Stereo AC3 transparency [Archive]

rosivaldo

9th December 2007, 03:41

Mp3 or AAC will give you better quality than AC3 at 192kbps, but it is still a good bitrate for AC3 encoding of stereo music.

AFAIK, I can't use mp3 nor aac (or can I? Am I wrong?), since I want to put the music onto DVDs that should be playable by any standalone DVD player (something like Audio DVD Creator does). It limits my choice to LPCM@48Khz (no quality loss, but huge files), mp2 or ac3 (which let me trade size for quality).

I need to do it for my entire CD collection and I'd like to preserve quality as much as possible and still save all the space I can, provided quality is not sacrificed. This is why I'd prefer not using LPCM. To me, it is enough to have a transparent sound. I know that transparency is subjective, but anything equivalent to what "Lame --preset standard" do for mp3 must be enough, I think.

Test it yourself.
(...)
If it still show artifacts,

Er... I'm afraid I'm not skilled enough to do such a kind of test. :o Aren't out there any public test about this encoder at these bitrates? I've found no one at Hydrogenaudio. :-(

u can take it up to 224kbps (it's more recommended than 192kbps, but the lowpass filters at 192kbps stereo are pretty much the same than 5.1 at 448kbps)

Should I forget 192kbps and use 224kbps instead, for quality sake? Going higher, do I *perceivably* (as they say: for *most* music/equipment/people) improve quality, or would I only waste space for no audible improvement?

Some one has already said that 5.1@448kbps / 5ch = 89.6kbps/ch, which means that 2.0@192kbps / 2ch = 96kbps/ch, which is *more than enough* to stereo music. Is this reasoning correct?

At last (please be patient :o): should I leave ac3 alone and stick to mp2? If yes, with which encoder and bitrate?

Thanks for your time,

Rosivaldo.

Skelsgard

9th December 2007, 07:54

AFAIK, I can't use mp3 nor aac (or can I? Am I wrong?)...

I mentioned MP3 and AAC as basis for comparison.

Er... I'm afraid I'm not skilled enough to do such a kind of test. :o Aren't out there any public test about this encoder at these bitrates? I've found no one at Hydrogenaudio. :-(

I don't know of any public test regarding AC3 vs. MP3 vs. AAC vs etc, as AC3 is pretty old as a codec to put it in the same bag.
And you don't need to be "skilled", just go with your ears, man.
Check out any song, encode it to AC3 at 192kbps, listen the original uncompressed song several times in a row and then listen to the AC3 encoded one. If there are important artifacts, you will hear them, don't worry about it.

Should I forget 192kbps and use 224kbps instead, for quality sake? Going higher, do I *perceivably* (as they say: for *most* music/equipment/people) improve quality, or would I only waste space for no audible improvement?

I personally can't hear any improvement above 192kbps, but that's just me. Others might claim to hear the difference but only believe in them if they provide you with proper blind tests.

Some one has already said that 5.1@448kbps / 5ch = 89.6kbps/ch, which means that 2.0@192kbps / 2ch = 96kbps/ch, which is *more than enough* to stereo music. Is this reasoning correct?

No. It would be correct if all channels would carry the same info, but 5.1 mixes have a lot less audio info in the surround channels than the front ones, and the LFE doesn't take 89.6 kbps but less, and also movies audio info is less complex or high-frequency-intense than music.
Actually, IIRC the number would come from 5 1/3 (5 for full-range and 1/3 for LFE)
So 448 would be: 84 x 5 + 84 x 1/3 = 448
Or 384: 72 x 5 + 72 x 1/3 = 384

At last (please be patient :o): should I leave ac3 alone and stick to mp2? If yes, with which encoder and bitrate?

Nooooo, a big NO to the MP2 thing.
MP2 is worse than AC3 and the perceptual clear is above 224kbps. You can use MP2 if you want to, but not as if it were better codec than AC3.

I would say that Ac3 stereo at 192kbps will be good enough for your backups.

Here I leave you two samples, the original PCM uncompressed file (ripped with EAC) and the AC3 192 kbps encoded file (using Aften 0.05 + AftenGUI 1.4) for you to see for yourself.
http://rapidshare.com/files/75317322/Rammstein-Links123.wav
http://rapidshare.com/files/75317356/Rammstein-Links123.ac3

Later, man

Skelsgard

10th December 2007, 06:12

I checked your Ac3 file, man, and it sounds really well, but why is the volume so low?

Later

Terranigma

10th December 2007, 15:19

I checked your Ac3 file, man, and it sounds really well, but why is the volume so low?

Later

:eek:, how were you able to view my post? It's hidden. :p
It's not that it's low. The reason why it sounds so low is because it uses "DRC", and you'll need a decoder that supports the "Dialog Normalization" level I specified. But yea, ac-3 is awesome, but only when used properly ^^

Skelsgard

10th December 2007, 19:09

Terranigma

10th December 2007, 20:09

@Terranigma

DRC is a dynamic normalization while DialNorm is a fixed normalization.
DRC can be turned off and the sound will be at it's original volume the entire track.
I think you mean you used DialNorm in your encoding.

I was using ffdshow and the volume was way too low, but changing back to AC3filter the volume is OK. Still, AC3filter reports a -9 DialNorm, which if true, is too much of a decrease in gain. It's a 22 dB reduction.
I think the reason why AC3Filter is playing it at the same volume is beacuse it's ignoring the DialNorm setting while ffdshow is respecting and enforcing it. SAPs will respect the setting and volume will be low when played.
If you want to use a 9dB reduction, use DialNorm -22 dB.
What encoder did you use?

I used both. I followed this (http://forum.doom9.org/showthread.php?t=56020) guide to the tee.
I measured the RMS Level just like it's written here (http://pages.sbcglobal.net/wilsondr/ddexsfrms.gif) and here (http://pages.sbcglobal.net/wilsondr/ddexacid1.gif).
and your audio was at -9.2db, so using -9db would be correct. The level should match that of the rms level, or as close as possible. And yes, I meant DRC and applied the proper setting for your audio source (which in your case would be music: light).
http://pages.sbcglobal.net/wilsondr/ddcompprof.gif
So if you're saying my audio was improperly encoded, then that guide needs to be unpinned/corrected. ;)

The encoder was the latest dolby encoder from sony sound forge 9.0 (the professional payware version)

Skelsgard

10th December 2007, 22:17

I'm not saying that the guide is wrong, but 0dB gain = -31 dB DialNorm and -31 db gain = 0db DialNorm.
In order to reach a -9 db gain, you need to set it at 31 - 9 = 22 dB
DialNorm = -22dB

This are samples for encoding DialNorm at -9dB and at -24 dB (no DRC)
-24dB: http://www.flyupload.com/?fid=2431123
-9dB: http://www.flyupload.com/?fid=7743419

Later

Terranigma

10th December 2007, 22:40

I see you didn't get what he was saying. :( You might wanna re-read it. :)
Also, sorry to tell you, but the settings you used for your encodes wouldn't make any difference, since you didn't use Dynamic Range Compression

==========================================================
File to analyze:
C:\RAMMST~1.AC3
SampleRate 0 : 48000 KHz.
BitRate 10 : 192 Kb/s.
Version (bsid) 8 : Standard.
Bit Stream mode (bsmod) 0 : main audio service: complete main (CM)
Audio coding mode (acmod) 2 : 2/0 - L, R
Dolby Surround Mode 0 : not indicated
Low frequency effects channel 0 : Not present
Dialogue normalization - 25 dB
RF atenuattion -0.28 dB Frame: 1
Languaje 0 : Not present
Audio Production Info 0 : Not present
CopyRight bit 1
Original bit 1
Timecode1 0 : Not present
Timecode2 0 : Not present
Additional Bsi 0 : Not present
Block switch flags 0
Dither flags 3
Dynamic Range Info 0 : Not present
----------------------------------------------------------
File length : 563712 bytes.
Valid Frames : 734
Duration : 23.488 seconds. ( 0 h. 0 m. 23.488 s.)
RF Ov. Pr. min/max : -6.59 /-0.28 dB
----------------------------------------------------------

If you compared my earlier sample to your wave, you wouldn't be able to tell any difference in the volume level. For kicks, I recoded your source using a dialnorm of -24 with drc (which can be downloaded Here (http://www.zshare.net/download/552299191ee7e8/)) and you can easily tell that it's not of equal loudness as it should be like the source.
The goal is to use the exact same, or almost same value you receive from scanning the audio's rms level, not by the formula you used above. -P

---
Also, I noted that this time you resampled the audio. That's a good thing, because dolby perfoms poorly at 44100khz, but at 48000khz, it shines.
(Your encodes could be better though.)

Skelsgard

11th December 2007, 00:29

The first was a fast encode, didn't give it much thought as it was to provide rosivaldo with a sample.
My encodes are perfect the way they are. The sound great in my SAP and PC, and I¿ve never had any problems with them.

I see you didn't get what he was saying. :( You might wanna re-read it. :)
Also, sorry to tell you, but the settings you used for your encodes wouldn't make any difference, since you didn't use Dynamic Range Compression

DialNorm might be relevant to DRC but DRC can be set to None and still have DialNorm working as intended.
DRC is not necessary to have DialNorm.
This is very basic info that you should know by now.

If you compared my earlier sample to your wave, you wouldn't be able to tell any difference in the volume level.

Yes, I do. I can hear a very distinctive volume decreasing as tested in ffdshow, Cyberlink's PowerDVD and my SAP.
So distinctive that I have my PC speakers to full volume and I can hear it very easily without it being defeaning.

The goal is to use the exact same, or almost same value you receive from scanning the audio's rms level, not by the formula you used above. -P

I honestly don't understand anymore what you're talking about, but here, http://www.dolby.com/assets/pdf/tech_library/18_Metadata.Guide.pdf
Page 6
That clearly states what I've already told you.
You need to read better

Terranigma

11th December 2007, 00:43

Yes, I do. I can hear a very distinctive volume decreasing as tested in ffdshow, Cyberlink's PowerDVD and my SAP.
So distinctive that I have my PC speakers to full volume and I can hear it very easily without it being defeaning.

Then your ffdshow must be broken, because I use ffdshow as well to decode, but don't hear a bit of difference in the volume level. The only way for us to be certain, is for others to volunteer and test as well to see who's in the right.

Actually, I learned what I learned from an audio/video expert (actually has some of the best encodes i've ever seen PERIOD!), and they say i'm in the right (and I trust them highly). Maybe somejoe could clear this up for you, because I know i'm right about this, and you're in the wrong. :P

Here (http://www.zshare.net/download/5525225c739a5a/)'s my source for the experts here.

DRC is not necessary to have DialNorm.
This is very basic info that you should know by now.

For the highest quality possible, drc is a must, and I care about quality. For me, it's the highest priority.

Skelsgard

11th December 2007, 00:51

Then your ffdshow must be broken, because I use ffdshow as well to decode, but don't hear a bit of difference in the volume level. The only way for us to be certain, is for others to volunteer and test as well to see who's in the right.

So your saying that ffdshow, PowerDVD's audio decoder and a SAP are all broken at the same time.
That's a bold statement.

Dude, a -9dB DialNorm is making your encode to drop 22dB in gain, that's a fact, not some random number calculated thru fantasy settings.
Ac3Filter reports the -9dB DialNorm and the file sounds like such.
Why don't you listen to your encode in other decoders and compare it to the original wav (like your own SAP, cyberlink or intervideo decoders, VLC, etc, etc)

Now this is the same wav but with a -22dB gain in Audition, and then encoded to AC3 using a -31dB DialNorm.
http://www.flyupload.com/?fid=8986563

DRC is a feature for comfort (so that you won't bother ppl with sudden loud noises), not a quality-related option.

Terranigma

11th December 2007, 01:10

Why don't you listen to your encode in other decoders and compare it to the original wav (like your own SAP, cyberlink or intervideo decoders, VLC, etc, etc)

I listened to my encoded file on my modded xbox (using xbox media center), and your source wav, and it sounds equal in loudness. So, my xbox is wrong too? :D

Lets not take this persons thread ot with this debate and let someone else reply please. :)
This's getting out of hand.

tebasuna51

11th December 2007, 01:26

I used both. I followed this (http://forum.doom9.org/showthread.php?t=56020) guide to the tee.
I measured the RMS Level just like it's written here (http://pages.sbcglobal.net/wilsondr/ddexsfrms.gif) and here (http://pages.sbcglobal.net/wilsondr/ddexacid1.gif).
and your audio was at -9.2db, so using -9db would be correct. The level should match that of the rms level, or as close as possible. And yes, I meant DRC and applied the proper setting for your audio source (which in your case would be music: light).
http://pages.sbcglobal.net/wilsondr/ddcompprof.gif
So if you're saying my audio was improperly encoded, then that guide needs to be unpinned/corrected. ;)

The Guide "How to Properly Encode Dolby Digital Audio" is correct, but maybe the Dialog Normalization purpose is obsolete.

The Dolby idea is maintain a constant volume between programs with a standard dialog at -31 dB. Then if you measure a rms of -9 dB you must use a DialNorm of -24 dB and when the signal is played the decoder must apply -22 dB (attenuate 22 dB).

There are two problems:

1) Don't exist a standard dialog to measure in this music sample, then don't exist a possible reference to equate with others programs.

2) The modern music CD's, TV commercials, ... don't respect the Dolby idea about -31 dB dialog normalization and try to reach the max volume allowed (-9 dB rms is really a very high volume).

If you listen the same sample encoded to mp3 and after 'properly' encoded to Dolby Digital the volume difference is very big.

Only if you only listen Dolby Digital 'properly' encoded sounds I can recommend you use other value to DialNorm than -31 dB.

Terranigma

11th December 2007, 01:36

2) The modern music CD's, TV commercials, ... don't respect the Dolby idea about -31 dB dialog normalization and try to reach the max volume allowed (-9 dB rms is really a very high volume).

If you listen the same sample encoded to mp3 and after 'properly' encoded to Dolby Digital the volume difference is very big.

Only if you only listen Dolby Digital 'properly' encoded sounds I can recommend you use other value to DialNorm than -31 dB.
oh my, dolby is more complicated than I originally imagined, so I guess the way I were encoding, is the same like you mentioned in 2. Why do you think the volume's so high though ?

According to this image
http://img239.imageshack.us/img239/9200/metadataca0.png
Setting the db to 9 will shift it by 22 db.

So in theory, skelgard's in the right here (according to your second statement). I'm sorry, but that's just too low for me. Looks like i'll be sticking to 2. :p

What throws me off here though, is you recommending me to go by another value than -31db. :scared:

---
nvm. tebasuna51, I used that program to read the statistics of dolby files, to look at how studio professionals encodes their audios, and I noticed that they don't use DRC. I looked at about 6 different ac-3 files, and were shocked by what I seen. So I decided not to encode using DRC, and encode the way you and skelgard were trying to tell me about, and the audio came out more transparent. In fact, now I can't differentiate from the original at all, heh.

Thanks guys. :)
skelgard, seems I was being a bit arrogant. I was honestly waiting for tebasuna51 to reply for reassurance (because I know that this is his area of expertise.)

tebasuna51

11th December 2007, 02:05

I listened to my encoded file on my modded xbox (using xbox media center), and your source wav, and it sounds equal in loudness. So, my xbox is wrong too? :D

Lets not take this persons thread ot with this debate and let someone else reply please. :)
This's getting out of hand.

But is interesting know how ac3 differ from others encoders (mp2, mp3, ogg, aac,...) because of Dialog Normalization (and Dynamic Range Compression).

How the encoder work:
The encoder encode the full input signal (wav/raw input) but add to each frame (32 ms in 48 KHz) a DialNorm value. And add to each block (5.333 ms) a DRC value (if DRC is in use).

The decoder can be instructed to apply or not the DialNorm and the DRC. First decode the full signal and after apply the attenuation resultant from the DialNorm and DRC values.

There are decoders than always apply the DialNorm then if we want the same volume in wav file (or mp3, ogg,...) and in ac3 file the better method is encode with DialNorm -31 dB.

You can see in the popular eac3 thread the tendency to don't use the DialNorm concept (Sony also).

tebasuna51

11th December 2007, 02:52

Dynamic Range Compression is another concept and is really interesting but may be not like Dolby Digital uses.

The idea is, if a sound have a high dynamic range (a whisper with a big explosion, a flute solo with a full orchestra sound, ...) and we need reduce the volume difference because neighbors or sleeping childs is the moment to apply the DRC.

This effect can be make by the ac3 decoder applying the values calculated by the encoder but also by Ac3Filter itself even if the ac3 have DRC info or not, or Ac3Filter is used to play other format than ac3.

The DRC values calculated dynamically and for any kind of source (Ac3Filter) is the best solution. The DRC values precalculated in ac3 stream permit more simple players but now there are enough power to do dynamically.

By the moment most of the software PC and standalone players only can use DRC if source is ac3 (or dts) then for movie tracks or classic/jazz music maybe is recommendable continue using DRC. For modern music with narrow dynamic range is not necesary.

