Let's Solve the Ac3 Volume issue

[primusmp]

Ok here's the deal I've read almost every thread there is concerning the AC3 low sound level. (In the Spanish and English Forums). I Understand the Ac3 coding guide and applied it to my Ac3 files, so this is what happens:

PCM track has a "normal" sound LEVEL (I mean by this that my TV sound volume must be around 10 to hear it properly)
AC3 track has a "low" sound LVEL (TV sound level about 50 to hear it properly)

Same sh--t with PowerDVD PCM track loud AC3 low... so I figured well not the DVD standalone player nor PowerDVD are decoding the Normalization Value correctly and in order to hear the AC3 track properly, I must get an AC3 Decoder but.... (there's always a but) I have another DVD that both AC3 and PCM ltracks play at similar volume (not exactly the same but both in the "normal" range).
So I figured that maybe it's the encoder (I've been using Scenarist AC3 encoder) so I changed and tryed Soft-Encode but the resulting AC3 i's still low.

So now I don't know what to do.. I know it's possible to create AC3 and PCM with "similar" levels but HOW?????

Maybe this DVD that plays "OK" has a "pumped" AC3 but that should clip the sound... so I don'know..

If there's a player problem (i.e. player not normalizing properly the AC3 track) why is this correct DVD track played at decent volume (same players).

In case you want to know original tracks ara 48k DV=> WAV PCM. I even tried normalizing the tracks in Cool edit before encoding to Ac3 (not just measuring them but normalizing and re-measure them)

I'm inviting you to design a series of tests or receipes to see if I/We can solve this...

Thanks.

[primusmp]

Ok here's more info:

I ripped the AC3 file from the dvd, the one that has the correct levels and here's the audio stream INFO according to Soft Encode (Audio OK params.jpg)
this file alone plays ok in Soft Encode and in Power DVD.

I exported the file to WAV PCM (Encoding settings in "ENC settings SOFTENC.jpg")

Then opened the file in Cool edit and mesured it (results in "DECODED MEASURE.jpg")

Then re-encode it with AC3 ENC (the one from scenarist) with this settings: "Reencode Setting AC3 ENC.jpg"

As you can see although cool edit measured -24 dB I kept the -27dB in the setting.

This newly created Ac3 plays with LOW sound level both in PowerDVD and SoftEncode.

Here are the new Parameters shown by Soft Encode ("Reencoded Params.jpg") adn both Waveforms as shown by Soft Encode, the original Ac3 and the new Ac3. Will add post reply cause exeded 5 attaches

[primusmp]

OK here are the waveforms:

Finally I must say that I tried to encode the Wav file with Soft Encode but it hanged up although I previously tried encoding the same Wav file (original 48k form DV=> PCM wav) with both AC3ENC and Soft Encode and both ended with LOW level AC3 so I suspect that if I could have exported it with soft Encoder it would have been a low level Ac3 file anyway.(Don't know if that if sentence is correct, but if it is; my english teacher can be proud of me). If someone need me to confirm this I'll try other experiments...

Hope to find some answers soon.

[Video Dude]

I previously tried encoding the same Wav file (original 48k form DV=> PCM wav) with both AC3ENC and Soft Encode and both ended with LOW level AC3
Encoding with dialnorm set to -31dB should not change the volume level.

Also, disable all extra settings except for the low pass filter.

[primusmp]

I know that the actual Volume level will not be changed, what I meant to say was that I took a different Wav PCM file (One that didn't hanged up the Soft Encoder) and the result in PLAYBACK audio level was the same with the SofEncode Ac3 as with the AC3 encoded with AC3Encoder.

Althoug Actual Volume is not changed by setting a diff. audio norm value it sholud affect the Playback Volume becuse de Decoder should amplify the signal according to this param, right???

Then why is it doing it correctly with the original one, and not with the reencoded one??

The extra settings were left as default cause it should not affect the problem I was trying to test.

Sorry Video Dude but your answer didn't add anything new.....

