Hello people, Should an N-tap FIR operating on continuous (infinite) analog input signal be called (executed) on every input sample or can be called much less like every (perhaps even separate) block composed of e.g. N (or some other number of) samples? please detail. The term N-tap will always be equivalent to the number of input, output or coefficient values? (There are FIR functions in which those are 3 different numbers) Is it correct to say that given coefficients remain the same regardless if used for FIR or IIR? Thanks, Rony J. _____________________________ Posted through www.DSPRelated.com
Analog signal filtering
Started by ●June 28, 2013
Reply by ●June 29, 20132013-06-29
On Fri, 28 Jun 2013 21:54:27 -0500, Ro_ny wrote:> Hello people, > > Should an N-tap FIR operating on continuous (infinite) analog input > signal be called (executed) on every input sample or can be called much > less like every (perhaps even separate) block composed of e.g. N (or > some other number of) samples? please detail. > > The term N-tap will always be equivalent to the number of input, output > or coefficient values? (There are FIR functions in which those are 3 > different numbers) > > Is it correct to say that given coefficients remain the same regardless > if used for FIR or IIR?I think you need to get one or more books and start reading. "Understanding Digital Signal Processing" is good, and Rick Lyons frequents this group. Generally your FIR filter would operate on every sample -- but after sampling the signal isn't continuous time. The 'N' in N-tap refers to the number of coefficients. Generally FIR filters are one input, one output -- but as you mention, there are exceptions. The meaning of the word "coefficients" is markedly different for an IIR filter than for a FIR filter, but by any interpretation of the meaning of the word in the context of an IIR filter, the coefficients are _not_ the same as a comparable FIR filter. And I use the term "comparable" loosely. Get that book. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●June 29, 20132013-06-29
I wrote continuous instead of discrete just to emphasize the infinite samples number. The theoretical point was if there’s a way not to operate an FIR (with 1I-1O) the sampling rate times. My practical case however is an FIR function with x inputs, y outputs (I assume it’s a block processing) and h coefficients so I need to know the optimal/appropriate way to use it. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 29, 20132013-06-29
"Ro_ny" <95309@dsprelated> writes:> Hello people, > > Should an N-tap FIR operating on continuous (infinite) analog input signal > be called (executed) on every input sample or can be called much less like > every (perhaps even separate) block composed of e.g. N (or some other > number of) samples? please detail.First of all, an N-tap FIR filter is not "called." It is a mathematical beast: y[n] = \sum_{j = 0}{N - 1} x[n - j] * h[j] You can _implement_ that equation in a computer program function in at least two different ways: 1) processing sample-by-sample, or 2) processing a block at a time.> The term N-tap will always be equivalent to the number of input, output or > coefficient values?Coefficient. "tap" and "coefficient" are synonymous in FIR filters.> (There are FIR functions in which those are 3 different > numbers)There are? Not that I've ever seen.> Is it correct to say that given coefficients remain the same regardless if > used for FIR or IIR?No. As Tim said, you've got some reading to do. Get a book - Rick's is good. I would suggest a more general Linear System Theory book like [signalsandsystems]. --Randy @BOOK{signalsandsystems, title = "{Signals and Systems}", author = "{Alan~V.~Oppenheim, Alan~S.~Willsky, with Ian~T.~Young}", publisher = "Prentice Hall", year = "1983"} -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 29, 20132013-06-29
“First of all, an N-tap FIR filter is not "called."” – When the said formula (which I know from previous Reading about FIR) is contained within a function then being a function it is most definitely called. Second, I already wrote above a clarification question. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 29, 20132013-06-29
>I wrote continuous instead of discrete just to emphasize the infinite >samples number. > >The theoretical point was if there’s a way not to operate an FIR (with >1I-1O) the sampling rate times. > >My practical case however is an FIR function with x inputs, y outputs (I >assume it’s a block processing) and h coefficients so I need to knowthe>optimal/appropriate way to use it. > > >_____________________________ >Posted through www.DSPRelated.com >Yes, you can totally do that. It would be equivalent to filtering then decimating. Look up polyphase filtering. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 29, 20132013-06-29
"Ro_ny" <95309@dsprelated> writes:> “First of all, an N-tap FIR filter is not "called." When the said > formula (which I know from previous Reading about FIR) is contained within > a function then being a function it is most definitely called.And so you miss the point. The filter is not the implementation.> Second, I already wrote above a clarification question.If by "above" you mean your second post, I read it. It is just as confusing as your first. I'm wondering what "I1-10" means, for one. Perhaps I have the wrong fonts or something but it makes no sense. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 29, 20132013-06-29
Randy Yates <yates@digitalsignallabs.com> writes:> "Ro_ny" <95309@dsprelated> writes: > >> “First of all, an N-tap FIR filter is not "called." When the said >> formula (which I know from previous Reading about FIR) is contained within >> a function then being a function it is most definitely called. > > And so you miss the point. The filter is not the implementation. > >> Second, I already wrote above a clarification question. > > If by "above" you mean your second post, I read it. It is just as > confusing as your first. I'm wondering what "I1-10" means, for one. > Perhaps I have the wrong fonts or something but it makes no sense.Ah, I see now. 1I-1O - one input / one output. But what a cryptic way to write. Oh yeah - you're probably writing this on a smartphone. Well find a desk with a computer and TYPE for criminy sake! Write things out; make it clear. You'll probably get more help if you show folks you're willing to expend some effort and not expect them to pander to you. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 29, 20132013-06-29
Randy, I see that the same abbreviation caused you much inconvenience, next time if anything is not clear to you simply ask. I am actually writing from a Desktop PC with Windows 7 English, the Admin of this forum may want to consider fixing the incorrect display here of those characters. “Yes, you can totally do that. It would be equivalent to filtering then decimating. Look up polyphase filtering.” - Given a “black box” of FIR with x inputs, y outputs and h coefficients, how should it be performed on infinite number of input samples? Do you mean a loop of <collecting x new samples, calling it and forget these samples>? (If yes, what will be the decimation ratio which you mentioned (y/x?), and what if I prefer without rate change?) Rony J. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 29, 20132013-06-29
On Sat, 29 Jun 2013 15:48:59 -0500, "Ro_ny" <95309@dsprelated> wrote:> >Randy, I see that the same abbreviation caused you much inconvenience, next >time if anything is not clear to you simply ask.We've already sorted out that your questions take quite a few "clarifications" before we can tell what you mean, so people are starting to get terse with you. If YOU want help, it's up to YOU to be clear about what you want and what you mean and what YOU are willing to do in the process, or people are just going to tire of dealing with you. I can see this happening already.>I am actually writing from a Desktop PC with Windows 7 English, the Admin >of >this forum may want to consider fixing the incorrect display here of those >characters.There is no admin on this forum. It's up to YOU to make sure that your message comes across.>“Yes, you can totally do that. It would be equivalent to filtering then >decimating. Look up polyphase filtering.” - > >Given a “black box” of FIR with x inputs, y outputs and h coefficients, >how should it be performed on infinite number of input samples? > >Do you mean a loop of <collecting x new samples, calling it and forget >these >samples>? (If yes, what will be the decimation ratio which you mentioned >(y/x?), and what if I prefer without rate change?)I think your questions are a bit too fundamental to expect people to help you starting from first principles. The fact that your posts are incredibly confusing doesn't help. For example, you're asking in Digital Signal Processing group about FIR filters in a thread you started with the title "Analog signal filtering". In other words, you need to understand enough about what you're doing to be able to ask questions that get you useful answers without overly taxing the people you're asking to help you. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
