in [DSP]

Prev: How Rune Allnor can attend the COMP.DSP Conference
Next: Inertial navigation (anyone familiar with this stuff here?)
From: Rune Allnor on 23 Mar 2010 19:51 On 23 Mar, 21:06, Eric Jacobsen <eric.jacob... (a)ieee.org> wrote:> Actually, bottom posting is the preferred method, since a single entry > can be read logically in order. I'm top-posting here just because > mixing top and bottom is worse than top posting. > > It seems to me that you're not grasping what people are trying to tell > you. Jerry mentioned a relevant article, but I'll post a link for you: > > http://www.dsprelated.com/showarticle/54.php No, this is far simpler than that. Andor's example was a IIR function with poles. Ther is no feedback in the dipole. This thread is about wave physics 101 stuff that anyone messing with array processing or wavefield analysis needs to know. And is expected to know. > You're not the first to be lured down this path and you won't be the last.. It's a matter of education. Or lack of such. Below is a crude *simulation* I made for matlab. Call it as FasterThanLightMovie(60,16); % Oblique angle at 60 degrees, % 16 frames in animation and see the simulation I hinted at a couple of days ago: The wave 2D field propagates in the positive direction along the x axis. There are two observations made of the field, one along the propagation axis (the blue graph / line) and one at an oblique angle (the red graph / line). In the upper plot the snapshot along the two lines are plotted. Do note the apparent speed of the zero crossing as it propagates donw the observation throughout a cycle (marked as a circle in the top plot and a cross in the lower plot). It is seen that the apparent speed along the oblique observation is far higher than the true, free field speed at which the wave travels down the x axis. If our friend WW splits up his simulation in monopole sources, he will be able to see exactly the same kind of effect but in a cylindrical or spherical coordinate system. So following WW's logic, all we need to do to obtain faster-than-light communication, is to observe the wave field along an axis oblique to the actual axis of propagation, thus requiring the information to travel a longer distance. Yeah. Right. Again, this is trivial material. It's only a matter of a bare minimum of knowledge about wave physics, simulation design, and data analysis that is needed to fully pull this stuff apart and see what is actually going on. Rune %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function FasterThanLightMovie(phi,Nframes) Phi = phi/180*pi; xv = [-1:0.01:2]; yv = [-1;3]; xv = reshape(xv,1,length(xv)); Lref = [0:0.01:1]; Loblique = Lref*cos(Phi); c = 3e8; f = 3e8; T = 1/c; tv = (0:Nframes-1)*T/Nframes; for n=1:Nframes clf s = ones(2,1)*sin(2*pi*f*(tv(n) - xv / c)); subplot(2,1,1) plot(Lref,sin(2*pi*f*(tv(n)-Lref/c)),'b') hold on plot(Lref,sin(2*pi*f*(tv(n)-Loblique/c)),'r') plot(c*tv(n),0,'ob') plot(c*tv(n)/cos(Phi),0,'or') ax = axis; plot(ax(1:2),[0,0],'k') subplot(2,1,2) imagesc(xv,yv,s) set(gca,'dataaspectratio',[1,1,1]) set(gca,'ydir','normal') hold on plot([0,1],[0,0],'b','linewidth',2) plot([0,1],[0,tan(Phi)],'r','linewidth',2) plot(c*tv(n),0,'xb','markersize',5,'linewidth',4) plot(c*tv(n),c*tv(n)*tan(Phi),'xr','markersize',5,'linewidth', 4) colormap(gray) drawnow pause(0.3) end end
From: Eric Jacobsen on 23 Mar 2010 20:20 On 3/23/2010 4:51 PM, Rune Allnor wrote: > On 23 Mar, 21:06, Eric Jacobsen<eric.jacob... (a)ieee.org> wrote:>> Actually, bottom posting is the preferred method, since a single entry >> can be read logically in order. I'm top-posting here just because >> mixing top and bottom is worse than top posting. >> >> It seems to me that you're not grasping what people are trying to tell >> you. Jerry mentioned a relevant article, but I'll post a link for you: >> >> http://www.dsprelated.com/showarticle/54.php > > No, this is far simpler than that. Andor's example was a > IIR function with poles. Ther is no feedback in the dipole. > > This thread is about wave physics 101 stuff that anyone > messing with array processing or wavefield analysis needs > to know. And is expected to know. > >> You're not the first to be lured down this path and you won't be the last. > > It's a matter of education. Or lack of such. > > Below is a crude *simulation* I made for matlab. Call it as > > FasterThanLightMovie(60,16); % Oblique angle at 60 degrees, > % 16 frames in animation > > and see the simulation I hinted at a couple of days ago: > The wave 2D field propagates in the positive direction > along the x axis. There are two observations made of > the field, one along the propagation axis (the blue > graph / line) and one at an oblique angle (the red graph / > line). > > In the upper plot the snapshot along the two lines > are plotted. Do note the apparent speed of the zero > crossing as it propagates donw the observation throughout > a cycle (marked as a circle in the top plot and a cross > in the lower plot). It is seen that the apparent speed > along the oblique observation is far higher than the > true, free field speed at which the wave travels down > the x axis. > > If our friend WW splits up his simulation in monopole > sources, he will be able to see exactly the same kind of > effect but in a cylindrical or spherical coordinate system. > > So following WW's logic, all we need to do to obtain > faster-than-light communication, is to observe the > wave field along an axis oblique to the actual axis > of propagation, thus requiring the information to travel > a longer distance. Yeah. Right. > > Again, this is trivial material. It's only a matter of a > bare minimum of knowledge about wave physics, simulation > design, and data analysis that is needed to fully pull > this stuff apart and see what is actually going on. > > Rune > > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% > function FasterThanLightMovie(phi,Nframes) > Phi = phi/180*pi; > xv = [-1:0.01:2]; > yv = [-1;3]; > xv = reshape(xv,1,length(xv)); > Lref = [0:0.01:1]; > Loblique = Lref*cos(Phi); > c = 3e8; > f = 3e8; > T = 1/c; > tv = (0:Nframes-1)*T/Nframes; > for n=1:Nframes > clf > s = ones(2,1)*sin(2*pi*f*(tv(n) - xv / c)); > subplot(2,1,1) > plot(Lref,sin(2*pi*f*(tv(n)-Lref/c)),'b') > hold on > plot(Lref,sin(2*pi*f*(tv(n)-Loblique/c)),'r') > plot(c*tv(n),0,'ob') > plot(c*tv(n)/cos(Phi),0,'or') > ax = axis; > plot(ax(1:2),[0,0],'k') > > subplot(2,1,2) > imagesc(xv,yv,s) > set(gca,'dataaspectratio',[1,1,1]) > set(gca,'ydir','normal') > hold on > plot([0,1],[0,0],'b','linewidth',2) > plot([0,1],[0,tan(Phi)],'r','linewidth',2) > plot(c*tv(n),0,'xb','markersize',5,'linewidth',4) > plot(c*tv(n),c*tv(n)*tan(Phi),'xr','markersize',5,'linewidth', > 4) > colormap(gray) > drawnow > pause(0.3) > end > end You sim doesn't run very well under my version of Octave, but the bit about the phase velocity on oblique angles is fundamental. I haven't been able to sort out what WW is doing well enough to know for certain that's the issue, but in his paper the waveforms he compares don't prove anything. I was at least trying to get him to see, as Jerry has been, that if a transient is introduced it'll expose his error. Andor's paper does that marvelously. I was hoping the idea would stick. -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
From: Vladimir Vassilevsky on 23 Mar 2010 21:34 WWalker wrote: > Eric, > > Interesting article, but I don't see how it applies to my system. Prior art: http://www.google.com/patents?id=csYDAAAAEBAJ&printsec=abstract&zoom=4#v=onepage&q=&f=false Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
From: Eric Jacobsen on 23 Mar 2010 21:34 On 3/23/2010 6:06 PM, WWalker wrote: > Eric, > > Interesting article, but I don't see how it applies to my system. The > system described in the paper is a bandpass filter in a feedback loop, > where the bandpass filter phase function is altered by the feedback. The > feedback forces the endpoints of the phase to zero, creating regions of > possitive slope, which yield negative group delays for narrow band signals. > This causes narrow band signals at the output of the circuit appear to > arrive earlier than signals at the input of the circuit. Because the > information in the signals is slightly redundant, the circuit is able to > reconstruct future parts of the signal from the present part of the > signal. Snipped context to allow bottom-posting. Feedback is not necessary to produce negative group delay. Here's another example with a passive notch filter that exhibits negative group delay. http://www.radiolab.com.au/DesignFile/DN004.pdf It doesn't matter what's inside a black box if it has a negative group delay characteristic if the transfer function is LTI. Whether there's feedback or not in the implementation is inconsequential. Consider that the passive notch filter could also be implemented as an active circuit with feedback, and if the transfer functions are equivalent they are functionally equivalent. This is fundamental. I don't think the feedback has anything to do with it. You're argument on the redundancy, though, is spot-on. Note that, as others have already pointed out multiple times, the signals you're using in your experiment are HIGHLY redundant, so much so that they carry almost no information. These signals are therefore not suitable for proving anything about information propagation. > First of all, this is a circuit which alters the phase function with > respect to time and not space, as it is in my system. The phase function in > the circuit is not due to wave propagaton, where mine is. As far as I've been able to tell, your evidence is based on a simulation, in which case dimensionalities are abstractions. You are not performing anything in either time or space, you're performing a numerical simulation. Space-time transforms are not at all unusual and it is likely that a substitution is easily performed. Nothing has propagated in your simulation in either time or space. > Secondly,unlike the circuit, my system is causal. The recieved signal in my > system arrives after the signal is transmitted. It just travels faster than > light. Uh, the circuit is causal. That was the point. You have not demonstrated that your system is causal or not causal. That cannot be concluded using the waveforms you show in your paper due to the high determinism and narrow band characteristics. > Thirdly, the negative group delay in the circuit was accomplished by using > feedback which does not exist in my system. As I stated above, this is inconsequential. > Information (modulations) are clearly transmitted using narrowband AM radio > communication, just listen to an AM radio. The simulation I presented > simply shows that random AM modulations arrive undistorted across space, in > the nearfield, earlier than a light speed propagated signal. Your simulation does not demonstrate that. Turn the signal off, even at a zero crossing if you want to minimize perturbations, and see what happens. > Signal purturbations can not be used to measure the signal propagation in > the nearfield because they distort in the nearfield, and group speed has no > meaning if the signal distorts as it propagates. > > William If you cannot use a perturbation (i.e., information transmission) to measure signal propagation then you cannot demonstrate the speed of information propagation. Until you can actually demonstrate something other than phase velocity (which is NOT information transmission and many here have acknowledged can be faster than c, as do I), then you cannot make the conclusions that you are claiming. -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
From: WWalker on 23 Mar 2010 22:02
Vladimir, Interesting patent but the idea presented is very different from the one I am proposing. Many of the ideas being discussed in this thread are published in my Ph.D. thesis submitted in 1997 at ETH Zurich, Switzerland. Since I have published most of what I have presneted, I doubt a patent would be possible. William > > >WWalker wrote: >> Eric, >> >> Interesting article, but I don't see how it applies to my system. > >Prior art: > >http://www.google.com/patents?id=csYDAAAAEBAJ&printsec=abstract&zoom=4#v=onepage&q=&f=false > > > > >Vladimir Vassilevsky >DSP and Mixed Signal Design Consultant >http://www.abvolt.com > |