BeSweet's ac3enc creates extremly low-volume ac3

[Koke]

Is it possible that reencoded AC3 (down to 224 kbps)
sounds too quiet. Or is it my audio setup?
I used AC3machine default settings.
Help please.

[DSPguru]

:logfile:

[stargazer]

Is it possible that reencoded AC3 (down to 224 kbps)
sounds too quiet. Or is it my audio setup?
I used AC3machine default settings.
Help please

I have the same problem (the lower the bitrate, the quieter the sound), and I've tried to solve it with extra gaining, but even 12 db gain on 448 to 320 transcoding was enough only to reach 70-80% of the original volume level.

here is the log (this is 448 to 320, but when output bitrate is lower, sound is much quieter)


BeSweet v1.4RC7 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using AC3enc.dll v0.2 by Gerard Lantau & Dg (http://ffmpeg.org).

Logging start : 09/13/02 , 12:07:40.

G:\mali_programi\BeSweetv1.4RC7\BeSweet.exe -core( -input G:\Documents and Settings\XliQ Stargazer\My Documents\avi\pottie AC3 T01 3_2ch 448Kbps DELAY 0ms.ac3 -output G:\Documents and Settings\XliQ Stargazer\My Documents\avi\pottie AC3 T01 3_2ch 256Kbps DELAY 0msnew.ac3 -logfilea BeSweet.log ) -azid( -g 12db ) -ota( -d 0 ) -ac3enc( -b 320 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : G:\Documents and Settings\XliQ Stargazer\My Documents\avi\pottie AC3 T01 3_2ch 448Kbps DELAY 0ms.ac3
[00:00:00:000] | Output: G:\Documents and Settings\XliQ Stargazer\My Documents\avi\pottie AC3 T01 3_2ch 256Kbps DELAY 0msnew.ac3
[00:00:00:000] | Floating-Point Process: No
[00:00:00:000] +-------- AZID -------
[00:00:00:000] | Total Gain: 12.0dB, Compression: None
[00:00:00:000] | LFE levels: To LR -INF, To LFE 0.0dB
[00:00:00:000] | Center mix level: BSI
[00:00:00:000] | Surround mix level: BSI
[00:00:00:000] | Dialog normalization: No
[00:00:00:000] | Rear channels filtering: No
[00:00:00:000] | Source Sample-Rate: 48.0KHz
[00:00:00:000] +------- AC3ENC ------
[00:00:00:000] | Bitrate method : CBR
[00:00:00:000] | AC3 bitrate : 320
[00:00:00:000] | Channels Mode : 5.1
[00:00:00:000] | Error Protection: Yes
[00:00:00:000] +---------------------

the rest of log file looks something like this:
...
[00:00:18.282] W7: Downmix overflow (1: +1.5dB)
[00:00:18.288] W7: Downmix overflow (1: +1.3dB)
[00:00:18.293] W7: Downmix overflow (1: +1.7dB)
....

[Bobby]

Hi guys,
I fix this low volume problem by using ac3machine and enabling "Dynamic compression" @ normal, and also enabling "Auto find maximum gain".
I must add though that I've only tested this when transcoding a 5.1 ac3 track to a lower bitrate (192kbps) 2.0 ac3 track, But I'm guessing that it is the same principle's for transcoding 5.1 to lower bitrate 5.1
Hope that helps ya ;)

[DSPguru]

in fact, i've noticed that property of ac3enc too.
i believe i can change that, but it will take me some time. i'm now working on other stuff..

[theReal]

I also noticed this behaviour on a re-encode from 448kbit 5.1 to 320kbit 5.1 - in the end I was using azid's auto-find max. gain and normal dynamic compression, but I think it resulted in some distortions in various scenes (I don't trust the azid auto-gain anyways, it always seems to distort the sound - is that possible?)

[stargazer]

*Originally posted by Bobby *
Hi guys,
I fix this low volume problem by using ac3machine and enabling "Dynamic compression" @ normal, and also enabling "Auto find maximum gain".


I don't like "Dynamic compression" or any sort of compressors that reduce dynamic range, but i tried "Auto find.." and it seems that it really _only_ search maximum gain and not apply it.


I must add though that I've only tested this when transcoding a 5.1 ac3 track to a lower bitrate (192kbps) 2.0 ac3 track, But I'm guessing that it is the same principle's for transcoding 5.1 to lower bitrate 5.1
Hope that helps ya ;)

No, it's not the same principle because 2 channel 192 kbps audio is not so low bitrate (96 kbps/channel) , and in 448 5.1 you have 448/6=74.6 per channel which is lower bitrate than yours used in 2.0 ac3 track...

[DSPguru]

maximal gain is being applied (you can simply test it by not using '-g max'), the issue belongs to ac3enc.dll. i even sorta find it in the source-code, but i don't have the time to fix it currently.
DRC, Gain, Normalize and other methods are always relevant, but in this case there's a more basic problem.

[MaTTeR]

I turn every feature off and just do a straight transcode, otherwise I too have heard the distortions. Glad to hear the low volume issue was tracked down.

To avoid distortion and/or artifacts my command line usually looks like-

"C:\Program Files\BeSweetv1.4RC8\BeSweet.exe" -core( -input "d:\false AC3 T01 3_2ch 448Kbps DELAY 0ms.ac3" -output "z:\false_256kbps.ac3" -logfile "C:\BeSweet.log" ) -azid( --maximize ) -ac3enc( -b 256 ) -profile( ~~~~~ Default Profile ~~~~~ )

[quantum]

I've got this same low volume problem. I downsampled a 5.1 channel to 224 or 256 using default settings in AC3machine and the resulting volume is much lower, almost inaudible when played at the same volume as the original.

[knight_inferno]

Hello ^^
For the first time last night i used AC3Machine to create a 5.1 ac3 448, down to 256 5.1..i hadn't done anything like it before (Using AC3's and all in AVI's) after testing out the sound by putting it into the avi, indeed it was extremely quiet! I followed the doom9 guide on this, and even used the original audio just to check and that was at a decent volume, still a little quiet, but i find DVD's tend to be a little on the quiet side anyhow :rolleyes:

Anyhow, i know that you said it was a problem with azid (I think that was it) and this post was started a couple of months back, has there been some kind of "Fix" since this time that i have been unable to find? :(

Any advice would be great! If i can't do it i might as well do the usual MP3's i always do :D

[MaTTeR]

Originally posted by knight_inferno
has there been some kind of "Fix" since this time that i have been unable to find? :( AFAIK the problem still exists somewhat, one movie I did last week required me to crank my amp almost all the way up just to hear the dialog. However, some moves re-encode with better loudness/gain for some reason. Then you have my extreme case from last week where it seemed I had lost almost 50% of the original volume:confused:

[quantum]

My fix for this was to use Softencode to perform the downsampling following an older guide posted here. It may be slower, but the resulting AC3 file has the same volume.

I can't help wondering, if the volume is wrong with the ac3machine/besweet technique, are there other problems that are not so obvious?

[brett]

After seeing this thread, I compared some trasncoded AC3 tracks I'd done. Tracks I have transcoded from 192kbps Dolby 2.0 to 80 or 96kbps stereo are actually louder. Not many of them are very significantly louder, but they're definitely louder.

I use AC3Machine 0.41 & BeSweetv1.4 to transcode the audio for director's commentary and special features when ripping DVD -> DVD-R. I always leave the main Dolby Digital 5.1 audio track as is, so maybe the quietness people are experiencing is because of the 5.1 -> 2.0 downmix. My settings are usually:

Channels Mode: stereo
Dynamic Compression: normal
Auto Find Maximum Gain: ON
Bitrate: 80 for commentary, 96 for special features

I leave everything else at the default settings. The only problem I've noticed is that if I try to set the bitrate lower than 80, BeSweet gives me a lot of errors while transcoding.

[aryan]

I fix this low volume problem by using ac3machine and enabling "Dynamic compression" @ normal, and also enabling "Auto find maximum gain".

@Bobby
I must agree with you that this is the solution for solving that problem. I've made my own tests and ONLY when using DC@normal and azid autofind max gain, the results are ok. In all other cases the sound is MUCH quieter. I've used for my test BeSweet 1.5b6 and AC3 from Shrek.

@DSPGuru
I'm watching @ that thread from more than 2 months, and I still waiting for expert advice from you - which settings are really right to get maximum possible quality, and if there is a bug when it will be fixed (and where actually is the bug - in BeSweet or in ac3enc.dll).

