-
#1380 by rfmwguy on 30 Aug, 2015 14:22
-
One more, with areas of interest for this flight.
Kwertyops, very impressed with your work...thank you. Stickiness is there making it an imperfect balance but useable for short term changes as you said. I can also reset balance points to unused portions of blades which I did not do in ft2a. I'm afraid the oscillations at start of run are my keypad inputs on microwave. It is electromechanicly connected to nsf-1701 through galinstan, but this is high viscosity and think it transmits small vibrations like a solid coupling. I've got to remote the power on signal and will try and do that over the next several days. Small improvements bit by bit thanks to your analysis. -
#1381 by SeeShells on 30 Aug, 2015 14:44
-
...On the other side with the meep analysis of the frustum I've come to the conclusion that without modeling the loop or square antenna any layouts will be lacking on exciting the TE mode....
I agree, I've come to the same conclusion that without modeling the loop antenna you will not be properly exciting an axi-symmetric TE mode. (See emphasis on "axi-symmetric")
Still, this shows that Meep has been very useful in showing how important is the antenna's shape and its location Even when Meep excited a TE mode with a parallel dipole (for the Yang/Shell 6 degree cone angle geometry) , the mode shapes were not axi-symmetric. The only way that appears possible to excite an axi-symmetric TE mode is with a loop antenna. A square antenna will not do, I think, because the square shape is incompatible with circumferential axi-symmetry.
Even when Meep excites a TM mode with a dipole, the mode shape is not axi-symmetric either. The antenna really distorts the mode shape into two perpendicular axes: parallel to the dipole and perpendicular to the dipole.
The other thing we have learnt from Meep(post processed with Wolfram Mathematica) is that the EM Drive cannot be modeled solely with standing waves (as done for example by Greg Egan, as an eigenvalue problem, ignoring the effect of the RF feed). It is crucial to model the RF feed: it changes the conditions inside the cavity, particularly the stresses at the big base
Absolutely!
I couldn't agree more. This is also the case with injecting the magnetron directly into the cavity whether it's with a Z-matched hole in the top or bottom or sidewall injection, axisymmetrically injections is the key for the correct mode generation. Only then the cones asymmetrical shape will lead to the highest stress generation on the plates and walls.
I believe some at EW understood this just from looking at some of the layouts and I wish they could comment but sadly they can not.
Loop or square. I was thinking if areo could model a very short cylinder why couldn't he model one inside the frustum with a small section cut away where the feeds are? I'm probably over simplifying it and it is a tough nut to crack.
Shell -
#1382 by X_RaY on 30 Aug, 2015 15:59
-
Good idea...On the other side with the meep analysis of the frustum I've come to the conclusion that without modeling the loop or square antenna any layouts will be lacking on exciting the TE mode....
I agree, I've come to the same conclusion that without modeling the loop antenna you will not be properly exciting an axi-symmetric TE mode. (See emphasis on "axi-symmetric")
Still, this shows that Meep has been very useful in showing how important is the antenna's shape and its location Even when Meep excited a TE mode with a parallel dipole (for the Yang/Shell 6 degree cone angle geometry) , the mode shapes were not axi-symmetric. The only way that appears possible to excite an axi-symmetric TE mode is with a loop antenna. A square antenna will not do, I think, because the square shape is incompatible with circumferential axi-symmetry.
Even when Meep excites a TM mode with a dipole, the mode shape is not axi-symmetric either. The antenna really distorts the mode shape into two perpendicular axes: parallel to the dipole and perpendicular to the dipole.
The other thing we have learnt from Meep(post processed with Wolfram Mathematica) is that the EM Drive cannot be modeled solely with standing waves (as done for example by Greg Egan, as an eigenvalue problem, ignoring the effect of the RF feed). It is crucial to model the RF feed: it changes the conditions inside the cavity, particularly the stresses at the big base
Absolutely!
I couldn't agree more. This is also the case with injecting the magnetron directly into the cavity whether it's with a Z-matched hole in the top or bottom or sidewall injection, axisymmetrically injections is the key for the correct mode generation. Only then the cones asymmetrical shape will lead to the highest stress generation on the plates and walls.
I believe some at EW understood this just from looking at some of the layouts and I wish they could comment but sadly they can not.
Loop or square. I was thinking if areo could model a very short cylinder why couldn't he model one inside the frustum with a small section cut away where the feeds are? I'm probably over simplifying it and it is a tough nut to crack.
