Author Topic: EM Drive Developments - related to space flight applications - Thread 3  (Read 1804339 times)

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
Clearly the most important thing to do in Meep simulations is to


model a loop antenna to excite TE (transverse electric) modes.

The presently used dipole antenna has only excited Transverse Magnetic modes whether at the big end or at the small end, whether oriented in the transverse or the longitudinal directions, whether for rfmwguy/NSF-1701 or for Yang/Shell geometry.

If unable to model a loop antenna to excite TE modes, what could be done as the next step is to place the dipole antenna in the transverse direction near the small end, to be able to compare the two runs made for Yang/Shell with the dipole antenna at the big end.

________________

After modeling the loop antenna, the next most important thing is to:


model a magnetron as the RF source

model a waveguide entering the EM Drive, as done by Prof. Yang
« Last Edit: 07/23/2015 07:00 PM by Rodal »

Offline aero

  • Senior Member
  • *****
  • Posts: 2744
  • 92129
  • Liked: 705
  • Likes Given: 239
@Rodal - I'm not sure where you are getting  the data indicating that Q varies by 3 orders of magnitude, perhaps a miscommunication somewhere. If your source is meep Harminv data, then the correct values as I recorded them at the time of the runs are:

 Q   87,830,861
 Q   87,830,729
 Q   5,068,251
 Q   59,477,392

where the first two were using, I believe, a shorter dipole antenna (29 mm) and driving with the previously determined resonance frequency to hone in on the actual cavity resonance frequency as model. It is very consistently 2.46316012E+009 Hz within kHz. The final two were from the model with drive frequency = 2.45 GHz and antenna length = 58 mm as used to generate the 2 Yang-Shell data sets uploaded to Google drive.
« Last Edit: 07/23/2015 06:41 PM by aero »
Retired, working interesting problems

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
@Rodal - I'm not sure where you are getting  the data indicating that Q varies by 3 orders of magnitude, perhaps a miscommunication somewhere. If your source is meep Harminv data, then the correct values as I recorded them at the time of the runs are:

 Q   87,830,861
 Q   87,830,729
 Q   5,068,251
 Q   59,477,392

where the first two were using, I believe, a shorter dipole antenna (29 mm) and driving with the previously determined resonance frequency to hone in on the actual cavity resonance frequency as model. It is very consistently 2.46316012E+009 Hz within kHz. The final two were from the model with drive frequency = 2.45 GHz and antenna length = 58 mm as used to generate the 2 Yang-Shell data sets uploaded to Google drive.

aero, the information comes from your message quoted again below:

...I wondered why that went quicker last night. Not quick but a little quicker. I guess you found out.

Check it again, they are there now.
I'm looking at the Yang/Shell Axial Antenna at Big Base case now: very unusual: the stress, and hence the force at the small base is practically zero.  The stress at the big base is a central point stress from the antenna.  Close inspection of this mode looks like another TM11 transverse magnetic mode but with drastically different amplitude.

QUESTION1: was the mesh kept the same as in the previous csv Yang/Shell case, and you are sure this is the stress at the small base and not outside it?

Most important: QUESTION2: did Meep give you a Q value for this case ?

Thanks

Everything about the run was identical except the antenna. The csv files are the same size aren't they? If something were changed likely they would change size. And really, the bases should be in the same place they were previously. I looked at this data set with HDFview. But note that the row numbers I gave you I had 1 added, to start at 1 like the csv matrices, instead of 0 as HDFview uses. If you also added 1, that would be the problem. The model skin is three matrix rows thick, adding an extra 1 would make the small base row be inside the skin.

It was also the same 58 mm antenna centered quarter wavelength from the inside face of the big base but rotated 90 degrees to an axial orentation. Note that 1/4 wave length is only slightly more than half of 58 mm, so the end of the antenna near the big base was about 1.5 mm away from the base, and excited with ez component although hy would have been more natural.

Q? Yea, Q was ridiculously high, like 60 million and the resonant frequency was like 2.463 GHz, which I ignored and made the run at 2.45 GHz.

In my post I was discussing your statement <<Q? Yea, Q was ridiculously high, like 60 million >>

referring to the last run for Yang/Shell with the antenna oriented along the longitudinal x axis

Q=60 million is more than 1,000 times greater than Q=50,000 reported by Shawyer

===> I thought you were getting reasonable Q's for your other runs.  My understanding from Shell is that your prior model for Yang/Shell was giving Q = 87,000+
Are you getting Q's with Meep of millions for all the runs?

===> The program I wrote gives a Q=50,175 for rfmwguy/NSF-1701, I have not had the chance to calculate it for Yang/Shell yet.

If Meep is giving a Q=80 million for rfmwguy/NSF-1701 there is something wrong somewhere...
Are you getting Q=80 million numbers with the Drude model ?

Or perhaps the interpretation of Q in Meep is wrong: it is 3 orders of magnitude larger than reasonable
« Last Edit: 07/23/2015 07:15 PM by Rodal »

Offline aero

  • Senior Member
  • *****
  • Posts: 2744
  • 92129
  • Liked: 705
  • Likes Given: 239
@Rodal - I'm not sure where you are getting  the data indicating that Q varies by 3 orders of magnitude, perhaps a miscommunication somewhere. If your source is meep Harminv data, then the correct values as I recorded them at the time of the runs are:

 Q   87,830,861
 Q   87,830,729
 Q   5,068,251
 Q   59,477,392

where the first two were using, I believe, a shorter dipole antenna (29 mm) and driving with the previously determined resonance frequency to hone in on the actual cavity resonance frequency as model. It is very consistently 2.46316012E+009 Hz within kHz. The final two were from the model with drive frequency = 2.45 GHz and antenna length = 58 mm as used to generate the 2 Yang-Shell data sets uploaded to Google drive.

aero, the information comes from your message quoted again below:

...I wondered why that went quicker last night. Not quick but a little quicker. I guess you found out.

Check it again, they are there now.
I'm looking at the Yang/Shell Axial Antenna at Big Base case now: very unusual: the stress, and hence the force at the small base is practically zero.  The stress at the big base is a central point stress from the antenna.  Close inspection of this mode looks like another TM11 transverse magnetic mode but with drastically different amplitude.

QUESTION1: was the mesh kept the same as in the previous csv Yang/Shell case, and you are sure this is the stress at the small base and not outside it?

Most important: QUESTION2: did Meep give you a Q value for this case ?

Thanks

Everything about the run was identical except the antenna. The csv files are the same size aren't they? If something were changed likely they would change size. And really, the bases should be in the same place they were previously. I looked at this data set with HDFview. But note that the row numbers I gave you I had 1 added, to start at 1 like the csv matrices, instead of 0 as HDFview uses. If you also added 1, that would be the problem. The model skin is three matrix rows thick, adding an extra 1 would make the small base row be inside the skin.

It was also the same 58 mm antenna centered quarter wavelength from the inside face of the big base but rotated 90 degrees to an axial orentation. Note that 1/4 wave length is only slightly more than half of 58 mm, so the end of the antenna near the big base was about 1.5 mm away from the base, and excited with ez component although hy would have been more natural.

Q? Yea, Q was ridiculously high, like 60 million and the resonant frequency was like 2.463 GHz, which I ignored and made the run at 2.45 GHz.

Again, in my post I was discussing your statement <<Q? Yea, Q was ridiculously high, like 60 million >>

referring to the last run for Yang/Shell with the antenna oriented along the longitudinal x axis

Q=60 million is 1,000 times greater than Q=60,000

The ~60 million is correct, it is the ~60,000 that I am wondering about. The closest number that I recorded was ~5 million, only 1 order of magnitude difference.

As for
Quote
If unable to model a loop antenna to excite TE modes, what could be done as the next step is to place the dipole antenna in the transverse direction near the small end, to be able to compare the two runs made for Yang/Shell with the dipole antenna at the big end.
That is what the first Yang-Shell data set was. See attached as downloaded from Google drive just now. Only one data set (the latest uploaded) had the longitudinal dipole antenna at the big end.

