DSPRelated.com
Forums
Online · Live + on demand
February 9-11, 2027
Price goes up Jan 1: register now to lock in today's rate
Register Now →

Why Side Lobes appear in magnitude response?

Started by Rune Allnor ●September 29, 2004
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message news:<7Z-dnVq8nbYAtfrcRVn-jw@centurytel.net>...
> Sandeep, > > Was there something wrong with my serious answer? > > Fred
I can't see much wrong with it, except for that it requires time and contemplation to understand, and can not be condensed to a seven-word universally true sentence. Rune
"Rune Allnor" <allnor@tele.ntnu.no> wrote in message 
news:f56893ae.0410090255.4e7ee27a@posting.google.com...
> "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message > news:<7Z-dnVq8nbYAtfrcRVn-jw@centurytel.net>... >> Sandeep, >> >> Was there something wrong with my serious answer? >> >> Fred > > I can't see much wrong with it, except for that it requires time and > contemplation to understand, and can not be condensed to a seven-word > universally true sentence.
Rune, Ah! A challenge. OP wants "physical interpretation" ..... of why sidelobes appear. One might interpret this in two ways: 1) "Why is there spectral leakage (as in: between samples)" and 2) "Why is there frequency spreading (as in: extending the spectrum)" The interpretation is necessary because "sidelobes" is a term that refers to energy outside the main lobe of very narrow antenna patterns "pencil beams". Translating into DSP terms it's best equivalent to characteristics of a Finite Fourier Transform - which I think is the center of the OP's question. So, what is a physical interpretation of why sidelobes appear in seven words or less? Don't you think that the old small optical aperture physics experiments tell it all? i.e. Light on a slit results in bands. 7 words..... The bands result from constructive / destructive interference. 6 words..... What you're seeing of course is the Fourier Transform of the slit and the cross-section of its illumination. As the slit becomes wider, the bands become narrower - as the slit becomes narrower, the bands become wider. The light bands outside the main beam of light are the "sidelobes". The dark bands are the zeros. Exercise for the student: Starting with the optical experiment, compose an analogy that helps visualize the physical reason why there are sidelobes of the Finite Fourier Transform in signal processing - in seven words or less. Hint: it's at a single frequency..... and, then, at each frequency. Fred
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message news:<j5WdnU2eh_ITifXcRVn-uQ@centurytel.net>...
> "Rune Allnor" <allnor@tele.ntnu.no> wrote in message > news:f56893ae.0410090255.4e7ee27a@posting.google.com... > > "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message > > news:<7Z-dnVq8nbYAtfrcRVn-jw@centurytel.net>... > >> Sandeep, > >> > >> Was there something wrong with my serious answer? > >> > >> Fred > > > > I can't see much wrong with it, except for that it requires time and > > contemplation to understand, and can not be condensed to a seven-word > > universally true sentence. > > Rune, > > Ah! A challenge.
It wasn't intended that way, I only tried to make a parody of what students usually want. But since you interpreted it as a challenge...
> OP wants "physical interpretation" ..... of why sidelobes appear. > One might interpret this in two ways: > 1) "Why is there spectral leakage (as in: between samples)" > and > 2) "Why is there frequency spreading (as in: extending the spectrum)" > > The interpretation is necessary because "sidelobes" is a term that refers to > energy outside the main lobe of very narrow antenna patterns "pencil beams". > Translating into DSP terms it's best equivalent to characteristics of a > Finite Fourier Transform - which I think is the center of the OP's question.
That's how it seems to me, too.
> So, what is a physical interpretation of why sidelobes appear in seven words > or less? > > Don't you think that the old small optical aperture physics experiments tell > it all? i.e. > > Light on a slit results in bands. 7 words.....
Very nice. It's "physical", it's straight to the point, 7 words...
> The bands result from constructive / destructive interference. 6 words.....
Ah! You beat me at my own game! Excellent!
> What you're seeing of course is the Fourier Transform of the slit and the > cross-section of its illumination. As the slit becomes wider, the bands > become narrower - as the slit becomes narrower, the bands become wider. The > light bands outside the main beam of light are the "sidelobes". The dark > bands are the zeros.
I think this is what is known as "Fourier optics", or is it "merely" Huygens' principle?
> Exercise for the student: > > Starting with the optical experiment, compose an analogy that helps > visualize the physical reason why there are sidelobes of the Finite Fourier > Transform in signal processing - in seven words or less. > > Hint: it's at a single frequency..... and, then, at each frequency.
Pass. My own attempt to answer my own 'challenge' goes as: "DFTs project arbitrary vectors onto 'sinusoidal' manifolds." 7 words. Rune
> My own attempt to answer my own 'challenge' goes as: > > "DFTs project arbitrary vectors onto 'sinusoidal' manifolds." > > 7 words. > > Rune