[Rockaria]

To conclude, you assume, there are problems in the encoders you've tested with all the parameters suggested in the forum, only to get the 50% of the original volume when decoded on the receiver.
There are some more encoders you can try such as besweet and ffmpeg... But I don't think there are much things you can handle to get the better results.
Another very differrent method is using the ffdshow/ac3filter spdif out capturing method in the graphedit. Also you can try is to capture(it takes full decoding time) the spdif-out of the soundstorm(DICE) encoded ac3 if you have those boards.

I suggest you try the FFDShow in the graphedit with many DSPs including gain/DRC control & previews before encoding :
Any multi channel source-> FFDShow decode & volume gain & spdif encoding -> file writer
Adjust the volume controls & preview with the direct sound out device attached at the FFDShow out pin with the AC3(upto 640k) out enabled in the output setting if you have a digital receiver attached. If not, use any pcm out for the preview.
If the max attained volume(with the DRC) sounds OK with no distortion, you can start recording.

If previewed with a digital receiver(spdif-passthrough), the recoded ac3 will play exactly as previewed, on any media.
Also you can check the encoded parameters with the FFDShow/AC3Filter stream info and reflect the value to your preferred encoder.

http://forum.doom9.org/showthread.php?p=666230#post666230

[primusmp]

If I understand correctly the above method you describe would be useful in case of a "Backup" or if you want to extract an AC3 and then re-encode it for let's say reauthoring it...

I ultimately want to know how to produce an AC3 Track that would decode properly on standalones and/or software DVD players.

I mostly want to produce this tracks from 48k PC Wav files.. I know it can be done I just don't know how to do it.

The avobe Experimente was made in order to track down the problem I started with a "correct" AC3 got a "correct" WAV PCM and then reencoded to a "Incorrect (LOW)" AC3 this should suggest AC3 Encoding issues but, don't ask me why, I suspect this is not the problem exactly, mainly because it seems that other people are able to produce "correct" AC3 tracks with the same encoders.

Needles to say, I definitely think It's not PC system related mainly because I've had this problem ever since I started.. and had changed OS and Hardware and Programs several times along the way (In at least 3 or 4 computers that I normally use).

I'm trying to do further test with other Encoders as soon as I have some time

If anyone want to suggest something else to test in order to better identify problem, please let me know.

Thanks

[Rockaria]

If I understand correctly, you've not tested the mentioned method but already concluded you need to test something else when it contains everything you are inquiring....

[tebasuna51]

@primusmp
See my post in spanish forum:
http://foro.doom9.org/viewtopic.php?t=6699

[Skelsgard]

I had the same problem a few years ago until I figured out how the encoding/decoding process worked for AC3 files. This is acquired thru experience so there are no technical papers that will suport the statements made here by myself, so if u find any technical info telling me wrong, don´t come back just to bitch about it (since I´ve already admitted that I might be wrong).
The main parameters that controls the volume levels on a AC3 file are the DRC, and the DialNorm. When the DRC is disabled, the DialNorm overall power to control the volume is diminished. DRC on the other hand is the most powrfull way to handle wih the volume level of the entire track.
-31 dB has been set as the min volume level that will be taken for the mixing source for the voice (this means that basically it is asumed that all sources have the dialog at -31 dB maximum compared to the loudest sound). The standard is -27 dB wich assumes that all sources have normally a -27 dB level compared to the loudest sound. But this parameters affects the entire track, not just where dialog is. Since -31 dB represents the base or floor for sound, it can be interpreted as 0 dB, then the difference between the selected level and the floor is the amount the volume will be decreased in the output file (I.e. if the A is played at 0 dB and B is played at 4 dB, in order to maintain a proper maximum normalized level the decoder will set a -4 dB gain to the B track to achieve that. When the floor is -31 dB, wich will be the lowest volume level, and a track is encoded with a -27 dB parameter (+4 dB above the floor) the decoder sets a -4 dB gain to ensure that the loudest sounds in the track don´t causes distortion of whatever kind).
DRC on the other hand has been created to warrant that the soundlevel won´t play so loud that will damage the speakers if these are not prepared to handle an specific amount of energy. This way if DRC is enabled, it will boost or lower the volume to maintain and overall normalized level throughout the entire track. All AC3 decoders can apply DRC (and most do it by default). Therefore the sound level will be dramatically decreased.
At the same time, most AC3 decoders have DRC as a feature that can be disabled, avoiding this issue if wanted. But to guarantee that DRC will never be applied, the Ac3 file must be encoded without it (for those decoders that don´t have the disable-DRC option). Also, the RF overmodulation protection should be disabled.

[primusmp]

If I understand correctly, you've not tested the mentioned method but already concluded you need to test something else when it contains everything you are inquiring....

I intend to test this method once I get the suggested programs and some time available.

I understood that this was a method for re-encoding AC3 (because I read it to fast), so sorry.


Thanks.

[primusmp]

I had not tested the FFdshow Grahedit method yet (I must learn about it first).

What I did was this:
I encoded the same track with no DRC and Norm at -3 and -31 I definitely found a differnece in playback volume between them (as I expected) but even the one encoded with -31 wasn´t as "loud" as the original AC3.

What I will try next is the following:
Cool edit gives you an AVG RMS and a max value so I will try to set my AC3 in order to the Max Value to be near 0dB instead wich I should be able to do if I set the diag norm param to AVG - MAX this will leave my AC3 track peaking at 0dB.

I will try also the FFDShow mehod once I figure how to use those utilities (MAybe we have results tomorrow)

[Skelsgard]

but even the one encoded with -31 wasn´t as "loud" as the original AC3
if u are using a AC3 --> WAV --> AC3, are u sure the source AC3 is not being decoded to WAV with DRC or DialNorm? cause this will give u a WAV that is quieter than the source AC3, thus the re-encoded AC3 will actually sound as loud as the WAV, but not as the source AC3.

[primusmp]

are u sure the source AC3 is not being decoded to WAV with DRC or DialNorm? cause this will give u a WAV that is quieter than the source AC3, thus the re-encoded AC3 will actually sound as loud as the WAV, but not as the source AC3.

I've checked this and WAV Level is between original AC3 and reencoded with -31 Diagnorm (Let's say in the middle).

Supposedly if I follow the AC3 Encoding guide, and according to cool edit's EQ-LOUD value it should have been encoding with -24 wich would have ended in an even lower volume track, but MAX value is-12 so if I was to apply my new criteria to leave the sound peaking at 0dB I should use a Diagnorm Value of -36 (wich I can't by the way) but in this case maybe I should pre-amplify the Wav in CoolEdit, or normalize it to a higher value...

I'm trying to understand how this sh**t works but don't quite figure it out completely, I'm pretty sure the problem is with diagnorm (Since I almost never apply DRC) I know that with lower values the sound get's played at higer volume but I still don't know how to make it sound at the same level as a PCM track. I also have questions about what would happen with audio clipping in case I encode everything with -31.

Certainly the method described in the AC3 Encoding guide doesn't achieve this.

Maybe I should, as suggested in other forums/threads, try to normalize the WAV file to a certain Value and even modify it so that the amplitud of the signal covers the hole space between that value and 0dB, before encoding it to AC3

[Skelsgard]

I also have questions about what would happen with audio clipping in case I encode everything with -31.
The DialNorm nor the DRC actually alter the AC3 file itself but they tell the decoder how the file should be played.
If DialNorm is set to be ignored by the decoder, the sound will be at the same volume than the source WAV.

[primusmp]

I know that both parameters are metadata intended for instructing the decoder how should it play it, but if the metadata says "play it 31dB louder" it may cause clipping if the original sound file has levels avobe -31dB (actually a little more because most systems can handle some overhead), that was what I meant...

In any case I don't think dolby's idea of this is something like "if it's low just put de Diag Norm at -31" I suppose there's a way to calculate the correct params, In the AC3 encoding guide is a method that ends with low volume playback in all of my encodings. I used to think that it was supposed to be that way and that this wass supposed to be corrected by a "Hardware AC3 Decoder" but I found some DVDs whose Playback sound level (the Ac3 tracks of course) are similar to those of the PCM ones, so I begun this thread in order to try to find why and how to make one of this marvelous AC3 tracks with high Playback volume levels :)