Heya Folk,

I was hoping that the lovely people who put together listening test could start planning for a multiformat multichannel listening test, which included at least AC3, AAC, WMA and WMA Pro.

Most of the encoding I do is multichannel and I'm looking for the best codecs and the best encoders of those codecs...

• WMA has traditionally 'brightened' audio with many artefacts, but noone seems to have tested it in multichannel.
• AC3 is naturally the standard, but how does it fair against the latest AAC?
• In my experience many AC3 -> AAC transcodes using Nero result in a file similar in size or larger than the AC3. What is the best multichannel AAC encoder? Do AAC+ technologies work well on multichannel?

Thanks!

Dom.

kinda undoable. You can't use headphones (for somewhat obvious reasons), material is hard to come by (CDs don't offer multichannel, DVDs already come compressed and transcoding is a no-no) and last time I checked, ABC/HR wouldn't play multichannel.

That is a shame, then, because I am visiting here from an AVS Forum thread where the controversy is over data and sampling rates of the new optical formats and how this codec, DD+ for example, is as good as lossless Dolby TrueHD. There are a lot of people who would like guidance on this.

I think a multichannel test would be very useful, and I disagree that it's "hard to come by". In foobar2000 you can downmix the channels for headphone listening - That would do most of the process, but I'm sure you are also able - with some plugin (channel mixer?) - to map other channels to 1-2 channel headphone listening.

Personally I'm interrested if the bitrate could be taken down way futher than the default 384kbit AC3. From my personal experience sub-384kbit AC3 sounds really bad.

You can downmix, but you also introduce the possibility for extreme phase issues. In fact, it's pretty much a guarantee that you're going to encounter seriously degrading phase-coherence problems that would make such a test practically pointless.

The best you could do would be to split each stream into two stereo streams and one mono stream (for the center), and you'd have to figure out the "LFE issue" as well (sum it into the three streams...but how much?). Even then, you're not fully realizing the purpose of the 5.1/7.1 formats.

This might be something that's doable at a local level. Book a THX-certified mixing stage, for instance, and have an operator run the test one person at a time. Of course, this ain't a cheap way of going about it, and it'd be difficult to find a method for playing all of these various formats (can't really install apps+decoders on studio machines). Failing that, you'd have someone sponsoring a test in their home, but the sponsor would have to ensure that a number of bases are covered.

Why does it have to be so complicated, don't most audiophile have a 5 or 5.1 setup on their TVs at home? I'm not an audiophile and I upgraded last year with my new TV.

Why would it make it pointless? I assume the phase-problems would be equal for all the encoded streams


Why make it more complicated than this? Eigher you can ABX any channel or you can't.

Edit: Do anyone know some good multichannel sources? E.g. music and some movie scenes with effective soundtrack. I assume DTS or some of the new lossless sources from Blu-ray/HDDVD are best suited as source for such a project.

Multichannel audio is not meant for listening over headphones. Only test in a real conditions (listening on 5.1 or 7.1 system) would make sense.

Simple solutions for down mixing to stereo produce very bad results. To preserve surround sound, something like Dolby Headphones should be used. But if this is used then you are testing how Multichannel+Dolby Headphones sound when listening over headphones, not how multichannel would sound in 5.1 system.

In public listening tests for stereo every listener used conditions which he had at home (soundcard, headphones, background noise). So I don't see that a listener using 5.1 system available at home would be a problem. Of course in this case room acoustics and position of speakers and position of a listener make big differences. If this kind of test is going to be organized then some conditions for position of speakers and listener should be predefined.

How does this matter? My goal for a listening test would be to find how low bitrate you can use without noticing artefacts. This should be possible to distinct no matter which channel is being ABX'ed, right?

There is lot of inter-channel masking going on when playing in 5.1 system. So you can introduce lot of quantization noise in one channel which will be masked from a signal from another channel.
Then I can construct you multi channel files with music which would sound totally different when down-mixed using standard matrix method. I believe that it is possible to create multichannel example which would create total silence when down mixed, but I would have to check that.
Then there are some cases where you almost can not hear differences from original when playing on a set of speakers and with headphones compressed signal sounds awful (such an example is double talk which some encoders introduce in famous German speaker sample used often in listening tests).

European Broadcasting Union has performed evaluation on multichannel audio codecs. There are differences in the bit rates so results are not directly comparable. Hope you find it interesting.

http://www.ebu.ch/CMSimages/en/tec_doc_t33..._tcm6-53801.pdf

Are there any multichannel test samples (problematic or not)? I'm doing some tests with lossy multichannel coding but lossless sources are hard to come by.

EDIT: Thanks for the paper on the EBU test of multichannel codecs, JP_1970.