Author Topic: EM Drive Developments - related to space flight applications - Thread 3  (Read 1795564 times)

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
I attached with the message above the σxx stress component normal to the big base for the Yang geometry.  Attached below is the σxx stress component normal to the small base for the Yang geometry.
« Last Edit: 07/19/2015 12:37 AM by Rodal »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
EM DRIVE FORCE vs. TIME

1) The compressive force for what is believed to be Yang's EM Drive (Db=0.201m,Ds=0.1492m,L=0.24m) against the small base is larger than the compressive force against the big base.  This is contrary to what was observed for the model for rfmwguy.  One big difference is that for the geometry of rfmwguy (Db=0.280m,Ds=0.1588m,L=0.259m), with the antenna near the small base, the pressure at the big base had two large crescent shaped regions at the periphery due to the TM11 mode.  This crescent shape regions are gone for the Yang model, we suspect because of the placement of the antenna near the big base.  The antenna replaces the natural mode shape TM11  (pressure along the periphery in two crescents) with a central pressure that results in a smaller overall force.

EDIT: This reminds me of when I asked Paul March whether he ever measured a force in the opposite direction.  and he said yes, that when he placed the dielectric at the big end instead of the small end, the force was in the opposite direction (and weaker).

2) the time at which the maximum force occurs at the big base is NOT phase-shifted with respect to the time at which the maximum force occurs at the small base. This is contrary to what was observed for the model for rfmwguy.  The lack of phase-shifting indicates perhaps more of a standing wave distribution for this model with the antenna near the big base.

3) The net force consistently, all of the time, points towards the small base (positive direction). This is contrary to what was observed for the model for rfmwguy, where the force pointed towards the big base most of the cycle time, but also pointed towards the small base for a small amount of the cycle time.

4) The magnitude of the net force calculated is significantly smaller (less than 1/3) than the net force calculated for rfmwguy's dimensions.

5) The fact that the force is smaller (less than 1/3), that it is pointing in the opposite direction and that the forces at the bases are not phase-shifted, may most likely be due to the placement of the antenna near the big base or it could be due to Yang's geometry (we suspect strongly the former).

6) We naturally expect that force on the overall copper itself should sum up to zero in order to satisfy the momentum equilibrium equation implied by Maxwell's equation.   We expect that the imbalance in net force between the bases should be compensated by the electromagnetic stress on the lateral surfaces, leading to a component on the direction of the big base to result in a net overall force of zero.  This means that a net tensile force, a suction would be needed on the lateral surfaces.  This is the contrary to the geometry of rfmwguy where a presssure was needed on the lateral surfaces [Although the presence of suction sounds strange at first sight, this had already been pointed out by Greg Egan for the case of standing waves].  Greg Egan justifies this tension as follows:  "the tension on the walls due to the Coulomb force acting on the charge distribution induced by the electric field meeting the walls"



We don't have access to the electromagnetic fields at the lateral surfaces computed by Meep, in order to calculate the stress tensor at the lateral surfaces and integrate it to get the force on the lateral surface.

7) We still need to calculate the Poynting vector field for this case, to see whether it is pointing towards the small base for the antenna located near the big base. The net force imbalance between the bases (not taking into account the lateral surfaces), pointing towards the small base, would mean a reaction acceleration of the EM Drive in the opposite direction, towards the big base, as a result of a recoil force. 

8) A fitted model of the time variation of the force (with excellent R^2 = 0.999353), shows that the present Finite Difference model (from which the force has been computed at the last two cycles ending at 0.013 microseconds from the time at which the Microwave feed was turned on), would have to be marched forward for 1,000 times longer, to a total of 10 microseconds, for the force to be magnified by the calculated exponential growth to a value of 10 microNewtons (for an inputPower of 43 watts).   Given the fact that the present Meep model takes an hour to run on a good PC modern computer, 1,000 hours of computer time represents over 41 days of computing time.  Thus running the Meep model to steady state is impractical.  Rather than using a supercomputer to perform such a computation, I suggest to use an implicit (unconditionally stable) Finite Difference model in time (rather than the explicit time difference model presently used that is subject to stability problems that limit the maximum finite difference time step).  Such implicit finite difference models are well known (I developed a version of them in my PhD thesis 35 years ago) and can be run much faster than explicit FD models.  There are also numerous alternative numerical schemes that are more accurate than Finite Differences.

__________________________

Quote from: aer

NSF-1701 - 245x261x261
Yang-Shell - 229x196x196

This is the final summary output from the log file.

run 0 finished at t = 13.054 (6527 timesteps)

Total number of slices 14, the last 14 of 32 full cycles, or periods at 0.1 period intervals. That is, at 30.7, 30.8 and so forth to 32.0 periods of the drive center frequency.
Number of time steps, 6527 and total meep time = 13.054 time units.

Quote from: aero
Same antenna, 58 mm in the y direction, Ez excitation.

(set! antlongx 0)                               ; direction vector of dipole antenna SI units
(set! antlongy 0.058)                           ; = 58 mm
(set! antlongz 0)
« Last Edit: 07/20/2015 08:10 PM by Rodal »

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
What about the momentum back-reaction on the antenna feed?

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
Radiation reaction will occur as a result of antenna action.

Online SeeShells

  • Senior Member
  • *****
  • Posts: 2324
  • Every action there's a reaction we try to grasp.
  • United States
  • Liked: 2952
  • Likes Given: 2586
Could you clarify why in the Wiki page it shows NWPU Prof. Juan Yang's test showed TE012 mode and in this you state mode TM11?

Nice piece of work Aero and Dr. Rodal!!!
« Last Edit: 07/19/2015 01:18 AM by SeeShells »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
Could you clarify why in the Wiki page it shows NWPU Prof. Juan Yang's test showed TE012 mode and in this you state mode TM11?

Nice piece of work Aero and Dr. Rodal!!!
Was the antenna placed to excite a TM (transverse magnetic) mode instead of trying to excite a TE (transverse electric mode)?


I'm not sure about the M here, I'm pretty sure about the 11


My understanding is that the antenna is identical to rfmwguy except the placement

Quote from: aero
Same antenna, 58 mm in the y direction, Ez excitation.

(set! antlongx 0)                               ; direction vector of dipole antenna SI units
(set! antlongy 0.058)                           ; = 58 mm
(set! antlongz 0)


Many modes nearby, which mode you excite has a lot to do with the antenna placement.

« Last Edit: 07/19/2015 02:04 AM by Rodal »

Online SeeShells

  • Senior Member
  • *****
  • Posts: 2324
  • Every action there's a reaction we try to grasp.
  • United States
  • Liked: 2952
  • Likes Given: 2586
I did want to see the actions in a parallel situation Exciting a TM mode on the large endplate.

We tried to do a TE mode for the small endplate (need to refresh) but it wasn't quite what we wanted the TM worked well for the small.

Tm looks like a bust for the large end, I suspected it but you even stated it wasn't going to be quite what we were looking for. Needed to test anyway.

So it looks like the final will be the TE012 dipole vertical centered Yang's dim on the large plate and compare the TM and TE modes and stress for both.



Offline tidux

  • Member
  • Posts: 11
  • Sol III
  • Liked: 6
  • Likes Given: 3

Attempting to follow this on Ubunto, which is, I think, Debian jessie (8.1).

I don't wind up with a libhdf5_openmpi.so, but rather x86_64-linux-gnu/libhdf5.so -> libhdf5.so.7.0.0

There's no directory /usr/include/hdf5/openmpi, but there is /usr/include/openmpi.  The hdf5* files are in /usr/include.

meep-1.3 requires libctl version 3.2 or later, but version 3.1 got installed.

These don't seem insurmountable, but it sounds like I am not driving against the same repository you are.  Can you give me a pointer as to how to update my repository list to match yours for, e.g., meep-openmpi ?

Thanks

Ubuntu stopped being a straight rebuild of Debian a while ago, and usually corresponds to frozen snapshots of the unstable repository, rather than a 1:1 match of stable releases like CentOS and Red Hat Enterprise Linux.  You can search for what versions of a package are in a given version of Ubuntu at http://packages.ubuntu.com/

It looks like 14.04 was the last version with the old version of libctl, so unless you specifically need an LTS build I'd suggest updating to 15.04 (current supported non-LTS).  Ubuntu 15.10 will have meep 1.3 packaged as well, but that won't hit a stable release until early October.  Otherwise, you'll have to build the newer libctl from source yourself too.

When it comes to the CPPFLAGS environment variable, make sure that's set to wherever the hdf5 openmpi .h files are, (you should be able to find them with "dpkg -L libhdf5-openmpi-dev | grep .h\$").  The -I just tells the C Preprocessor (hence CPP) that the directory contains header files.  For the symlink, just make sure the libhdf5_openmpi.so.something.something with the longest version string is linked to libhdf5.so.  The specific version doesn't matter, this is just a kludge to get the configure script to find -lhdf5.

UPDATE: it looks like for the NSF-1701 meep job, I can get it done in about 45 minutes with 8 threads.  12 threads definitely seems to be overkill.
« Last Edit: 07/19/2015 02:17 AM by tidux »

Online aero

  • Senior Member
  • *****
  • Posts: 2735
  • 92129
  • Liked: 704
  • Likes Given: 235
@Rodal.
Quote
the Finite Difference mesh is coarser for the smaller big diameter of the Yang model (0.201m) than the big diameter of rfmwguy's model (0.2794 m): a less precise FD mesh with less nodes in the cross-section.  The actual physical distance between nodes was kept practically the same.  But what matters for a numerical discretization of a partial differential equation is the number of nodes to characterize the fields in the solution to the partial differential equation.  For the same natural frequency mode shape it would be best to keep the same mesh regardless of the actual physical size. ]

I don't know why you say that. It is not correct. The actual Meep pixel separation is identical at 0.004 and the time step is identical at 0.002 between the Yang-Shell model and rfmwguy's NSF-1701 model. I use the same code, setting a switch to select the Yang-Shell model dimensions and antenna location. When creating the lattice, the control file is designed to use only a large enough lattice to include the model and a fixed space around the model. Because the Yang-Shell frustum has a significantly smaller big base diameter, the lattice is significantly smaller in the y and z coordinate directions, but the step size and node separation within the cavity is identical between the two models. The drive center frequency and noise bandwidth is identical between models so there are an identical number of nodes per wavelength and an identical number of time steps per period. The mesh is NOT courser. It is identical. The difference is that the Yang-Shell frustum is smaller, so the lattice surrounding the frustum is smaller.
Retired, working interesting problems

Offline rfmwguy

  • EmDrive Builder (retired)
  • Senior Member
  • *****
  • Posts: 2165
  • Liked: 2681
  • Likes Given: 1124
NSF-1701 thanks everyone!

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
@Rodal.
Quote
the Finite Difference mesh is coarser for the smaller big diameter of the Yang model (0.201m) than the big diameter of rfmwguy's model (0.2794 m): a less precise FD mesh with less nodes in the cross-section.  The actual physical distance between nodes was kept practically the same.  But what matters for a numerical discretization of a partial differential equation is the number of nodes to characterize the fields in the solution to the partial differential equation.  For the same natural frequency mode shape it would be best to keep the same mesh regardless of the actual physical size. ]

I don't know why you say that. It is not correct. The actual Meep pixel separation is identical at 0.004 and the time step is identical at 0.002 between the Yang-Shell model and rfmwguy's NSF-1701 model. I use the same code, setting a switch to select the Yang-Shell model dimensions and antenna location. When creating the lattice, the control file is designed to use only a large enough lattice to include the model and a fixed space around the model. Because the Yang-Shell frustum has a significantly smaller big base diameter, the lattice is significantly smaller in the y and z coordinate directions, but the step size and node separation within the cavity is identical between the two models. The drive center frequency and noise bandwidth is identical between models so there are an identical number of nodes per wavelength and an identical number of time steps per period. The mesh is NOT courser. It is identical. The difference is that the Yang-Shell frustum is smaller, so the lattice surrounding the frustum is smaller.

I see what you are saying that you expected perhaps to excite a lower mode with less of a field variation.  That you expected perhaps to excite a 01 mode instead of a 11 mode.  But one doesn't know the solution ahead of time, otherwise one would not be using a numerical solution to solve the problem.  In this case the indication (the two crescent shapes characteristic of mode 11) is that mode 11 was excited, not 01.  Even if you expected to excite m=0 (constant in the circumferential direction), TE01 still has n=1 requiring the same discretization in the diameter direction as TM11 for example, they both have n=1.

The FD mesh you used is coarser:

NSF-1701 - 245x261x261
Yang-Shell - 229x196x196


245 is larger than 229
261 is larger than 196
261 is larger than 196

Multiplying all the numbers, the mesh is 1.89 times coarser .   The mesh for RFMWGUY had 1.89 times more nodes.  Using the same FD operator.  That's what matters.

The FD mesh is the number of nodes per side.  That's it.

<<The actual Meep pixel separation is identical at 0.004 >><< node separation within the cavity is identical between the two models.>> yes, that's what I said.  That's the physical distance between the nodes (in meters) .

What matters is the discretization of the fields.

A smaller EM Drive, operating at the same mode shape at higher frequency should require the same mesh discretization not the same physical distance between the nodes. If one carries this to the absurdum: if one insists on the separation between nodes in meters as being significant, you would be saying that something smaller than the distance between the nodes would require no nodes to be modeled.

To model something that is nanometers long still requires a fine mesh if it has strong field variation.

Partial differential equations are solved by the FD method at the FD nodes.  (Keeping the same type of FD operator, for example for a central-difference operator being used in both cases) The number of nodes is what matters.
« Last Edit: 07/19/2015 03:12 AM by Rodal »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
...
Regarding SeeShell's question?

Was the same antenna dimensions and orientation and Io=1 amp used for Yang-Shell as for RFMWGUY NSF-1701 ? Was all you changed (regarding the antenna) the location to be near the big end?
« Last Edit: 07/19/2015 02:58 AM by Rodal »

Offline TheTraveller

Ref TheTraveller Post #4570

I don't know your critical specs, so I don't know if these are 'better' or not:

Insulated Wire:   http://www.iw-microwave.com/cable_specs

The cable series numbers are their nominal diameter in mils.  To get specs for a specific cable just click on the cable number.  The 280 series provides around .25 dB/m @ 2.5 GHz and will handle around 1 kw, for example.

IW brags about their low insertion loss but the ones I have used tend to be relatively stiff for a given diameter.

Or

W. L. Gore:  http://tools.gore.com/gmcacalc/#/

The Gore link is to their cable calculator, which provides specs for connectorized cables of the length specified at the freq of interest.  ( .32 db/1 m @ 2.5 GHz, with a power rating of 1532 watts, for example.)

Gore is known for extreme flexibility, low insertion loss, good VSWR, and tolerance for small radius bends  They also define the term 'expensive cables'.

Thanks for that.

The EcoFlex15PLUS rates at 0.15dB loss per mtr at 2.5GHz and can handle 350W. My max is 100W. Price from one supplier is around $15/mtr 10mtr min. Very doable.
"As for me, I am tormented with an everlasting itch for things remote. I love to sail forbidden seas.
Herman Melville, Moby Dick

Offline TheTraveller

I really don't see the point of the posts describing again and again how shorter travel times would be with a working EMDrive tech. Everyone knows wonderful things could be achieved. There is no scientific value in such "projections".

Eagleworks' Dr. White was the 1st to show a series of planetary journey times, based on 0.4N/kW and 4N/kW propollentless drives. They can be found here by scrolling down a bit.

http://emdrive.wiki/Potential_EMDrive_solar_system_explorer_ship

I then did a few journeys at 1N/kW and for that action a few folks here want to tar & feather me.

Go figure?

Guys if you have an issue with those examples, take it up with Dr. White.
"As for me, I am tormented with an everlasting itch for things remote. I love to sail forbidden seas.
Herman Melville, Moby Dick

Offline rfmwguy

  • EmDrive Builder (retired)
  • Senior Member
  • *****
  • Posts: 2165
  • Liked: 2681
  • Likes Given: 1124
I really don't see the point of the posts describing again and again how shorter travel times would be with a working EMDrive tech. Everyone knows wonderful things could be achieved. There is no scientific value in such "projections".

Eagleworks' Dr. White was the 1st to show a series of planetary journey times, based on 0.4N/kW and 4N/kW propollentless drives. They can be found here by scrolling down a bit.

http://emdrive.wiki/Potential_EMDrive_solar_system_explorer_ship

I then did a few journeys at 1N/kW and for that action a few folks here want to tar & feather me.

Go figure?

Guys if you have an issue with those examples, take it up with Dr. White.
I think we all cringed at whites quick assumptions. So did nasa perhaps. First things first...more tests...more builds...more data...more theory...rinse and repeat.

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
Yes, we cringed.

Offline D_Dom

  • Global Moderator
  • Full Member
  • *****
  • Posts: 456
  • Liked: 154
  • Likes Given: 105
I'm excited.

Found Prof Yang has achieved my long term goal of 1N/kW as attached.


I am happy to read about progress in any effort. Thanks to all who share their knowledge!
Space is not merely a matter of life or death, it is considerably more important than that!

Offline TheTraveller

Yes, we cringed.

Experimenters can do as much static testing as they wish but it will not reveal the very important dynamic characteristics such as the rate that Power is consumed to the rate of Acceleration and Velocity increase.

So far there is only one published data set of Power usage versus Velocity increase and it makes me cringe because it shows Power consumed dropping as Velocity increases. I think I understand why that happened but need to confirm it from my test data.

Which is another reason I decided to abandon doing Static testing as I want to explore the Power to Acceleration & Velocity relationships as if there is any new physics, that is where it may show up.

Doing such dynamic testing is not easy, quick nor simple as you need to be able to dynamically measure and record Rf amp power usage, velocity, rate of angular acceleration, cavity bandwidth change, VSWR, tracking frequency change, cavity temperature and internal pressure to name a few.

Add to that Raspberry Pi 2B software engineering development load, having the control requirements to continually track best cavity resonant frequency and keep the Rf gen right in the middle of the cavity bandwidth plus developing monitoring and data logging and the PC side software. Overall a major IT project in itself.

Static testing is so much easier to do but in reality produces little useful data and the data it does produce may be subject to scale compression ability versus the magnetron duty cycle and rep rate per second.

In my opinion any serious EMDrive experimenter should be doing rotary testing as that is where the sweet meat is, despite it needing around 50x times the man hours to develop, debug and get working correctly. I have great respect for the EMC issues that will need to be solved to get quite baselines and really good signal to noise ratios. Just look at all the time Paul March puts into getting good signal to noise ratio data and his test rig was static.

Bottom line is good testing is not simple, quick or easy. It is bloody difficult, especially when you have NSF looking over your shoulder. ;) Which is all good!
« Last Edit: 07/19/2015 07:20 AM by TheTraveller »
"As for me, I am tormented with an everlasting itch for things remote. I love to sail forbidden seas.
Herman Melville, Moby Dick

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5831
  • USA
  • Liked: 5897
  • Likes Given: 5245
According to this graph (original by Roger Shawyer):




Yang's EM Drive should exhibit a force in the opposite direction to the force direction exhibited by NASA's truncated cone.  This is in agreement with these Meep/Wolfram Mathematica results ( http://forum.nasaspaceflight.com/index.php?topic=37642.msg1406309#msg1406309 ) , as the geometry of RFMWGUY has the same diameters as NASA's truncated cone, and the Meep/Wolfram Mathematica results show the force to be in opposite direction between Yang-Shell and NASA-rfmwguy.

It was a very good idea to place the antenna in the opposite direction, this time next to the big base, to see its effect. 

Very different results were obtained as a result of these changes.

One learns a lot from big changes in output.

Many changes were made at once in the input: different geometry (Yang-Shell vs NASA-rfmwguy) and antenna placement (near the big base and small base respectively) so we are not sure what is responsible for the different results.  Now, if only one change is made at once in the input (either just the placement of the antenna: for example same Yang-Shell with antenna at small base or for example NASA-rfmwguy with antenna placed at big end) we will learn what change in input was responsible for the big change in output (force in opposite direction).
« Last Edit: 07/19/2015 08:47 AM by Rodal »

Offline graybeardsyseng

  • Full Member
  • *
  • Posts: 114
  • Texas, USA
  • Liked: 159
  • Likes Given: 636
Ref TheTraveller Post #4570

I don't know your critical specs, so I don't know if these are 'better' or not:

Insulated Wire:   http://www.iw-microwave.com/cable_specs

The cable series numbers are their nominal diameter in mils.  To get specs for a specific cable just click on the cable number.  The 280 series provides around .25 dB/m @ 2.5 GHz and will handle around 1 kw, for example.

IW brags about their low insertion loss but the ones I have used tend to be relatively stiff for a given diameter.

Or

W. L. Gore:  http://tools.gore.com/gmcacalc/#/

The Gore link is to their cable calculator, which provides specs for connectorized cables of the length specified at the freq of interest.  ( .32 db/1 m @ 2.5 GHz, with a power rating of 1532 watts, for example.)

Gore is known for extreme flexibility, low insertion loss, good VSWR, and tolerance for small radius bends  They also define the term 'expensive cables'.

Thanks for that.

The EcoFlex15PLUS rates at 0.15dB loss per mtr at 2.5GHz and can handle 350W. My max is 100W. Price from one supplier is around $15/mtr 10mtr min. Very doable.

You might want to check Universal Radio:
http://www.universal-radio.com/catalog/cable/2499.html

They seem to have -15plus for about $8.2/meter (they quote per foot) and IIRC their min is 10 feet.   Pretty good prices on N connectors for the EcoFlex15Plus (around $13.00 each).   No connection to them - have bought other items from them in the past.   Another place with good prices (especially if you can order without VAT) is
http://www.thedxshop.com/ecoflex/ecoflex-15-plus-cables-connectors.html
Never have used them.

BTW -  Traveller, I'm sure you are well aware of this, but for the others out there who might be using some EcoFlex15 or similar -connections are absolutely critical, especially for hi power hi freq RF. 

If one is not quite comfortable attaching these specialty connectors (or any coax for that matter) - buy some extra and practice.   It is quite possible to get a connection which looks good and even feels secure but in fact has a lot of loss.   A poorly executed connector installation can add several dB of loss or even an open circuit.  Is it worth "several" x $13.00 for practice - YES !  Poor connections can not only totally screw up your results but potentially damage your RF source.

Also, I agree completely with rq3 on reply 4619 - in these ranges, if possible avoid coax for hi power and use direct coupling or short waveguides.
EMdrive - finally - microwaves are good for something other than heating ramen noodles and leftover pizza ;-)

Tags: