Top 3 DSP/SP books
in [DSP]
|
Prev: Reg:Design of Butterworth approximated filter using FIR filter design technique
Next: Independent Component Analysis
From: Al Clark on 12 Sep 2009 11:39 Vladimir Vassilevsky <nospam(a)nowhere.com> wrote in news:etadnXrCMaNn6TfXnZ2dnUVZ_oxi4p2d(a)giganews.com: > > > Dirk Bell wrote: > >> On Sep 10, 8:08 pm, "steveu" <ste...(a)coppice.org> wrote: >> >>>The kinds of things people do in military systems tends to be a lot >>>different from most non-military work. I guess that in the 70s, 90% of >>>all copies of O&S and Rabiner and Gold sold to people in military work, >>>who had not studied anything about DSP at college - because very few >>>colleges had begun DSP courses - and who needed to get a clue what they >>>were doing. Its a while since I looked at either book, but I remember >>>once reflecting on how the contents seemed to reflect that heritage. >>> >> >> Having read both books in the early 80's, I didn't get that >> impression. > > I don't have Rabiner and Gold at hand right now, but IIRC it is biased > towards the radar signal processing. When they give an example of a > technique, they often mention some kind of military application. And > yes, R&G used to be a handbook of people who actually worked on radars. > > > Vladimir Vassilevsky > DSP and Mixed Signal Design Consultant > http://www.abvolt.com With the exception of all the various O&S variations that have been mentioned, I think I have every book mentioned. I remember R&G as one of my first good books. I think the "best" books are going to depend heavily on where you are starting. In my case, I started as an analog engineer. I was veery good with s-plane stuff. There were no DSP chips when I read my first DSP book. I could do the math but I really had no idea what a digital filter would really do. The first big challenge for me when I started using DSP chips (first gen DSP - NEC 7725) was that I didn't know how to translate the math to twos complement fixed point based assembly code. I remember O&S as all math and no bridge to real problems. This was very typical of most of the early books. Jerry Purcell of Momentum Data Systems sent my a few pictures and notes on a fax that really help me get started. I still use QEDesign (not the original Fortran based DOS version) that I bought from him. Everone has mentioned books that were mostly written as course books for college students. I think I learned as much from the later manufacturer books that would explain soem basics and then show the code that targeted their particular DSPs. Analog Devices' Digital Signal Processing Applications using the ADSP-2100 Family (Vol 1 & 2) was probably the best example. Today's books have more examples written in either C or Matlab. I think this helps tie the theory and practice together. I would also suggest doing a simple project with a real target. It doesn't need to be fancy. Even a simple FIR filter is OK to start. Just don't copy someone's code, write it yourself and sweat out the bugs. Most of us that participate in this group probably forget how much we didn't grok when we started. Our backgrounds are also very different. Some of us like Jerry Avins (and me) are really generalists that learned some DSP along the way. Others are algorithm guys; Robert Bristow-Johnson probably doesn't know which end of a soldering iron to hold, but his math and signal processing skills are enviable. I probably have learned as much from the guys that frequent this site as I have from all the books in my library and I have certainly learned from designing lots of products. So the bottom line is "IT DEPENDS". It won't be three books, its at least 20 and they won't be enough either. Al Clark Danville Signal Processing, Inc. -- One of the grey beards......
From: Rune Allnor on 13 Sep 2009 06:59 On 12 Sep, 17:39, Al Clark <acl...(a)danvillesignal.com> wrote: > I remember R&G as one of my first good books. I've been actively collecting DSP books for 15 years now, but I don't think I have even seen one copy of R&G. I saw a copy of one of the classics I haven't been able to find, in the bookshelf of one of the engineers with a former employer of mine. It could have been Rabiner & Gold, or it could have been something with Rabiner with co-author. Whatever it was I saw: The Rabiner and Gold book has not been avalable for at least a couple of decades, already. > I think the "best" books are going to depend heavily on where you are > starting. Agreed. > Today's books have more examples written in either C or Matlab. I think > this helps tie the theory and practice together. I would also suggest doing > a simple project with a real target. It doesn't need to be fancy. Even a > simple FIR filter is OK to start. Just don't copy someone's code, write it > yourself and sweat out the bugs. Agreed. If I were to design a course for DSP, I would include programming excercises in something else than matlab for just about everything: - The FFT - IIR Filter designs - Window FIR filter design - Parks-McClellan filter design These kinds of problems don't require much more than basic programming skills, and knowledge of the task at hand. Once you get to stuff like MUSIC and ESPRIT, you need a linear algebra library, so for these problems wone would need matlab. I don't know too much about adaptive filters - maybe they can be implemented without optimization libraries; maybe not. The key is to start doing hands-on work and develop e.g. filter design routines you will use later. The FFTs are good for pedagogic purposes - once you have implemented crude versions yourself, you appreciate the time and effort spent by those who implemented e.g. FFTW. > Most of us that participate in this group probably forget how much we > didn't grok when we started. Well - I didn't. Forget what I did *not* do. I had the (mis)fortune to do a lot of R&D work on various algorithms myself - find published algorithms, implement them, test them, compare them. All this was done in the context of a PhD scholarship, but I always maintained that my degree was irrelevant, because I did not do anything else than I expect any engineer to do: Research alternatives, implement, test, verify and evaluate them. My problem was, of course, that I came from a background where that way of work was second nature to people. No one in the universities or R&D institutions I have ever been in touch with, work that way. In retrospect, I should have been at the level I ended up after the PhD at least 4 years earlier, if the people whose job it was to prepare us students for our crafts had done their jobs. Or known anything about what people outside academia actually do and how they work. If somebody really wants to help, then one needs to know exactly how and why one self was screwed by those whose job it was to tutor, and try and do better. So if somebody urges that 'you need to look into this', the reason is that one knows why it is improtant: Either because of the benefit one self had from looking into the problem. Or because of the problems one encountered from not knowing enough. 'Teaching' is as much about avoid repeating mistakes as it is about replicating successes. Rune
From: steveu on 13 Sep 2009 07:19 > > >Dirk Bell wrote: > >> On Sep 10, 8:08 pm, "steveu" <ste...(a)coppice.org> wrote: >> >>>The kinds of things people do in military systems tends to be a lot >>>different from most non-military work. I guess that in the 70s, 90% of all >>>copies of O&S and Rabiner and Gold sold to people in military work, who had >>>not studied anything about DSP at college - because very few colleges had >>>begun DSP courses - and who needed to get a clue what they were doing. Its >>>a while since I looked at either book, but I remember once reflecting on >>>how the contents seemed to reflect that heritage. >>> >> >> Having read both books in the early 80's, I didn't get that >> impression. > >I don't have Rabiner and Gold at hand right now, but IIRC it is biased >towards the radar signal processing. When they give an example of a >technique, they often mention some kind of military application. And >yes, R&G used to be a handbook of people who actually worked on radars. Ah, those were the days. With Rabiner and Gold in one hand, and Skolnick in the other, we fought the evils of communism. :-) Steve
From: Vladimir Vassilevsky on 13 Sep 2009 10:27 steveu wrote: >> >>I don't have Rabiner and Gold at hand right now, but IIRC it is biased >>towards the radar signal processing. When they give an example of a >>technique, they often mention some kind of military application. And >>yes, R&G used to be a handbook of people who actually worked on radars. > > > Ah, those were the days. With Rabiner and Gold in one hand, and Skolnick > in the other, we fought the evils of communism. :-) Don't know what Skolnick is, however we defended the glorious teachings of Lenin using the Russian translations of R&G and the AOE of Horowitz and Hill. That was all because of those belligerent books. The originator of the thread could very well be a warrior of Allah... VLV
From: Richard Owlett on 13 Sep 2009 14:52
steveu wrote: >> >> Dirk Bell wrote: >> >>> On Sep 10, 8:08 pm, "steveu" <ste...(a)coppice.org> wrote: >>> >>>> The kinds of things people do in military systems tends to be a lot >>>> different from most non-military work. I guess that in the 70s, 90% of > all >>>> copies of O&S and Rabiner and Gold sold to people in military work, who > had >>>> not studied anything about DSP at college - because very few colleges > had >>>> begun DSP courses - and who needed to get a clue what they were doing. > Its >>>> a while since I looked at either book, but I remember once reflecting > on >>>> how the contents seemed to reflect that heritage. >>>> >>> Having read both books in the early 80's, I didn't get that >>> impression. >> I don't have Rabiner and Gold at hand right now, but IIRC it is biased >> towards the radar signal processing. When they give an example of a >> technique, they often mention some kind of military application. And >> yes, R&G used to be a handbook of people who actually worked on radars. > > Ah, those were the days. With Rabiner and Gold in one hand, and Skolnick > in the other, we fought the evils of communism. :-) > > Steve > IIRC NORAD radar system once reported massive incoming missile attack from "nowhere" (or more definitively 'no place') OOOPS, moon rise Later blamed on "computer error" related to anomalous return time of echoes. Might it have more with moving from a 'continuous' to 'discrete' domain? IE aliasing problem? |