eac3to - audio conversion tool

[tebasuna51]

Originally Posted by Terrachild

1) I've recently gotten into a debate with some people online that use the Surcode plug-in. I need some facts. If I use eac3to to encode an ac3 file, is the quality substantially different then Surcode?

Read '>' like 'better quality than':
AC3 640 > DTS 1536 > AC3 448 > DTS 768

Is FFmpeg or Aften better?.

The same guy (Justin Ruggles) wrote Aften and ac3 code in ffmpeg, but Aften have more options to tune the encode.

[Terrachild]

Originally Posted by tebasuna51

Read '>' like 'better quality than':
AC3 640 > DTS 1536 > AC3 448 > DTS 768


The same guy (Justin Ruggles) wrote Aften and ac3 code in ffmpeg, but Aften have more options to tune the encode.

So are they equally up to date? In other words, if I'm doing 5.1 ac3 without settings anything other than the bitrate, is there any reason to chose one over the other?

Audacity uses FFmpeg for encoding ac3. Should I stick with that, or is there an advantage to eac3to with Aften?

Do they come close to Surcode for quality?

[leeperry]

Originally Posted by tebasuna51

AC3 640 > DTS 1536

got any source about this please? the more I upgrade my sound system, the less I stand 1.5mbit DTS

[Terrachild]

To answer another debate I've entered into, could someone tell me if it's legal to use an open-source ac3 encoder for non-commercial use. Just a guy sitting at home testing out surround sound. Does the act of simply making an ac3 break any laws in the U.S.?

[tebasuna51]

Originally Posted by Terrachild

So are they equally up to date? In other words, if I'm doing 5.1 ac3 without settings anything other than the bitrate, is there any reason to chose one over the other?

Audacity uses FFmpeg for encoding ac3. Should I stick with that, or is there an advantage to eac3to with Aften?

Well, ffmpeg have many forks, branches, compiles, ... then I can't know what is your ffmpeg and can't say nothing about.

I can't test all ffmpeg versions, I always use Aften.

Do they come close to Surcode for quality?

I don't know blind test between Aften and Surcode.
For me, with my audio equipment and my ears, sound the same.

Originally Posted by Terrachild

Does the act of simply making an ac3 break any laws in the U.S.?

I don't know in U.S. but we are talking about audio encoded (A/52) AC3 compatible, not Dolby Digital.

[MokrySedeS]

Originally Posted by Terrachild

if I'm doing 5.1 ac3 without settings anything other than the bitrate, is there any reason to chose one over the other?

I've made a test but still... I'm not sure what is better.


Here it is:

Original wav............................................ Slowed down wav.................................... eac3to encode......................................... Aften encode


Original wav is encoded using eac3to from 448kbps 25fps AC3 track.
Slowed down wav is made the same way original wav was, just slowed to 23.976fps.
eac3to encode is made from slowed down wav and the only used setting was bitrate at 448kbps.
Aften encode is made from slowed down wav, bitrate at 448kbps and tweaked -w option (bandwidth).

When I did an aften encode leaving -w set at default, the result was exactly the same as eac3to's.

So what do you guys think... should I use aften and tweak bandwidth or is it a bad idea?

[jruggle]

Originally Posted by tebasuna51

Read '>' like 'better quality than':
AC3 640 > DTS 1536 > AC3 448 > DTS 768


The same guy (Justin Ruggles) wrote Aften and ac3 code in ffmpeg, but Aften have more options to tune the encode.

That is not true. Though I am listed as the current maintainer of it, I did not write the AC3 encoder in FFmpeg. It was written by Fabrice, who started FFmpeg. I used his code as the starting point for Aften. The original FFmpeg AC3 encoder has not changed much since it was first written. I made a few tweaks and some speed improvements, but nothing significant.

Originally Posted by Terrachild

So are they equally up to date? In other words, if I'm doing 5.1 ac3 without settings anything other than the bitrate, is there any reason to chose one over the other?

FFmpeg is a simpler encoder than Aften. It only allows 16-bit input, and it uses 16-bit fixed-point for the MDCT and other calculations. It does not have any options other than bit rate and frequency cutoff. Aften allows higher resolution input such as 24-bit and floating-point, it uses floating-point calculations internally, and it has many options. Aften is also multi-threaded.

That said, the quality is not significantly different. There is only so much you can do with the format. The core encoding features are basically the same, the 2 encoders just use different algorithms in several places.

[jruggle]

Originally Posted by MokrySedeS

So what do you guys think... should I use aften and tweak bandwidth or is it a bad idea?

The only downside of increasing the bandwidth is that the encoder has to spread the available bits across more frequency coefficients, so it can decrease overall quality.

[tebasuna51]

Originally Posted by jruggle

That is not true. Though I am listed as the current maintainer of it, I did not write the AC3 encoder in FFmpeg... I made a few tweaks and some speed improvements, but nothing significant...

Sorry, my fault. I was thinking your tweaks like current maintainer was more important.

