what does everyone think about this process for testing the accuracy of a lossy codec:

http://www.microsoft.com/windows/windowsme...lo/moreinfo.asp

im interested to see what you all think.

the fact that someone, and I think it's fair to say that everyone would, will be able to clearly hear the sounds in this "difference" file does not automatically mean that the codec-produced file would sound different than the original. Obviously any song passed through a lossy codec is going to differ from the original. That in no way means that you would absolutely be able to tell the original from the lossy-encoded file. I fail to see any merit in such a test, unless your goal is to prove that a lossless codec is lossless, and clearly that's not what is being spoken of.

Ironically, a codec that knows more about masking is going to do worse on this 'test'.

Of course, coming from a reliable source like the makers of WMA, what would you expect?

--
GCP

Yes, the fact that there is more difference doesn't mean that the sound is worse, because all that a lossy encoder is doing, is finding the maximum amount of info that can be removed without the difference being audible. The problem is not the amount of info removed, but how well this info has been chosen.

However, reading further the Windows page, it seems quite serious : they link to "Questions Raised Concerning Real Audio 8 Test Results from Exodus Performance Labs (EPL, formerly KeyLabs)", unfortunately, the link is an EXE, and I won't run it for safety.

Then, they supply audio samples : http://www.microsoft.com/windows/windowsmedia/wm8/audio.asp

I don't have the time to listen to them right now, but what we have to do instead of criticizing blindly is downloading them and doing better (they provide the original).

It's a selfextracting zip containing a Word file. They criticise a test by Real Software, which (I guess) showed Real to be better than WMA.

You have to understand you will not find any objective information about WMA on Microsoft's own pages. I did a review of one of their 'tests' that demonstartes WMA to be better than Real once, and you might be able to find it in the archives (could be on r3mix.net as well)

--
GCP

A nice quote from Microsoft in that document:

To that end, we encourage you to take a look at the following white paper entitled, “Windows Media Audio 7 and RealAudio 8: A Technical Comparison”. Included in this white paper are steps on how you can evaluate the output of Windows Media Audio and RealAudio objectively using a spectrum analyzer.


Need I say more?

--
GCP

No, no, it's enough !

Yucks! That's all I have to say!

The fact that Micro$oft states "WMA8 saves more of the original wav file than blah blah blah" demonstrates thier complete lack of knowledge in the perceptual coding field. Lossy compression is not about saving more/less of a physical wav file. . that is more a data entropy issue, and we're so far below the entropy point (the point at which data cannot be compressed any further without loss) that it's a moot point. It's more an issue of extracting the "right stuff" from PCM data to get a similar-sounding output with less data. WMA may claim that thier format saves more of the PCM data, but it sure as hell doesn't help the sound quality in any way!

If the idiots at Micro$oft would use bandpass filters, their codec might sound a little better. . .I've edited wav's in Sound Forge and manually bandpassed them at 12 and 16Khz. When encoded to wma8, the 12Khz lowpassed file lacks the ringing artifacts that plague WMA (among other things), and sounds just like. . a shitty-quality low-bitrate mp3. Perhaps "telephone quality" at 64kbps is a more truthful statement for WMA!

I think the people over at MS know very well, that their test doesn't say anything about codec quality. But to the majority of the people this approach sounds very plausible. "Hum, WMA must be better because it is closest to the original WAV." And how many people still judge the quality of a codec by looking at a spectrum analyzer? You can ask many "experts" that this is the way to go. MS is just exploiting the lack of knowledge of the masses in the best way possible. :spank: