Author Topic: EM Drive Developments - related to space flight applications - Thread 4  (Read 1245521 times)

Offline X_RaY

  • Full Member
  • ****
  • Posts: 852
  • Germany
  • Liked: 1146
  • Likes Given: 2479
Here are the two png sets scaled to the maximum range of the E and H fields for the t=0 time slice. The ranges are from field components as follows:

# ez all data range from -0.000245405 to 0.000245405.     E field big end
# hx all data range from -0.000538633 to 0.000538633.     H field big end
# ez all data range from -4.04811e-05 to 4.04811e-05.     E field small end
# hy all data range from -0.000287974 to 0.000287974.     H field small end


Please look these png's over closely to be sure that this is what you want to see as it will be considerable effort to scale the png files in this way for the general case.

Or do you need to see the side views for this t = 0 case?

Thank you for going through the trouble to do this.

1) Assuming that the longitudinal axis is the z axis, this looks indeed like a TE mode (transverse electric) as the Hz mode (magnetic field in the longitudinal direction) is strong and the Ez (electric field in the longitudinal direction) is weak

however...

2) The blue areas in the fields: they look like the magnitude has been clipped. 



They look like ContourPlots in Mathematica when I use PlotRange with a range smaller than the full magnitude, so the contours get cliped.  I would double check whether magnitudes have not been clipped. 

3) Is there some means to control the number of contour values in the contour plots?  It looks to me like there are more contour levels than the number of colors available, as a result colors keep repeating themselves.  What one wants is to have blue represent the lowest magnitude and red the highest magnitude, instead of red/green/blue/yellow patterns being repeated cyclically.


ex all data range from -3.67269e-05 to 3.67269e-05.
ey all data range from -1.42517e-05 to 1.42517e-05.
ez all data range from -0.000245405 to 0.000245405.


Of course they have been clipped.  Look at the data ranges for the three E field components. Anything below the data range (or above, but that won't happen) are well ....from the manual:  "Data values below or above this range will be treated as if they were min or max respectively." I suppose I could use the range from the minimum of the field  to the maximum. Not minimum/ maximum of the most energetic component. To do so would be only a little more difficult, but would that be better? I can't guess without looking.

I have a silly question. If the data values are clipped why am I still seeing artifacts on the side walls?

shell
Maybe artefacts from the grid but there are strong currents in the wall and corresponding fields...
« Last Edit: 08/11/2015 08:18 pm by X_RaY »

Offline aero

  • Senior Member
  • *****
  • Posts: 3629
  • 92129
  • Liked: 1146
  • Likes Given: 360
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.
Retired, working interesting problems

Offline X_RaY

  • Full Member
  • ****
  • Posts: 852
  • Germany
  • Liked: 1146
  • Likes Given: 2479
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.
Today (morning) someone pointed out that it may be helpful to use a logarithmic scale, did you try that in the actual case? (one try to compare with the mag. scale?)
Other color map for the mag. scale like you sad could be also helpful i think.

BTW: great work, thanks for spend your time!
« Last Edit: 08/11/2015 08:48 pm by X_RaY »

Offline aero

  • Senior Member
  • *****
  • Posts: 3629
  • 92129
  • Liked: 1146
  • Likes Given: 360
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.

bluered (opaque blue to transparent white to opaque red). I have several color maps available, this is just the first try. I personally don't like it very much.
Retired, working interesting problems

Offline aero

  • Senior Member
  • *****
  • Posts: 3629
  • 92129
  • Liked: 1146
  • Likes Given: 360
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.
Today (morning) someone pointed out that it may be helpful to use a logarithmic scale, did you try that in the actual case? (one try to compare with the mag. scale?)
Other color map for the mag. scale like you sad could be also helpful i think.

BTW: great work, thanks for spend your time!
I saw that and frankly it gave me a nice chuckle. Two points on a log scale?

See: http://ab-initio.mit.edu/h5utils/h5topng-man.html Look really, really closely for the log scale output, then tell me what you find.
Or maybe it would be more fruitful to look here: http://www.hdfgroup.org/products/java/hdfview/UsersGuide/ug05spreadsheet.html#ug05subset_transpose

My point is that logarithmic seems to be an alien concept to meep and it's support software.
Retired, working interesting problems

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5911
  • USA
  • Liked: 6124
  • Likes Given: 5564
....

bluered (opaque blue to transparent white to opaque red). I have several color maps available, this is just the first try. I personally don't like it very much.