Shell
Another way may be the useage a larger number of dipoles to form a loop like antenna structure? -
#1383 by Rodal on 30 Aug, 2015 16:03
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no straight long dipole antennas: straight long antennas distort the symmetry of the modes ! -
#1384 by SeeShells on 30 Aug, 2015 16:29
-
From what I've heard and been PMed by an old friend (retired high energy physics). You can generate a stable mode in a frustum, not easy, but doable and this is what I'm looking for!...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
Got my coffee and back to the shop!
Shell -
#1385 by aero on 30 Aug, 2015 16:44
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
I can probably do something like that but I sure wish there were another meeper who would make and run some models of their own. Someone working the same problem to compare notes with.
As for artificially changing the Drude model so that the calculated Q value is realistic, I don't know. Could probably do it but I believe it would need to be done on a case by case basis. I don't think I would find a "One size fits all" solution. And I am reluctant to add the series of runs required for realistic Q to every change in the model. As all can see, calculated Q changes by changing the antenna and it also changes for the same cavity, same antenna but different EM field component excitation. The current model, CE2 - 8 gives a Q ranging from nothing to 47 million depending on which of the six field components is used to excite the antennas. I'm pretty sure that adding resistance to the copper without a sound theoretical basis would be a lost cause. The current model does have a sound theoretical basis at least. There may be something missing (no idea what) but what is there is soundly based. -
#1386 by Rodal on 30 Aug, 2015 16:59
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
I can probably do something like that but I sure wish there were another meeper who would make and run some models of their own. Someone working the same problem to compare notes with.
As for artificially changing the Drude model so that the calculated Q value is realistic, I don't know. Could probably do it but I believe it would need to be done on a case by case basis. I don't think I would find a "One size fits all" solution. And I am reluctant to add the series of runs required for realistic Q to every change in the model. As all can see, calculated Q changes by changing the antenna and it also changes for the same cavity, same antenna but different EM field component excitation. The current model, CE2 - 8 gives a Q ranging from nothing to 47 million depending on which of the six field components is used to excite the antennas. I'm pretty sure that adding resistance to the copper without a sound theoretical basis would be a lost cause. The current model does have a sound theoretical basis at least. There may be something missing (no idea what) but what is there is soundly based.
I have not seen any basis to justify the Drude constants to be realisitc for microwave frequencies.
Q's of millions are completely unrealistic, and show that the Meep models are not realistically incorporating power losses. I don't see that as being "soundly based".
Not addressing the Drude constants in the model just ensures that the Q will continue to be unrealistic in future runs.
We may all choose not to do certain things, for example because we have better things to do, for example I could write my own Finite Difference code to analyze this problem, but I chose not to do so, because I have better things to do. But that doesn't mean that we cannot discuss and agree on what are the right things to do if anybody would chose to do so. Using these Drude material constants are not a good model of reality.
________
PS: Did you address the fact that the user said that turning off eps averaging decreases the Q by a factor of 100 ?
-
#1387 by SeeShells on 30 Aug, 2015 17:00
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
I can probably do something like that but I sure wish there were another meeper who would make and run some models of their own. Someone working the same problem to compare notes with.
As for artificially changing the Drude model so that the calculated Q value is realistic, I don't know. Could probably do it but I believe it would need to be done on a case by case basis. I don't think I would find a "One size fits all" solution. And I am reluctant to add the series of runs required for realistic Q to every change in the model. As all can see, calculated Q changes by changing the antenna and it also changes for the same cavity, same antenna but different EM field component excitation. The current model, CE2 - 8 gives a Q ranging from nothing to 47 million depending on which of the six field components is used to excite the antennas. I'm pretty sure that adding resistance to the copper without a sound theoretical basis would be a lost cause. The current model does have a sound theoretical basis at least. There may be something missing (no idea what) but what is there is soundly based.
In for another cuppa coffee.
Question, how does the meep software calculate the Q? it it a one simulated pulse into the cavity or does it look at a set series. The reason is if we are collapsing a mode every full wavelength and regaining it in the next. How can you get any Q from the decaying modes so frequently? Can it be specified the number of pulses meep uses in calculating the Q?
Back out.
Shell -
#1388 by Rodal on 30 Aug, 2015 17:04
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
I can probably do something like that but I sure wish there were another meeper who would make and run some models of their own. Someone working the same problem to compare notes with.
