On Sun, 2 May 2021, David Kastrup wrote:
From a signal processing perspective, it could make sense to do the mixing and IIR-based filtering after the immediate A/D conversion stage _without_ digital antialiasing filtering and downsampling and skip the respective inverse stages for D/A conversion. That could cut down on conversion delay and phase artifacts. Whether this is actually done, I am skeptical: high-quality A/D converters have the filtering and downsampling stuff integrated and I doubt that they have a fast unfiltered parallel high-rate side-channel available.
Processing power would be a factor as well. Both speed and heat. I think the general idea is to make a digital mixer that is cheaper than a similar featured analog mixer. All the digital mixers I have looked at support many more channels than there are knobs to control and only has a full set of knobs for a channel at a time (well sorta full, the sends and eq frequencies are often banked).
There seem to be two ways of doing things. The X32, Xair are 32bit float processing while the A&H mixers seem to be 48 or 64 bit int processing with multiple DSP blocks depending on channel count (24bit ADC/DAC). The reason (I think) the int based mixers use higher bit depth is that two 24 bit channels mixed would be 25 bits... which could be scaled back each time... but adding bits may be cheaper, at least in processing power. I think 24bit int processing would also give audible artifacts. I have read complaints about the sound of int based mixers, while others seem happy with them.
-- Len Ovens www.ovenwerks.net _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx https://lists.linuxaudio.org/listinfo/linux-audio-user
