Talk:Filter (signal processing)

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Wiki Education Foundation-supported course assignment[edit]

This article was the subject of a Wiki Education Foundation-supported course assignment, between 22 January 2020 and 14 May 2020. Further details are available on the course page. Peer reviewers: Wintersfire.

Above undated message substituted from Template:Dashboard.wikiedu.org assignment by PrimeBOT (talk) 21:24, 16 January 2022 (UTC)[reply]

Changes to the lede[edit]

I have partially reverted this edit. Describing a filter as "elaborating" a signal is so vague that it could be applied to about any signalling processing block. Filtering means removing something. The original and most common thing to remove is a band of frequencies. In that context, changing "act in the frequency domain" to " act on signals related to time" is quite baffling. SpinningSpark 18:22, 12 March 2013 (UTC)[reply]

I disagree (But in the mean time I had done some other editing; for this reason and for "etiquette" reasons, I will not revert to my previous edition for the time being.), *but*:
Simply it is not always true that "filtering means removing something": non linear filter (quoted in the same article) are able (and used) to introduce previously non existing frequencies (see, for example, filters to distort guitar sounds); the main purpose of the widely used matched filter (sonar, demodulators etc) is (apart the auxiliary function of providing noise suppression) is to recognize a signal pattern, etc. That said, I am open to accept suggestions for replacing "elaborate", but the filter concept can not be so restricted as it was.
The original author of the article used the locution "act in the frequency domain" to mean that the concept of "signal" is not limited to something that evolves in time, i.e. - as clarified by him in the following - we may speak of "signal" to refer to, for example, a scan line of a picture, where different shadows of grey develop along a spatial dimension. In the signal processing jargon, however, the expression "frequency domain" is customarily used in contrast to "time domain" to indicate different mathematical representations and tools, regardless of the physical dimensions - speaking, for example, of "spatial frequency" (cycles per meter).

In that context saying "act in frequency domain" is misleading; then the correction to "act on signal related to time", that more correctly reflects the true meaning.

I would greatly appreciate a second thought by Spinningspark and commnents from him and (may be) others. Thanks. — Preceding unsigned comment added by Ginsi (talkcontribs) 19:31, 12 March 2013 (UTC)[reply]
I have removed your new heading and indented your reply according to our usual WP:TALK conventions. A typical use of a non-linear filter is to remove noise. I don't know about guitar amp filters - that sounds like an abuse of terminology to me. I cannot see that "signal related to time" adds any more clarity and I challenge you to produce reliable sources that classify filters to such a scheme. SpinningSpark 22:24, 12 March 2013 (UTC)[reply]

What is a filter?[edit]

There are many uses of the word filter but in signal processing I believe that it means a functional block that alters the relative amplitude and phase of components of the frequency spectrum of a signal. That means that most image processing filters are not signal processing filters. Attempting to broaden this definition results in almost anything being called a filter. A clamper circuit is not a filter, even though a black level clamp was used to remove mains hum from analog television signals. A clipper circuit is not a filter, even though it can be used to remove noise from FM signals. A mixer is not a filter even though it alters the frequency of a signal.

To support my case for a frequency filter definition, here are some definitions from books on the subject of signal processing filters:

  • A filter is a system that can be used to modify or reshape the frequency spectrum of a signal according to some prescribed requirements.[1]
  • Filtering is the most common form of signal processing used [...] to remove the frequencies in certain parts and to improve the magnitude, phase, or group delay in some other part(s) of the spectrum of a signal.[2]
  • An electric analog filter is typically designed to pass certain things and attenuate if not completely block other things [...] it is typically designed to pass certain wavelengths, or frequencies, and atttenuate or block others.[3]
  • The digital filters perform the frequency related operations such as lowpass, highpass, band reject (notch), bandpass and allpass etc.[4]
  1. ^ Miroslav D. Lutovac, Dejan V. Tošić, Brian Lawrence Evans, Filter Design for Signal Processing Using MATLAB and Mathematica, page 241, Miroslav Lutovac, 2001 ISBN 0201361302.
  2. ^ B. A. Shenoi, Introduction to Digital Signal Processing and Filter Design, page 19, John Wiley & Sons, 2005 ISBN 0471656380.
  3. ^ L. D. Paarmann, Design and Analysis of Analog Filters: A Signal Processing Perspective, page 1, Springer, 2001 ISBN 079237373.
  4. ^ J.S.Chitode, Digital Signal Processing, page 2.1, Technical Publications, 2009 ISBN 8184316461.

Many other books on filters fail to give a definition, but the scope of the coverage plainly indicates that they are working to the frequency filter definition. I intend to remove from the article all material that does not conform to this definition, partly because a lot of stuff has been added to the lede that is not discussed in the article, but mostly so that the article actually follows the sources. Comments? SpinningSpark 17:33, 18 January 2014 (UTC)[reply]