I was just wondering how mpc (standard or extreme setting) compares to ac3 (at 192kbs 44.1 kHz). A friend of mine has ac3 encoding software (soft encode) and has been demonstrating how well it handles encoding his CDs.

Tes

AC3 at midrange bitrates (192kbps) doesn't even perform as well as AAC, let alone MPC.

If you try encoding some of the more difficult clips this should be fairly noticeable.

I have a rather hard to encode song by Orbital in AC3 (192Kbs 48kHz). My listening ability may be lacking at the moment due to the fact that I am just getting over a cold but I have been unable to distinguish the difference between the AC3 file and the original WAV.

If a anyone would care to listen for themselves I have uploaded the file here: http://ac3demo.tripod.com/testfiles/Orbital.zip
(this is a large file: 34 MB)

The zip file incudes the AC3 file, and the original WAV compressed using the latest beta version of monkey's audio. I would be very interested if someone with "golden ears" can point out any artifacts in the AC3 file.

BTW the WAV file was resampled to 48kHz using ssrc before it was encoded to AC3 with Soft Encode. This was done to enable the file to be played back in Windvd(seems to only want to playback 48kHz AC3 samples via s/pdif). Denon AVR-1801 used for AC3 decoding.

Tes

The download link doesn't work for me.

--
GCP

The link doesn't work for me either.

And also, what I meant by "hard to encode" tracks are the well known problem case samples for various codecs. Fatboy, castanets, then some of my personal favorites like death2, drone, 2nd_vent_clip, etc.

When it comes to higher bitrate encoding (>160kbps) I favor using more difficult tracks for tuning/testing because the results they show are usually indicative of how finely tuned the psymodel is, or how well a codec can perform when pushed to it's limits.

Not incidentally, MPC seems to outperform every other codec I know of currently in these situation and judging by the very low number of clips (I can probably count them on 1 hand) which have been proven to show (usually slight) audible differences it seems to be a safe bet that this carries over into "more realistic" music.

For that matter however, I have actually tested AC3 at those bitrates in the past and do not remember being particularly impressed (which is probably the reason I don't use it now), but if you have links to an encoder or something (legally of course), I'd check it out if you want my opinion.

Sorry guys, Tripod deleted the account already. I could try to find somewhere else to host the files if anyone is still interested.

Dibrom: Were you using a software decoder for the AC3 files you tested in the past? I am using a hardware decoder, this might make a small difference.

Tes

Yes, but a compliant decoder, be it hardware or software, is not going to make up for format deficiencies, and I'd find it unlikely that negative experiences I may have had with the format in relation to sound quality were due to bugs in the software decoder either.

I had a chance to to do some more listening tests today and was finally able to notice some slight artifacts in the AC3 encoded version. The song used in the encoding is fairly intense (as far as electronic music goes) and I believe it to be a good sample of the encoders ability.

On another note, I also encoded the song using Frank's 0.90l mpc encoder. The results were impressive.

Although I was not able to find any freely distributable AC3 encoders, I did notice some files that looked like source code for an AC3 encoder in the avifile 0.6 source. I don't know how well it works if it is even working at this point.

Dibrom: Thank you for your insight on the subject of AC3 quality.

Tes

Np. I'm actually going to be able to post some AC3 files encoded via Soft Encode here in a few minutes so you can decide for yourself how you think these particularly difficult files sound (and then compare to MPC).

http://www.animus-facticius.org/ac3stuff/

Some "codec killers" (yes, all from real music too, no sine sweeps or the like) encoded to 192kbps AC3 via Soft Encode then decoded back to .wav files via ac3dec, along with the originals in the same directory.

They are encoded via LPAC:

http://www-ft.ee.tu-berlin.de/~liebchen/lpac.html

(binaries towards the bottom)

As you can see on these files, AC3 does quite poorly. Try encoding the originals with mpc -xtreme if you like for comparison. I'd even say that VBR MP3 (LAME rev7 --alt-preset standard for sure) outperforms AC3 at 192kbps, given these clips at least.

Wow, those samples clearly bring out major problems with AC3. The most obvious difference I noticed are that the AC3 versions seem muffled (they lack some of the high frequencies that occur the original).

If you are interested in my sample I can upload it If you know anywhere that will keep the files alive for a while.

Tes

It seems to me that those samples reflect more of a problem with Sonic Foundry's ac3 implementation than they do with ac3. I encoded death2 with Liquifier and got MUCH better results.

BTW, I LPAC'ed the decoded Liquifier sample of death2. If I can find a server to upload it to, I'm more than happy to share it...

I don't think so. I've encoded AC3 files via Liquifier before also and the results are generally similar (in that I remember sound quality not being impressive) even if a file or two may sound different. Try encoding some of those other files if you like (drone for example)..

AC3 just really isn't that great of a format to use for music encoding at these kind of bitrates.

At any rate, I'll encode some of these files via Liquifier and put them online also.

I will. While you're encoding those files with Liquifier, see if you don't also see a HUGE difference (at least in high frequency loss) with death2. I abx'd the death2 Sonic Foundry encoded sample versus the Liquifier encoded sample 20 out of 20 times...

Yeah. I'd say the muffling was one of the big problems, but even worse is the absolutely horrible pre-echo artifacts everywhere. There's also some weird rattling sounds and all kinds of other anamolies..



Well I'd be interested in listening to it otherwise, but I'm not sure its necessary at this point.

OK, I just did drone with Liquifier. Same thing. HUGE difference between the Liquified ac3 and Sonic Foundry's. 20 out of 20 abx... If you have an FTP address, I'll happily upload the samples. Or, can you reccomend a free server I can post a link to?

OK, I uploaded the death2 and drone samples to this site:

http://www.geocities.com/layer3maniac_greg/

This lame server forced me to zip the pac files, but it's the best I could do...

Hrmm.. I was only able to download drone, now the site won't load up. Eventually I'll put up the clips encoded with liquifier myself but I won't be able to for a little while.

Anyway, it's arguable whether or not Liquid Audio's file is "better", the bass certainly sounds worse than Soft Encode though in general it doesn't sound as "muffled" or have that "pumping" quality. Neither sound good at all and both are very far from MPC quality.

So I still conclude that AC3 is nowhere near the level of quality of MPC, especially at these bitrates. I don't think anyone would really argue with that.

Edit:

I was able to get the death2 file. This one might actually sound "better" than the Soft Encode one. There's not as much of a "hollow" sound I suppose due to hf content not being cut out so much, but the pre/post-echo is much more audible in this file and there is a generally unusual sound to the background pulses.. a bit of a noisy almost rumbling quality. So it's the same thing again, like I originally said.. it may sound different, but it still sounds bad. In the case of death2, VBR MP3 (--alt-preset standard) is clearly better.. let alone AAC or MPC. I'd imagine it'd be similar in the other cases as well since it is with Soft Encode, and so far these Liquifier encodings, while producing different artifacts, do not sound significantly better.

Also, not directly related to this, but there appears to be a bug with Liquifier in that it places a few samples from the end of the file at the beginning... I've noticed this before but I'm not sure if there's anything you can do about it.

Sorry to bring this subject up again but after a short while of reading this thread I came to a shocking conclusion.

So, does this mean that at high-fidelity point of view all my "crystal clear, hi-fi surround quality" DVDs actually have crappy soundtracks compressed with an inferior codec, probably containing tons of artifacts? Or do producing studios use much better software/hardware for AC3 creation?

And wouldn't that also mean that AC3 -> MP3 transcode would be right out of the question considering transcoding rules (or a single rule; don't do it!)?

Well I believe with most DD5.1 encoded material as well as audio stuff encoded with AC3, the bitrate is much higher than 192kbps. Usually 384kbps instead. I have not tested at those bitrates but I'd say its likely that quality is significantly better at those bitrates than at 192kbps. However.. I think it's safe to say that MPC or AAC at those bitrates would still outperform AC3, probably by quite a margin even.

About the quality of the actual AC3 encoding itself, I don't think it would be much better than Liquifier, since that is actually supposed to be using Dolby's algorithms which should be the best.

Also it's probably worth noting that the cases I presented are "codec killers". They give MP3 hell also.. but most people don't have many qualms about using that for high quality encoding... so maybe that puts it into perspective some. Still, a highly tuned LAME setup seems like it will outperform AC3 (at those bitrates) on these really samples.



Maybe. It certainly depends on the bitrate I think. There's not much point in going from 192kbps AC3 to 160kbps mp3 (vbr or not) at all. 128kbps might make the loss more worth it because of the space savings but still..

The problem is however that there are quite a bunch of movies with 2ch 192kbps soundtracks available, especially older movies prior to Dolby Surround revolution. Now there's a surprise for that audiophile with his $100,000 THX rated surround system. "What? Pre-echo!? AAARRRGG!"

And even greater problem is that I've seen numerous music video/live performance DVDs which also have 2ch 192kbps compressed audio. This is quite enraging actually, if I pay 199FIM ($30) for a music DVD I would think I'm entitled to at least CD-quality audio!

Fortunately at least few of those DVDs have uncompressed PCM audio. I wish it would be realistic to to hope for a better compressed audio format for DVDs but with current licensing and hardware support AC3 is here to stay...

To confrim (from experience, and test results): two channel AC-3 is inferior to mp3 at an equivalent bitrate. Two channel AAC is superior to mp3 at an equivalent bitrate.

So for stereo material, AC-3 is much worse than AAC.


AC-3 is spec'd up to 640kbps. The 384kbps "limit" exists on optical film soundtracks (due to the physical space on the film), but not on DVDs. I think the widespread use of 384kbps on DVDs is due to lazyness, where it's just mastered from the film encoding, rather than going back to the original 6-track masters and generating a 640kbps track. Also, some decoders fail to handle higher bitrate streams, but most do.


One thing bemuses me: I've heard some terrible artefacts on music at 384kbps surround on DVD discs - but using the Sonic Foundry Dolby AC-3 encoder I can't recreate these serious problems, so I don't know how the films came to sound so bad! Probably transcoding, or an earlier encoder version.

AC-3 is the DVD standard because it was already available, and it is marketed by a big American company. As we are hearing, sound quality has nothing to do with it.

David.

My experiencec with ac3 are quite good!!! Also for 192Kbit 2 channel audio....

If you use Softencode(sonic foundry) make sure you turn off all the normalization options and filters!!!!!!!!!!!!!!
Movies normally need a much bigger reach in volume level and I think AC3 besides for multiple channel was made to do this as well......

I just started making mini-DVD's containing several audio albums, complete with graphics and easy menu's.... I do use 224Kbit AC3, just to be sure, but I can't really hear the difference.
I can't wait to start stuffing music on dvd's this way..


Mikez

Actually, a lot of DVD movies are coming out with a higher bitrate...448 versus 384.

You want high quality? Buy some DVD-Audio discs... Uncompressed 24-bit, 96kHz, 6 channel LPCM audio... Now we´re talking quality... (Unless the person/persons in charge of the mastering sucks... )