Note that the small end antenna .csv files are in the same folder as the big end antenna csv files. Sorted alphabetically all of the "small" are after all of the "Big" file names. And I note that the data seems to have been generated using a lattice size different from the current lattice size. I don't have a good way to check end row or antenna location using the uploaded data, I hope you do. I will run it again if needed.
Retired, working interesting problems

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
I just calculated Yang/Shell TM113 with my program:

Q =45,039 Yang/Shell (copper)  natural frequency for TM113 = 2.4941 GHz
Q =50,175   rfmwguy/NSF-1701(copper)


Used the following material properties:

epsilon0 = 8.854187817*10^-12);
mu0 =  0.999991(*copper*)*4*Pi*10^(-7);
resistivity =  1.678*10^(-8)(*copper*);

For  using same material constants, I get 

If I use brass or bronze I will get a lower Q.  For example, for:

resistivity = 1.437*10^(-7)  high strength brass

giving:

Q=  15,391  (Yang/Shell)  high strength brass
Q=  17,146 (rfmwguy/NSF1701)  high strength brass



what material model are you using to get Q's in the millions?  are you using the Drude model ?

I would look for something responsible for orders of magnitude off:  Meep units conversion ?


____________

PS: I edited my prior message, eliminating reference to the 60 million Q, as it appears that all the Q's from Meep runs are suspect.
« Last Edit: 07/23/2015 08:08 PM by Rodal »

Offline tchernik

  • Full Member
  • **
  • Posts: 247
  • Liked: 297
  • Likes Given: 577
This was discussed in thread 2, thing is for a "propulsive effect" of constant thrust it relies on a non stationary ever increasing current :
http://forum.nasaspaceflight.com/index.php?topic=36313.msg1350655#msg1350655

They also have a more recent paper called "Relativistic Engine Based on a Permanent Magnet" (from July 13, 2015). The link is: http://lib-arxiv-008.serverfarm.cornell.edu/pdf/1507.02897v1.pdf

Remarkable paper. Does this paper suggest a new kind of thruster we could build in practice? or is this un-physical due to it assuming things like Terawatts or input power or some such?

Because it sounds mechanically simple. So, where's the catch?
« Last Edit: 07/23/2015 08:03 PM by tchernik »

Offline b0nafide

  • Member
  • Posts: 4
  • Liked: 1
  • Likes Given: 0
...what do people think about reporting NULL results of experimenters ? ...

I'm just a random person following this thread, but I have been waiting for kml's null tests to appear on the experimental results section of the wiki ever since kml reported his findings. It has been bothering me that nobody else seems to think kml's test is notable. I learned something new. We can alter the reading of a digital scale without touching it.

Offline flux_capacitor

  • Full Member
  • ****
  • Posts: 552
  • France
  • Liked: 654
  • Likes Given: 909
Remarkable paper. Does this paper suggest a new kind of thruster we could build in practice? or is this un-physical due to it assuming things like Terawatts or input power or some such?

Because it sounds mechanically simple. So, where's the catch?

The catch is maybe in the following sentence in their paper:

"hence for any momentum that is acquired by matter an opposite momentum is attributed to the electromagnetic field."

Does that sentence preclude the device from being used as a thruster?

Offline aero

  • Senior Member
  • *****
  • Posts: 2744
  • 92129
  • Liked: 705
  • Likes Given: 239
I just calculated Yang/Shell TM113 with my program:

Q =45,039 Yang/Shell (copper)  natural frequency for TM113 = 2.4941 GHz
Q =50,175   rfmwguy/NSF-1701(copper)


Used the following material properties:

epsilon0 = 8.854187817*10^-12);
mu0 =  0.999991(*copper*)*4*Pi*10^(-7);
resistivity =  1.678*10^(-8)(*copper*);

For  using same material constants, I get 

If I use brass or bronze I will get a lower Q.  For example, for:

resistivity = 1.437*10^(-7)  high strength brass

giving:

Q=  15,391  (Yang/Shell)  high strength brass
Q=  17,146 (rfmwguy/NSF1701)  high strength brass



what material model are you using to get Q's in the millions?  are you using the Drude model ?
The Drude model.
Quote
____________

PS: I edited my prior message, eliminating reference to the 60 million Q, as it appears that all the Q's from Meep runs are suspect.