Terranigma

11th December 2007, 03:16

I was reading this (http://forum.doom9.org/showthread.php?p=369342#post369342) by SomeJoe; seems he recommends the method I thought earlier (and in a few more posts).
lol, this's quite confusing. :p

Skelsgard

11th December 2007, 03:19

As tebasuna says, basically music is as loud as possible, most of songs are at near or at 0dB, right on the verge of digital clipping.
This issue becomes more noticeable when comparing music ripped from CDs to music recorded from vynils.
The dynamic range in vynils was wider than of CDs as the tracks were not dynamic-range-compressed and gain-maxed and that allowed for a richer sound most of the times.

By just disabling DRC on the decoder, you'll experience the full dynamic range of an AC3 track. This option comes in most decoders (virtually all of software and mosts SAPs) these days.
DRC is an option more than a mandatory feature, as I said before and tabasuna confirmed.

Edit: also remember that music is loud all the time. Unlike movies, which have only passages of loud sounds (explosions, accidents, etc), music tends to keep the same levels throughout its entire length.
Because of this, the rules of "speech at -31dB floor" cannot be applied the same way. It becomes something to be done "by ear". If a loud song happens to have a passage of silence or a whispering voice solo, the average RMS value will be affected by this (and its duration), and the overall DialNorm value will be biased by a non-representative portion of the song.
This Rammstein's song happens to have a low gain intro which puts the calculated average RMS at -8.20 dB for those 30 seconds at the beginning, while the first 10 seconds are about -28dB.
But the entire song's RMS value is -7.85dB, with some passages going up to -10.5dB and even to -14dB.
What I'm trying to say is that music is mostly loud while movies are mostly quiet, therefore the sample principles for DialNorm calculation can't be applied in a strict way.

tebasuna51

11th December 2007, 10:29

I was reading this (http://forum.doom9.org/showthread.php?p=369342#post369342) by SomeJoe; seems he recommends the method I thought earlier (and in a few more posts).
lol, this's quite confusing. :p

If you need a Dolby Digital certification to play your ac3 in a environment than only play Dolby Digital material you can use the DialNorm feature, but remember, Dialog Normalization must be calculated in a fragment with a volume equivalent to a normal dialog.

Like SomeJoe say:
"...dialnorm is supposed to be based on the level of the dialogue, not the entire soundtrack."
in this (http://forum.doom9.org/showthread.php?p=846394#post846394)post at the referred thread "How to Properly Encode Dolby Digital Audio"

If you want play your ac3 (I don't say Dolby Digital) in a environment with music CD, mp3, TV audio and so on, forget the DialNorm feature or you will need turn up/down the volume constantly. Sorry but the TV commercials and modern music don't respect the Dialog Normalization propossed by Dolby at -31 dB, now the goal is have the max volume possible.

Terranigma

11th December 2007, 15:49

If you need a Dolby Digital certification to play your ac3 in a environment than only play Dolby Digital material you can use the DialNorm feature, but remember, Dialog Normalization must be calculated in a fragment with a volume equivalent to a normal dialog.

Like SomeJoe say:
"...dialnorm is supposed to be based on the level of the dialogue, not the entire soundtrack."
in this (http://forum.doom9.org/showthread.php?p=846394#post846394)post at the referred thread "How to Properly Encode Dolby Digital Audio"

Thanks tebasuna51, so I was only halfway right. I was scanning entire files (even if it had soft sections like with the rammstein audio) instead of a portion of the loudest areas, or when they at least look equaled. :)

Here's an example:
http://img519.imageshack.us/img519/949/audiowavespectrumgg3.png

By scanning the entire file, I ended up with a rms level of -14.1db, but when scanning the loud area only, the new db was -13.0db

tebasuna51

11th December 2007, 16:10

By scanning the entire file, I ended up with a rms level of -14.1db, but when scanning the loud area only, the new db was -13.0db

But you need scan a quiet area equivalent to normal dialog and not loud area equivalent to shouts.

rosivaldo

16th December 2007, 16:42

Thank you very much for all the information. I thanked a week ago or so, but I must have done something wrong, since the post hasn't reached the forum.

(...)
I would say that Ac3 stereo at 192kbps will be good enough for your backups.

Indeed my purpose is not having a backup. I have this by other means. I really want to have as many music as possible put on some few DVDs, but keeping the quality as perceptually high as possible.

