds poor,without "life"...
do you have recommendations to encode AC3 musics?
Using -31 or -16 the signal is encoded at 100%, with -31 the decoder is instructed to don't attenuate and with -16 the decoder is instructed to attenuate 15 dB.
Using -31 the volume is similar to modern music CDAudio, mp3 normalized 100%, commercials in TV, ...
Using -16 the volume is similar to others ac3 Dolby compliant.
You always can turn the volume of your amplifier down or up.

ps: adjusting in AftenGui dialogue normalization 0(zero) means "none" ?
Means 0 dB (the decoder attenuate the signal until -31 dB).

Rockaria

3rd August 2006, 15:31

Then now the baseline is :
. the attached sample wav with max-gain should have no problems in ac3 encoding & decoding
. the SomeJoe's and Dolby's dialnorm explanations are basically SAME with slight different expressions
?

raquete

3rd August 2006, 17:47

That 2ch wav file is just max-gained, the peaks with no sign of clipped or hard-limited when expanded in Audacity, also played all-green(no red)
A integer wav like the sample can't have more than 100%(0db). Is a maximized wav (or normalized at 0 dB) with peaks at 0 dB.
the attached sample wav with max-gain should have no problems in ac3 encoding & decoding

or my eyes and ears are too bad,or audition or ... the sample (lol)

look in all screenshots the bar levels in red(clip) after 0dB.
whole sample wav playing:
http://img110.imageshack.us/img110/1157/sampleclippingbd5.th.png (http://img110.imageshack.us/my.php?image=sampleclippingbd5.png)

clipping in right channel 0:04.321
http://img110.imageshack.us/img110/5872/clip1ez4.th.png (http://img110.imageshack.us/my.php?image=clip1ez4.png)

clipping in left channel 0:14.356
http://img105.imageshack.us/img105/1663/clip2cq4.th.png (http://img105.imageshack.us/my.php?image=clip2cq4.png)

clipping (double) in left channel 0:28.840 and 0:28.841
http://img105.imageshack.us/img105/1472/clip3wx7.th.png (http://img105.imageshack.us/my.php?image=clip3wx7.png)

don't have sound in the blue lines! ? :confused: (zoom the pictures if needed)
more than 0dB and audition can't show it?

You always can turn the volume of your amplifier down or up.
yes,it change the volume but...
For typical broadcast material (speech and popular music), the setting lies more often in the range of -15 to -20.
... it don't change the quality of the sound.the problem is: or low and faded at -16,or too loud and crispy at -31.
both sounds ugly!

Means 0 dB (the decoder attenuate the signal until -31 dB). all right.

about the sound: i hear the cymbals too crispy,seems(are)...shiver.
please,comments about "everything" will be appreciated(about my ears include).

thank you boys,you are very specials in the team!
:)

Rockaria

3rd August 2006, 20:38

Hi raquete,

The audacity expanded images(by the appearance) clearly show the symptoms of the clipped wav forms to me also, so you don't have to worry about your EYES.
The problem still remaining is there can be some different interpretations(although I expressed as mis-) between the tools.
So I tested with one more popular FREE tool : foobar2k v0.92->replaygain scan!
name : 41_30sec
track peak : 0.999969
track gain : +0.54dB

So it is not max-gained yet(0.54dB gain room left) and the issue seems to be which tools to use not to have the clips clipped.

[edit] I notice the proportional variable dB scales on the Y-axis making it hard to assume the 0dB positions. Maybe you need to change the scale view.

raquete

3rd August 2006, 22:15

so you don't have to worry about your EYES.
lol. now about my ears,audition and the sample...

Maybe you need to change the scale view. done.

http://img311.imageshack.us/img311/8265/clip2aww2.th.png (http://img311.imageshack.us/my.php?image=clip2aww2.png)

(zoom please)...the thin blue line is in 0dB.the sound after the blue line is above 0dB.

about the sound: i hear the cymbals too crispy,seems(are)...shiver.
as i hear and audition advice that the sound is clipping,i can trust in my ears and in audition too.
:p
trust,when i was playing the sample my first impression was the sound quality("crispy" as i posted)

thanks!

edit: see the bargraph in the pictures showing clips in red (full scale)

http://img403.imageshack.us/img403/2471/clip4gs0.png
in this post http://forum.doom9.org/showpost.php?p=858952&postcount=121

Rockaria

3rd August 2006, 23:44

Oh yes, I now find the two clipped out-of-phase positions @ : 14.35700 & 14.35710, :thanks:
I assume the audacity's(and foobar 2k, ffdshow...) interpretation & expression : hard-limit on the 0dB, the audition's one : rebuild the clipped area.