Of course they are. There are no losses, not because of the material model, but because of the sampling resolution, (node separation). They are a valid indication of whether or not the cavity model will resonate but more than that it is hard to say. I use them for that purpose, interpreting high Q's as strong resonance of the model and low or no Q's to indicate low or no resonance at that frequency/antenna configuration as modelled.

We have known for a long time that Q's in the millions are not realistic for room temperature cavities.
Retired, working interesting problems

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
...what do people think about reporting NULL results of experimenters ? ...

I'm just a random person following this thread, but I have been waiting for kml's null tests to appear on the experimental results section of the wiki ever since kml reported his findings. It has been bothering me that nobody else seems to think kml's test is notable. I learned something new. We can alter the reading of a digital scale without touching it.

I think that one reason they have not been listed yet is because:

IMPORTANT:  The other EM Drive listed have variable cross-section.  KML is testing a constant-cross section cavity.

It looks like if it is listed it should be listed separately, as it does not conform to the "tapered cross-section" concept.

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
...

Of course they are. There are no losses, not because of the material model, but because of the sampling resolution, (node separation). They are a valid indication of whether or not the cavity model will resonate but more than that it is hard to say. I use them for that purpose, interpreting high Q's as strong resonance of the model and low or no Q's to indicate low or no resonance at that frequency/antenna configuration as modelled.

We have known for a long time that Q's in the millions are not realistic for room temperature cavities.

 I don't agree that the losses should be modeled with nodes through a micrometer skin depth.  That's not the way that losses are modeled in COMSOL FEA or in most modeling programs, or in my computer program, nor do I think that's how Meep should do it.  It would not be feasible to model the losses that way nor do I think it would be a good way to do it.  It would be very inaccurate, and computer-time wasting.  The skin depth problem has an exact solution, well-known and verified in too many experiments.

All you have to do to model the Q is to perform a volume integral and a separate surface integration.  If not built-in in Meep maybe someone has already done it.

No need for nodes:  the losses are just the skin depth times the surface integral of the square magnitude of the B field.  The losses are surface losses.

Q should be calculated as follows:



The numerator is a volume integral (of the energy density), while the denominator is a surface integral.

do you have a link for how (what equation it uses) Meep calculates the Q quality factor ?

I feel that there is some conversion issue or a wrong constant somewhere...

« Last Edit: 07/23/2015 08:48 PM by Rodal »

Offline notarget

  • Member
  • Posts: 18
  • San Carlos, CA
  • Liked: 17
  • Likes Given: 9

Of course they are. There are no losses, not because of the material model, but because of the sampling resolution, (node separation). They are a valid indication of whether or not the cavity model will resonate but more than that it is hard to say. I use them for that purpose, interpreting high Q's as strong resonance of the model and low or no Q's to indicate low or no resonance at that frequency/antenna configuration as modelled.

We have known for a long time that Q's in the millions are not realistic for room temperature cavities.
For interest, I've attached a movie of a mid-plane Y direction slice of the NSF frustum playing the Ez field.  The movie covers about 400 cycles right before then after the Gaussian pulse ends - so shows the resonating cavity as the energy diminishes.

I've also attached the ctl file I used to generate this case - note that I'm using quarter symmetry and a reduced resolution of 150 as I was centering in on the pulse behavior.  In the movie, I've clamped the Ez values to +/- 0.028665 so that the effect of the source isn't overwhelming in the beginning when it's on.

Note that in the ctl file, I've included an alternative Cu model that is taken from the reference:   http://falsecolour.com/aw/meep_metals/investigation.html#SECTION00011600000000000000 where the author validated the Cu model inside meep.

I'm running higher res (250) now and bracketing the interesting time period on output after the Gaussian pulse ends.  I think we should probably naturally terminate the Gaussian source pulse instead of artificially terminating it to avoid the transients.

WRT running a loop source, should be pretty easy, though the aliasing into the cartesian mesh is going to introduce some weird frequencies...

Online A_M_Swallow

  • Elite Veteran
  • Senior Member
  • *****
  • Posts: 8154
  • South coast of England
  • Liked: 249
  • Likes Given: 104
{snip}
I've found it interesting and keep coming back to the red flags in the tests.
Air pressure=thrust vs Vacuum=small thrust
This is the 800 pound gorilla sitting in the middle of the lab and nobody wants to poke it. :D

{snip}

That is why I suggested to Traveller that he design his device with an off resonate frequency mode. Heat and magnetic effects would still cause movements in air but not the EM Drive itself. We can test vacuum effects when we have a design that works reliably in air.

Offline CraigPichach

  • Member
  • Posts: 59
  • Calgary Alberta
  • Liked: 42
  • Likes Given: 4
New question for design purposes - For commercial magnetrons at 100kW need Solenoid, Waveguide Launcher and Isolator.  I am proposing to feed the copper frustrum after  the end of the isolator (where a waterload is used to absorb RF energy); this is different than what Eagleworks proposed.  I guess one could attach the thruster at the end of the waveguide launcher but that would require a perfect impedance match between the magnetron and the thruster otherwise any reflected energy would immediately do damage to the magnetron. If we do that the EM-Drive/Q-Thruster "thrust" should still occur as in theory at resonance there should be no return to the magnetron ? ? ? ? (is this correct???)
Also has anyone worked with CST software? Anyone think that's the software to use to confirm dimensions for resonance frequency at 930MHz (it seriously looks like >2GHz is out for any magnetron application over 10kW)?
« Last Edit: 07/23/2015 09:34 PM by CraigPichach »

Offline notarget

  • Member
  • Posts: 18
  • San Carlos, CA
  • Liked: 17
  • Likes Given: 9
Number of time steps, 6527 and total meep time = 13.054 time units.
Wow - got it - so we're showing an exponential growth of force on the large base for the Yang model, but we need to run a lot further in time to get where we want to be. Are these the forces output by meep or calculated by you afterward? Can you please summarize what you need for the force calculation / output?

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
Number of time steps, 6527 and total meep time = 13.054 time units.
Wow - got it - so we're showing an exponential growth of force on the large base for the Yang model, but we need to run a lot further in time to get where we want to be. Are these the forces output by meep or calculated by you afterward? Can you please summarize what you need for the force calculation / output?
1) I calculate the forces using Wolfram Mathematica (https://www.wolfram.com/) post-processing .csv files

2) I need all 6 electromagnetic fields at a surface to calculate the stress.  For example, for the Big Base, with normal "x" (*) Cartesian axis of axi-symmetry of the truncated cone and y and z transverse axes perpendicular to x, I need:
Exx, Exy, Exz, Hxx, Hxy and Hxz fields at the Big Base

where the first subscript refers to the axis perpendicular to the surface, and the second subscript refers to the component of the electromagnetic field vector

Ditto for the 6 fields at the small base

3) aero has been outputting this information for the last 14 time slices (of the total of 320 time slices), the 6 fields for the Small Base and for the Big Base. 

4) aero has been using eps averaging done automatically by Meep, so the location of the bases (and other nodes for the material) may not be where you expect them to be: have to check the location of the bases ahead of time, and output the csv files such as to have the 6 fields in contact with the base

5) I need to have the output information: total Meep run time (in aero's case it was 13.054 as I recall), the total number of Meep time slices (it was 320) and the total number of Finite Difference times steps ( 6527 ).  I use this information for scaling purposes, since Meep units need to be scaled to real time, etc.

6) I need all Meed scaling information: Meep length (for aero is 0.3), Meep Current (aero uses default input Io=1 Amp) etc.


(*) I would use z as the longitudinal axis of axi-symmetry of the cone, but aero has been using x.  The choice is arbitrary of course
« Last Edit: 07/23/2015 09:53 PM by Rodal »

Offline kml

...what do people think about reporting NULL results of experimenters ? ...