[Abradoks]

I have recently made an objective comparison between proprietary Dolby encoders and aften. Aften has much worse SNR than DD. Even 224 kbps AC3 encoded with DD beats aften 448 kbps in SNR. When using PEAQ metric aften is closer to DD, but still worse.
Also, if you want to have frequency cutoff similar to DD encoders, you should use w=48 for 448 kbps and w=8 for 224 kbps. This tweak gives a little worse SNR but better PEAQ on 448 kbps (w=8 gives worse PEAQ than default settings).
If somebody is interested, I can repost here that comparison. But, it seems that development of AC3 encoders is dead.

[tebasuna51]

Originally Posted by Abradoks

...
If somebody is interested, I can repost here that comparison. But, it seems that development of AC3 encoders is dead.

Two post before you have jruggle the Aften developper.

Maybe he need some support to improve the encoder, everybody is interested in a good free ac3 encoder.

But, please, use the Aften thread for this.

[shon3i]

Originally Posted by Abradoks

When using PEAQ metric aften is closer to DD, but still worse.

Nothing than usual ABX comparing is not optimal for making decision which encoder is better, none of methods PEAQ or Spectogram like MokrySedeS posted. Because encoder made decision to cutoff certain frequency range to keep overall quality. That means audio still can be transparent to source, but if you look at spectogram and see that something missing, because forcing to not cutoff (leave frequency range) is not good.

Many people here have a bad decision, rather than listen, they watch the some values and graphs, which is completly wrong.

[Terrachild]

Abradoks,

very interesting.
Could you tell me what commercial encoder you tested. (Surcode?)

How did you measure signal to noise ratio, is there a free program you can point me to that can do that? What about measuring PEAQ?

You said use W=8, what setting is that. Im using "WAV to AC3 Encoder 4.1" where is that setting?

Thanks

[raziel666]

I have a certain DTS-MA audio track from a bluray disc of mine and I want to compress it to FLAC.

I own Sonic Cinemaster Audio Decoder 4.3 and have installed the .dll's of ArcSoft using the trial version as described in other posts in doom9.

Now the problem is this: when trying to encode to FLAC, the ArcSoft decoder gives a file twice the size of that from the Sonic Audio Decoder.

After examining the files with MediaInfo, I found out that the bitrate varied between these two decoders. The ArcSoft one produced a file with 2833kbps and the Sonic one with 1631kbps. Moreover, the file produced with ArcSoft is supposedly 14 minutes longer than the original duration and when I tried to play it back, it sounded slowed down and like crap...

Am I doing something wrong? Is it some bug that one of the decoders has?

Thanks in advance for your answers.

[rapscallion]

It's a True-HD Blu-ray track and I'm trying to create lossless wavs. (using eac3to_more) and I'm running into a strange problem.

First, when I create wavs, from a regular AC-3 track, Nero decoder (direct show) does the processing and maintains 24 bit.

However, when I try to do this to the HD track, ffmpeg/Libav somehow does the decoding and after the first pass Libav reports superfluouus zero bytes, eliminates them and drops down to 16 bits.

Is there a way create lossless wavs, maintaining 24bit, from True-HD ?

[Inspector.Gadget]

If eac3to using LAVC says a lossless codec is actually 16-bit, it is. 24-bit decoding for a lossy format like AC3 makes sense because presumably you're going to transcode it later and the additional accuracy is desirable.

[rapscallion]

Yes, however, at the first pass it's reported as 24 bit. And, as I stated, libav is dropping it down from 24 to 16 bit, NOT reporting that it originally is.
What is lavc btw, or did you mean libav?

My goal is to re encode the wavs to DTS-MA.

[Inspector.Gadget]

LAVC = libavcodec. The first pass calls it 24 bit because eac3to has to make sure that the 8 bits of padding lasts the length of the file and there's no actual 24-bit content in there. Once eac3to has verified this, it can discard the padding during decoding.

[rapscallion]

Ok, thanks, I got it now....just two last questions.

Is Nero docoder (from ver 7.10.1) not able to decode TrueHD and that's why eac3to is choosing libav/ffmpeg ?

When I re encode to DTS-MA will the resulting track be back to 24 bit or does it even matter ?

[Abradoks]

I have posted comparison in aften thread.

Originally Posted by shon3i

Nothing than usual ABX comparing is not optimal for making decision which encoder is better, none of methods PEAQ or Spectogram like MokrySedeS posted. Because encoder made decision to cutoff certain frequency range to keep overall quality. That means audio still can be transparent to source, but if you look at spectogram and see that something missing, because forcing to not cutoff (leave frequency range) is not good.

Many people here have a bad decision, rather than listen, they watch the some values and graphs, which is completly wrong.

I'm not going to argue about subjective vs. objective tests. If you have enough resources to make ABX comparison, it'll be very interesting to see how they correlate with SNR and PEAQ.
Besides, my tests were started because somebody reported bad listening quality of AC3 track encoded with aften compared to DD encoder. It was suggested that default frequency cutoff causes such effect.

Originally Posted by Terrachild

You said use W=8, what setting is that. Im using "WAV to AC3 Encoder 4.1" where is that setting?

You can control encoded frequency range through commandline option "-w" when using aften. I don't know if any GUI can do this.
You can find information about used encoders and measurement tools in the post I linked.