@all
I agree with quantum that for now the best result which can be given is with Softencode. If someone wondering where to get this fine software (Sonic Foundry discontinued offering it) here is the link:
link to warez removed
But for christ sake its damn slow :(

And there is something strange I've noticed. When I reencode Shrek AC3 with Softencode, and then load reencoded or original AC3 track back, there are stream settings of the track which are same for both:
Dialog normalization: -27dB
Surround mix: -3dB
Center mix: -3dB
Ofcourse you can set these params when reencoding, and the result is just what u r setting, and I prefer to save original parameters

BUT if I load reencoded with BeSweet Shrek AC3 track I see that stream settings are changed:
Dialog normalization: -31dB
Surround mix: -6dB
Center mix: -4.5dB
That result is no matter what settings u use in BeSweet (for example I've tried to change them in azid, but the result is always the same

I think DSPGuru is the only one here which can explain that strange phenomenon:confused:

And I must say that I cant find difference using my own ear between track reencoded in Softencode, and the same track reencoded in BeSweet(using normal DC, auto find max gain and hybridgain).
Thats y i'm asking all of u do u know some program where i can load them both and compare them graphically (Sonic Foundry 5.0b for example cant open AC3 at all)

And ofcourse if U think my english is bad, dont be angry :cool: ... nevertheless I'm from Bulgaria :cool:

[DSPguru]

1. sorry guys, i'm too busy to currently work on this.
2. there is no bug with BeSweet.exe
3. true, ac3enc's Dialog normalization,Surround mix,Center mix values are static and never changes. (currently!)

Originally posted by DSPguru
the issue belongs to ac3enc.dll. i even sorta find it in the source-code, but i don't have the time to fix it currently.

[aryan]

Well, since yesterday (29.11.02) there is a new BeSweet release- 1.5b8.
And guess what ... there is new ac3enc.dll in it!:devil:
I pray that finally DSPGuru fix it:cool:
I'll give it an immediate try and inform you about results.

Also about removing link to Sonic Foundry Soft Encode as warez link, I think that this is DSPGuru last decision right and I'm not going to debate it.
But for myself I cant identify dvd.box as warez site. All the programs you can see there are freeware, and if you look at theirs first page there is poll especially asking for if you want to see shareware programs in the future.
And especially for Soft Encode, Sonic Foundry no longer support it and no longer even offer it. Here is the link you can see this:
http://www.sonicfoundry.com/Products/NewShowProduct.asp?PID=12

I'm not a licensing guru, but in my oppinion when you refuse your own children, they become noones, and thus everyone owns.
The other example is with MS Dos. They no longer support it or sell it and thus owning it is not a crime.
Am I not right?

[DSPguru]

the new ac3enc.dll is just a faster compile. that's all :(

[jamiro]

It's been 4 months and this problem is still unsolved, so there's my question: DSPguru are you able to fix this problem or not?

"(...) the issue belongs to ac3enc.dll. i even sorta find it in the source-code, but i don't have the time to fix it currently.
DRC, Gain, Normalize and other methods are always relevant, but in this case there's a more basic problem."

If it's a "basic problem" then please solve it finally because many ppl are waiting for solution.

[MaTTeR]

No need to be demanding, Dg's free time is obviously limited. Someone already pointed out a working solution above which is to use Sonic Foundry's Soft Encode. It work's great but only slower than BeSweet:)

[theReal]

And especially for Soft Encode, Sonic Foundry no longer support it and no longer even offer it. I think it is allowed by international law to spread music which has a certain age and which is not published anymore. There has been a site publishing only old, rare and unavailable reggae albums - and nobody would sue the owner because he stated clearly that if anyone wanted to make use of the songs' copyrights, he would remove the mp3s from his site immediately.
But unfortunately, something else stopped this poor guy, have a look at his page now: http://reggaescientist.org/

:( :( :(

[DSPguru]

@jamiro
i appreciate your concern, but you should know that lots of people are also waiting for other features/bug-fixes, which, to my understandment, are more important. (like the fact that some FRCs ends up with clicking stream).
so, i guess, i have to prioritize.

lately, i have been very busy in my personal life, and i hardly been online, not to mention having the time to code.

on a second note,
i recall lots of people trying to convince me to publish BeSweet's source with claiming that the development would be faster.
but as you can see, this is not true. until today, 4 monthes already, no-one had fixed this bug in ac3enc (being open-sourced), and no-one had written a fade-out plugin (which also has open-source samples).

[jamiro]

@matter @dspguru
I'm not demanding, it would be rude. I just pointed out that this problem is still unsolved and i thought that solving it won't take a long time since dspguru know's where's the bug. I'm also a programmer and i know that sometimes 90% of time takes finding the bug and only 10% or less fixing it. Please dont't take me wrong i'm not demanding, it's your spare time and it's your decision what to do with it.

[Rombaldi]

tried using BeSweet (and AC3 Machine) to convert the MP2 track to AC3. Conversion no problem. Remux OK. Sync, on the nose.

but it's way too low. noticably lower than it was before.

How can I boost the AC3 up and try to match the original volume.


... and please DON'T say "don't convert it to AC3". I need to for compatibility with another application. Not an option, needs to be AC3.

[DJ Bobo]

Set "Compression characteristic" on "None"

[quantum]

Is there any progress or additional information on this issue?

[bira]

Again,
Is there any progress or additional information on this issue?

[DSPguru]

i suggest you to send your comments to : ffmpeg-devel@lists.sourceforge.net

[Idefixus]

Hi,

i tried to reduce the bitrate to 224 kBit, but after this step, the ac3 is extrem quietly. I used the options Auto Find Maximum Gain from ac3machine, but it seems not to work.

Is it possible to normalize the ac3-stream after transcoding without transcode it again?

[michaelsez]

Here read this.

http://www.geocities.com/transcodeac3/#_Toc522546849

[ihgl]

1. Have an 23,967 AVI with Ac3 5.1 audio. Want to make a DVD PAL with AC3 Audio - authorizing program SpruceUp/Maestro
2. Rips the audio with Nandub, get a wav file
3. "Clean" the wav up with BeSlice to an AC3.
4. Uses BeSweet to convert from 23,967 to 25 fps, gives an AC3
5. The new Ac3 plays great in Winamp/Mplayer
6. Compile the m2v and Ac3 in SpruceUp/Maestro...

Problem: The amplitude gets far too low, sound hardly audible..

Have tried a lot of options in BeSweet without success. Even if I use BeSweet without framerate conversion the amplitude gets far too low. With the volume control on max you can hardly hear the sound.

My question: What does happen to the sound in BeSweet? Am I doing some fundamental mistake? The original audio file, produced with Nandub/BeSlice sounds great if you mux it with the m2v file in SpruceUp/Maestro. It is just that I have to convert the file into a 25 fps one in BeSweet - and something goes wrong there. Does anyone know what mistake I am doing here???

[DJ Bobo]

nothing happens, it's ac3 behaviour to lower the volume ¬.¬

You've got to decode your ac3 to 6 separate mono wav files, amplify them equally: look which file has the largest peak, and gain by that peak. For example, the largest peak is -11db (all others have for example -12, -15, and such), make a gain of 11db on all files.
Adjust the length of each wav.
Encode those 6 wavs to AC3 5.1 @ 448kbps.
The volume should be comparable to your original ac3 file.

[ihgl]

Originally posted by DJ Bobo
nothing happens, it's ac3 behaviour to lower the volume ¬.¬

You've got to decode your ac3 to 6 separate mono wav files, amplify them equally....

Thanks a lot. I will give it a try. Have not been dealing with AC3 audio before and knew I must have missed something.... ;-)

[ihgl]

Originally posted by DJ Bobo
nothing happens, it's ac3 behaviour to lower the volume ¬.¬



Well, there is one thing I do not understand though. The original AC3 that came out from Nandub/BeSlice (fps 23,967) works perfectly when muxed with the m2v in SpruceUp - no amplitude change at all. It is when I try to do the framerate conversion in BeSweet that the amplitude almost decreases to "zero". So something must happen in BeSweet. And I think I have tested all options ??

[barking_mad]

I appreciate everyones time is valuable here especially DSPguru as I've taken on board what's he's said about other issues which have priority over this. But has there been any progress on this matter?
I've read the posts regarding other methods, but like everyone else I suppose I'm looking for a 1 stop shop to re-encode an AC3, without the resulting audio volume being considerably lower.
My amp usually opperates at a volume seeting of about 20%. After re-encoding the AC3 it has to be pushed waaay up to 85%.
If anybody knows of a quick workaround, that doesn't involve multiple proggies I'd be extremely greatful.
Thanks for reading.:confused:

[Mikel]

I did some test doing AC-Trancoding as well.
Indeed the sound level decreases quite a bit.
For me, AC transcoding is useless in this way. A lot of sound quality is lost that way.
I will send e email to the above adress and i hope that others will follow my example so something might be done about it.

The more emails the higher the chance it will be fixed.

