should i let eac3to remove dialog normalization?

heliocentric · 24th November 2010, 15:04

i have noticed that eac3to does this automatically...

is this a good thing? why would you want to make the dialog harder to hear?

i have noticed with certain films that the dialog is very quiet and the action scenes alot louder, this is very annoying. is this because eac3to has removed the dialog normalization from my source file before i go on and encode with handbrake?

(in handbrake i always encode the resulting DTS/DD audio file to Dolby Pro Logic II for playback on a apple tv)

thanks

tebasuna51 · 24th November 2010, 18:02

Quote:

Originally Posted by heliocentric

i have noticed that eac3to does this automatically...
is this a good thing?

Yes. To avoid different volume between AC3 tracks and others with DTS, AAC, MP3, ...

Quote:

why would you want to make the dialog harder to hear?

i have noticed with certain films that the dialog is very quiet and the action scenes alot louder, this is very annoying. is this because eac3to has removed the dialog normalization from my source file...?

Seems you mistake Dialog Normalization with Dynamic Range Compression.

The attenuation applied by DN (typically -4dB) is the same for low (dialogs) and high (explosions) volume sounds.
With DRC you apply different attenuation to low and high volume sounds.

Quote:

(in handbrake i always encode the resulting DTS/DD audio file to Dolby Pro Logic II for playback on a apple tv)

You can use eac3to parameter -down2 to do the same downmix.

heliocentric · 24th November 2010, 23:51

if its not a good thing why do they add it to the soundtrack of the bluray?

if you played back the original mkv created by eac3to in say xmbc would that add drc to avoid volume spikes?

tebasuna51 · 25th November 2010, 04:01

1) Because is mandatory in Dolby Digital specs. Here we use ac3 not Dolby Digital compliant.

2) Sorry I don't know xmbc. Many receivers have functions equivalent to drc (Night mode). Ac3Filter can also add DRC to any sound.

leeperry · 27th November 2010, 23:35

Quote:

Originally Posted by heliocentric

i have noticed with certain films that the dialog is very quiet and the action scenes alot louder, this is very annoying.

Dialogs come from the center channel, Explosions from the front stereo AFAIK...A proper calibration will counter-balance this problem. I've finetuned the 5.1 coeffs for headphones use and calibrated them, and everything's fine. No inaudible dialogs, no overwhelming explosions. I don't use DRC either.

yetanotherid · 3rd December 2010, 20:09

heliocentric,
The dialog normalisation level has nothing to do with normalising between the dialog and the rest of the audio as such, it's a value which is used to adjust the entire audio... all channels equally. If you find the dialog can be too quiet compared with the loud bits it'll still be the same with the value present or removed. The purpose of the dialog normalisation is to ensure the average dialog level in the audio of one movie is the about the same as the average dialog level in another, supposedly to help you avoid having to reach for the volume control between movies.
If like me, you use stereo speakers because you find surround sound to be a gimmick which just reminds you the video isn't, then it's of course hard to adjust the centre speaker level. If you're using a PC though and something like ffdshow to decode the audio then you could use it's output matrix to output 6 channels, boost the centre channel (or cut the rest) and then let your sound card mix it all back to stereo. That might help.

I prefer MeGUI over Handbrake, but I only mention that because it doesn't remove the dialog normalisation.
I tend to use the ffdshow decoder's volume normalisation/regain volume function as it works on all audio. It's not ideal (compression is a better method) but it's effective. It normalises the volume "on the fly" slowly amplifying it until it's peak level reaches a set maximum. If a louder bit comes along it drops the level a bit. Normally it'd then just stay at that level which means it's just turned the volume up, but the regain volume function makes it continually try to keep turning the volume up until it reaches that set level again, so in the quiet parts (dialog) it keeps slowly turning the volume up until it hits the maximum level or another loud bit comes along. Some people hate that sort of normalisation but I don't mind it. You don't really notice it working and if you're watching video at night when others are trying to sleep it's much better than constantly having to turn the volume up and down yourself.

Sorry if I'm explaining what you already know, but for the record here's a brief explanation as to how the dialog normalisation works and it's effect on using AC3's dynamic range compression. Basically if you want to use the dynamic range compression to quiet down the load bits so you can turn the dialog up, removing the dialog normalisation level is probably a bad idea. If you're converting the AC3 to something else then it probably doesn't matter.

