However, on the course web site Prof. Zoltowski has made the 3rd edition of the Proakis & Manolakis textbook available as a downloadable PDF file. This DSP book is NOT the best book for a DSP beginner because it's difficult to read. But once you build up your DSP knowledge this book can be quite useful.
That 3rd edition PDF file can be accessed at the following web site:
many thanks for the link. But unfortunately this kind of books is not really suitable for people like me, where the study of mathematics was 40 years ago. It is like the "Landau Lifschitz" in my profession: not easy to read and rather hard to understand - but very elegant in the view on the topics, seen from a very high viewpoint.
My very old problem: while calculating a sin(x)/x FIR Filter kernel (I do know, you do not like the word "kernel") is a easy task, I would like to ask you, how to calculate a similar filter kernel, which is not mirror-symmetrical (f(x) = f(-x)), but rather point symmetrical (f(x) = -f(-x)). Do you have a nice procedere for this?
Yes, the "DSP Principles Algorithms and Applications" is NOT for beginners, that's for sure. Many professors use that book for their 1st-semester DSP classes. Trying to learn DSP, from scratch, using that book is like trying to drink Kentucky bourbon from a fire hose. And that's why I say the most inefficient, most painful, way to learn DSP is by way of a college DSP class.
However, if you first go through my "Understanding DSP" book or Steven Smith's "Scientist and engineer's Guide to DSP" book (also online) to build up your knowledge of DSP, then the "DSP Principles Algorithms and Applications" book will be of considerable value to you.
As for your "not mirror-symmetrical" filter impulse response, I appears to me you're trying to design a "digital differentiating" filter or a Hilbert transformer (both of which have asymmetrical, also called "anti-symmetrical", coefficients). If I'm correct here are the equations (from my DSP book) to generate the coefficients for those two tapped-delay line networks:
where M = (N–1)/2, and 0≤k≤N–1. For odd N we set h_gen((N–1)/2), the center
coefficient, to zero. To start experimenting, set w_c = 3*pi/4.
where n = ...,–3,–2,–1,0,1,2,3,... . And h(n=0) = 0.
I really DO like the book from Steven, I printed it out a long time ago and it was very helpful for me.
You are right, it is all for the problem of receiving SSB (singe sideband signals) on shortwave without making the necessary filtering and phase turning clumsy. First bandpass-filtering by a FIR and then phase turning by a Hilbert allpass is the clumsy way, it eats too much processor cycles.
The much better way is to use a better filter kernel and do both with one act. But those better filter coefficients need to be calculated by me - this is my problem. 4 month's ago I retired, so I thougth, now I would have a lot of time to spend to such problems - but thinking so I failed.