Maybe we can say the audition is more intelligent but artificial as well. So it also depends on the personal interpretations?
I personally think in my environment, it has very little(ignorable) negative effect to be used as a transcoding/playing source...

jruggle

3rd August 2006, 23:59

Hi,
I need a favor. :)
I am trying to setup a cross-compile environment so that I can build Windows binaries for others to use. Could someone test these executables to see if either works?

aften.exe (http://jbr.homelinux.org/aften/aften.exe)
aften_g.exe (http://jbr.homelinux.org/aften/aften_g.exe)

Thanks!
-Justin

Rockaria

4th August 2006, 01:08

jruggle

4th August 2006, 01:21

Both work (correctly) with -b mode but not correctly in -q mode.
i.e, aften -b 640000 -m 0 -acmod 2 41_30sec.wav aaa1.ac3
...
. no effect in -m mode change
. waiting for the v0.04 window version.

Thank you for the feedback. I'm just glad it works at all (runs).

What do you mean by "not correctly in -q mode"? Does it produce an ac3 file, but it just doesn't decode right? or does it crash when you're encoding?

I'm hoping that "no effect in -m mode" means that the file isn't much different, not that the output file is bit-identical. The sample file doesn't have a lot of channel correlation, so I would expect rematrixing not to do much...if it doesn't do anything at all that's a bug.

As far as version 0.04...that was pretty much it. Unless I run into any major bugs I need to fix, the current version will become 0.04 in a day or two. The VBR problem worries me...but I'll wait for more feedback before freaking out. ;)

Thanks again,
-Justin

tebasuna51

4th August 2006, 02:56

Rockaria

4th August 2006, 03:52

What do you mean by "not correctly in -q mode"? Does it produce an ac3 file, but it just doesn't decode right?
Exactly.. When I verified the encoded ac3 with softencode, it was normal. But it always showed wrong but different playbacks if I change the -q values.

The -m mode change with CBR had no effects on the bit rate of course, but also I could hardly notice the quality differernce(to measure this short clip to my ears).
When used with the VBR(-q), I guess it will clearly show the bit efficiency...

jruggle

4th August 2006, 04:40

@jruggle
Your windows binary work ok.

I tried the test suggested in aftenblog with:
aften -b 192000 -w -1 -bwfilter 0 41_30sec.wav z41_-1_0.ac3
aften -b 192000 -w -1 -bwfilter 1 41_30sec.wav z41_-1_1.ac3
aften -b 192000 -w -2 41_30sec.wav z41_-2.ac3
but my ears can't find any difference, sorry.

wonderful! thank you.

I gave the test a fresh listen just to reassure myself that it wasn't a placebo effect. I did still notice a difference. However, I also realized that the variable bandwidth cutoff ended up quite a bit less than the fixed adaptive bandwidth, so I tried the test with "-w 32 -bwfilter 1". I still noticed a difference...enough to convince me, but I really need to get or make a working ABX test program so I can be less subjective about it.

Anyway, I'm excited that my cross-compile worked. Version 0.04 will be released, probably on Saturday, as both source and binary.

raquete

4th August 2006, 19:08

jruggle

4th August 2006, 20:28

from 42_DDFAQ.pdf (Dolby Inc.)
"Dolby Digital can process up to 24-bit digital audio signals over a frequence range from 20Hz to 20KHz on the full-range channels..."

Aften can encode 20 or 24bit? if don't,...why not as new feature for test?
(if yes,let me out to buy some fireworks :p )

Actually the format can do more than that. I think the doc from Dolby is referring to what Dolby certified encoders are required to do.

Aften can handle 8-bit/16-bit/24-bit/32-bit/float/double wav input. All of these are converted to double-precision floating point samples, which is what libaften expects as input. The encoder runs the MDCT in floating-point then quantizes the coefficients to the AC-3 internal floating-point format (separately coded exponents and mantissas). The exponents do allow for a full 24-bit range. Actually it's more like 25-bit, but I won't get into that. Any precision above that just gets lost in the quantization...it is a lossy codec after all.

One thing I can add is 20-bit wav support (or 17 or 12 or whatever). This should not be too difficult.

ps: anyone is testing/using VBR?
I am considering relegating the VBR mode to non-default until I can do more testing myself or get lots of feedback. The default CBR bitrate will depend on the number of full-bandwidth channels being encoded. I'm thinking:
1 = 96kbps
2 = 192
3 = 256
4 = 384
5 = 448

If I decide to make VBR default again it will depend on DVD/HD-DVD/Blu-ray support (actual, not spec). If players will play it, then it should get more use and be the default...otherwise it will still be there for those who wish to use it.

BigDid

4th August 2006, 20:55

The default CBR bitrate will depend on the number of full-bandwidth channels being encoded. I'm thinking:
1 = 96kbps
2 = 192
3 = 256
4 = 384
5 = 448

Hi,

Please consider adding 320 for CBR, as it is in the specs (from a Tebasuna post): http://forum.doom9.org/showthread.php?p=854685#post854685
and a multi-channels compromise between 384-good- and 256-acceptable to bad- (I have been told). Thanks.

Did

Rockaria

4th August 2006, 21:20

That looks like enough CBR bit rates for backup purposes. I also think the 64k increments from the full 2ch(2.x~) as default bit rates looks more reasonable.
My previous feedbacks were for just brief verification that it runs on Xp. Now I did some more tests and analysis on the below result which might be useful before the release of v0.04.
Encoded with Aften (last binary from Kurtnoise), ffmpeg or SoftEncode (and with dnorm 17 or 31), when is decoded to wav with NicAc3Source (Bepipe/BeHappy) crash, and decoded with BeSweet-Azid there are more than 20 errors like:
[00:00:13.235] W7: Downmix overflow (0: +1.7dB)
<condition>
. 44_30sec.wav 2ch wav source : has some soft clippings(around 10(20?)~) not noticeable(to me)
. also tested with another 2ch wav(no clippings) and 6ch wav to verify
<results>
. softencode gave no warning, encoded & decoded well
. aften encoded (well) with no warning but failed to play resonably on -q mode encodings : both 2ch & 6ch
- aften encoded 2ch clips failed to decode with nicac3source avisynth plugin regardless of the clippings : 6ch is ok
<conclusion>
. 2ch encoding issue : compatibility problem in some decoders
. -q mode issue : seek or sync problem ?, not proper decoding(&encoding)
. clipping issue in decoding/reading : some tools(like audition) are rebuilding the clipped area, may require some pre-gaining(~-3dB) process when you get warnings.

[edit]
The NicAc3Source avisynth plugin(the original version) also failed to play the softencode encoded 2ch ac3 regardless of the clippings.
So it seems not the aften's problem.

raquete

4th August 2006, 22:49

jruggle

Aften can handle 8-bit/16-bit/24-bit/32-bit/float/double wav input.
of course,i know and use. i used 2 big sources(.wav 6-channels 48000Hz, 32-bit with 3,24Gb 50:25.200 and 3,19Gb 42:340.360)in Aften and works great!
i want to know about AC3 48K 20 or 24 bit output. :stupid:

I'm thinking:
1 = 96kbps
2 = 192
3 = 256
4 = 384
5 = 448
Please consider adding 320 for CBR, as it is in the specs
I also think the 64k increments from the full 2ch(2.x~) as default bit rates looks more reasonable.
jruggle,i encode min 512K,sometimes 640K...always 5.1(never 2.0)
please,extend the list, don't forget me. ;)

best regards and thanks for explanations.

Rockaria

4th August 2006, 23:14

he he! I forgot the conditions(and those are just default values per -acmod that can be overridden).;)
When the 'rematrixing, channel coupling and optionally VBR/ABR' get extremely effective/efficient as intended, it won't need that much bit rates to please your sensitive ears. I personally prefer no transcoding at home though(I use 6ch 640k dd live for the codec compatibility to my receivers).

raquete, you are one big AC3 fan and will never be forgotten by jruggle, I bet.:cool: :)
/me on a long travel.

tebasuna51

4th August 2006, 23:39

[edit]
The NicAc3Source avisynth plugin(the original version) also failed to play the softencode encoded 2ch ac3 regardless of the clippings.
So it seems not the aften's problem.
Yes, but i think is a NicAc3Source problem with 44.1 KHz, not with 2 or 6 channels. Can you confirm this with your samples?.
HPlease consider adding 320 for CBR, as it is in the specs (from a Tebasuna post): http://forum.doom9.org/showthread.php?p=854685#post854685
and a multi-channels compromise between 384-good- and 256-acceptable to bad- (I have been told)
I agree with defaults proposed by jruggle, you can use 320 or any other valid value in command line.

BigDid

5th August 2006, 00:16

defaults[/B] proposed by jruggle, you can use 320 or any other valid value in command line.
Hi detractors :)

I have already stated I am really not at ease with command line apps :o

@ jruggle
[blatant advertising on]
Please imagine all the potential AC3 encoding people (hundreds, surely thousands) just waiting to learn there is a free and performant tool available, either directly, with a GUI or included in a powerfull all-in-one audio app (Behappy).
I'm sure the first reaction will be:
-Is it easy to use? -> Yes there is a GUI
-Can I do more but still with a GUI -> Yes with Behappy
[blatant advertising off]

Specialists are praising your tool; I believe it's just a matter of time before non-specialists will also use it.
Don't forget the "PAY PAL donate" button and the 320-CBR :D

Did

Rockaria

5th August 2006, 00:33

Yes, but i think is a NicAc3Source problem with 44.1 KHz, not with 2 or 6 channels. Can you confirm this with your samples?.
Is it a known problem? If not, it must be a small but useful discovery by us all.

Indeed, the previous 6ch clip was 48k, so I made 2 wav files to test it fully: 48k 2ch wav & 44.1k 6ch wav
==>All the 44.1k ac3 encoded crashed with NicAc3Source, but 48k played OK regardless of the channels(-acmod).

Some other things that may require your confirmations :
. I interpreted the audition performs 'rebuilding' on the clipped area making it to have wider range(beyond 0dB), you may be able to confirm the BeSweet-Azid if it becomes OK with -3dB pre-attenuated 44_30sec.wav. (the attenuation option might be useful for aften)
. the behappy environment(avs wrapper) seems to downsize(16bit) the avs input stream making the aften's wider input capability an overkill, you might have the latest source to confirm.
. some other things not related to aften...

/good to share

jruggle

5th August 2006, 01:37

jruggle
of course,i know and use. i used 2 big sources(.wav 6-channels 48000Hz, 32-bit with 3,24Gb 50:25.200 and 3,19Gb 42:340.360)in Aften and works great!
i want to know about AC3 48K 20 or 24 bit output. :stupid:

Ah...well Aften is not a decoder, but the AC3 files produced by Aften can take advantage of the full AC3 accuracy range...depending on the depth of the source material. Or am I misunderstanding your question...?

jruggle,i encode min 512K,sometimes 640K...always 5.1(never 2.0)
please,extend the list, don't forget me. ;)

You can still encode at 512 or 640. I chose 448 as default because that is what Dolby recommends for 5.1 content.

As far as the issues with that sample I've been using...I didn't notice the clipping before, so maybe I should find a better sample to use.

I don't think that Aften should do any fancy audio processing (clipping rebuild/attenuation). I'll leave that up to the professional audio apps. I do hope to provide resampling though, as it will be a huge benefit to those needing 48kHz for DVD (and apparently NicAc3Source) compatibility.

Thanks,
Justin

tebasuna51

5th August 2006, 02:46

. I interpreted the audition performs 'rebuilding' on the clipped area making it to have wider range(beyond 0dB), you may be able to confirm the BeSweet-Azid if it becomes OK with -3dB pre-attenuated 44_30sec.wav. (the attenuation option might be useful for aften)
I make a previous test at 80% (-2dB) without Azid warnings. But this levels of clip at a few peaks don't affect the overall quality.
. the behappy environment(avs wrapper) seems to downsize(16bit) the avs input stream making the aften's wider input capability an overkill, you might have the latest source to confirm.
With AviSynth 2.57 (alpha) and Bepipe i make some test, and work with 32 bit wavs (int and float). From 2.57 docs:
* WavSource() accept audio streams of type WAVE_FORMAT_IEEE_FLOAT.
* Adding global OPT_AllowFloatAudio=True to your script enables WAVE_FORMAT_IEEE_FLOAT audio output.

Rockaria

5th August 2006, 03:31

Then the Besweet-azid seems to perform kinda 'rebuilding' on decoded ac3 encoded originally clipped area(or some other interpretations on the streams).

I am less familiar with the behappy wrapper than the original open avisynth environment.
So you can regard I am well aware of the avisynth's wide range capability of the bit sizes.
However, in the older(20060226) source of the AvisynthWrapper.cpp I read :

if (inf.HasAudio())
{
*originalSampleType = inf.SampleType();
if( *originalSampleType != SAMPLE_INT16)
{
res = pstr->env->Invoke("ConvertAudioTo16bit", res);
pstr->clp = res.AsClip();
infh = pstr->clp->GetVideoInfo();
if(infh.SampleType() != SAMPLE_INT16)
{
strncpy(pstr->err,"Cannot convert audio to 16bit",ERRMSG_LEN-1);
return 6;
}
}
}
which forces any audio stream process beyond 16bit not that useful including aften ac3 encoder through pipe.
There might be some reasons for this restriction or already excluded(which I cannot confirm) that behappy users with aften certainly want?
(I am not sure if this issue is discussed already)