This is much better.

I would definitely use this from now on.

This way we can tell what is high amplitude and what is low amplitude.  The previous way....just say no good

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5911
  • USA
  • Liked: 6124
  • Likes Given: 5564
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.
Today (morning) someone pointed out that it may be helpful to use a logarithmic scale, did you try that in the actual case? (one try to compare with the mag. scale?)
Other color map for the mag. scale like you sad could be also helpful i think.

BTW: great work, thanks for spend your time!
I saw that and frankly it gave me a nice chuckle. Two points on a log scale?

See: http://ab-initio.mit.edu/h5utils/h5topng-man.html Look really, really closely for the log scale output, then tell me what you find.
Or maybe it would be more fruitful to look here: http://www.hdfgroup.org/products/java/hdfview/UsersGuide/ug05spreadsheet.html#ug05subset_transpose

My point is that logarithmic seems to be an alien concept to meep and it's support software.
For very good reasons: IMHO a logarithmic scale is the complete opposite of what one wants to do to display mode shapes for these purposes: it will increase the number of contours in the low range and it will decrease them in the high range.

Quote
A logarithmic scale is a nonlinear scale used when there is a large range of quantities. Common uses include the earthquake strength, sound loudness, light intensity, and pH of solutions.

It is based on orders of magnitude, rather than a standard linear scale, so each mark on the scale is the previous mark multiplied by a value.

It will distort all the shapes.  Logarithmic scales are suitable when you have an exponential increasing behavior.  Not suitable for the fields.

May be suitable to plot attenuation, not for discriminating mode shapes from the electromagnetic fields.

Suitable for earthquake strength, sound loudness

« Last Edit: 08/11/2015 09:25 pm by Rodal »

Offline X_RaY

  • Full Member
  • ****
  • Posts: 852
  • Germany
  • Liked: 1146
  • Likes Given: 2479
Dr. Rodal,

I was trying to figure out a way to communicate what I thought was going on, but I think I was wrong. It's simple -

Draw 3 sine waves on a sheet of paper, x-y axis, each symmetric about the x axis, but with different amplitudes. Then draw horizontal lines tangent to the extreme positive and negative values of each sine wave. I used the y value of the top and bottom horizontal line.

So the weaker sign waves are not clipped, they just don't have enough power to show up on the scale of the strongest.

I'll see about using a different color map.
Today (morning) someone pointed out that it may be helpful to use a logarithmic scale, did you try that in the actual case? (one try to compare with the mag. scale?)
Other color map for the mag. scale like you sad could be also helpful i think.

BTW: great work, thanks for spend your time!
I saw that and frankly it gave me a nice chuckle. Two points on a log scale?

See: http://ab-initio.mit.edu/h5utils/h5topng-man.html Look really, really closely for the log scale output, then tell me what you find.
Or maybe it would be more fruitful to look here: http://www.hdfgroup.org/products/java/hdfview/UsersGuide/ug05spreadsheet.html#ug05subset_transpose

My point is that logarithmic seems to be an alien concept to meep and it's support software.
For very good reasons: a logarithmic scale is the opposite of what one wants to do to display mode shapes for these purposes: it will increase the number of contours in the low range and it will decreased them in the high range.

Quote
A logarithmic scale is a nonlinear scale used when there is a large range of quantities. Common uses include the earthquake strength, sound loudness, light intensity, and pH of solutions.

It is based on orders of magnitude, rather than a standard linear scale, so each mark on the scale is the previous mark multiplied by a value.

It will distort all the shapes.  Logarithmic scales are suitable when you have an exponential increasing behavior.  Not suitable for the fields.
May be suitable to plot attenuation, not for the fields
OK i can follow your explanation (high vs. low field strength), but i am not sure in view of the resulting pictures. It would be soften the edges we look since the rescaling.
You are the expert here, if you think i isn't helpful then forget it, no problem with that.
And yes it's much better with the new color map!!
« Last Edit: 08/11/2015 09:24 pm by X_RaY »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5911
  • USA
  • Liked: 6124
  • Likes Given: 5564
...
OK i can follow your explanation (high vs. low field strength), but i am not sure in view of the resulting pictures. It would be soften the edges we look since the rescaling.
You are the expert here, if you think i isn't helpful then forget it, no problem with that.
And yes it's much better with the new color map!!
Well it is just my opinion of course. 
Let's clarify it with an example:  a logarithmic scale is eminently suited to rank earthquakes because each level changes by an order of magnitude.  The difference between an earthquake with a level of 9 from a level of 8 is much greater than the difference between an earthquake between a level of 5 and a level of 4.
Ditto for sound.