-31db is the standard for the "average dialog level" in AC3 audio. Without a dialog normalisation level the decoder leaves the volume alone. With a dialog level.... say -20db... the decoder reduces the level by 31db, less 20db, to give you a 11db reduction. As -31db is supposed to be the average dialog level for all AC3 and you have AC3 audio with a dialog level of -20db, the 11db reduction in the previous example gives you dialog at -31db. If you remove the -20db dialog normalisation, the decoder doesn't attenuate the audio so the average dialog level stays as it really is, -20db. The decoder will play the audio (all channels) 11db louder.
If the dialog normalisation level is wrong for some reason then the decoder will adjust the audio by the wrong amount, but the dynamic range compression works under the assumption the average dialog level is at -31db, so an incorrect normalisation level can cause the dynamic range compression to not work, or to compress the audio too much.

What I don't know is what happens when you use dynamic range compression on AC3 audio which has had the dialog normalisation level removed. As the decoder doesn't reduce the volume with it removed then theoretically you're likely to end up with the audio compressed way too much. Logically to prevent that problem, you'd assume even with dynamic compression enabled a decoder would only apply compression if the dialog normalisation level is present, but I have no idea if it works that way. Maybe someone else will know.

Hagbard23 · 4th December 2010, 01:32

@yetanotherid:
Thanks really for explanation - i've waited for something like this. I'll copy and paste it to a text file on my homePC...just for the case the question is asked a second time...

i really appreciate explanations from users, who got deep into the materia by the time...it is much more worth than generic FAQ's...

more of that please...

Groucho2004 · 4th December 2010, 01:39

I think this post explains it very well.

tebasuna51 · 4th December 2010, 02:26

Quote:

Originally Posted by yetanotherid

...
What I don't know is what happens when you use dynamic range compression on AC3 audio which has had the dialog normalisation level removed. As the decoder doesn't reduce the volume with it removed then theoretically you're likely to end up with the audio compressed way too much. Logically to prevent that problem, you'd assume even with dynamic compression enabled a decoder would only apply compression if the dialog normalisation level is present, but I have no idea if it works that way. Maybe someone else will know.

The Dialog Normalization (DN) can't be removed, but can be set to -31 dB then the decoder don't apply a fix attenuation to the full range of sounds. Of course you can obtain dialogs with high volume than -31 dB.

The Dynamic Range Compression (DRC) are different values, calculated at encoder time, of atenuattion (or gain), to be applied for each block of 256 samples, and stored in the bitstream.

The decoder don't analyze the volume of samples and don't know if the DN was modified ot not, it apply the read value (or a %) to the block of samples.

Then change DN don't modify the previously calculated values of DRC and the effect is only a gain to the full range of sounds.

The DN-DRC ac3 system was a great idea long time ago, and can be great now ... if all the audio was compressed in dolby digital compliant mode. But there are different formats and:

- The DRC with calculated values (usefull for old systems with low power of process) isn't needed now, modern receivers (Night mode, ...) or PC soft (ffdshow, Ac3Filter, ...) can apply similar functions to all audios (not only ac3) in real time.

- The DN assumption, dialogs must have -31 dB, isn't respected but many other audio sources and formats, TV commercials, CD Audio (Loudness war), ..., and the conclusion is always: the ac3 audio play with low volume (than the rest).

For all that eac3to encode with DN=-31dB and DRC=none.
And put DN=-31dB, by default, in extracted ac3 tracks.

laserfan · 4th December 2010, 03:28

Quote:

Originally Posted by tebasuna51

The Dialog Normalization (DN) can't be removed, but can be set to -31 dB then the decoder don't apply a fix attenuation to the full range of sounds. Of course you can obtain dialogs with high volume than -31 dB....

For all that eac3to encode with DN=-31dB and DRC=none.
And put DN=-31dB, by default, in extracted ac3 tracks.

For all this I am more confused than before. If DN is applied to DD tracks to achieve intended playback levels, then why does eac3to remove it by default?

tebasuna51 · 4th December 2010, 04:24

Quote:

Originally Posted by laserfan

For all this I am more confused than before. If DN is applied to DD tracks to achieve intended playback levels, then why does eac3to remove it by default?

To obtain the same volume than the same track encoded to aac, dts, ogg, mp3 if stereo ...

We know this isn't DD compliant, you can always use -keepDialnorm at your choice.

