Thanks, Paul. I thought so.
Until and unless Mr. Boulet supplies a link to this alleged admission by Roger Dressler, I would not take it as gospel. And I would want to know how much (if any) was directly from Dressler and how much was extrapolated from what he said (a phrase such as "just like running your audio data through a sampling rate converter or noise-shaping filter would re-write data" leaves room for all sorts of conceptual slippage).
Adam Barratt and Cees Alons addressed this argument in another thread where it was made about dialnorm and Dolby TrueHD:
Quote:
|
Originally Posted by Adam Barratt
This isn't rocket science we're talking about. The mathematics are black and white, cut and dried, and extremely basic wherever they occur: attentuation via subtraction. These mathematics aren't a problem for any DSP that can decode Dolby Digital (or any other compressed format). Transparency in this sense implies mathematical precision, and this sort of task can be done with total precision (given that DSPs are mathematical calculating devices I'm sure this is of no surprise to any observers).
|
http://www.hometheaterforum.com/htf/...ml#post3279880
Quote:
|
Originally Posted by Cees Alons
Changing the loudness of a digital audio signal doesn't involve determining a new "place" in the sampled signal, or the combining of two adjacent values: just the sampled value at hand is adjusted. All that's involved are individual signal values (more comparable to colour or brightness, video wise). If computations on these are done in the 24-bit domain (like in TrueHD decoders), it could affect the lowest bit(s) indeed. But affecting the lowest bits is hardly significant at all!
In fact, more than the 3 lowest bits of any 24-bit value will take no part in the waveform that will eventually reach the ears of a home HiFi-enthusiast (given a properly scaled signal as well as the noise and harmonic distortion levels of even the best non-professional HiFi-equipment on the market.)
Example: halving the value of the signal involves shifting the whole 24-bit word 1 bit down. The lowest bit will disappear, but that bit-value wasn't to be trusted in the first place! The resultant 23 bits (completed with an upper 0) will still represent all audible data.
|
http://www.hometheaterforum.com/htf/...ml#post3280031
M.