I am interested for other reasons that it will be fixed. There have been some tests on a German page comparing the audio quality of the ac3enc.dll with SoundForge SoftEncoder. The comparison show frequency responses of the signals.
However, ac3enc.dll is superior in that way.

So hopefully, someone can fix it soon.

Cheers

Mikel

[gizmau]

its the fault of ac3enc, not yours. take a look at this thread http://forum.doom9.org/showthread.php?s=&threadid=33535.

[Kilyan]

Originally posted by aryan
[B@all
I agree with quantum that for now the best result which can be given is with Softencode. If someone wondering where to get this fine software (Sonic Foundry discontinued offering it) here is the link:
[/B]

That's ok. But I do have softencode 5.1 from sonic but no way to reencode an ac3 from dvd. I mean I extract an ac3 from the dvd then load it to softencode then no buttons to reencode no menus to do that, neither can I disassemble to 6 separate mono waves, how do you reencode directly or inderectly with softencode?

[Mikel]

There is a function in BeSweed to Decode AC3 to 6 mono wav files.
After that you can reencode it with any AC3 encoder software.

Cheers

Mikel

[Kilyan]

Thanks Mikel!
Some other questions:

Some settings flags are shown by softencode for the original ac3, but
how do I know that besweet does invers this process? Can you post your batch?
I mean there is for example the 90 degree shift for the rear left and rear right(By the way that sounds good but means nothing I think, It would be stupid to delay by 90 for each frequency), what is besweet doing with that after decoding it to 6 mono waves? Does the 90 stay in the waves or this is really just a flag in the ac3 stream, and is neglected?
And please tell me also what parameters to set in softencode or scenarist ac3 encoder, because it is not described.
I mean digital deemphasis lpf/bpf/hpf and other stuff (I know what a lpf is so that's not the question == the question is that if the signal is in digital format(sampled/quanted) it had to be sampled at the input but then an lpf must be applied)

I've heard of some not free decoding tool,..name was...eer..Digigram ac3 encoder/decoder, someone knows them?

Thanks

One more:
When you decode an ac3 stream, the waves are always much below -3dB, sometimes below -20dB. Why do they do that? I accept that they should group normalize the 6 waves (One of the 6 has a 0dB or -3dB value, and the others are levelled relatively to that. And this also reduces dynamics and S/N), but they Dont!!

I think this is a trick to make you buy bigger and bigger amplifiers, it's like windows...more and more ram...cpu...

[Mikel]

Your question about the settings i cannot really answer. If you look further down in the forum I started another thread with more or less the same questions.

It seems that transcoding AC3 is something that is not done very often.

In fact I have only done it once so far with the movie "Green Mile" which is a 3 hour movie. Since I wanted to keep both German and English audio with all subtitles and extras, I had to find the best solution for a good 4.7GB disk. So I reencoded the audio in 320kbs which gave me a reasonable movie stream of around 2500kbs instead of 2300 kbs.

When reencoding, i set the phase of the surround to 0° to leave everything the same.

I did not do any further comparisons of the reencoded sound.
I would have liked to use Besweet for reencoding the AC3 stream, but there is a amplitude problem with the AC3end.dll.

So i using the AC3 encoder built into Scenarist, which has about the double speed of softencoder.
The parameter seem to be the same.

Cheers

Mikel

[jsoto]

@DSPGuru
In ffmpeg 0.4.6 (Released on 2002-12-27) changelog it can be read:

version 0.4.6pre1:
- fix quantization bug in AC3 encoder.

Only for information:
Have you compiled ac3enc.dll with this ffmpeg version in any BeSweet release?. I was trying to compile it (obviously with your sources), but I am too newbie in programming, and I am not able to success.
Could be this "quantization bug" the reason of the gain problem?

Thanks.

[DSPguru]

Originally posted by jsoto
Could be this "quantization bug" the reason of the gain problem?i believe not.

[pago cruiser]

OK, I give up. Went thru a search and found several similar threads, but nothing that my tired brain could see solves the problem.

Gnot 0.23, DivX 5.02 pro, 2 pass encoding, the usual stuff. The same old LOW LOw Low low volume MP3's in my final avi. So now I try using BeSweet like lots of folks said. Do my DVDTOAVI, do my COMP Test, and THEN I boot up a 2nd copy of Besweet. Load the AC3 file, select AC3 to AC3, set boost to 3 db, then select this new "boosted" AC3 file as my audio source when encoding/converting to MP3. No difference. Try again at 6db. Nothing. 9db. 27 db. 54 db. NADA ZIP NO DIFFERENCE.

What the heck is happening? I've now encoded this #*&@! thing 5 times, and I ain't a happy camper... :angry:

[bitz4brainz]

Okay. I went through all the same trials and tribulations as the other people on this thread about the way too soft AC3 encodes. Same result. I tried everything. VERY soft.

All I am trying to do is convert a PCM 48KHz soundtrack (captured analog) to 2 channel AC3 so that it will produce a STANDARD DVD (vs MP2) without chewing up a lot of space. Nothing fancy. The track already (purportedly) is Dolby Surround, and I'd like to preserve that. In a Wave editor the track looks fine and uses the full 16-bit dynamic range.

I got hold of a copy of Soft Encode and tried that. A lot better but STILL a lot softer than the WAV. Maybe by half. Why? I turned off every option I could find, especially pre-processing. All I need is an AC3 compression.

Why is this so hard???

[pago cruiser]

Don't know if this will help with AC3's, but I finally solved my problem using MP3 Gain. However, I had to set default volume to 100db average. The program now boosts all audio by 12 to 20 db! It does report clipping, but it is only apparant at high volumes, which I never listen to anyway. It only adds about 10 minutes post processing time, so it is acceptable to me.

[mookyking]

When encoding/transcoding with sonic foundry soft encode, I always set dialog normalization to -27db and turn off all input filtering and dynamic compression. The volume is about the same as the original. Dialog normalization can also be set all the way to -31db, which makes the results even louder.

Soft encode can directly transcode an ac3 stream without having to first decode to seperate .wav files; when using the File->Open dialog, select dolby digital(decode to PCM).ac3 as the file type. Takes a while to decode though! After it's done decoding, it's ready to re-ecode at a different bitrate.

[FreQi]

I just ran a 2 second 2Ch 48kHz WAV file through Soft Encode and the AC3 Encoder that comes with Scenarist and BeSweet to do a little testing. You can download all the files here (http://freqi.net/forums/TIABMA.zip) (985k). Here's what I did...

(a) Soft Encode -- TIABMA_a.ac3
Audio service config
Data rate: 192k
Sample rate: 49 kHz :: (auto checked)
Audio mode: 2/0 (L,R) :: LFE enabe was grey'd out
Bit stream: Main: Complete
Dialo norm: -27 dB
Audio bw: 20.30 kHz
Save frames in Intel byte order: UNchecked
Bit Sream tab
Center mix: grey'd out
Surround mix: grey'd out
Dolby surround mode: Not indicated
Copyright bit: checked
Orig bit stream: checked
Info exists: checked
Mix level: 105 dB SPL
Room type: Small
Preprocessing tab
Digital deemphasis: UNchecked
DC high pass filter: checked
bw low pass filter: checked
LFE low pass: grey'd out
90 degree phase: grey'd out
3 dB attenuation: grey'd out
Compression characteristic: Film: standard
RF overmodulation protection: checked

(b) Soft Encode -- TIABMA_b.ac3
Audio service config
(same as a)
Bit Sream tab
(same as a)
Preprocessing tab
(same as a, except...)
DC high pass filter: UNchecked
bw low pass filter: UNchecked
Compression characteristic: None
RF overmodulation protection: UNchecked

(c) Soft Encode -- TIABMA_c.ac3
(same as b, except...)
Audio service config
Dialo norm: -1 dB

(d) Soft Encode -- TIABMA_d.ac3
(same as b, except...)
Audio service config
Dialo norm: -31 dB

(e) Scenarist AC3Enc -- TIABMA_e.ac3
Audio Coding mode: 2/0: L,R
2-Ch Interleaved Input: checked
LFE Enable: UNchecked
Start time: (all zero's)
End time: (all zero's)
AC3 Settings, Edit
Data rate: 192k
Sample rate: 48
Dialog Normalization: -27dB
Bit Stream Mode: main, complete
Audio BW: 20.30
Center mix: grey'd out
Surround mix: grey'd out
Dolby surround mode: not indicated
Copyright bit: checked
Original bit stream: checked
Audio Production Info exist: checked
Mix level: -105dB
Room type: small, flat
Digital De-emphasis: UNchecked
DC highpass filter: checked
BW Lowpass filter: checked
LFE lowpass filter: checked
Surround channel 90 phase shift: grey'd out
Surround 3 dB Attenuation: UNchecked
Compression Characteristic: Film: Standard
RF Overmodulation Protection: checked