Both for sound and earthquakes it makes sense to have a scale that has more levels at low magnitude and less levels at high magnitude.

Now, for the electromagnetic fields, we do not have a scale of electromagnetism where we want to minimize the number of levels at high magnitude and maximize the number of levels at low magnitude.

A linear scale for the mode shapes in EM Drive cavities is fine because we don't have singularities.

A logarithmic scale would make sense for this kind of shape: see the attachment for attenuation where the attenuation goes to Infinity at each end, the plot is clipped at the ends because a linear scale cannot show the singularities at each end.
« Last Edit: 08/11/2015 09:36 pm by Rodal »

Offline X_RaY

  • Full Member
  • ****
  • Posts: 852
  • Germany
  • Liked: 1146
  • Likes Given: 2479
...
OK i can follow your explanation (high vs. low field strength), but i am not sure in view of the resulting pictures. It would be soften the edges we look since the rescaling.
You are the expert here, if you think i isn't helpful then forget it, no problem with that.
And yes it's much better with the new color map!!
Well it is just my opinion of course. 
Let's clarify it with an example:  a logarithmic scale is eminently suited to rank earthquakes because each level changes by an order of magnitude.  The difference between an earthquake with a level of 9 from a level of 8 is much greater than the difference between an earthquake between a level of 5 and a level of 4.
Ditto for sound.

Both for sound and earthquakes it makes sense to have a scale that has more levels at low magnitude and less levels at high magnitude.

Now, for the electromagnetic fields, we do not have a scale of electromagnetism where we want to minimize the number of levels at high magnitude and maximize the number of levels at low magnitude.

A linear scale for the mode shapes in EM Drive cavities is fine because we don't have singularities.

A logarithmic scale would make sense for this kind of shape: see the attachment for attenuation where the attenuation goes to Infinity at each end, the plot is clipped at the ends because a linear scale cannot show the singularities at each end.
Yes you should be right, its a little bi t late and i am tired today.
I just take a look of a random sonnetEM plot  with both scalings... have to agree.
Wish a good night (later for you).

Offline WarpTech

  • Full Member
  • ****
  • Posts: 1407
  • Do it!
  • Statesville, NC
  • Liked: 1453
  • Likes Given: 1925
Ok since the photons inside the cavity are traveling in a medium at a group velocity <c, we can technically define a frame where the photon is at rest relative to the cavity walls. I can't justify a reason to not treat them as massive particles in calculations.

This is a special case since we're not dealing with photons in vacuum. Does this sound controversial to anyone? That means switching to relativistic momentum.

http://hyperphysics.phy-astr.gsu.edu/hbase/relativ/relmom.html

This ref was shared by someone else recently and I think it is significant.

http://arxiv.org/abs/0708.3519

I have found this paper useful and the mathematics makes sense. The relativistic treatment of photons in a waveguide gives them a rest mass, that is proportional to the cut-off frequency of the waveguide.

I've used this in a tapered cylinder waveguide approximation to derive the thrust equation. I've just been too busy to work on my paper for the past couple of weeks.
Todd

Offline jmossman

  • Member
  • Posts: 73
  • San Jose, CA
  • Liked: 58
  • Likes Given: 180
My limited experience with the comsol EM module is it works fine!
Without calculations it is difficult(not possible) to say a specific peak in the S-parameter plot is the target resonance.

Sure understand that. But if you do a scan on a cavity and the predicted resonance is not there, then what?

As far as I know, there were no scans to back up all the mode and freq calculations that EW did.
I have a simple VNA I plan to use to find resonance frequencies and it's simply a verification to the calculations.
...
Even NASA should have used VNAs in air and vacuum just to check and I'm not sure they did. 
...

