from the sci.physics.research newsgroup. i tried cross-posting my reply to comp.dsp, but i think it must get approved by the moderator before it goes to either newsgroup. i am *very* dubious of the claim here and wonder what the rest of you guys think? r b-j On Apr 5, 6:48 pm, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote:> [ Mod. note: A reminder to please keep the discussion on topic, as > discussing engineering details of analog computer technology would not > be. -ik ] > > On Apr 5, 7:47 am, robert bristow-johnson <r...@audioimagination.com> > wrote: > > > > > On Mar 31, 9:53 am, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: > > > > On Mar 30, 6:16 pm, Rock Brentwood <federation2...@netzero.com> wrote: > > > > > What are the ways of doing Fourier Analysis with time-dependent > > > > spectra (e.g. for quasi-particle representations in quantum theory or > > > > acoustic analysis/engineering, or music mixing -- or better yet: > > > > unmixing)? > > > > > The obvious ideas that come to mind are: Coherent State Transform > > > > (which I see is apparently also included under the header "smoothed > > > > Wigner Transform"), Wigner Transform, wavelet analysis; > > > the most obvious is the Short-Time Fourier Transform (STFT). > > > > > although I > > > > don't see either (a) how the third is related to the other two > > > > directly, nor (b) how the first two can be efficiently done in > > > > numerical calculations (the best I've seen in my survey of the field > > > > is O(n^2 log n) in the size, n, of the data to be transformed). > > > > I've worked with dedicated analog computers that did FFT's,http://en.wikipedia.org/wiki/Fast_Fourier_transform > > > (Fast Fourier Transforms) > > > *analog* computers computing the FFT??? can you describe this a > > little, Ken? how many points? how are you doing the adding and > > scaling? with resistors and summing junctions? op-amps? > > Yes that's right, a dedicated analog computer is very quick, > up to the quality of the components and design, here's some links, > > http://en.wikipedia.org/wiki/Ultrasound#Diagnostic_sonography > > I'll provide a quote from this site. > > http://www.drgdiaz.com/tables.shtml > > "FFT. Fast Fourier Transform analyzer is a common device that performs > spectral analysis in ultrasound instruments. In this case, it displays > different quadrature Doppler frequencies, or reflector velocities when > a sample volume cursor is used (Doppler frequency is proportional to > reflector velocity) along time" > > The reason for 'dedicted analog computer' was to modify the information > from the ultrasonic sonar pulse in 'real time' so the person doing the > scan didn't need to hesitate, like scanning pregnant womens babies. > Regards > Ken S. Tucker
hey, have you guys ever heard of this?
Started by ●April 6, 2011
Reply by ●April 6, 20112011-04-06
On Apr 6, 8:07�am, robert bristow-johnson <r...@audioimagination.com> wrote:> from the sci.physics.research newsgroup. > > i tried cross-posting my reply to comp.dsp, but i think it must get > approved by the moderator before it goes to either newsgroup. > > i am *very* dubious of the claim here and wonder what the rest of you > guys think?I suppose you mean this claim:> > > > I've worked with dedicated analog computers that did FFT's,http://en.wikipedia.org/wiki/Fast_Fourier_transform > > > > (Fast Fourier Transforms)This is a matter of definitions: 1) Does the term 'FFT' mean what we think it does in DSP? 2) What does the term 'analog computer' mean? I can imagine how somebody who is not familar with DSP *and* have worked as a techie with some analog devices, and have a vague notion of what task is silved and have heard mentioned some terms about how it is done, but without gaining detailed insights, might confuse the lingo to the point that he talks 'the truth' (in the sense that he attempts to describe something he has once done) and yet come up with a statement like the above. But if one stick with the strict intrepretations as I suppose any of the comp.dsp regulars would understand these terms, the statement doesn't make much sense. Rune
Reply by ●April 6, 20112011-04-06
I suspect that it's similar to an Optical Fourier Transform, in which a lens computes the Fourier Transform of light. The same would hold true for any wave phenomenon that was somehow focused. Off-hand I'm not certain exactly how the analog computer fits into this, but the concept is not outlandish. (One of my undergrad professors said, "You talk about Fast Fourier Tranforms? A lens computes the Fourier Transform at the speed of light!") Greg
Reply by ●April 6, 20112011-04-06
On Tue, 05 Apr 2011 23:07:21 -0700, robert bristow-johnson wrote:> from the sci.physics.research newsgroup. > > i tried cross-posting my reply to comp.dsp, but i think it must get > approved by the moderator before it goes to either newsgroup. > > i am *very* dubious of the claim here and wonder what the rest of you > guys think?I suppose that it would be *possible* to build an FFT-shaped analog circuit with op-amps and what-not, but I suspect that what is being discussed regarding analog "FFT" systems is probably analog Fourier Transforms, or perhaps even analog spectrum analysis. An old-school spectrum analyser (swept sinusoidal modulation and matching notch filter) may very well be fast enough for real-time ultrasound or some radar imaging applications. I've known radar guys to use optics to do their real-time Fourier transforms too. I imagine that you could also do spectrum analyisis with surface-wave filters. Not done any of those myself, though. Cheers,> > r b-j > > On Apr 5, 6:48 pm, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: >> [ Mod. note: A reminder to please keep the discussion on topic, as >> discussing engineering details of analog computer technology would >> not be. -ik ] >> >> On Apr 5, 7:47 am, robert bristow-johnson <r...@audioimagination.com> >> wrote: >> >> >> >> > On Mar 31, 9:53 am, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: >> >> > > On Mar 30, 6:16 pm, Rock Brentwood <federation2...@netzero.com> >> > > wrote: >> >> > > > What are the ways of doing Fourier Analysis with time-dependent >> > > > spectra (e.g. for quasi-particle representations in quantum >> > > > theory or acoustic analysis/engineering, or music mixing -- or >> > > > better yet: unmixing)? >> >> > > > The obvious ideas that come to mind are: Coherent State Transform >> > > > (which I see is apparently also included under the header >> > > > "smoothed Wigner Transform"), Wigner Transform, wavelet analysis; >> >> > the most obvious is the Short-Time Fourier Transform (STFT). >> >> > > > although I >> > > > don't see either (a) how the third is related to the other two >> > > > directly, nor (b) how the first two can be efficiently done in >> > > > numerical calculations (the best I've seen in my survey of the >> > > > field is O(n^2 log n) in the size, n, of the data to be >> > > > transformed). >> >> > > I've worked with dedicated analog computers that did >> > > FFT's,http://en.wikipedia.org/wiki/Fast_Fourier_transform (Fast >> > > Fourier Transforms) >> >> > *analog* computers computing the FFT??? can you describe this a >> > little, Ken? how many points? how are you doing the adding and >> > scaling? with resistors and summing junctions? op-amps? >> >> Yes that's right, a dedicated analog computer is very quick, up to the >> quality of the components and design, here's some links, >> >> http://en.wikipedia.org/wiki/Ultrasound#Diagnostic_sonography >> >> I'll provide a quote from this site. >> >> http://www.drgdiaz.com/tables.shtml >> >> "FFT. Fast Fourier Transform analyzer is a common device that performs >> spectral analysis in ultrasound instruments. In this case, it displays >> different quadrature Doppler frequencies, or reflector velocities when >> a sample volume cursor is used (Doppler frequency is proportional to >> reflector velocity) along time" >> >> The reason for 'dedicted analog computer' was to modify the information >> from the ultrasonic sonar pulse in 'real time' so the person doing the >> scan didn't need to hesitate, like scanning pregnant womens babies. >> Regards >> Ken S. Tucker-- Andrew
Reply by ●April 6, 20112011-04-06
http://www-03.ibm.com/ibm/history/exhibits/attic3/attic3_157.html was built a long time ago. There might be improvements available today. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●April 6, 20112011-04-06
On Apr 6, 2:07�am, robert bristow-johnson <r...@audioimagination.com> wrote:> from the sci.physics.research newsgroup. > > i tried cross-posting my reply to comp.dsp, but i think it must get > approved by the moderator before it goes to either newsgroup. > > i am *very* dubious of the claim here and wonder what the rest of you > guys think? > > r b-j > > On Apr 5, 6:48 pm, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: > > > > > [ Mod. note: A reminder to please keep the discussion on topic, as > > � discussing engineering details of analog computer technology would not > > � be. �-ik ] > > > On Apr 5, 7:47 am, robert bristow-johnson <r...@audioimagination.com> > > wrote: > > > > On Mar 31, 9:53 am, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: > > > > > On Mar 30, 6:16 pm, Rock Brentwood <federation2...@netzero.com> wrote: > > > > > > What are the ways of doing Fourier Analysis with time-dependent > > > > > spectra (e.g. for quasi-particle representations in quantum theory or > > > > > acoustic analysis/engineering, or music mixing -- or better yet: > > > > > unmixing)? > > > > > > The obvious ideas that come to mind are: Coherent State Transform > > > > > (which I see is apparently also included under the header "smoothed > > > > > Wigner Transform"), Wigner Transform, wavelet analysis; > > > > the most obvious is the Short-Time Fourier Transform (STFT). > > > > > > although I > > > > > don't see either (a) how the third is related to the other two > > > > > directly, nor (b) how the first two can be efficiently done in > > > > > numerical calculations (the best I've seen in my survey of the field > > > > > is O(n^2 log n) in the size, n, of the data to be transformed). > > > > > I've worked with dedicated analog computers that did FFT's,http://en.wikipedia.org/wiki/Fast_Fourier_transform > > > > (Fast Fourier Transforms) > > > > *analog* computers computing the FFT??? �can you describe this a > > > little, Ken? �how many points? �how are you doing the adding and > > > scaling? �with resistors and summing junctions? �op-amps? > > > Yes that's right, a dedicated analog computer is very quick, > > up to the quality of the components and design, here's some links, > > >http://en.wikipedia.org/wiki/Ultrasound#Diagnostic_sonography > > > I'll provide a quote from this site. > > >http://www.drgdiaz.com/tables.shtml > > > "FFT. Fast Fourier Transform analyzer is a common device that performs > > spectral analysis in ultrasound instruments. In this case, it displays > > different quadrature Doppler frequencies, or reflector velocities when > > a sample volume cursor is used (Doppler frequency is proportional to > > reflector velocity) along time" > > > The reason for 'dedicted analog computer' was to modify the information > > from the ultrasonic sonar pulse in 'real time' so the person doing the > > scan didn't need to hesitate, like scanning pregnant womens babies. > > Regards > > Ken S. Tucker- Hide quoted text - > > - Show quoted text -Most definitely optical processing has and is being done. Since the far field and near field distributions of E-M waves are related via fourier transforms, then a simple lens may be used to compress the physical distance between these two regions to a procatical working distance. Additionally a mask may be placed into the optical fourier plane to perform correlation. This has actually been done to perform optical fingerprint recognition. The disadvantage is it is so sensitive, that poor print impressions get rejected even though they are a match. And you may think of holograms as optical processors. Holographic optical elements (HOEs) were used by IBM in grocery store cash registers for bar code scanners. There are also many miltitary uses of optical target reconition where you can make a holographic optical element as a superposition of a multitude of targets each with its own reference beam at unique orientations, thus when the HOE is illuminated with the target (fourier image), a plane wave is launched in a unique direction corresponding to which target was seen. Clay
Reply by ●April 6, 20112011-04-06
On Wed, 6 Apr 2011 07:02:42 -0700 (PDT), Jerry Avins <jya@ieee.org> wrote:>http://www-03.ibm.com/ibm/history/exhibits/attic3/attic3_157.html was built a long time ago. There might be improvements available today. > >Jerry >-- >Engineering is the art of making what you want from things you can get.Yeah, there's not much new under the sun. Most decent Spectrum Analyzers are still mostly analog. An analog computer could do "FFT"s either by sweeping, with banks of sweeping sections (to increase speed), or with a separate LO for each output. Or optically, or mechanically. ;) Eric Jacobsen http://www.ericjacobsen.org http://www.dsprelated.com/blogs-1//Eric_Jacobsen.php
Reply by ●April 6, 20112011-04-06
On Wed, 6 Apr 2011 07:22:52 -0700 (PDT), Clay <clay@claysturner.com> wrote:>On Apr 6, 2:07=A0am, robert bristow-johnson <r...@audioimagination.com> >wrote: >> from the sci.physics.research newsgroup. >> >> i tried cross-posting my reply to comp.dsp, but i think it must get >> approved by the moderator before it goes to either newsgroup. >> >> i am *very* dubious of the claim here and wonder what the rest of you >> guys think? >> >> r b-j >> >> On Apr 5, 6:48 pm, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: >> >> >> >> > [ Mod. note: A reminder to please keep the discussion on topic, as >> > =A0 discussing engineering details of analog computer technology would = >not >> > =A0 be. =A0-ik ] >> >> > On Apr 5, 7:47 am, robert bristow-johnson <r...@audioimagination.com> >> > wrote: >> >> > > On Mar 31, 9:53 am, "Ken S. Tucker" <dynam...@vianet.on.ca> wrote: >> >> > > > On Mar 30, 6:16 pm, Rock Brentwood <federation2...@netzero.com> wro= >te: >> >> > > > > What are the ways of doing Fourier Analysis with time-dependent >> > > > > spectra (e.g. for quasi-particle representations in quantum theor= >y or >> > > > > acoustic analysis/engineering, or music mixing -- or better yet: >> > > > > unmixing)? >> >> > > > > The obvious ideas that come to mind are: Coherent State Transform >> > > > > (which I see is apparently also included under the header "smooth= >ed >> > > > > Wigner Transform"), Wigner Transform, wavelet analysis; >> >> > > the most obvious is the Short-Time Fourier Transform (STFT). >> >> > > > > although I >> > > > > don't see either (a) how the third is related to the other two >> > > > > directly, nor (b) how the first two can be efficiently done in >> > > > > numerical calculations (the best I've seen in my survey of the fi= >eld >> > > > > is O(n^2 log n) in the size, n, of the data to be transformed). >> >> > > > I've worked with dedicated analog computers that did FFT's,http://e= >n.wikipedia.org/wiki/Fast_Fourier_transform >> > > > (Fast Fourier Transforms) >> >> > > *analog* computers computing the FFT??? =A0can you describe this a >> > > little, Ken? =A0how many points? =A0how are you doing the adding and >> > > scaling? =A0with resistors and summing junctions? =A0op-amps? >> >> > Yes that's right, a dedicated analog computer is very quick, >> > up to the quality of the components and design, here's some links, >> >> >http://en.wikipedia.org/wiki/Ultrasound#Diagnostic_sonography >> >> > I'll provide a quote from this site. >> >> >http://www.drgdiaz.com/tables.shtml >> >> > "FFT. Fast Fourier Transform analyzer is a common device that performs >> > spectral analysis in ultrasound instruments. In this case, it displays >> > different quadrature Doppler frequencies, or reflector velocities when >> > a sample volume cursor is used (Doppler frequency is proportional to >> > reflector velocity) along time" >> >> > The reason for 'dedicted analog computer' was to modify the information >> > from the ultrasonic sonar pulse in 'real time' so the person doing the >> > scan didn't need to hesitate, like scanning pregnant womens babies. >> > Regards >> > Ken S. Tucker- Hide quoted text - >> >> - Show quoted text - > >Most definitely optical processing has and is being done. Since the >far field and near field distributions of E-M waves are related via >fourier transforms, then a simple lens may be used to compress the >physical distance between these two regions to a procatical working >distance. Additionally a mask may be placed into the optical fourier >plane to perform correlation. This has actually been done to perform >optical fingerprint recognition. The disadvantage is it is so >sensitive, that poor print impressions get rejected even though they >are a match. And you may think of holograms as optical processors. >Holographic optical elements (HOEs) were used by IBM in grocery store >cash registers for bar code scanners. There are also many miltitary >uses of optical target reconition where you can make a holographic >optical element as a superposition of a multitude of targets each with >its own reference beam at unique orientations, thus when the HOE is >illuminated with the target (fourier image), a plane wave is launched >in a unique direction corresponding to which target was seen. > >ClayI know of a number of military applications where optical processors were used to do "fast" (i.e., frequency domain) correlation/convolution from the 50s through the 70s. One of the big pains for reconnaissance that was solved by moving to digital processing was the hassle and time lag in processing the film for results. Digital systems didn't require a film lab and produced results more quickly. Eric Jacobsen http://www.ericjacobsen.org http://www.dsprelated.com/blogs-1//Eric_Jacobsen.php
Reply by ●April 6, 20112011-04-06
On 4/5/2011 11:07 PM, robert bristow-johnson wrote: snip....>> >>>> I've worked with dedicated analog computers that did FFT's,http://en.wikipedia.org/wiki/Fast_Fourier_transform >>>> (Fast Fourier Transforms) >> >>> *analog* computers computing the FFT??? can you describe this a >>> little, Ken? how many points? how are you doing the adding and >>> scaling? with resistors and summing junctions? op-amps? >>I think it's simply a misnomer where FFT is being used in place of "spectrum analyzer" or something similar. A time-doppler correlator, implemented with analog technology would be an example of something like an analog spectrum analyzer (with discrete bins). So, yes, this sort of thing has been done but there's no way I'd call it an "FFT". I thought that the Spectral Dynamics spectrum analyzers of the 70's were analog but the recent web page cites an "FFT analyzer" of 1968 - which is likely the device I'm referring to. The FFT was published in 1965. Microprocessors started being generally useful in the early 70's. So, I rather doubt there was an FFT box as early as 1968. These things were a box just a bit bigger than the CRT display that they used. Maybe Dale or someone else remembers.... Fred
Reply by ●April 6, 20112011-04-06
Clay wrote:> Most definitely optical processing has and is being done.Before the era of DSPs, the synthetic aperture radars used to do the processing by optical means; that is recording the radar data to photographic film and then exposing it like a hologram. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
