|
Most of my DSP work was in the early 70's using discrete chips. It was not even called DSP then. Built both FIT and IIR filters. At the time chips and DACs and ADCs were slow. That was OK as we were working on detecting and classifying submarine signals at acoustic frequencies. In fact, the first couple of filters that I built used serial arithmetic using chips from Rockwell Collins. Things got a lot nicer when AMD came out with parallel multipliers. All my filters were built with cascaded second order sections of the Elliptic design. Also, we used the Weaver method to make them tunable. Regards - Mike
toggle quoted message
Show quoted text
Mike B. Feher, N4FS 89 Arnold Blvd. Howell NJ 07731 908-902-3831 -----Original Message-----
From: [email protected] <[email protected]> On Behalf Of Reginald Beardsley via groups.io Sent: Saturday, July 6, 2024 12:37 PM To: [email protected] Subject: Re: [qex] Some questions on digital filter design This is a general reply to Jeff, Mike and Dave. I initially looked at: Digital Design Handbook Fred J. Taylor Marcel Dekker 1983 Digital Signal Processing Oppenheim and Shafer Prentice-Hall 1975 Theory and Application of Digital Signal Processing Rabiner and Gold Prentice-Hall 1975 Digital Signal Processing with Kernel Method Rojo-Alvarez et al Wiley/IEEE 2018 Advanced Digital Signal Processing and Noise Reduction Vaseghi Wiley 4th ed 2008 Subsequent to the responses I checked: Spectral Analysis and Time Series Priestly Academic Press 1981 Continuous a Signals and System Analysis and Discrete McGillem and Cooper Holt, Reinhart and Winston 1974 None of them properly discuss the issue of causality. I did NOT look at any of the many monographs by Enders Robinson, Robinson & Treitel or John H. Karl as I know they have the matter stated very clearly. Robinson and Treitel were members of Norbert Weiner's Geophysical Analysis Group at MIT and founded the entire basis of DSP. I spent 4 years at UT Austin under another member of the GAG, Milo Backus. Robinson performed the first digital deconvolution problem in 1952 using pencil, paper and a desk calculator. Robinson has pride of place as the first person to apply DSP to data. Causality requires that the real and imaginary parts be a Hilbert transform pair. This is well stated in: The Fourier Integral and Some of Its Applications Ronald Bracewell McGraw-Hill 2nd ed 1978 Causality simply states that there is no output prior to the input. If a filter is not causal it produces output before the event. The EE community appears to consistently label a zero phase signal as linear phase with a phase delay. As Bracewell provides a proof and I spent 2 semesters studying Churchill's "Operational Mathematics" under Bill Guy in "Integral Transforms" at Austin in addition to the semester I spent with Bracewell in "Linear Systems", if the EE community wants to redefine the math they may. I shall stick to what the mathematicians wrote. Have Fun! Reg |