I have two issues with it: 1) I don't understand it - which is not a valid measure for anyone else! But it sure sounds cool! 2) Where's the "physical"? My initial attempt used math too. I really rather liked looking at the transform pair of the time-differentiated gate > resulting in a frequency sinusoid. Then, that integrating that would have a sinusoidal component seems satisfying - to me at least. Going from a sinusoidal component to sidelobes seemed an easy observation. Regarding light and the slit: Yes, I guess Huygen's Principle. I imagine you don't really get a Fourier Transform that way, you get the magnitude of it if all you do is observe the intensity of the light in the bands. That's a more correct description. Y'know, this got me to thinking: I've always had the notion that math should reflect a physical understanding ... that math should follow the understanding and isn't worth as much all by itself and can actually be confusing or just plain wrong. After all, we use mathematical models, even sometimes calling them "laws", for physical things. So, when the model breaks down and we're in the middle of pondering mathematical manipulations - how do we know? That was the sense of it .... and, I still put a lot of weight on it. [I once took an existing 6-degree of freedom hydrodyamic model and examined each and every coefficient from a physical point of view: "if it's spinning this way then the rotational drag will be in this direction.....", etc. We probably found 3 errors in signs in the coefficients in the original model. We found one nonlinear coefficient that was modeled to vary as the square of the velocity when it had to be an odd function like a cube or at least sgn[v]*v^2 if that was a better fit.] Then here, I used the time-derivative discussion as if it had some sort of physical reality. I guess when we're comfortable with some math thingys then they transcend into a kind of "reality" for us. Well, you could start with 2+2=4 so I'm not sure I'm expressing the thought very well - I guess it has a lot to do with our individual ability to comfortably abstract. Obviously, different minds will have different spans of this kind of reality. My "manifolds" are still on engines.... :-) Fred
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message news:<c42dnSptLd1y6PTcRVn-iQ@centurytel.net>...
> > My own attempt to answer my own 'challenge' goes as: > > > > "DFTs project arbitrary vectors onto 'sinusoidal' manifolds." > > > > 7 words. > > > > Rune > > I have two issues with it: > > 1) I don't understand it - which is not a valid measure for anyone else! > But it sure sounds cool!
A "manifold" is some sort of continuous structure in a vector space, like a line or surface. It is more general than the 'usual' lines and plane surfaces, in that the manifold can be twisted, i.e. have a curvature. If you use the vector e(w) = 1/sqrt(N)*[1, exp(jw), exp(j2w), ... , exp(j(N-1)w)] as a parametric curve in N-dimensional space (w being the parameter), the points that satisfy {x : x=e(w), 0 <= w < 2pi } will describe a 1-manifold. As far as I know, manifolds are most often used in differential geometry.
> 2) Where's the "physical"?
Eh.... I focused more on the "universally true" part...
> My initial attempt used math too. I really rather liked looking at the > transform pair of the time-differentiated gate > resulting in a frequency > sinusoid. Then, that integrating that would have a sinusoidal component > seems satisfying - to me at least. Going from a sinusoidal component to > sidelobes seemed an easy observation.
Well, yes, I guess that's the basis of my undrestanding of it, too.
> Regarding light and the slit: Yes, I guess Huygen's Principle. I imagine > you don't really get a Fourier Transform that way, you get the magnitude of > it if all you do is observe the intensity of the light in the bands. That's > a more correct description.
I never knew much optics. The last couple of years I have learned some 'the hard way' because of my intest in telescopes for birding and stargazing. It's amazing how sensitive telescopes are to convection in the air due to the sun heating the ground.
> Y'know, this got me to thinking: > I've always had the notion that math should reflect a physical understanding > ... that math should follow the understanding and isn't worth as much all by > itself and can actually be confusing or just plain wrong.
Well, yes... and no. I have commented on this before in this thread. To me, at least the basic principles of physics should have a mathemathical formulation. I am not so sure about that all maths should have a physical interpretation. Unless the maths is used in the context of mathemathical physics, where I think only end results (as opposed to the computations themselves) should be attempted interpreted.
> After all, we use > mathematical models, even sometimes calling them "laws", for physical > things. So, when the model breaks down and we're in the middle of pondering > mathematical manipulations - how do we know? That was the sense of it .... > and, I still put a lot of weight on it. > [I once took an existing 6-degree of freedom hydrodyamic model and examined > each and every coefficient from a physical point of view: "if it's spinning > this way then the rotational drag will be in this direction.....", etc. We > probably found 3 errors in signs in the coefficients in the original model. > We found one nonlinear coefficient that was modeled to vary as the square of > the velocity when it had to be an odd function like a cube or at least > sgn[v]*v^2 if that was a better fit.] > > Then here, I used the time-derivative discussion as if it had some sort of > physical reality. I guess when we're comfortable with some math thingys > then they transcend into a kind of "reality" for us. Well, you could start > with 2+2=4 so I'm not sure I'm expressing the thought very well - I guess it > has a lot to do with our individual ability to comfortably abstract. > Obviously, different minds will have different spans of this kind of > reality. > My "manifolds" are still on engines.... :-) > > Fred
Rune
Fred Marshall wrote:

> Light on a slit results in bands. 7 words.....
> The bands result from constructive / destructive interference. 6 words.....
> What you're seeing of course is the Fourier Transform of the slit and the > cross-section of its illumination. As the slit becomes wider, the bands > become narrower - as the slit becomes narrower, the bands become wider. The > light bands outside the main beam of light are the "sidelobes". The dark > bands are the zeros.
My favorite undergrad lab experiment was from an optics lab using that described above. An object (which was in a 2x2 slide mount, and usually made up of transparent and opaque regions) is illuminated by a laser through some lenses to make a wide beam approximating a plane wave. A lens is put on the other side of the object, and a TV camera with its lens removed is placed at the focal point of the lens. The image, as seen by the TV camera, is the 2D Fourier transform of the object. The TV camera is then moved away, and a second lens set up farther away. This second lens will do an inverse Fourier transform and reconstruct an image of the object. Additional 2x2 slide mounts with clear and opaque regions are available to construct spatial filters. That is, filtering in the transform domain, where higher frequencies are farther from the center. A low pass filter is just a circular hole, and a high pass filter is an opaque disk. It is then not hard to see the connection between lens size and resolution for an optical system. I suppose today it would all be done with a computer to do the FFT and display the results, though not quite as much fun as doing it with a few lenses and a TV camera. -- glen
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message
news:c42dnSptLd1y6PTcRVn-iQ@centurytel.net...
> > Regarding light and the slit: Yes, I guess Huygen's Principle. I imagine > you don't really get a Fourier Transform that way, you get the magnitude
of
> it if all you do is observe the intensity of the light in the bands.
That's
> a more correct description. >
Hello Fred and others, Huygen's principle is involved, but it is not the whole story. Huygen's principle is a neat way of drawing a new wavefront given an existing one. However, his principle is implemented by drawing circles with radius of 1 wavelength for a huge number of points along the wavefront. The overlap of the circles allows you to see the new wavefront. The problem with this aproach is light can go backwards! The circles represent an equal probability in all directions. So Kirchoff introduced his obliquity factor. Basically for each point on the source wavefront instead of a circle, he placed a cardiod whose axis is aligned normally to the wavefront. This factor takes care of the going backwords problem. Next with the light going through the slit problem (Young's experiment), we make a couple of assumptions. One is the screen is sufficiently far enough away, so all of the light components travel approximately the same distance so the inverse (amplitude) law applies nearly equally to all. Secondly, the scattering angles are small enough that Kirchoff's obliquity factor is effectively unity. If these assumptions fail to be met, then we end up with Fresnel diffraction. In the limit of large distances and small scattering angles, then we end up with Fraunhofer diffraction which may be found with the Fourier transform. A neat way to avoid large distances is the use of a lens. The spatial distribution at a focal length away from a lens on one side is the fourier transform of the spatial distribution one focal length away on the other side of the lens. Glen mentioned a fun experiment based on this. Just use two lenses separated by 2 focal lengths. One may place apatures and such into the Fourier plane in between the lenses. Those familar with optices know that two lenses spaced 2f apart simply inverts an image. Remember the identity that the Fourier transform of the Fourier transform of f(x) is f(-x). Just thought I'd add a little detail to this famous and popular experiment. Clay
glen herrmannsfeldt wrote:

> Fred Marshall wrote: > >> Light on a slit results in bands. 7 words..... > > >> The bands result from constructive / destructive interference. 6 >> words..... > > >> What you're seeing of course is the Fourier Transform of the slit and >> the cross-section of its illumination. As the slit becomes wider, the >> bands become narrower - as the slit becomes narrower, the bands become >> wider. The light bands outside the main beam of light are the >> "sidelobes". The dark bands are the zeros. > > > My favorite undergrad lab experiment was from an optics lab > using that described above. An object (which was in a 2x2 slide > mount, and usually made up of transparent and opaque regions) > is illuminated by a laser through some lenses to make a wide > beam approximating a plane wave. A lens is put on the other > side of the object, and a TV camera with its lens removed is > placed at the focal point of the lens. The image, as seen by > the TV camera, is the 2D Fourier transform of the object. > > The TV camera is then moved away, and a second lens set up > farther away. This second lens will do an inverse Fourier > transform and reconstruct an image of the object. > > Additional 2x2 slide mounts with clear and opaque regions > are available to construct spatial filters. That is, filtering > in the transform domain, where higher frequencies are farther > from the center. A low pass filter is just a circular hole, > and a high pass filter is an opaque disk. It is then not > hard to see the connection between lens size and resolution > for an optical system. > > I suppose today it would all be done with a computer to > do the FFT and display the results, though not quite as > much fun as doing it with a few lenses and a TV camera. > > -- glen >
Know of any good web pages which describe the experiment and show the various images obtained? [Particularly any that associate results with Fourier transforms.] Memories of introductory college physics from >40 years ago are being tickled. I'm wondering if reading such would be my roesetta stone for grokking Fourier transforms. [ was that ever a mixed metaphor 2 b or - ]

On 7 Oct 2004 20:37:40 -0700, sandeep_mc81@yahoo.com (Sandeep
Chikkerur) wrote:

>Hi, > >THis thread is going haywire. >First 4-5 threads discussions were in the path. >Later it went to oigin of Maths and what not. >Pls do not use this thread to say how much you know abt other things. > >Since there are many DSP gurus, they can pitch in & answer to the qn. > >Thanks >Sandeep
Sandeep, I appreciate your sentiment but this is an unmoderated newsgroup. There is some general etiquette expected, but just like verbal conversations the threads here occassionally digress. Some are more conscientious than others about starting new threads and labelling things as off topic, but many (like myself) aren't quite so diligent. The digressions often lead to new topics and improved understanding on a wide variety of things, so personally I think they're very useful and often entertaining. I think they're part of what makes this newsgroup an interesting place. I don't know what newsreader you use, but mine (Forte) makes it quite easy to skip or ignore portions of threads. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Clay Turner wrote:

> Those familar with optices > know that two lenses spaced 2f apart simply inverts an image. > Remember the identity that the Fourier transform of the Fourier > transform of f(x) is f(-x). > Just thought I'd add a little detail to this famous and popular > experiment.
Aha. I like that. Thanks for this contribution. Bernhard