Also eac3to don't encode to DD, encode to ac3 with a52 specs.

yetanotherid · 4th December 2010, 07:18

Quote:

Originally Posted by laserfan

For all this I am more confused than before. If DN is applied to DD tracks to achieve intended playback levels, then why does eac3to remove it by default?

Because AC3 audio is almost always quieter than other types of audio due to the DN level, getting the decoder to turn it down. By removing the DN (or more correctly as it's been pointed out, by changing the value to -31db) the decoder will play the audio at it's real volume. Mind you if the conversion program normalises the audio when converting it, the resulting MP3/AAC/etc file will probably still be somewhat louder anyway.
The DN was intended to make the average dialog level in all AC3 audio roughly the same, but as other audio types don't use it there's probably no reason to keep it if the audio in your video files is generally a mixed bag rather than being AC3 only.

yetanotherid · 4th December 2010, 08:00

Quote:

Originally Posted by tebasuna51

The decoder don't analyze the volume of samples and don't know if the DN was modified ot not, it apply the read value (or a %) to the block of samples.

Then change DN don't modify the previously calculated values of DRC and the effect is only a gain to the full range of sounds.

I've probably got that aspect of it wrong.... I've only got a basic knowledge as to how it works, but I did look at the link provided by Groucho2004. As it turns out, I'm pretty sure that post is one of the pages I read a long time ago when trying to understand how the DN thing works myself, and after reading it again now, then banging my head on the desk to jolt my brain into working and reading it again, I finally realise I seem to have misunderstood some of what was written.

Hagbard23,
Everything I wrote earlier about DN and DRC.... well most of it's right.... but out of habit I've just thought of the the DN value as having been removed when technically it's changed to -31db.... which gives you the same result anyway I guess.... the decoder doesn't alter the volume on playback.
I sometimes use the Aften encoder to convert audio to AC3 and I always assumed it wouldn't add a DN value, but maybe it just specifies a -31db value too? I've never really thought about it too much before.

It seems the part of my post referring to the effect changing the DN value to -31db has on DRC was completely wrong. Sorry about that.
It appears the info in the post Groucho2004 linked to was referring to encoding AC3 and the effect of then applying DRC if the DN value is wrong. I think I read it a long time ago and assumed it was referring to decoding the AC3 too, but no.....
As tebasuna51 pointed out (and I assume he's correct) once the DRC values are added to the AC3 file they're still present to be used regardless of later changes to the DN value.

laserfan · 4th December 2010, 15:15

I have been extracting audio with eac3to using default i.e. DN removed, but it appears now that I want instead to use -keepDialNorm given my decoding/playback is always Dolby Digital via BD disc/settop player/decoder-amplifier. I will try it anyway.

It is interesting that I have never noticed any audio problems with my setup when playing my converted (i.e. not original BD) discs, though I always do fiddle with the main volume control of my amplifier once the movie has begun. It seems by the comments here that perhaps I will need to do less volume adjustment from disc-to-disc if I keep DN the way it was from the original.

Do I understand this right now, d'ya think?

24th November 2010, 15:04	#1 \| Link
heliocentric Registered User Join Date: May 2010 Posts: 24	should i let eac3to remove dialog normalization? i have noticed that eac3to does this automatically... is this a good thing? why would you want to make the dialog harder to hear? i have noticed with certain films that the dialog is very quiet and the action scenes alot louder, this is very annoying. is this because eac3to has removed the dialog normalization from my source file before i go on and encode with handbrake? (in handbrake i always encode the resulting DTS/DD audio file to Dolby Pro Logic II for playback on a apple tv) thanks

25th November 2010, 04:01	#4 \| Link
tebasuna51 Moderator Join Date: Feb 2005 Location: Spain Posts: 6,914	1) Because is mandatory in Dolby Digital specs. Here we use ac3 not Dolby Digital compliant. 2) Sorry I don't know xmbc. Many receivers have functions equivalent to drc (Night mode). Ac3Filter can also add DRC to any sound. __________________ BeHappy, AviSynth audio transcoder.

