How to get envelope from AM signal without phase shift
in [DSP]
|
Prev: How Rune Allnor can attend the COMP.DSP Conference
Next: Inertial navigation (anyone familiar with this stuff here?)
From: WWalker on 1 Apr 2010 08:51 Eric, I appreciate this discussion and I do understand the points all of you have been making very clearly. I teach advanced analog and digital signal processing, mathematics, as well as RF technique, and EM theory, and I have been looking at this problem for 20 years. I simply do not agree with some of the conclusions in this discussion for very logical reasons. I think maybe the problem is that we are all having difficulty understanding what is the information in the simulations being discussed, where is it located, and how does it propagate. Information is not well understood and I think it needs to be discussed to see if it can be defined better, as it applies to these simulations. As I mentioned before, I agree that if the information is the edge of a sharp pulse, which is passed through a narrowband nearfield dipole AM transmission and detection system, then the time delay will be the time delay of the narrowband filter plus the freespace propagation time of the pulse edge, which propagates at speed c in both the nearfield and farfield. The pulse will distort in the nearfield but the edge will be clearly defined and can be used to trigger a bomb. With this type of setup, the overall time delay of the pulse edge will clearly be less than a light speed time delay. But, if the information signal is directly input, bandlimited, in a narrowband nearfield dipole AM transmission and detection system, then the time delay of the bandlimited signal will only be the freespace propagation delay, which is less than light speed as shown in my simulations. For example, if the signal is created by a voltage source that is manualy slowly adjusted, and if the signal is then mixed with a carrier and sent though a nearfield dipole system, then the detected envelope will arrive undistorted earlier than a light propagated signal, as I showed in my simulations. Each voltage point on the voltage vs time curve of the voltage source is information about what the voltage was at that time. If that pattern is reproduced exactly a distance away, then the time delay of each information voltage point is the propagation time of the information. William >On 3/31/2010 4:09 PM, WWalker wrote: >> r b-j, >> >> The dicussion here has been very helpful with signal processing and >> understanding the information in the signals I use in my simulations. The >> topic has been discussed theoretically over dacades without proving or >> disproving the superluminal behavior of the the dipole system (i.e. Speed >> of Gravity etc). I do not believe theoretical evidence will ever prove it >> one way or another. >> >> My hope is that an experiment might be able to prove or disprove the >> superluminal behavior of this system, that is why I am discussing >> simulations here to see if a good experimental setup and signal processing >> method can be developed. >> >> William > >It's going to be difficult to prove or disprove it either way. I think >it would have been done by now if it was straightforward, and the >discussions here seem to indicate how that comes to be. A physicist >who doesn't understand signal processing can be (and have been) tripped >up by misinterpreting the results, as you seem inclined to do. > >To use your button-push bomb trigger example again, a step or impulse >that has been bandlimited spreads out in time. The impulse (or step, >but the impulse is easier to follow), which can be defined very >specifically in time, gets spread out over time by the bandlimiting >filter. This can be seen easily in your plots, in Andor's paper, >anywhere a plot of a bandlimited "impulse" exists. In high-snr cases >the temptation is to use the peak of the spread pulse to indicate the >arrival of the impulse. At low SNR it may be necessary to integrate >over the entire pulse length time in order to reliably detect the >"arrival of the impulse". > >Clearly the rising edge of the spread pulse doesn't anticipate the >arrival of the peak, so from a causality point of view the beginning >traces of the initial arrival of the leading edge of the pulse may be >most indicative of the actual arrival time of the earliest portion of >information associated with the impulse. > >Unless, as I mentioned, the SNR is low, in which case one has to wait >longer to integrate the energy for reliable detection. > >So when does the actual, narrow, "impulse" arrive? It is ambiguous. >At high SNR it could be argued that detection of initial energy (which >is why Vladimir suggested you start with zero-input-zero-output, put in >energy, and see when energy comes out) defines the actual propagation, >since the system is necessarily causal. > >I'm hoping you're beginning to see why a small phase advance that is >much narrower than the pulse length is NOT reliably indicative of >accelerated propagation. It is just a phase advance, and the >mechanisms by which those can happen are real. It does not indicate >propagation faster than c, but people who see the phase advance are >sorely tempted to continue to point to it and claim either noncausality >or propagation faster than c. > >If you measure from the initial stimulus, i.e., the button push, or the >arrival of the wideband impulse into the bandlimiting filter, then you >have a hope of measuring the actual propagation through the system. >Filter delays can then be observed reliably. If you just compare the >relative phases of the signals after bandlimiting and the difference is >small compared to the length of the impulse response, then it is >extremely difficult to distinguish phase advance due to dispersive >effects or negative group delay from accelerated propagation. Phase >advance due to bandlimited prediction is the far more likely explanation >than propagation faster than c, and continuing to point to the phase >differences as evidence does nothing to resolve the issue. > >You could still simulate disconnection of the signal from the Tx antenna >input by interrupting the signal and see how that affects the output. >That's a wideband stimulus and it should be much easier to see how fast >that propagates through the system. > >-- >Eric Jacobsen >Minister of Algorithms >Abineau Communications >http://www.abineau.com >
From: Jerry Avins on 1 Apr 2010 10:34 On 4/1/2010 6:22 AM, WWalker wrote: > Jerry, > > In my last post (argument pasted again below) I presented an analysis which > showed in the nearfield dv>>c in the nearfield and dv<<c in the farfield. > Once the photon is created, it is propagating in one direction away from > the creation point with, lets assume, a possitive velocity. Lets say in the > nearfield dv=10c therefore, the velocity of the photon will range between: > 0-10c, with an average of 5c, which is much faster than light. > > "Lets calculate the uncertainty of the velocity of a photon that > propagates > one wavelength after it is created: According to the Heisenberg > uncertainty > principle, the relation between the uncertainty in Energy (dE) and the > uncertainty in time (dt) is: dE*dt>= h. The time for a photon to cross > one > wavelength distance is: dt = lambda/c. Since dE = h*df and df=dv/lambda > then dE*dt=h*dv/c, but dE*dt<= h therefore: dv>= c > For smaller distances the uncertianty will be greater and for larger > distances the uncertainty will be much smaller. > " That argument has a certain amount of plausibility at first hearing, but it raises some questions. How far does the photon get from its source before the velocity uncertainty becomes very small? What justifies the assumption of one wavelength? After all, as the photon's energy varies, so does its wavelength. You wrote dt=lambda/c. shouldn't that be t=lambda/c? When dt/t is much greater than unity (you picked 10 in your example) can we still write in terms of differentials? Jerry -- "It does me no injury for my neighbor to say there are 20 gods, or no God. It neither picks my pocket nor breaks my leg." Thomas Jefferson to the Virginia House of Delegates in 1776.
From: Jerry Avins on 1 Apr 2010 10:37 On 4/1/2010 7:10 AM, WWalker wrote: > Jerry, > > It is true that in my simulation model I assume the receiving antenna does > not change the field it is detecting. This is possible if the reflections > from the receiving antenna are small and the detected signal is completely > absorbed and not reflected. Since the nearfields decay at 1/r^3 the > reflected signal will be very small by the time they reflect back to the > receiving antenna. This is observed in my network analyser dispersion > measurement test, where the nearfield part of the curve is not affected by > metal plates placed near the antennas. Only the farfield (1/r) fields are > observed to be affected by the metal plates. If the resistor used to detect > the received signal matches the impedance of the antenna, then the signal > will be completely absorbed and not reflected. I may not understand the mechanism, but it seems to me that pulling power out of the field that heats the resistor necessarily alters the field. Jerry -- "It does me no injury for my neighbor to say there are 20 gods, or no God. It neither picks my pocket nor breaks my leg." Thomas Jefferson to the Virginia House of Delegates in 1776.
From: Clay on 1 Apr 2010 10:52 On Apr 1, 5:23 am, glen herrmannsfeldt <g...(a)ugcs.caltech.edu> wrote: > Clay <c...(a)claysturner.com> wrote: > > (snip) > > > The quick and easy way is via 4-vectors. > > Here the scalar electric potential and the magnetic vector potential > > form a 4-vector. > > It isn't that quick and easy, but it does work. Feynman does > it after he does it the other way. Or maybe in the middle. > > > It is interesting to note that the fields themselves do not transform > > nicely. In fact it is the potentials that affect things and not the > > fields. This is well demonstrated by the Aharonov-Boehm effect. Yes > > there are cases where you have non zero potentials with zero fields > > and can observe the quantum interference being affected by varying the > > potential! > > Anyway, in Feynman's description there are three terms, one due > to the position of a charge (Coulomb's term except delayed by r/c), > the second is a velocity dependent correction to the first. > The result of the second term is that in the near field the > electric field vector is radial from (or to) the current position > of the charge in the constant velocity case. It is just as Jerry > mentioned a long time ago: in the case of a predictable motion, > nature knows how to fix it. It might be that is necessary for > special relativity and frame invariance, but it is surprising. > > Using that term, in the case of a slowly oscillating charge, > it wouldn't be surprising if the field wasn't what you would expect > from an appropriately delayed Coulomb field. > > The third term is the acceleration term, the only one you see > in the far field, especially for neutral sources. > > > But start with a single stationary charge in one frame of reference > > (in this frame the potentials are trivial) and then view the charge > > from a moving frame and using 4-vector calculus you get the new > > potentials. The curl of the magnetic vector potential will give you > > the B fields resulting from a single moving charge. A lot of Physics > > books will start with the Biot-Savart law and work from there avoiding > > the relativity approach. But it makes it much easier to calculate. > > My college class used the Berkely book, which does get into the > relativistic form pretty fast, but not quite the same as Feynman > does it. We did have the whole 4-vector explanation, but I don't > think we had homework problems using it. > > -- glen I had this in two different courses. Obviously this comes in E&M where we used the book by Jackson. Apparently most grad programs in physics use Jackson. So much so, that some of our foreign students had knock off copies of the text provided to them by their home country printed up with many errors not in the official text! 4-vectors were again demostrated in relativistic astrophysics, which again would be more appropriate for physics guys than EE guys. And we had homework with 4- vectors in both courses. I found them to be so useful, that looking back to other methods seems a bit primative. But that is the advantage of using a well honed theory. For example deriving the relativistic doppler shift is simple when one notes that for all observers the phase of the wave is invariant and therefore the scalar temporal frequency and the vector "wavenumber" become the transformed quanties. Even Fresnel's equation for the velocity of light through moving media becomes trivial to derive once Einstein's velocity addition formula is applied. Einstein's formula may be easily derived from two applications of 4-vec transformations. I don't see many EEs taking this subject this deep (gets kinda far afield for thier studies) but the salient point there is Maxwell's eqns are invariant under the Lorentz transformation. Particular features of the theory may be taken from there by various means. But to get back to Mr Walker's problem, we see in atomic physics most transitions involve the dipole approximation which says other transitions can't happen. But they do, so these are called "forbidden transitions." Since they occur with lower likelihood the dipole theory is mostly correct, but like most theories some approximations and simplications are applied to make the theory tractable. With nearfield stuff involving antennas one really needs to resort to something like L & R's theory of retarded potentials. The math quickly gets hairy and encourages one to approximate and this is where some unreal things show up in the approximate theory but a thorough and complete application resolves the quirks. Clay p.s. I think Feyman's 3 volume lecture set should be required reading for all undergraduate physics majors.
From: Clay on 1 Apr 2010 11:11
On Mar 31, 7:44 pm, robert bristow-johnson <r...(a)audioimagination.com> wrote: > On Mar 31, 6:03 pm, Clay <c...(a)claysturner.com> wrote: > > > > > > > On Mar 31, 4:17 pm, robert bristow-johnson <r...(a)audioimagination.com> > > wrote: > > > > to simplify things, the thought experiment i prefer is instead of a > > > single moving charge moving on the Z-axis, what one should consider is > > > an infinite line of uniform charge (also having a non-zero lineal mass > > > density) moving along the Z-axis. actually, it should be two parallel > > > infinite lines of uniform charge (parallel to the Z-axis) of known > > > spacing moving together in the Z direction. > > > > for the observer moving along with the two parallel lines of charge, > > > there is no motion relative to that observer and the problem is simple > > > electrostatics and, knowing the distance between the two lines, the > > > repulsive acceleration (sideways) of the two lines can be determined > > > purely from electrostatics. > > > > now there's a "stationary" observer that watches the two lines of > > > charge (and the "moving" observer) whiz by him and also notices that, > > > due to time dilation, the moving observer's clock is ticking more > > > slowly, so the outward acceleration of the moving lines of charge > > > appears to be slower than what is observed if they are not moving (as > > > the first observer sees). > > > > the rate of outward acceleration of the moving lines of charge, when > > > considering *only* electrostatics together with special relativity is > > > exactly the same outward acceleration of the same two moving lines of > > > charge when considering both static and magnetic forces in a classical > > > context (no relativistic effect). > > > > that thought experiment, first introduced to me by a physics prof (who > > > now is at Analog Devices) in the 70s, was sufficient to convince me > > > that the electromagnetic interaction (in the classical context) is > > > none other than just the sole electrostatic interaction with > > > relativistic effects applied. > > > > i s'pose the same can be done with lines of mass and the static > > > Newtonian gravitational interaction, also applying special relativity > > > and what you'll get out would be consistent with GEM (gravito-electro- > > > magnetism) where Maxwell's and Lorentz equations have mass (or mass > > > density) replacing charge (or charge density) and the Coulomb > > > constant, 1/(4*pi*epsilon_0), is replaced with -G (the minus sign is > > > because like-signed charges repel while like-signed masses attract). > > > for some reason (that i don't really get), the gravito-magnetic force > > > has an extra factor of 2 tossed into it (at least that's what the lit > > > seems to say). > > > > but, i think either classical EM or GEM can both be sorta understood > > > by considering the simple Coulomb or Newtonian static model with > > > relativistic effects. that's why we know that the magnetic > > > interaction is really just a consequence of the static interaction and > > > not a separate fundamental interaction. > ... > > > The quick and easy way is via 4-vectors. > > i might call that "the formal and general and legitimate way". and i > don't see how you would teach that to college sophomores after they > have first learned about classical EM and later about special > relativity. engineering/physics/chem majors in their sophomore year > should know how to derive the electrostatic field due to an infinite > line of charge and how that field will act on a little segment (of > given length) of another infinite line of charge (where nothing is > moving). and they should know how to derive the electromagnetic field > of an identical line of charge that is moving (co-linearly) at some > known velocity and how that magnetic field would act on a short > segment of another similar line of charge that is moving with a know > velocity. and, once they accept the postulates (i really think that > only one postulate is needed) of special relativity, they should > understand where time dilation comes from and how to apply that to an > observer in motion relative to another observer. > > it's a special case. it does not prove it in the general case, but i > think it can be used to persuade a student who hasn't yet (and may > never) learned about Minkowski spacetime, tensors, and 4-vectors, that > the classical magnetic interaction is nothing more than the > electrostatic interaction with special relativity considered. > > and, being an EE into DSP and being a third of a century away from any > formal physics class, it's about where my atrophied neurons regarding > all of this are stuck. and i never had a physics class where anything > other than the basics of special relativity had been taught (in > "General Physics"). i never had a course in formal SR (with Minkowki > constructs) or in GR (but i think i am okay about the postulates of > both). > > r b-j- Hide quoted text - > > - Show quoted text - I understand everyone comes at this from different backgrounds and perspectives. Rigorous nearfield stuff with antennas is going to be mathematically hairy no matter the approach and awaits those in grad school. Kraus's book on antennas is very good, and it is from a EE's approach. He also wrote a more general book on E & M. You may have had one of these, since you went through a EE program. I went 1st through a mathematics and then a physics program so my viewpoint is different from that of many others. This is sometimes good and sometimes not so good ;-) And I agree one of the perils of aging is a slowing of the brain (mental obtundation) As an aside, the math for general relativity is so obtuse that Schwarzschild's solution for the spherically symmetric case (the simplest one) was worked out about a dozen times by people not realizing that other worked out solutions were the same solution! We spent a fair bit of time in the GR course working through the details of the Schwartzschild case. The Schwarzschild Radius for blackholes comes from this. Clay Clay |