Here's a question for the people who know their shit on this board... what yeilds higher quality sound, a lower sample rate with a higher sample size or vice versa? (ie. 22050hz 32bit vs. 44050hz 16bit)

I heard about a new audio format that is being developed for either cd or dvd... it relies on an extremely high sample rate, but has a very low bit rate... I think it might have been 8 bits or even lower if my memory serves me. Joel (I think) said that dynamic range is 6.02 * number of bits.... so wouldn't this audio format have a pathetic dynamic range? Is there any redeeming value to having a ridiculously high sample rate?

Definitely 44100 16bit ! There was already a discussion along these lines; with 22050Hz sampling the highest frequency you would be able to reproduce accurately (without aliasing) would be 11 KHz ... and that would sound horrible (to hear what it would sound like, just take any random wave file and run a lowpass filter on 11000 Hz in CoolEdit).

About the new "super audio" format developed for DVD, I think its specifications are 96KHz sampling at 24bit ... hence giving you a dynamic range of 144dB (which is totally useless, since the sound of the air current in a room is louder than the noise floor of a CD's 96dB dynamic range).

As far as I know, CD format (ie. 44 / 16 / stereo) is already good enough to surpass the capabilities of human hearing, so anything better than that is just hype (I'm not 100% sure about that though, if somebody can come up with some more info on this please post it).

___________________
djskywarp.com
:: Ways of the Wikkid :: Canada's premiere Artist, DJ, Promoter, and Club Portal ::

I'll have to answer the second to explain the answer to the first.

That technique is called Oversampling and uses some clever mathematics to reconstruct to original data, but the concept is very simple:

If you can transfer very quickly, you only need to transfer 1 piece of information at a time compared to someone who transfers slowly but does multiple pieces of information separately.

eg 1 sec, 8 transfers of 1 info == 8 info per second
1 sec, 1 transfer of 8 info == 8 info per second

With this oversampling technique, rather than transfer the ABSOLUTE sample value (eg 32767, digital max), they transfer the DIFFERENCE in value between this and the previous value eg 20000 to 23500 == +3500.

If you decide that your limit is going to be 20 kHz and you don't mind losing anything above, then the wave between samples MUST approximate a straight line at 40kHz sample rate. Now, as sample rate increases with fixed bit rate, the wave between samples gets closer and closer to a straight line, and the difference between samples becomes closer to zero (look up Fourier analysis, sync x curves and reconstruction filters for more detail on this paragraph).

Once you have increased the sample rate high enough, you can see that you only need to transmit small numbers (low bit rate) at high speed (high sample rate) to achieve the same quality.

Now you're probably thinking "What's the point?".

Oversampling is a solution to an expensive hardware problem. It is very hard to precisely quantise a voltage (high bit rate) in terms of the electronics involved (I don't want to go into detail because it's very complex and I'm not an electronics expert). It is also cheap to run electronic components at high speed (the average cpu is 800 MHz!).

So the answer to the original question is that it depends on the sampling technique. Using the traditional technique it becomes a subjective issue of whether you'd prefer to lose dynamic range or top end. Personally, I'd choose to lose dynamic range for dance music as it's ultra compressed anyway. With the oversampling technique, there is no theoretical difference.

Finally, the dynamic range of the ear is about 150 dB, but as skywarp said there's no point as you just end up masking the lower bits with your own body noise (like heart beats and breathing etc).

Joel out.......

Wow, thanks for the in-depth answers!