3rd December 2010, 20:09	#6 \| Link
yetanotherid Banned Join Date: Apr 2008 Posts: 723	heliocentric, The dialog normalisation level has nothing to do with normalising between the dialog and the rest of the audio as such, it's a value which is used to adjust the entire audio... all channels equally. If you find the dialog can be too quiet compared with the loud bits it'll still be the same with the value present or removed. The purpose of the dialog normalisation is to ensure the average dialog level in the audio of one movie is the about the same as the average dialog level in another, supposedly to help you avoid having to reach for the volume control between movies. If like me, you use stereo speakers because you find surround sound to be a gimmick which just reminds you the video isn't, then it's of course hard to adjust the centre speaker level. If you're using a PC though and something like ffdshow to decode the audio then you could use it's output matrix to output 6 channels, boost the centre channel (or cut the rest) and then let your sound card mix it all back to stereo. That might help. I prefer MeGUI over Handbrake, but I only mention that because it doesn't remove the dialog normalisation. I tend to use the ffdshow decoder's volume normalisation/regain volume function as it works on all audio. It's not ideal (compression is a better method) but it's effective. It normalises the volume "on the fly" slowly amplifying it until it's peak level reaches a set maximum. If a louder bit comes along it drops the level a bit. Normally it'd then just stay at that level which means it's just turned the volume up, but the regain volume function makes it continually try to keep turning the volume up until it reaches that set level again, so in the quiet parts (dialog) it keeps slowly turning the volume up until it hits the maximum level or another loud bit comes along. Some people hate that sort of normalisation but I don't mind it. You don't really notice it working and if you're watching video at night when others are trying to sleep it's much better than constantly having to turn the volume up and down yourself. Sorry if I'm explaining what you already know, but for the record here's a brief explanation as to how the dialog normalisation works and it's effect on using AC3's dynamic range compression. Basically if you want to use the dynamic range compression to quiet down the load bits so you can turn the dialog up, removing the dialog normalisation level is probably a bad idea. If you're converting the AC3 to something else then it probably doesn't matter. -31db is the standard for the "average dialog level" in AC3 audio. Without a dialog normalisation level the decoder leaves the volume alone. With a dialog level.... say -20db... the decoder reduces the level by 31db, less 20db, to give you a 11db reduction. As -31db is supposed to be the average dialog level for all AC3 and you have AC3 audio with a dialog level of -20db, the 11db reduction in the previous example gives you dialog at -31db. If you remove the -20db dialog normalisation, the decoder doesn't attenuate the audio so the average dialog level stays as it really is, -20db. The decoder will play the audio (all channels) 11db louder. If the dialog normalisation level is wrong for some reason then the decoder will adjust the audio by the wrong amount, but the dynamic range compression works under the assumption the average dialog level is at -31db, so an incorrect normalisation level can cause the dynamic range compression to not work, or to compress the audio too much. What I don't know is what happens when you use dynamic range compression on AC3 audio which has had the dialog normalisation level removed. As the decoder doesn't reduce the volume with it removed then theoretically you're likely to end up with the audio compressed way too much. Logically to prevent that problem, you'd assume even with dynamic compression enabled a decoder would only apply compression if the dialog normalisation level is present, but I have no idea if it works that way. Maybe someone else will know. Last edited by yetanotherid; 3rd December 2010 at 20:17.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

24th November 2010, 23:51	#3 \| Link
heliocentric Registered User Join Date: May 2010 Posts: 24	if its not a good thing why do they add it to the soundtrack of the bluray? if you played back the original mkv created by eac3to in say xmbc would that add drc to avoid volume spikes?

4th December 2010, 01:32	#7 \| Link
Hagbard23 23sKiDdOo! Join Date: May 2010 Location: Germany Posts: 182	@yetanotherid: Thanks really for explanation - i've waited for something like this. I'll copy and paste it to a text file on my homePC...just for the case the question is asked a second time... i really appreciate explanations from users, who got deep into the materia by the time...it is much more worth than generic FAQ's... more of that please...

4th December 2010, 01:39	#8 \| Link
Groucho2004 Join Date: Mar 2006 Location: Barcelona Posts: 5,034	I think this post explains it very well.

4th December 2010, 15:15	#14 \| Link
laserfan Aging Video Hobbyist Join Date: Dec 2004 Location: Off the Map Posts: 2,461	I have been extracting audio with eac3to using default i.e. DN removed, but it appears now that I want instead to use -keepDialNorm given my decoding/playback is always Dolby Digital via BD disc/settop player/decoder-amplifier. I will try it anyway. It is interesting that I have never noticed any audio problems with my setup when playing my converted (i.e. not original BD) discs, though I always do fiddle with the main volume control of my amplifier once the movie has begun. It seems by the comments here that perhaps I will need to do less volume adjustment from disc-to-disc if I keep DN the way it was from the original. Do I understand this right now, d'ya think?