As for artificially changing the Drude model so that the calculated Q value is realistic, I don't know. Could probably do it but I believe it would need to be done on a case by case basis. I don't think I would find a "One size fits all" solution. And I am reluctant to add the series of runs required for realistic Q to every change in the model. As all can see, calculated Q changes by changing the antenna and it also changes for the same cavity, same antenna but different EM field component excitation. The current model, CE2 - 8 gives a Q ranging from nothing to 47 million depending on which of the six field components is used to excite the antennas. I'm pretty sure that adding resistance to the copper without a sound theoretical basis would be a lost cause. The current model does have a sound theoretical basis at least. There may be something missing (no idea what) but what is there is soundly based.
In for another cuppa coffee.
Question, how does the meep software calculate the Q? it it a one simulated pulse into the cavity or does it look at a set series. The reason is if we are collapsing a mode every full wavelength and regaining it in the next. How can you get any Q from the decaying modes so frequently? Can it be specified the number of pulses meep uses in calculating the Q?
Back out.
Shell
It is calculated by Harmonic Inversion (see http://ab-initio.mit.edu/wiki/index.php/Harminv ), as the 1/2 ratio of the real and imaginary parts of the frequencies calculated by Harmonic Inversion.
It is not analyzing a pulse or a set series of pulses but finitely-many sinusoids of varying magnitude and varying frequency.Quotegiven a discrete-time, finite-length signal that consists of a sum of finitely-many sinusoids (possibly exponentially decaying) in a given bandwidth, it determines the frequencies, decay constants, amplitudes, and phases of those sinusoids.
-
#1389 by aero on 30 Aug, 2015 17:16
-
...
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no long dipole antennas: long antennas distort the symmetry of the modes !
I can probably do something like that but I sure wish there were another meeper who would make and run some models of their own. Someone working the same problem to compare notes with.
As for artificially changing the Drude model so that the calculated Q value is realistic, I don't know. Could probably do it but I believe it would need to be done on a case by case basis. I don't think I would find a "One size fits all" solution. And I am reluctant to add the series of runs required for realistic Q to every change in the model. As all can see, calculated Q changes by changing the antenna and it also changes for the same cavity, same antenna but different EM field component excitation. The current model, CE2 - 8 gives a Q ranging from nothing to 47 million depending on which of the six field components is used to excite the antennas. I'm pretty sure that adding resistance to the copper without a sound theoretical basis would be a lost cause. The current model does have a sound theoretical basis at least. There may be something missing (no idea what) but what is there is soundly based.
In for another cuppa coffee.
Question, how does the meep software calculate the Q? it it a one simulated pulse into the cavity or does it look at a set series. The reason is if we are collapsing a mode every full wavelength and regaining it in the next. How can you get any Q from the decaying modes so frequently? Can it be specified the number of pulses meep uses in calculating the Q?
Back out.
Shell
Meep runs the model for a period of time using a Gaussian noise source, the source is turned off and the fields are propagated for a user specified time. At the end of that time the un-decayed fields are fitted (by Harminv) to a number of exponentially decaying sin waves(?) and those fits that exceed Q=50 are output as the frequency and quality factor. Q is calculated as - Re(w) / 2 Im(w) - real over twice the imaginary part of the fitted frequency.
Default is to search for 5 frequencies but can search for up to 100 frequencies. For the EM drive cavities it is rare for Harminv to return more than 1, sometimes 2 frequencies unless the noise bandwidth is set very wide.
Note: Sorry Dr. Rodal, we cross posted the same answer -
#1390 by flux_capacitor on 30 Aug, 2015 17:21
-
One more, with areas of interest for this flight.
Kwertyops, very impressed with your work...thank you. Stickiness is there making it an imperfect balance but useable for short term changes as you said. I can also reset balance points to unused portions of blades which I did not do in ft2a. I'm afraid the oscillations at start of run are my keypad inputs on microwave. It is electromechanicly connected to nsf-1701 through galinstan, but this is high viscosity and think it transmits small vibrations like a solid coupling. I've got to remote the power on signal and will try and do that over the next several days. Small improvements bit by bit thanks to your analysis.
Is the green portion of the test results showing the truncated cone moving in the direction of the small base (when the magnetron is on) much larger and significant than in previous test runs?
This is the first run for NSF-1701 where I see a very clear movement in the direction of the small base
If so, can you point out to what changes in this latest run may have been responsible (in your view) for the very noticeable movement toward the small base?
Rodal, as I said the two graphs are identical but one of them has been flipped upside-down, see the drawing below where I put them together on the same scale.
The small base in NSF-1701 Flight Test 2A points towards the floor, downwards. When the frustum goes upwards (a priori due to thermal effects) the laser dot goes downwards.
EmDriven (blue/brown curve) on Reddit claims he showed the weight, whereas kwertyops (red) shows laser position (?)
kwertyops's curve increases while EmDriven's brown portion of his blue curve decreases. There is something I don't understand in those curves: to me it a appears a weight reduction (EmDriven) should follow laser position (kwertyops) in the same trend, because both would track a frustum going upwards. But they are opposite. Why?
For now in my opinion there is no thrust signature in NSF-1701 Flight Test 2A graphs, only vibration and thermal artifacts.
-
#1391 by aero on 30 Aug, 2015 17:24
-
Quote
I have not seen any basis to justify the Drude constants to be realisitc for microwave frequencies.
@Deltamass - Do you want to respond to Dr. Rodal? As I recall you did devote significant effort to develop the Drude model and did use the correct range of microwave frequency in your derivation. -
#1392 by X_RaY on 30 Aug, 2015 17:38
-
I have a suggestion for the model....
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no straight long dipole antennas: straight long antennas distort the symmetry of the modes !
Cone dimensions (CrazyEddie).
Dipole orientation like in the picture(YangShell), but more and shorter antennas around the circumference near the big end.
Note that the YangShell design with only two dipoles in this configuration showed TE01 but deformed.
-
#1393 by Rodal on 30 Aug, 2015 17:44
-
An important issue is that Harminv (the harmonic inversion software for Meep) is being used to examine the response so far when the response is growing exponentially during the transient with the RF Feed on.QuoteI have not seen any basis to justify the Drude constants to be realisitc for microwave frequencies.
@Deltamass - Do you want to respond to Dr. Rodal? As I recall you did devote significant effort to develop the Drude model and did use the correct range of microwave frequency in your derivation.
The correct way to calculate Q (as we discussed before) should be:
1) Run Meep with an RF feed ON, for a geometry and RF feed where the Meep model excites the same mode that is excited at the natural frequency in an eigenvalue problem (with standing waves, no RF feed).
2) Turn the RF feed OFF in Meep after a reasonably amount of time has elapsed (from prior calculations, it looks like 128 cycles would be OK, as by that time the amplitude is close to being stable, as the exponential growth has decreased significantly)
3) Use Harminv to harmonically invert the frequency response with the RF feed OFF, and determine the Q for the RF feed off.
The present analysis for Q are being done with the RF feed on, and therefore do not represent a clean decay of the response, since with the RF feed ON one has an exponentially growing response instead of a decaying response.
In general, damping, and hence Q, is simpler to analyze with the source of oscillation being OFF.
This is crucial for a problem, like this one, where there is no closed-form solution for the RF feed ON -
#1394 by aero on 30 Aug, 2015 17:49
-
I have a suggestion for the model....
I fully agree, a large number (say every 10 degrees: 36 such antennas) of short antennas around the circumference, would ensure axi-symmetry .
Good idea
Another way may be the useage a larger number of dipoles to form a loop like antenna structure?
Please, no straight long dipole antennas: straight long antennas distort the symmetry of the modes !
Cone dimensions (CrazyEddie).
Dipole orientation like in the picture(YangShell), but more and shorter antennas around the circumference near the big end.
Note that the YangShell design with only two dipoles in this configuration showed TE01 but deformed.
That is a clear description of the general idea. Now could you give the engineering specification of what you'd like to see? Once SeeShells vets it, I can try to build it. No guarantees but I can guarantee nothing without real, clear specifications. -
#1395 by kwertyops on 30 Aug, 2015 18:01
-
NSF-1701 Flight Test 2A laser position graph posted on Reddit by @EmDriven.
@kwertyops: did you flip your graphs upside-down in order to show a downwards thrust of the frustum going with the curve going also downwards? I'm talking of your graphs in this post.
I actually didn't flip it myself, that's how the data came out of TrackMate. I expect that it's using the pixel coordinate system where the positive y-direction is downwards (laser goes down, y increases, graph goes upwards).
So in your image with the blue and red plot, the red one (mine) is actually reflective of the frustum position, while the blue one (EmDriven) is showing the laser position. But yes, they should be mostly mirror images of each other.
[Edit] So, just to be clear - we are both showing laser position, but mine (red) is flipped to the pixel coordinate system, and so it more directly reflects the frustum position. When the red line moves down, that is the frustum moving down. The blue graph reflects just the laser position, as you see it visually in the video, in a normal cartesian coordinate system. When the blue line goes down, the frustum is going up.can you upload the full trackmate data file plese, this just has displacement.
Also
Can you upload the video that's generated? I suspect that the change in the target made spotify splot.
The video output seems like it's taking prohibitively long, so I'm not going to try to do that, but I will include the cropped input video, and the full raw TrackMate data (XML). Please note that for this run I applied a slight Gaussian Blur to the video beforehand, to minimize the effect of that blue reflection, as well as due to the grid markers potentially causing aberrations.
-
#1396 by Rodal on 30 Aug, 2015 18:10
-
...Rodal, as I said the two graphs are identical but one of them has been flipped upside-down, see the drawing below where I put them together on the same scale.
Thank you so much for kindly taking the time to explain this so clearly and thoroughly.
The small base in NSF-1701 Flight Test 2A points towards the floor, downwards. When the frustum goes upwards (a priori due to thermal effects) the laser dot goes downwards.
EmDriven (blue/brown curve) on Reddit claims he showed the weight, whereas kwertyops (red) shows laser position (?)
kwertyops's curve increases while EmDriven's brown portion of his blue curve decreases. There is something I don't understand in those curves: to me it a appears a weight reduction (EmDriven) should follow laser position (kwertyops) in the same trend, because both would track a frustum going upwards. But they are opposite. Why?
For now in my opinion there is no thrust signature in NSF-1701 Flight Test 2A graphs, only vibration and thermal artifacts. -
#1397 by Mezzenile on 30 Aug, 2015 18:17
-
Does somebody know if a tentative has been made to analyse the RF cavity thruster phenomena at the light of the analysis by James Woodward of the Mach principle in the frame of General Relativity of Einstein ?
The thesis of Woodward is that the Mach Principle on the origin of Inertia forces is a logical consequence of both General Relativity and the fact that our space-time is flat (confirmed by WMAP and Planck Space experiments on the microwave radiation background). This thesis is presented in the book of Woodward "Making Starships and Stargates - The Science of Interstellar Transport and Absurdly Benign Wormholes" edited by Springer.
James Woodward explains that an object which is accelerated by a force while it accumulates energy will experience mass fluctuations which can be utilized to provide thrust on the object itself, without need of local momentum conservation (the momentum is conserved at the the whole universe level).
I have read this book this week (somewhat in diagonal as the mathematical part is not so benign to assimilate !). It is evident that our RF cavity is submited to the acceleration of the Earth gravitation, that it stores internal energy via a high frequency electromagnetic process. So following Woodward/Mach this cavity should present mass fluctuations. Now it should be sufficient to search and identified a force applied in quadrature with this mass fluctuations to be sure (according Woodward) that a "magic" thrust will appear on the RF cavity.
Could a phenomena of electrostriction play the role of this quadrature force ??
In any case the book of Woodward is worth to be read !!!
-
#1398 by SteveD on 30 Aug, 2015 18:17
-
Didn't TheTraveller estimate, with the projected Q of NSF-1701 that we would see a force of around 140 micronewton's (approx 14mg)? The upward force of thermal effects is going to swamp any downward thrust source. Run 2 had an interesting result where the system stabilizes on power off, then rmfwguy walks by it, causing it to move lower (drive up) as if inertia had been holding the beam laser at a higher position (device lower) than the direction the rising hot air wants to go.
Given the stickiness of the beam, I'm not sure if this is not simply an artifact of the rig. Without something to break the inertia of the rig, I would expect the slow decay of any downwward energy (device up) would be swamped by thermal cooling (device down).
My advice, walk by the thing on power off and see what happens. Also consider a build that feeds by coax in order to increase Q and separate a major heat source from the device.
-
#1399 by deltaMass on 30 Aug, 2015 18:21
-
Sad to say I threw away the papers containing that work.QuoteI have not seen any basis to justify the Drude constants to be realisitc for microwave frequencies.
@Deltamass - Do you want to respond to Dr. Rodal? As I recall you did devote significant effort to develop the Drude model and did use the correct range of microwave frequency in your derivation.
