I am unsure if it's OK to ask this here.

My question is: are the standard DSP methods still important in signal processing areas?(FIR, IIRF, etc) or have new techniques come out which have replaced these LTI(or pseudo LTI) operations? From a practical perspective are they still worth learning?

  • \$\begingroup\$ Before we tell you whether they are worth learning, how about if you tell us what you are going to study instead? My DSP knowledge, which was never deep to begin with, is also a decade or more out of date. But I am not aware of anything that universally replaces IIR and FIR and Fourier techniques in a wholesale fashion. \$\endgroup\$ – mkeith Jan 10 '18 at 6:00
  • \$\begingroup\$ I'm not sure, which is why I am asking. For example sparse signal processing is a newer thing. \$\endgroup\$ – FourierFlux Jan 10 '18 at 22:47

Linear Time Invariant (LTI) operations, filtering being one of them, whether IIR (Infinite Impulse Response) or FIR (Finite Impulse Response), are fundamental to DSP.

A few people who create DSP systems today do so without 'getting their hands dirty' with that 'low level stuff' by dragging and dropping higher level components on 'system builders'.

Much as I, a dyed-in-the-wool low level DSP man might bemoan that trend, it's not much different to the fact that very few people today write assembler, we all use higher level languages. Myself, I don't write assembler now, only use C when I have to, but for preference stay in Python.

It's just so much more productive to design at a higher level. But this move into higher levels of abstraction in software is only possible because of the speed of computation targets makes any code inefficiency due to generation from a high level irrelevant for many purposes.

While power constrains DSP applications, and many of them are hand-held and battery driven, most practitioners of the art will be using low level tools, designing FIR filters by the stage and the bit, to optimise power. When (if) the power constraint lessens, then more design will move to a higher level. But it will still be 'compiled' in some way to use LTI primitives at the core.

  • 1
    \$\begingroup\$ The move to higher levels is a good thing provided the underlying principles are still understood; based on much of the code I see, that is not necessarily always the case. \$\endgroup\$ – Peter Smith Jan 10 '18 at 8:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.