I have to reduce the bit-depth of a digital audio signal from 24 to 16 bit.
Taking only the 16 most significant bits (i.e. truncating) of each sample is equivalent to doing a proportional calculation (out = in * 0xFFFF / 0xFFFFFF)?
+2
A:
You'll get better sounding results by adding a carefully crafted noise signal to the original signal, just below the truncating threshold, before truncating (a.k.a. dithering).
tdammers
2010-10-26 10:43:13
I know that by reducing bit depth I'm introducing quantization noise (and therefore a reduced 24 bit signal would sound worst than a 16 bit master), but how can I "carefully craft" the noise for the dithering?
Lorenzo
2010-10-26 11:45:49
The goal of dithering is to distribute the rounding error, but you want to minimize the amount of audible artifacts you are introducing. White noise would be a good starting point, but depending on the application, balancing the noise's spectrum one way or the other might deliver better results.
tdammers
2010-10-26 19:11:23
And of course you could simply shift the bits down: `out = in >> 8` (for unsigned samples).
andrewmu
2010-10-26 11:27:07
Because if you have signed samples you might be able to do: `out = in >>> 16`, but you maybe need to extend up the sign bit if only 24 bits are actually set in your original data.
andrewmu
2010-10-26 22:43:00
@andrewmu: I'm working in C++ so no need to >>>. I have 24 bit signed samples stored in 32 bit signed integers, so the value "-1" is the same for all the bit-depths.
Lorenzo
2010-10-27 09:10:09
+1
A:
Dithering by adding noise will in general give you better results. The key to this is the shape of the noise. The popula pow-r dithering algorithms have a specific shape that is very popular in a lot of digital audio workstation applications (Cakewalk's SONAR, Logic, etc).
If you don't need the full on fidelity of pow-r, you can simply generate some noise at fairly low amplitude and mix it into your signal. You'll find this masks some of the quantization effects.
Nick Haddad
2010-10-26 13:18:05