(f) Scenarist AC3Enc -- TIABMA_f.ac3
(same as e, except...)
DC highpass filter: UNchecked
BW Lowpass filter: UNchecked
LFE lowpass filter: UNchecked
Compression Characteristic: None
RF Overmodulation Protection: UNchecked

(g) BeSweetGUI -- TIABMA_g.ac3
In the pulldown menu to the left of the "Azid 1" "Azid 2" "Lame 1" "Lame 2" set of buttons, select AC3.
In the OTA section, check Auto Gain (Post Gain doesn't matter, it will go grey)
Click the "AC3 & OGG" button, check bitrate and select 192.
Click the BeSweet button and your command line will look something like "C:\....\BeSweet.exe" -core( -input "....\TIABMA.wav" -output "....\TIABMA_g.ac3" ) -ota( -g max ) -ac3enc( -b 192 )

(h) BeSweetGUI -- TIABMA_h.ac3
Same as g expect Auto Gain is unchecked.

(i) BeSweetGUI-- TIABMA_i.ac3
Same as g expect Auto Gain is unchecked, and Gain was checked and set to 50db.

Sadly, none of the resulting AC3 files have the same volume as the original WAV. Most of the AC3's actually sound a lot like each other, with the exception of (g). On that one, the logfile from BeSweet reports a Total Gain of 5.462dB was found, but the AC3 was practially mute. Unchecking Auto Gain yeilded similar results to all AC3's by Scenarist or Soft Encode, and forcing a Gain of 50 only hurt it.

So there you have it. Everything I know about encoding AC3. Here's a list of the files, their size in bytes and their checksums...

bytes file_name___ checksum
64512 TIABMA_a.ac3 4CFC89AC
64512 TIABMA_b.ac3 03EDD89F
64512 TIABMA_c.ac3 492A7865
64512 TIABMA_d.ac3 27CFC62F
63744 TIABMA_e.ac3 62DCE7B4
63744 TIABMA_f.ac3 0B8059D6
63744 TIABMA_g.ac3 BB9F6D11
63744 TIABMA_h.ac3 69B76C17
63744 TIABMA_i.ac3 8A3B531F

I find it interesting that BeSweet and Scenarist make files of the same byte count, but none of them generate the same file with what should have been the same settings.

[bitz4brainz]

Freaking interesting! Thanks for this!

If two supposedly professional encoders always produce lower volume and we eliminate the possibility of imbalance due to the player (since we are using the software's own player), then...

Maybe there is some technical "reason" why a proper AC3 encoder deliberately reduces the volume. Maybe it has something to do with a broadcast standard, or the way audio gear in a theatre operates. Or maybe its a silly or unthinking reason that got codified into the standard.

Like this one: Most audio CDs produced in the first decade or two of the technology do not use the full 16 bit headroom. Sometimes they use only 50% or even a third. Why? I am not a recording engineer, but I surmise that they either didn't understand the new technology (vs analog) or they didn't realize it mattered. CDs were evidently produced with the volume that happened to result from whatever setting they happen to like to use on their preamps. (?) It DOES matter because you have to boost the volume on playback, thus amplifying the noise and deficiencies of your analog playback system. It also matters, at least theoretically, because 16 bits is more accurate than 15 or 14 bits.

But audio CDs engineered in the past decade or so tend to use 90-95% or even 98% of the headroom. Even remasters of the same music on earlier CDs. Somebody woke up.

But if this is really part of the standard, then there will be no waking up. We will have to live with it.

Just conjecture and speculation, of course. I could be way off base.

I'm not doing 5.1 stuff so I'm afraid that there isn't much incentive to use AC3 on simple stereo. I think most DVD players of recent vintage, even in the US, will play back MPEG Layer II. For me, that's a known quantity.

[Julio]

Hi,

I'm trying to encode a WAV (16-bit stereo, 48 kHz) to a stereo AC3 file, using BeSweet 1.4.

Here's my original wave signal :

http://www.ifrance.com/InTheNuts/dvd/wav.jpg

I tried this very simple command :

besweet -core( -input country.wav -output country.ac3 ) -ac3enc( -b 384 )

... but I get this :

http://www.ifrance.com/InTheNuts/dvd/ac3.jpg

As you can see, the output level has been divided by 2 ... why ??

Is there any other way to encode a WAV to an AC3 without loss of information ?

Thanks...