Frank Raffaeli wrote:

Eric,

The really efficient dsp / digital algorithms come about with
time-varying processes. Another good bit of science is the use of
recursive filters to produce a finite impulse response ... That's one
nice way to squeeze a lot of functionality in a medium-sized IC / FPGA
/ ASIC ... or whatever.

Radio receiver (demodulation) and bandpass algorithm / code get a lot
smaller.

No doubt about it! Though my point was something different.

A recursive filter would probably count as LTI since the coefficients don't
change (at least in the kind I'm thinking of). An adaptive filter, though,
might not, though it'd be time-invariant between adaptations.

A DSP / math-oriented guy would look at a mixer or modulator and say "See,
it's linear!" because the stage is a-times-b. Look, no second- or third-order
terms! An EE guy would say "it generates new frequencies, it's not linear."
The EE guy means "function of one variable that only changes the scale or
intercept".

Just terminology differences, is all it is.