[Don Pearce[_3_]]

On Sun, 10 Feb 2019 13:04:31 -0800, Robert Orban
wrote:

In article ,
says...


The problem with clipping in the digital domain is that the resulting
square wave has harmonics all the way up to the sampling rate. That
can't happen with analogue clipping, because the lowpass input filter
will guarantee that they are suppressed well before the Nyquist rate.

Those harmonics surrounding the sampling rate alias back into the
audio band following decimation. They are nasty. Never ever allow
digital domain clipping.

d

You can clip in the digital domain, but it's difficult. It requires
oversampling the clipper, and aliasing decreases quite slowly as sample rate
increases.

Current Optimod processors use a combination of oversampling and other
tricks to get an equivalent sample rate of about 10 MHz for "clipper-like"
processes. At that frequency, aliasing distortion into the audio band is
negligible compared to the desired clipping distortion (i.e. distortion that
is equivalent to that produced by an analog clipper with the same
input/output gain shape).

--Bob Orban

OK, I think I've got that. I'll work it out later in Mathcad. But what
we are talking about here is accidental clipping by simply hitting
digital max. That can never be good news.

d