Log in

View Full Version : combine 6 mono wavs to 5.1ac3?


masterpug
22nd March 2005, 16:26
Can it be done? if so, how?


my source was dd5.1 ac-3 file that i broker/saved into 6 mono wav files, did a batch fade and now need a way to combine them back into a proper dd5.1 ac-3 with channels in the right order. please help...thanks!

SallyDog
22nd March 2005, 16:52
You need one of the Surcode Products, depending on what your final outputs needs are.

http://www.surcode.com/

ursamtl
22nd March 2005, 21:33
Or why not check out http://needfulthings.webhop.org where you can download the latest version of HeadAC3he for free. It features AC3 encoding that seems to work fine.

Sycho
22nd March 2005, 22:45
Originally posted by ursamtl
Or why not check out http://needfulthings.webhop.org where you can download the latest version of HeadAC3he for free. It features AC3 encoding that seems to work fine. but without channel coupling, correct?

planet1
22nd March 2005, 23:58
@masterbug

See 3rd reply:

http://forum.doom9.org/showthread.php?threadid=91239

btw HeadAC3he's "ac3enc.dll" is based on ffmpeg, so using the latest ffmpeg.exe might be "better".

bye :)

ursamtl
23rd March 2005, 00:46
Originally posted by Sycho
but without channel coupling, correct?

I haven't seen anything to indicate otherwise. Then again, the price is certainly quite nice.

ursamtl
23rd March 2005, 00:51
Originally posted by planet1
@masterbug

See 3rd reply:

http://forum.doom9.org/showthread.php?threadid=91239

btw HeadAC3he's "ac3enc.dll" is based on ffmpeg, so using the latest ffmpeg.exe might be "better".

bye :)

Daphy can probably fill you in on this better than I but I believe this verion of ac3enc.dll has been modified. I've tried it on several projects and was happy with the results.

violao
23rd March 2005, 14:11
Originally posted by ursamtl
Daphy can probably fill you in on this better than I but I believe this verion of ac3enc.dll has been modified...

...but it still has the "volume bug" and neither dialog normalization nor DRC implemented.

ursamtl
23rd March 2005, 17:09
Originally posted by violao
...but it still has the "volume bug" and neither dialog normalization nor DRC implemented.

No the volume bug seems to be gone in the modified ac3enc.dll that comes with the version on Needfulthings. As fpr dialog normalization and DRC, my main interest is music so I never use them anyway. I don't want anything messing with the dynamic range or levels of my audio!

Granted, ac3enc.dll does not have the features of expensive decoders, but it has one feature they have yet to match, it's legally free! :)

violao
24th March 2005, 12:03
Originally posted by ursamtl
No the volume bug seems to be gone in the modified ac3enc.dll that comes with the version on Needfulthings...
According to my tests new alpha headac3he (with supposedly modified ac3enc.dll) produced 6 dB lower volume than input. For input I used ac3 clip that was converted to wav with azid command line decoder without normalization. Comparing output ac3 to original shows -6 dB difference. I can se no option for increasing output volume in headac3he other than GlobalGain which I believe amplify input signal.

I think there must be some wrong shifting of the signal bits in ac3enc. Shifting 1 bit to the right would equal 6 dB.

As for compression, original audio is never compressed in ac3. These are just flags telling the decoder how to present the audio to the listener.

Regarding levels, that could also present a problem. ac3enc sets dialog normalization to a minimum level that is -31 dB. That effectively means that your decoder should apply gain to the signal such that -31 db becomes "normal" dialog level, that is the maximum gain encoder is allowed to use. For norm levels > -31 the gain would be smaller. Depending on the rest of the system you might get distorsions if the gain is too big.

ursamtl
24th March 2005, 14:01
Ok well test results generally aren't wrong. All I know is that when I processed 6-channel wave files with the old ac3enc.dll, it produced dramatically reduced volume in the resulting AC3 file, whereas the new one produced roughly the same level as my input wave file. I didn't measure the levels but I can tell you that I know what a -6dB change in level sounds like and this was definitely not the case with my experience. At most the difference was -1.5 to -2dB.

Could it be that there's something you did along the way that caused this? I know Daphy and others have reported good results with this modded ac3enc.dll. Perhaps they can share their experiences.

Regards,
Steve.

violao
24th March 2005, 14:32
Originally posted by ursamtl
... a -6dB change in level sounds like ...
50% in fact. As for what I did, I think explained everything. No gain, no normalization, compression, boosters, nothing that could affect volume whatsoever. Ac3enc configuration in Headac3he has only 1 parameter, that is bitrate. I would also like to see some other reports. I was too enthusiastic about having the possibility of open source ac3 encoding.

ursamtl
24th March 2005, 19:06
Originally posted by violao
50% in fact.
Actually this is not exactly so. From an electrical point of view, yes -6dB is 50% but it's a long established fact that psychoacoustically, -3dB is perceived by the human ear to be 50% as loud, so -6dB sounds like 25% 1/4 as loud. This is why I was surprised by your test result of -6dB because there is no way my resulting AC3s sounded 1/4 as loud as the source 5-channel wave file. The difference in level was very subtle. I'm at work right now so I don't have the software in front of me. I'll check it when I get home.

violao
25th March 2005, 10:22
What you are referring to is that -6dB (50%) signal voltage will produce -12dB power on the output, that is 25% power. 25% sounds like 25%, 50% sounds like 50%, just that for 50% less power you'll need -3dB voltage. This comes from tha fact that speaker input impendance is more or less constant and power is derived from:

voltage squared / resistance

Therefore power rises with squared voltage, hence the difference in dB.

The other thing is that human hearing has mostly logarithmic characteristcs, meaning that the difference between 25% and 50% power is percieved approximately the same as the difference between 50% and 100%. That's why logarithmic dB measures are used, and all this amplification thing, that is in fact multiplication, is represented by db offsets.

But we were talking about signal dynamic range, gain (or attenuation) and -6dB is certainly -6dB.

ursamtl
25th March 2005, 17:02
Exactly, and as I said, I definitely know what a -6dB level change sounds like: a perceived drop in sound level to 25% of the former level. I knew there was no way the projects I'd encoded using the latest version of HeadAC3he with the modded ac3enc.dll exhibited that much of a drop! I ended up with at least that much of a drop when using the old ac3enc.dll with default settings, but definitely not with this new version!

In fact, I've just tested my last surround project's 6-channel wave source file plus the output AC3 file from my conversion using HeadAC3he. There was no difference in output level. Both peaked at just below 0dB.

Could it be you were using the wrong version? According to the log file display from mine, the versions I used were:

  HeadAC3he v0.24-a13 by Dark Avenger 15.12.2004
  AC3enc.dll 1.00 (15.12.2004) engine 1.20

Upon opening HeadAC3he, one needs to click the Options button on the right side of the window to display the HeadAC3he Azid window. The settings that always need to be changed are: The 6 ch button in the bottom right corner of the Azid window. The Downmix Type dropdown box in the bottom left corner of the Azid window should be set to Surround.As long as you input a 16- or 24-bit/48kHz source file, you should have no problem with it. Let us know how you make out.

Regards,
Steve.

planet1
25th March 2005, 19:06
just one question

why not include ffmpeg.exe in your tests (after all those ac3enc.dlls are based on the ffmpeg project) ?

http://www.aziendeassociate.it/cd.asp?dir=/ffmpeg

http://www.videohelp.com/tools?toolsearch=gui4ffmpeg

ursamtl
25th March 2005, 20:33
Originally posted by planet1
just one question

why not include ffmpeg.exe in your tests (after all those ac3enc.dlls are based on the ffmpeg project) ?

http://www.aziendeassociate.it/cd.asp?dir=/ffmpeg

http://www.videohelp.com/tools?toolsearch=gui4ffmpeg
Because according to the information I have, it does not include the latest modified verions of ac3enc.dll.