Thanks.
[edit] I confirm the latest cpp source dated 05/09/2006 is unchanged in the mentioned area.

jruggle

5th August 2006, 03:31

I made the bitrate changes that have been discussed. This is the final change before the release of v0.04 tomorrow.

changes: CBR is default. Bitrate is given in kbps. Also, you can use both '-b' and '-q' to encode VBR with a maximum bitrate.

raquete

5th August 2006, 04:40

Or am I misunderstanding your question...?
Justin
maybe ...is my fault,excuse me(poor english and lots of typos)

first
what i do:
load any cd track,convert sample type 48k-32bit,extract center and surrounds,save this all (LR,CLFE and SLSR) as "32 bit normalized float(type 3) default" in adobe audition.
load this tracks in multichannel encoder and "export as one interleaved,6-channels wave file" windows PCM waveform audio - 32bit,normalized float (type 3) or as 32 bit,4-byte integer (type 1).
(sometimes exporting 32bit,normalized float (type 3) in audition give some clicks, then i export as 32 bit,4-byte integer (type 1) )
...i take this "one interleaved,6-channels wave file" to AftenGUI to get AC3 5.1 512K.

second:
now my answer (after this big road) and i'm not sure if this is possible or if works:
can Aften encode this AC3 5.1 512K in 20 or 24 bit?

thanks Justin
;)

jruggle

5th August 2006, 04:59

can Aften encode this AC3 5.1 512K in 20 or 24 bit?

I see. The short answer is "there is no such thing". :)

AC3 does not have a specific bit depth. The closest thing to it would be something like 21-bit floating-point. AC3 uses an exponential floating-point format with an exponent of 0 to 24 and a 0-bit to 16-bit variable-depth mantissa. This sort of means that it has the precision of 16-bit, but the range of 24-bit. Also, what is encoded is in the frequency-domain, not in the time-domain like wav.

I hope this answers your question.

-Justin

raquete

5th August 2006, 05:08

I see. The short answer is "there is no such thing". :)

AC3 does not have a specific bit depth. The closest thing to it would be something like 21-bit floating-point. AC3 uses an exponential floating-point format with an exponent of 0 to 24 and a 0-bit to 16-bit variable-depth mantissa. Also, what is encoded is in the frequency-domain, not in the time-domain like wav.

I hope this answers your question.

-Justin
(...living and learning)

thank you so much Justin,very clear.
:)

Kurtnoise

5th August 2006, 07:52

ouchhi. One week of vacation and some new stuff is already here. Great...:)

So, I uploaded a fresh compile here (http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften-0.03-dev.zip). I'll update the GUI as soon as 0.04 will be release. (I hope before Monday coz I go back in holidays next week...;))

@Justin : for -bwfilter/dcfilter/lfefilter in the command help, which value is default ? For the moment, both values (0/1) are sticked as default...Sorry, I've no time to check the code carefully.

@BigDid : ac3 <--> ac3 is useless imo. Keep in mind that this is a lossy format. And I really don't know why you reencode your audio stream coz dvds/cds are more and more cheaper nowadays.

jruggle

5th August 2006, 14:52

Glad to see you back Kurtnoise.

So, I uploaded a fresh compile here (http://kurtnoise.free.fr/index.php?dir=Aften/&file=aften-0.03-dev.zip). I'll update the GUI as soon as 0.04 will be release. (I hope before Monday coz I go back in holidays next week...;))

It will be later today (US Eastern time).

@Justin : for -bwfilter/dcfilter/lfefilter in the command help, which value is default ? For the moment, both values (0/1) are sticked as default...Sorry, I've no time to check the code carefully.

oops! will be fixed in 0.04. The default is 0 (no filter).

BigDid

5th August 2006, 18:02

@BigDid : ac3 <--> ac3 is useless imo. Keep in mind that this is a lossy format. And I really don't know why you reencode your audio stream coz dvds/cds are more and more cheaper nowadays.
Hi Kurtnoise and happy holidays,

We have had that kind of exchange in the past but you're the expert, so I will try to argument:
It's not ac3->ac3 for pleasure it's for:
1. playing multichannels audio on a SAP (AAC or OGG not playable on most models)
2. getting the low sound/dialogs (or too much expansion) to a better listening level.
3. Cheaper DVD media are yet to arrive here, but I can't have the butter ... and so on

I know, from other threads like Behappy that some people, like me, wants to DRC/normalize for reason 2. I also use a higher bitrate than before (384 or 320) to keep more bandwidth so, at least, some of your advices are getting through :D

Did