Talk:Digital signal processing
|This article is of interest to the following WikiProjects:|
External Links, Movable Feast
I've cleared the external links because they are unnecessary, and serve to attract linkspam. Please add content to the article and cite references. Jehochman (Talk/Contrib) 04:07, 11 December 2006 (UTC)
This recently created page is tied up in conflict. Extra eyes and opinions would be welcome; see the RFC on its talk page, and other talk sections; expert content input would also be of considerable value, but the topic is a bit narrow so I don't expect to find an expert here. Dicklyon 00:52, 23 October 2007 (UTC)
I like to point your attention to the following issue. According to this page autocorrelation has its own domain. Can somebody please explain why this is stated as such? Because the output values are a function of time! — Preceding unsigned comment added by 18.104.22.168 (talk) 20:02, 19 December 2011 (UTC)
- Why is the Autocorrelation domain flagged with "citation needed" when none of the other domains are? If autocorrelation, like the others, is its own link, then why should a citation be needed? rowley (talk) 20:44, 19 December 2011 (UTC)
I think it may be fumble fingers (I've assumed that and edited the article accordingly), or it could be a confusion in menaing between domain as in field of analysis, and domain as in units on x axis. Same question would be asked of wavelets, the units are seconds and Hz in the plots I look at. Greglocock (talk) 06:24, 20 December 2011 (UTC)
FIR vs IIR
The distinction between FIR and IIR is not recursion. for example, y[n] = y[n-1] + x[n] - x[n-k] is not an IIR filter as it could be rewritten as the sum of the last k samples. The actual distinction is based on the number of non-zero outputs for a non-zero input.
And also, the ideal lowpass "sinc" filter is an IIR filter, though it is commonly written using only inputs samples.
- In your example filter, the output can be nonzero forever with no input. It is not what is usually called an FIR filter, nor what is usually called stable. I reverted your edit to the article; I believe it is correct as stated. Dicklyon (talk) 04:12, 10 September 2012 (UTC)
- There is a more important difference between y[n] = y[n-1] + x[n] - x[n-k] and y[n] = x[n] + x[n-1] + ... + x[n-k+1].
- With floating-point arithmetic, the recursive implementation has arithmetic error accumulation (random-walk) as n increases. The FIR implementation does not.
- --Bob K (talk) 13:47, 18 September 2016 (UTC)
Further reading section is bloated
The further reading section should be trimmed down. Alternatively, some headings / sort comments should explain why the reference is significant. It should not be a bibliography of DSP texts. Glrx (talk) 16:26, 19 October 2012 (UTC) changed "sort" to "short" Glrx (talk) 19:15, 29 October 2012 (UTC)
- I agree. Further reading is not a list of all books covering the subject. Books in further reading should be notable for their quality, reliability, etc and should be about the subject of the article. Some of the books that are listed, like "Orthogonal Transforms for Digital Signal Processing" and that new addition that keeps getting re-added, are really too specific to be listed on this page, which is about DSP in general. Radiodef (talk) 17:44, 27 October 2012 (UTC)
- I've removed the following:
- Bimal Krishna, K. Y. Lin, Hari C. Krishna: Computational Number Theory & Digital Signal Processing, CRC Press, ISBN 0-8493-7177-5 Not easily found for purchase, descriptions suggest the book mainly covers new theories devised by the authors.
- Henrique S. Malvar: Signal Processing with Lapped Transforms, Artech House Publishers, ISBN 0-89006-467-9 Subject matter is pretty specific. Also fairly safe to assume book is dated (publication in 1992) with respect to the subject matter, descriptions suggest the book includes a fair amount of speculation, lapped transforms like the DCT are pretty ubiquitous now and I doubt the book is an accurate representation of the subject 20 years later.
- Artur Krukowski, Izzet Kale: DSP System Design: Complexity Reduced Iir Filter Implementation for Practical Applications, Kluwer Academic Publishers, ISBN 1-4020-7558-8 Subject matter is highly specific.
- I've left these on the list for now, but they could also be removed:
- Paul M. Embree, Damon Danieli: C++ Algorithms for Digital Signal Processing, Prentice Hall, ISBN 0-13-179144-3 Reviews of this book suggest it is geared more towards teaching the C++ and less towards explaining the DSP. Book is also primarily code examples in a single language, but the prevalence of C++ and breadth of the book's examples may still warrant it being on the list.
- N. Ahmed and K.R. Rao (1975). Orthogonal Transforms for Digital Signal Processing. Springer-Verlag (Berlin – Heidelberg – New York), ISBN 3-540-06556-3 Although the subject matter is somewhat restricted and the book's assumptions about computing may be outdated (publication in 1975), the author is apparently notable and respected.
- Those were obvious ones to me, but there are also a lot of general DSP books on the list that could be sorted through and pared down. Radiodef (talk) 19:04, 27 October 2012 (UTC)
- Yes, the section should focus on general texts. It may be OK to add some narrow/specific ones, but then there should be a short comment about why that text is useful. I recognize the Embree ref, but I don't recall its quality or breadth. Oppenheim is a classic (4.3 stars on Amazon). I'm not familiar with them, but Lyons (4.6 stars on Amazon) and Proakis are often cited; McClelland and Shaefer are names in the field. CRC Press often puts out huge encyclopedias about an area; the CRC for DSP might be good, but the radar/sonar/medical imaging might be too far afield. The Doug Smith / ARRL book has no reviews on Amazon.
- I encourage editors to add comments for good entries and delete weak entries.
- Glrx (talk) 19:15, 29 October 2012 (UTC)
- Is there a wiki guideline for further reading sections? It seems to me that they are open to abuse, even more than the usual See Also and External Links sections. If a book is so good it ought to be a ref, otherwise this list at best is just 'books I use', or 'books I wrote' or 'books I've seen on Amazon' at worst. I recommend deleting it entirely. Greglocock (talk) 21:10, 29 October 2012 (UTC)
- Wikipedia:Layout#Further reading (current guidelines) and Wikipedia:Further reading (proposed expanded guidelines). I don't think deleting it entirely is really the right thing to do, just pare it down a lot. It's useful to the reader to have a list of books they could continue studying with if they wanted to, but there's no point in having so many. Radiodef (talk) 02:24, 30 October 2012 (UTC)
- How do we decide? If we have Munchkin's classic 'DSP for dummies', and Zoperman's classic 'DSP for idiots', do we list both? neither? who gets to choose? there are so many books out there on DSP I can't see a wiki style way of choosing. I've read the links you gave and they don't really help much in this case.Greglocock (talk) 01:45, 31 October 2012 (UTC)
Deleting the whole section has the benefit of simplicity. We could then debate the entries that are reintroduced. I'm sympathetic to that viewpoint. Less radical would be agreeing what to keep on the talk page and then trimming the article accordingly. I would keep
- Proakis (the first one)
Intro reads like an app note, not a text book
- The goal of DSP is usually to measure, filter and/or compress continuous real-world analog signals. The first step is usually to convert the signal from an analog to a digital form, by sampling and then digitizing it using an analog-to-digital converter (ADC), which turns the analog signal into a stream of numbers. However, often, the required output signal is another analog output signal, which requires a digital-to-analog converter (DAC). Even if this process is more complex than analog processing and has a discrete value range, the application of computational power to digital signal processing allows for many advantages over analog processing in many applications, such as error detection and correction in transmission as well as data compression.
I think this paragraph shows evidence of making sense, but doesn't. This should stick to first princples (sampling and quantization), or discuss what could be described as "degrees of freedom over domains of operation" (the group of all applications).
Don't Talk - Write.
This article has languished for too long. Complaining about things you dislike will not make it better. Start writing. I will begin on the lead - which sucks btw. Codwiki (talk) 16:19, 7 July 2015 (UTC)