>Yes, you can just truncate. Yes, you can also apply a random >noise function to a truncated datafile, to smear the truncation >artefacts. No, you cannot apply optimized dithers and noise >shaping without going from the full wordlength to the reduced >wordlength in one step, or at least without referencing the >original data, because just looking at the truncated data you >have no way of determining the error that is caused by truncating >and that needs to be corrected with the noise shaping, which >is a way of distributing an error terms over time such that >they cancel out. Oh, okay, I didn't realize that dithering algorithms looked at the lost data & took the error amounts into account. I thought that they just twiddled the lowest bit using some algorithm. My mistake, sorry. >The three options sound rather different, amazingly big difference >in fact. >I fear that eMagic made this change to make things more foolproof, >similar to the fact that in Logic 4.x the dither/noise-shaper >plug-in *had* to be the last plug-in in the chain. Ah, that's true, that might be the reason. Although still, that wouldn't explain why the dithering would have to happen after mixdown finished. Emagic certainly could insert the process as the last step in the signal path while bouncing. This same result (foolproofness) could be accomplished by adding a little flip menu to the master fader objects for selecting the dithering & noise shaping algorithms, for realtime monitering... Marc -- [ Destroy FX - http://www.smartelectronix.com/~destroyfx ]