Here I leave you two samples, the original PCM uncompressed file (ripped with EAC) and the AC3 192 kbps encoded file (using Aften 0.05 + AftenGUI 1.4) for you to see for yourself.
http://rapidshare.com/files/75317322/Rammstein-Links123.wav
http://rapidshare.com/files/75317356/Rammstein-Links123.ac3
(...)

I've downloaded the files and the ac3 really sounds good but it is somewhat lower than the wav. I've also read the other posts in this thread and the guide to encoding to ac3, but the amount of information is to much to grasp so quickly. :o

The ac3 is being played by WMP, with AC3Filter decoder (I think - ffdshow has been installed in my machine and it seems to be the source of this decoder). Which decoder do you recommend? And how can I encode all my music without concern with this volume issue?

By the way... You win: I'd like to use foobar2000 in order to do an ABX test with your two samples. But... the different volumes turn it too easy to know which is which, so it will not be a blind test. :( And I don't know how to make foobar2000 to play an ac3 file. Do you know?

Besides, this music is very aggressive to my standards (classic, Vangelis, Kitaro, Enya, etc.). I am *really* not skilled enough to find artifacts in all that brawl. :)

A last word about this skill issue. In order to be sure that an encoder@bitrate is really transparent, shouldn't I test several samples of several genres, several times? It seems a frightening task, easily able to drive me madly obssessive. That's why I'd prefer to rely on the testimonies of experts. I hope I'm not asking to much. :)

Thanks again for your time.

Rosivaldo.

Skelsgard

16th December 2007, 22:25

Indeed my purpose is not having a backup. I have this by other means. I really want to have as many music as possible put on some few DVDs, but keeping the quality as perceptually high as possible.

I've downloaded the files and the ac3 really sounds good but it is somewhat lower than the wav. I've also read the other posts in this thread and the guide to encoding to ac3, but the amount of information is to much to grasp so quickly. :o

The ac3 is being played by WMP, with AC3Filter decoder (I think - ffdshow has been installed in my machine and it seems to be the source of this decoder). Which decoder do you recommend? And how can I encode all my music without concern with this volume issue?

AC3Filter in fact decodes AC3 at about 4 to 5dB lower gain. Don't know why, Vigosvky (ac3filter's creator) has never address this issue as a bug or a problem.
Use ffdshow to decode AC3 and the volume will be the same as the original WAV.

By the way... You win: I'd like to use foobar2000 in order to do an ABX test with your two samples. But... the different volumes turn it too easy to know which is which, so it will not be a blind test. :( And I don't know how to make foobar2000 to play an ac3 file. Do you know?

Foobar uses its own decoder so the volume would be OK with the ac3.
But I don't use foobar so i don't know if it allows blind tests.

Besides, this music is very aggressive to my standards (classic, Vangelis, Kitaro, Enya, etc.). I am *really* not skilled enough to find artifacts in all that brawl. :)

I used that song as example as hard riffs are more likely to produce artifacts than more calm music. If there were artifacts, you'd hear them more easily.

A last word about this skill issue. In order to be sure that an encoder@bitrate is really transparent, shouldn't I test several samples of several genres, several times? It seems a frightening task, easily able to drive me madly obssessive. That's why I'd prefer to rely on the testimonies of experts. I hope I'm not asking to much. :)

That would be the best idea.
To test on several samples to get a greater comparing pool.

Also, the first sample I send was encode as it was ripped from CD.
Some adjustments might need to be done, like reducing the overall gain, as CD tracks are normalized to full volume (closest to 0db as posible) and as such, the volume might be too high for your external decoder or DVD player.
I think a -12 to -16dB gain on all tracks might just be enough, but this is not a scientifically calculated number, just a random one that might be OK for your music.
Also, upsampling to 48kHz for better compliance with DVD standards. Use a high quality upsampler, like SSRC.

Later

rosivaldo

17th December 2007, 13:34

(...)
Use ffdshow to decode AC3 and the volume will be the same as the original WAV.

I coudn't find how to use ffdshow. I thought it was the responsible for making WMP to use AC3Filter (I'm very ignorant :) ).

(...)
Also, upsampling to 48kHz for better compliance with DVD standards. Use a high quality upsampler, like SSRC.
(...)

I should upsample the wav *before* encoding this to ac3. Is that right? I've noticed this advice in a previous post. Why does ac3 perform poorly at 44.1kHz? Even so, will it be playable in standalone DVDs?

Thanks once more.

Rosivaldo.