Fred Marshall wrote:> > "Jerry Avins" <jya@ieee.org> wrote in message > news:3F60FC4B.8C2314D5@ieee.org... > > > > > > First I've heard that taking the first difference was a way to get > frequency > > > information. > > > > I didn't mean to take differences. The frequency information is implicit > > in the changes of value from one sample to the next. DC happens when all > > samples are the same, etc. > > ***er.... "the changes of value from one sample to the next" *are* the first > differences. >Only if you so actual subtraction. All I'm saying is that if the samples don't vary, the spectrum is uninteresting, but that there's something to think about if they do. As to the rest, we agree -- that is, we understand one another and defend to the death .... No! Wait a minute! It hasn't come to that. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Bandpass FIR Filter
Started by ●September 10, 2003
Reply by ●September 12, 20032003-09-12
Reply by ●September 12, 20032003-09-12
Nitesh Gupta wrote:> > Thanx all for your co-operation, > > The reason behind asking this question was: > > The following block diagram, that I'm not able to understand. Can U > please help me in understanding this diagram? Why this "Freq to Value" > block is placed in between, if the output of 16 TAP BPF represents > amplitude component? > > ------------ ------------ ----- ---- > | | | | | | | | > ADC_OUT --->| 16 TAP BPF |-->|Freq to val |----> | LPF |->| -- |<-- > | | | | | | | | | > ------------ ------------ ------ ---- | > -------------------------------------------| | > | | > -----V------- ------------ ------- | > | | | | | | | > OUT <---| SINE |-->|Freq to val |----> | LPF |--------- > | GENERATOR | | | | | > ------------ ------------ -------- > > --NiteshYou raised this in another thread and got an answer that you've probably seen by now. Think of the "Freq to val" blocks as very complicated digitized frequency-to-voltage converters. It doesn't make sense to me either, because the inputs to them aren't frequencies, but numbers. You have to ask the author. I'll bet it's someone's ill-concieved pipe dream that never got built. Either that, or the top one is an FFT whose output is the bin number. The bottom one could be too, but why measure a frequency that's known because being generated? Let's be confused together. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●September 12, 20032003-09-12
Hello Nitesh If I was to guess, it looks like you are trying to build a specialized algorithm that will pass only the strongest single tone component of an input signal, be it a sine, square, triangle that is within the original pass band. I have just such a demo working on the C3x DSK's, but rather than using all of these blocks I have most of it crammed into one Sliding DFT. It is kind of cool to put anything in and always get a sine wave, but if you want to spice this up a bit, take the sine output and multiply its frequency, put in an AGC or something! Hope this helps, Best regards, Keith Larson ---------------------------------------- Nitesh Gupta wrote:> Thanx all for your co-operation, > > The reason behind asking this question was: > > The following block diagram, that I'm not able to understand. Can U > please help me in understanding this diagram? Why this "Freq to Value" > block is placed in between, if the output of 16 TAP BPF represents > amplitude component? > > ------------ ------------ ----- ---- > | | | | | | | | > ADC_OUT --->| 16 TAP BPF |-->|Freq to val |----> | LPF |->| -- |<-- > | | | | | | | | | > ------------ ------------ ------ ---- | > -------------------------------------------| | > | | > -----V------- ------------ ------- | > | | | | | | | > OUT <---| SINE |-->|Freq to val |----> | LPF |--------- > | GENERATOR | | | | | > ------------ ------------ -------- > --Nitesh >+------------------------------------------+ |Keith Larson | |Member Group Technical Staff | |Texas Instruments Incorporated | | | | 281-274-3288 | | k-larson2@ti.com | |------------------------------------------+ | TMS320C3x/C4x/VC33 Applications | | | | $150 TMS320VC33 DSK's ARE AVAILABLE NOW | | | | TMS320VC33 | | The lowest cost and lowest power | | floating point DSP on the planet! | | 500uw/Mflop | +------------------------------------------+
Reply by ●September 15, 20032003-09-15
Hey Fred, try thinking about it this way----it will illuminate a lot: Imagine a sampled sinusoid. Now, imagine an identical copy. Try delaying one copy of the signal by several samples, and adding it to the original. Notice how the amplitude is different? Try delaying it by a different amount. See how the amount is different? Now, try it with a different frequency sinusoid. Notice how the amount of delay of two signals causes the amplitudes of sinusoids of different frequencies to change by different amounts. You can also add additional identical copies of the signal delayed by different amounts, and amplitude scaled. FIR's are all about doing just that- adding delayed copies of a signal to itself at different amplitude levels. The delays and amplitudes are calculate with advanced methods to produce the best possible spectral attenuation and amplifcation pattern per amount of CPU time & memory. So, yeah, there is no explicit frequency information in a string of amplitude samples. However, FIR's don't need to "know" the frequency information; they simply attenuate or amplify based on delays and "tap" amplitudes through the principles of constructive and destructive interference... ...for her pleasure. -N In article <Leb8b.1304$v22.1093671@feed2.centurytel.net>, Fred Marshall wrote:> > "Jerry Avins" <jya@ieee.org> wrote in message > news:3F60FC4B.8C2314D5@ieee.org... >> > >> > First I've heard that taking the first difference was a way to get > frequency >> > information. >> >> I didn't mean to take differences. The frequency information is implicit >> in the changes of value from one sample to the next. DC happens when all >> samples are the same, etc. > > ***er.... "the changes of value from one sample to the next" *are* the first > differences. > > ....... > >> > Of course, if there were frequency information explicit in the sequence > then >> > your comment about the effect of the filter would be appropriate. It's >> > still appropriate if you keep the discussion in the time domain - > otherwise >> > there is implied analysis which, I suggest, makes the discussion more >> > complicated. >> >> We can't stay entirely in the time domain when discussing the effects of >> a band-pass filter. > > ***Well, you *could* but it would be difficult and not helpful. That's what > I said below. > >> > >> > The OP's question was "what do these numbers represent?" There's only > one >> > answer to this question. Anything beyond assumes analysis of those > numbers. >> > He didn't say that this was a lowpass filter with the lowest possible > cutoff >> > frequency so that the output effectively represents the dc component of > the >> > input - which would be a special case and frequency would be part of the >> > answer. >> >> He asked if they represented frequencies somehow. That indicates a >> confusion that we ought to try to clear away. > > ***Agreed. > >> > >> > So, one might say: the filter's output could have more rapid changes or > less >> > rapid changes depending on the type of filter: high-pass or low-pass >> > respectively. At that, there's some frequency jargon involved but even > that >> > could be put in the time domain if necessary. But, that would be > overkill >> > to make a point. >> > >> A little overkill when slaying misconceptions can be salutary. > > ***Yeah, except I meant that keeping the discussion purely in the time > domain would be overkill. That's not what your'e advocating and I've > agreed. > > ***I think we agree. My real point was that frequency information is > implicit in the time samples and not explicit in the time samples, that's > all. Some analysis is necessary to get more explicit re: frequency and this > can include using filters. > > Fred > >-- different MP3 every day! http://gweep.net/~shifty/snackmaster . . . . . . . . ... . . . . . . "Maybe if you ever picked up a goddamn keyboard | Niente and compiler, you'd know yourself." -Matthew 7:1 | shifty@gweep.net
Reply by ●September 16, 20032003-09-16
"Crackpot" <shifty@sidehack.sat.gweep.net> wrote in message news:slrnbmc188.1ncp.shifty@sidehack.sat.gweep.net...> > Hey Fred, try thinking about it this way----it will illuminate > a lot: > > Imagine a sampled sinusoid. Now, imagine an identical copy. > Try delaying one copy of the signal by several samples, and > adding it to the original. Notice how the amplitude is > different? Try delaying it by a different amount. See how > the amount is different? Now, try it with a different > frequency sinusoid. Notice how the amount of delay of > two signals causes the amplitudes of sinusoids of > different frequencies to change by different amounts. You > can also add additional identical copies of the signal > delayed by different amounts, and amplitude scaled. > > FIR's are all about doing just that- adding delayed > copies of a signal to itself at different amplitude levels. > The delays and amplitudes are calculate with advanced > methods to produce the best possible spectral attenuation and > amplifcation pattern per amount of CPU time & memory. > > So, yeah, there is no explicit frequency information > in a string of amplitude samples. However, FIR's don't > need to "know" the frequency information; they simply > attenuate or amplify based on delays and "tap" amplitudes > through the principles of constructive and destructive > interference...Well, yeah. No argument there. But that wasn't the question. The question was whether the output of a FIR represented time domain or frequency domain values - [with editorial emphasis added]. My response was that there are no explicit frequency domain values. No one has disagreed with this simple observation. That's all it was, a simple observation..... Fred
Reply by ●September 19, 20032003-09-19
In article <Vox9b.1439$v22.1342002@feed2.centurytel.net>, Fred Marshall wrote:> > "Crackpot" <shifty@sidehack.sat.gweep.net> wrote in message > news:slrnbmc188.1ncp.shifty@sidehack.sat.gweep.net... >> >> Hey Fred, try thinking about it this way----it will illuminate >> a lot: >> >> Imagine a sampled sinusoid. Now, imagine an identical copy. >> Try delaying one copy of the signal by several samples, and >> adding it to the original. Notice how the amplitude is >> different? Try delaying it by a different amount. See how >> the amount is different? Now, try it with a different >> frequency sinusoid. Notice how the amount of delay of >> two signals causes the amplitudes of sinusoids of >> different frequencies to change by different amounts. You >> can also add additional identical copies of the signal >> delayed by different amounts, and amplitude scaled. >> >> FIR's are all about doing just that- adding delayed >> copies of a signal to itself at different amplitude levels. >> The delays and amplitudes are calculate with advanced >> methods to produce the best possible spectral attenuation and >> amplifcation pattern per amount of CPU time & memory. >> >> So, yeah, there is no explicit frequency information >> in a string of amplitude samples. However, FIR's don't >> need to "know" the frequency information; they simply >> attenuate or amplify based on delays and "tap" amplitudes >> through the principles of constructive and destructive >> interference... > > Well, yeah. No argument there. But that wasn't the question. > The question was whether the output of a FIR represented time domain or > frequency domain values - [with editorial emphasis added]. > My response was that there are no explicit frequency domain values. No one > has disagreed with this simple observation. That's all it was, a simple > observation.....Ah, you want to get technical? The output of an FIR represents the same domain as the input. -- different MP3 every day! http://gweep.net/~shifty/snackmaster . . . . . . . . ... . . . . . . "Maybe if you ever picked up a goddamn keyboard | Niente and compiler, you'd know yourself." -Matthew 7:1 | shifty@gweep.net
Reply by ●September 20, 20032003-09-20
Crackpot <shifty@sidehack.sat.gweep.net> wrote in message news:<slrnbmc188.1ncp.shifty@sidehack.sat.gweep.net>...> Hey Fred, try thinking about it this way----it will illuminate > a lot:-snip-> attenuate or amplify based on delays and "tap" amplitudes > through the principles of constructive and destructive > interference... > > ...for her pleasure. > > -N > >You're just ribbing us... Robert
Reply by ●September 20, 20032003-09-20
"Crackpot" <shifty@sidehack.sat.gweep.net> wrote in message news:slrnbmm7vn.2qos.shifty@sidehack.sat.gweep.net...> In article <Vox9b.1439$v22.1342002@feed2.centurytel.net>, Fred Marshallwrote:> > > > "Crackpot" <shifty@sidehack.sat.gweep.net> wrote in message > > news:slrnbmc188.1ncp.shifty@sidehack.sat.gweep.net... > >> > >> Hey Fred, try thinking about it this way----it will illuminate > >> a lot: > >> > >> Imagine a sampled sinusoid. Now, imagine an identical copy. > >> Try delaying one copy of the signal by several samples, and > >> adding it to the original. Notice how the amplitude is > >> different? Try delaying it by a different amount. See how > >> the amount is different? Now, try it with a different > >> frequency sinusoid. Notice how the amount of delay of > >> two signals causes the amplitudes of sinusoids of > >> different frequencies to change by different amounts. You > >> can also add additional identical copies of the signal > >> delayed by different amounts, and amplitude scaled. > >> > >> FIR's are all about doing just that- adding delayed > >> copies of a signal to itself at different amplitude levels. > >> The delays and amplitudes are calculate with advanced > >> methods to produce the best possible spectral attenuation and > >> amplifcation pattern per amount of CPU time & memory. > >> > >> So, yeah, there is no explicit frequency information > >> in a string of amplitude samples. However, FIR's don't > >> need to "know" the frequency information; they simply > >> attenuate or amplify based on delays and "tap" amplitudes > >> through the principles of constructive and destructive > >> interference... > > > > Well, yeah. No argument there. But that wasn't the question. > > The question was whether the output of a FIR represented time domain or > > frequency domain values - [with editorial emphasis added]. > > My response was that there are no explicit frequency domain values. Noone> > has disagreed with this simple observation. That's all it was, a simple > > observation..... > > Ah, you want to get technical? > > The output of an FIR represents the same domain as the input.Crackpot, hehehe ... OK, let's get technical / literal: The output of a FIR can only be in the time domain because the name includes the term "impulse response" and we all know what that means. I guess the frequency domain equivalence would have to be called a "bandlimited lifter" or "finite frequency response" FFR lifter??? as you know: filter<->lifter and "<->" means "corresponds to" when dealing with transform pairs so this is an extension of that. :-) But, yes, a "FIR" in the frequency domain would be understood by most I suppose. Lots of correspondence here supports the assumption / context that the initial sequence is in the time domain. One has to be careful if that assumption doesn't hold. If the context is clear then it should be OK. Now I'm waiting for Jerry to come along and ask about ADCs that output frequency samples..... Fred
Reply by ●September 20, 20032003-09-20
"Robert Gush" <robert@suesound.co.za> wrote in message news:a0f35ea5.0309200322.69bf84b@posting.google.com...> Crackpot <shifty@sidehack.sat.gweep.net> wrote in messagenews:<slrnbmc188.1ncp.shifty@sidehack.sat.gweep.net>...> > Hey Fred, try thinking about it this way----it will illuminate > > a lot: > -snip- > > attenuate or amplify based on delays and "tap" amplitudes > > through the principles of constructive and destructive > > interference... > > > > ...for her pleasure. > > > > -N > > > > > > You're just ribbing us... > > RobertRobert, You give credit. Probably a wise thing to do. I've been much too serious.... Fred
Reply by ●September 21, 20032003-09-21
Fred Marshall wrote: ...>=20 > Now I'm waiting for Jerry to come along and ask about ADCs that output > frequency samples..... >=20 > Fred >=20Gimme a break, Fred! Do we want to talk about the instantaneous=20 frequency of a major triad? Or are we discussing samples Principle? And=20 what about Naomi? Y'r 'umble obd'nt s'v'nt, Jerry P.S, Is that "servant" or "savant"? --=20 Engineering is the art of making what you want from things you can get. =AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF= =AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF= =AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF