darrenyeats wrote: > mswlogo;543633 Wrote: >> The DSP processing is done in 48bit. > I can understand how 48 bits helps the DSP to be more transparent (less > rounding errors).
Its not just rounding errors. Nearly all DSP is done by converting the time-doman signal to the frequency domain using a FFT. DSP chips are notable because they do a combined multiply-and-add function quickly, usually massively parallel processing. Nearly all of the time, they are actually working on floating point numbers, not the 16 or 24 bit integers that most audiophiles are used to talking about. When you do as many multiply and add functions as a typical DSP does, and we are talking about tens of thousands, you need to be very careful with the numerical analysis. Its much more complicated than simple rounding of a few bits. This whole thread has been off in the weeds for weeks. A "digital volume control" is talking about a nearly trivial single multiply or shift function. DSP is more like dealing with quantum physics rather than Newtonian physics. -- Pat Farrell http://www.pfarrell.com/ _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/audiophiles