fertig said:
I'm working in ProTools and my session is 24bit 48Khz, obviously I have to bounce it to 16bit 44100 to burn it to a CD.
This is kind of a sidenote, but it's best to use multiples of your destination sample rate. For example, CD is 44.1kHz, DVD is 48kHz. If you want to use a higher sample rate for production stages, use 88.2kHz or 96kHz, respectively.
I say this because downsampling is a delicate thing. When trying to downconvert 48kHz or 96kHz to 44.1, the algorithm has to estimate and throw away data irregularly. You get differing results based on the quality of your convertor because it's not an even division. 48/44.1 = 1.088435374... The quality of this kind of conversion depends on how many digits of this fraction the algorithm pays attention to - the more the better, but no algorithm will ever perfectly divide this, resulting in quantization error (as you'll read on that Wikipedia article posted by our friend above).
Whereas, with a perfect multiple (88.2/44.1 = 2.000000...), your downconvertor will (evenly) throw away every other sample, or average 2 adjacent samples - this is easier to process (since it calls for a far simpler algorithm), and is *much* cleaner in the end than the guesswork involved in 48->44.1, for example.
fertig said:
If I use a plug in like the L1 Ultimax or the L2, they appear to have a dither built in. Should I click it to 16bits or should I just let Protools convert to 16 bit when it bounces?
The setting on the L1 or L2 refers to the bitlevel of dither you want applied. Protools will not select this for you; it knows (based on your bounce-to-disk settings) what bitdepth to convert to, but Protools itself does not handle dither. It only takes a 24bit output (which you want the L2 to have dithered already!) and converts to 16bit.
So let me rephrase for clarity. The *output* of whatever plugin adds your dither will actually be a 24bit stream, but will have the noise floor raised by dither to a 16bit level. Then Protools handles the actual conversion from 24bit to 16bit, your audio already having been prepped for conversion. Make sense?