I'd like to propose a logical progression to the VNA analysischaracterize each experimental frustum at multiple temperatures

@deltaMass recently proposed that thermal data could be used as an approximation to how well tuned the frustum was to the source RF feed.  To continue the line of thought, if one could characterize a frustum at discrete temperatures using a VNA , then one could theoretically build a control loop using frustum temperature that varied frequency and/or impedance controls without having to perform real-time VSWR or sample port analysis.

Effectively I'm proposing that a lookup table could be constructed that indicates the optimal experimental control settings for each temperature point.  As the frustum heats up, the optimal experimental control settings can be obtained from the table.  As an implementation detail, one could use curve-fitting to build an equation or use simplistic "if-then-else" to drive the control loop outputs.

While the temperature is likely to be the dominant frustum parameter than impact bandpass characteristics and impedance, other parameters may also be important to in order to implement a control loop that yields repeatable results (i.e. frustum orientation, humidity, etc).

EDIT: On further thought, one might need multiple temperature inputs from different areas on the frustum geometry.  The back-of-napkin reasoning being that temperature differences at certain key locations may have a more pronounced effect on bandpass characteristics and impedance than a single "average" frustum temperature.  I'd guess that the temperature of the "big end" and "small end" are the two dominant parameters.
« Last Edit: 08/11/2015 11:00 pm by jmossman »

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
I'm still confused. Are we looking at a single impulsive event, or a steady-state force?
The unraveling of the EM Drive "force measurement" this is what is great about replications.
The South African experiment does not show a constant steady-state force.

His fulcrum is undamped and will oscillate for some time before settling at the final value. You know that so why the comment?
Your Quote link to Rodal's comment (about "unraveling") does not work. Can you please provide a working link?

Offline rfmwguy

  • EmDrive Builder (retired)
  • Senior Member
  • *****
  • Posts: 2205
  • Liked: 2713
  • Likes Given: 1134
About PaulTheSwag's experiment:
1. What differentiates the thrust profile from one where only an initial impulse was produced?
2. What is a "double knife-edge fulcrum"? It makes no sense to me.
3. Do we know the power value?

2. Did you saw the pictures? http://imgur.com/a/iO7er
OK. I now understand the double knife edge (actually triple).
Same as mine, 3 blades, 2 balance points.

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5911
  • USA
  • Liked: 6124
  • Likes Given: 5564
About PaulTheSwag's experiment:
1. What differentiates the thrust profile from one where only an initial impulse was produced?
2. What is a "double knife-edge fulcrum"? It makes no sense to me.
3. Do we know the power value?

2. Did you saw the pictures? http://imgur.com/a/iO7er
OK. I now understand the double knife edge (actually triple).
Same as mine, 3 blades, 2 balance points.

Offline frobnicat

  • Full Member
  • ****
  • Posts: 518
  • Liked: 500
  • Likes Given: 151
How about my other two questions - the power? and why not just an impulse?

An impulse would give an undamped mass-spring* harmonic oscillator an initial velocity to the mass but would not change the equilibrium point : the oscillations would have a mean position around the same equilibrium (before excitation). This is not what we see, even if the time axis lacks of resolution it looks like the response to a step, the mean after excitation is above the pre-excitation position. Too bad the record is too short to see what happens after power-off (why ?) : the undamped oscillator should continue to oscillate, more or less depending on phase at power off, but back around the initial equilibrium point, at least if there is a steep step down back to 0 force at power-off (which somehow often proves elusive with the emdrive experiments).

* in this case the "springiness" is actually a centre of mass below axis (by what amount ?) restoring torque, but this is the same dynamic harmonic oscillator for small linearisable displacements.

Offline SeeShells

  • Senior Member
  • *****
  • Posts: 2442
  • Every action there's a reaction we try to grasp.
  • United States
  • Liked: 3186
  • Likes Given: 2708
Ok since the photons inside the cavity are traveling in a medium at a group velocity <c, we can technically define a frame where the photon is at rest relative to the cavity walls. I can't justify a reason to not treat them as massive particles in calculations.

This is a special case since we're not dealing with photons in vacuum. Does this sound controversial to anyone? That means switching to relativistic momentum.

http://hyperphysics.phy-astr.gsu.edu/hbase/relativ/relmom.html

This ref was shared by someone else recently and I think it is significant.

http://arxiv.org/abs/0708.3519

What do you think happens at the height of a mode switch, right at the phase change peak. Time freezes.

Kidding aside, I want to read those papers, but later tonight. Thanks.

Shell


Offline SeeShells

  • Senior Member
  • *****
  • Posts: 2442
  • Every action there's a reaction we try to grasp.
  • United States
  • Liked: 3186
  • Likes Given: 2708
About PaulTheSwag's experiment:
1. What differentiates the thrust profile from one where only an initial impulse was produced?
2. What is a "double knife-edge fulcrum"? It makes no sense to me.
3. Do we know the power value?

2. Did you saw the pictures? http://imgur.com/a/iO7er
OK. I now understand the double knife edge (actually triple).
Same as mine, 3 blades, 2 balance points.

Two balance points for me.

Offline rfmwguy

  • EmDrive Builder (retired)
  • Senior Member
  • *****
  • Posts: 2205
  • Liked: 2713
  • Likes Given: 1134
NSF-1701 Update 8/11/15. Assembly of frustum complete. Final weigh-in 2.383 kg. (grill) Thermometer installed, thanks to an idea from a pal here on NSF. Electrodes (solid copper wire) and terminal strip attached. Frustum is pictured upside down, as test will have small end closest to floor.

Final assembly on microwave (oven) controller just about done. Terminal strip on back and all wired up. Outside case reinstalled, fan and turntable motor disabled. All that's left is to attach copper wire and galinstan copper cups.

Just about at the end of the build phase!!!!!!!!!!!

Offline lmbfan

  • Member
  • Posts: 55
  • Phoenix
  • Liked: 42
  • Likes Given: 42
That looks very promising.
No, the -t is what it is, the -x -y and -z are parameters, (one set per png file). But it looks like the shell file will need to undergo major changes. I've attached my shell file (change the extension to .sh). It works for what I have been doing but I am forced by h5topng to copy the .png files into the sub-folders from the -out folder. The new desire to set min/max for each value of t, x, y, z will force us to discard this approach and adopt a naming convention something like I use for the csv files. That shell file is also attached (changed extension) but h5totxt handles the sub-directory correctly so no file moving is needed.

As for running it in bash? Not necessary until the h5topng commands are fully formulated and written to a shell file. You can do all of that in Python if that is your language of choice. Then run that shell file in bash.

You will likely need to create a set of h5 files to run your tests against. Don't go to to much work until you do that because I'm not sure that the single case of data that I generated and posted is representative of what h5topng outputs when multiple files are generated. It's sure not what the attached shell file logs, but then, the attached is probably not representative of the ultimate design, either.

and see the manual, it's short:
http://ab-initio.mit.edu/h5utils/h5topng-man.html

Please find attached a shell script generator file (rename from .txt to .py).  Usage:

python generate_shell.py --colormap dkbluered --z-big-end 15 --z-small-end 214 --out-dir ./output_dir --time 0:13

For now, the "-m" minimum, "-M" maximum, and "2-d-loop64-" prefix are hard coded.  You may change these manually on lines 42 and 53-57.  I have not completely coded different min/max values for the ex, ey, ez, hx, hy, hz directions, for now the e/h min/max are the same (e.g. ex min/max is the same as hx min/max).  The input files are assumed to be in out-dir and have the format of dir.h5 where dir is one of ex, ey, ez, hx, hy, hz.

The output is also attached.  Since I am on windows, the file separator is "\", however in unix it should switch to "/" automagically.

This may be useful to you for generating shell scripts using custom min/max values pending the creation of a script to extract the min/max from log files.  If you can post or PM me an actual log file, I can see about creating a new parser that can automate the whole process.  Also, maybe the "h5ls" command can extract the min/max values without having to generate a set of "throwaway" png files?

EDIT: I threw in the h5totxt for free, took about 30 sec once I had the h5topng coded.  Feel free to remove that part.  Also, the output.txt was generated using a slightly different command line than the example from above. :)
« Last Edit: 08/11/2015 11:46 pm by lmbfan »

Tags:
 

Advertisement NovaTech
Advertisement Northrop Grumman
Advertisement
Advertisement Margaritaville Beach Resort South Padre Island
Advertisement Brady Kenniston
Advertisement NextSpaceflight
Advertisement Nathan Barker Photography
0