In Creative Labs new USB based soundcard Extigy, they switch to ADPCM on the fly to achieve more channels than two to overcome the bandwith limit of USB. Otherwise, the limit would be just two channels.

My question is if anybody here knows how a 4:1 ADPCM audio coding affects the sound quality compared to ordinary PCM coding.

The technichal part of me says "wait until Creative releases a usb 2.0 version" to adress this issue, but if the sound quality for ADPCM is not an issue - it will probably only be needed for plaing back Audio DVDs and DVD movies - then why wait?

Peace,

Emanuel

You can test the effects of 4:1 ADPCM encoding yourself. All windows 9X version come with IMA ADPCM and M$ ADPCM installed as ACM codecs. Use an editing program to convert a wave file to ADPCM (Use IMA, it's slightly better than Microsoft's) and listen to both to see if you find any sound degradation.

Of course, there will probably be differences between Creative's ADPCM and IMA's. But that test already gives an idea of the pros and cons.

Concluding: Of course, there will be sound degradation. But you won't find there artifacts common to psychoacoustic encoders, like pre-echo.

Regards;

Roberto.

Thanks for your answer, Roberto. I have already tested the method, but for now only on a relatively cheap audio system - and it is very difficult to judge where the artifacts are. Until I upgrade the system and can judge for myself, I wonder what kind of known/common problems there are with ADPCM audio coding. If pre-echo is not an issue, does it cut in the frequency range?

Peace,

Emanuel

The only thing you'll probably notice is a slight increase in noise. This is because ADPCM uses prediction and a predefined index of sample values. These predefined are used instead of the real value.

So if you have, say, 283, 999, 391 as the input, ADPCM codecs may make them 290, 1000, 400.

It sounds pretty good, but it's not perfect

You won't get pre-echos, dropouts, etc, but mostly quantisation noise with ADPCM. How ADPCM works is that it looks at the last few samples and predicts the value of the next sample. The prediction error is then scaled (adapted to the average volume of the source), quantised and encoded. Since 4:1 is used, that's 4 bits per sample for a 16bit source, so the error can only take 16 values per sample, so that's where the quantisation noise comes in. How well this encoder performs depends on a lot of things, including the complexity of the source (higher complexity = worse predictability), the source volume, and other things.

Thanks everyone for clearing this out for me.

To whom it may interest: The times when this might be a problem may be limited afaik, since Extigy (when properly configured) handles the direct audiostream from lets say a DVD video with its internal DTS decoder. In that case it will not need to convert the audio channels to ADPCM to fit in the USB 1.1 interface.

Peace,

Emanuel