I'm just a random person following this thread, but I have been waiting for kml's null tests to appear on the experimental results section of the wiki ever since kml reported his findings. It has been bothering me that nobody else seems to think kml's test is notable. I learned something new. We can alter the reading of a digital scale without touching it.

Those early test results were confirmed to be the result of RFI affecting the scale, so they should be disregarded.
The original design with an adjustable copper covered plunger on one end was not very RF tight.   The current design with copper gaskets at each end is much better and most of the RFI issues since that change have been with the remote control transmitter and not the test unit.

I am continuing to run test and collect data.   I have been holding off on posting more test results until I am absolutely sure that the data is clean and not affected by RFI.    I can say the vast majority of the test runs have shown no deviation from the thermal slope while RF is on.   The few that did have either been confirmed or suspected as RFI related.   Early indications are that a dielectric alone is not able to produce the effect.

Here is an updated picture of the test rig with the 30W PA driven by an alinco DJ-G7T transciever in crossband repeat mode.  The fan on the PA is not used during test runs.  The new sample port location is on the left side.

« Last Edit: 07/23/2015 10:01 PM by kml »

Offline notarget

  • Member
  • Posts: 18
  • San Carlos, CA
  • Liked: 17
  • Likes Given: 9
3) aero has been outputting this information for the last 14 time slices (of the total of 320 time slices), the 6 fields for the Small Base and for the Big Base. 
Got it - BTW props to Aero (Steve) for learning meep and putting the input / geoms together - I'm drafting his work effectively.

I'll do some long runs and save a lot more time steps to get a feel for the transients. Pls stand by as I'm fumbling around.

Online Rodal

  • Senior Member
  • *****
  • Posts: 5838
  • USA
  • Liked: 5919
  • Likes Given: 5267
...what do people think about reporting NULL results of experimenters ? ...

I'm just a random person following this thread, but I have been waiting for kml's null tests to appear on the experimental results section of the wiki ever since kml reported his findings. It has been bothering me that nobody else seems to think kml's test is notable. I learned something new. We can alter the reading of a digital scale without touching it.

Those early test results were confirmed to be the result of RFI affecting the scale, so they should be disregarded.
The original design with an adjustable copper covered plunger on one end was not very RF tight.   The current design with copper gaskets at each end is much better and most of the RFI issues since that change have been with the remote control transmitter and not the test unit.

I am continuing to run test and collect data.   I have been holding off on posting more test results until I am absolutely sure that the data is clean and not affected by RFI.    I can say the vast majority of the test runs have shown no deviation from the thermal slope while RF is on.   The few that did have either been confirmed or suspected as RFI related.   Early indications are that a dielectric alone is not able to produce the effect.

Here is an updated picture of the test rig with the 30W PA driven by an alinco DJ-G7T transciever in crossband repeat mode.  The fan on the PA is not used during test runs.  The new sample port location is on the left side.



Great information !!!

2 questions:

1) Should we document your present results as NULL in the Wiki or do you want us to wait?

2) Have you "motivated" the EM Drive by tapping it or vibrating it? ==> Recall TheTraveller/Shawyer's conjecture that for EM Drive to register a force, it needs to be motivated by tapping it or vibrating it
« Last Edit: 07/23/2015 10:05 PM by Rodal »

Offline kml


2 questions:

1) Should we document your present results as NULL in the Wiki or do you want us to wait?

2) Have you "motivated" the EM Drive by tapping it or vibrating it? ==> Recall TheTraveller/Shawyer's conjecture that for EM Drive to register a force, it needs to be motivated by tapping it or vibrating it

Please hold off on listing it as NULL until I post plots that have all of the test parameters on them.   A NULL result without knowing the power level, Q etc will be less useful.

I have thought about inducing vibration in a controlled way.   I did some test runs with the PA fan operating specifically as a source of vibration but it induces 200mg-f swings in force that would obscure any real signal.   A small solenoid with an adjustable function generator would be ideal.

Tags: