Author Topic: EM Drive Developments - related to space flight applications - Thread 3  (Read 1805264 times)

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
A result from Test #1 of the Baby EmDrive
Some rather elementary algebra using the experimental data yields:

The apparent thrust is 15% of the frictional force and drag to an accuracy of about 2%

The frictional force is assumed constant at all rotational velocities plotted.
« Last Edit: 06/14/2015 08:43 PM by deltaMass »

Offline Prunesquallor

  • Full Member
  • *
  • Posts: 174
  • Currently, TeV Brane Resident
  • Liked: 157
  • Likes Given: 73
Data from second Baby EmDrive test

https://hackaday.io/project/5596-em-drive/log/19468-torsion-test-2-data

Using their data, I plotted maxima and minima, then averaged: ave(t) = [min(t)+max(t-1)]/2.  Thruster on/off times were interpreted from hackaday plot.

As others have said you would need to perform proper signal / time series analysis on this to be sure.
Maybe something that could be useful for a start is if to compute the average the Ys for the periods where the drive is "off" and then for the period where the drive is "on".  Is there any noticeable pattern ?

Since I already had this plotted, it was pretty easy.  And more interesting, especially the second half of the test.  Comments?
Retired, yet... not

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
I do see a half-decent anticorrelation now, on the 2nd half.
But I won't eat my hat until I see a similar set of data with the cavity reversed.
« Last Edit: 06/14/2015 08:48 PM by deltaMass »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5845
  • USA
  • Liked: 5927
  • Likes Given: 5270
I do see a half-decent anticorrelation now, on the 2nd half.
But I won't eat my hat until I see a similar set of data with the cavity reversed.

I see here a stock market that is looking pretty good right now, with a great trend, it looks like great momentum for trading:



Oh , no sorry, that's a random walk of 10,000 coin tosses   ;)
« Last Edit: 06/14/2015 09:07 PM by Rodal »

Offline deuteragenie

  • Member
  • Posts: 71
  • Germany
  • Liked: 22
  • Likes Given: 0
Data from second Baby EmDrive test

https://hackaday.io/project/5596-em-drive/log/19468-torsion-test-2-data

Using their data, I plotted maxima and minima, then averaged: ave(t) = [min(t)+max(t-1)]/2.  Thruster on/off times were interpreted from hackaday plot.

As others have said you would need to perform proper signal / time series analysis on this to be sure.
Maybe something that could be useful for a start is if to compute the average the Ys for the periods where the drive is "off" and then for the period where the drive is "on".  Is there any noticeable pattern ?

Since I already had this plotted, it was pretty easy.  And more interesting, especially the second half of the test.  Comments?

Thank you.  Here you have now a more difficult case to reject without proper data analysis, as it looks like in almost all case the average drops when the engine is "on".  I have said before that I am not a physicist nor an electrician.  Unfortunately, I am neither a statistician, so I cannot tell whether this is significant.  I would try to establish a correlation with "on", "off" and the average values for a start, but you do not have enough "on" "off" events I am afraid.  Another thing is to do some autocorrelation analysis on the TS.  R has what you need.
« Last Edit: 06/14/2015 09:07 PM by deuteragenie »

Online ThinkerX

  • Full Member
  • ***
  • Posts: 303
  • Alaska
  • Liked: 113
  • Likes Given: 59
So....

...would 'borderline net thrust' be a fair assessment for test 1 of the Baby EM drive?

...and would 'null' or 'signal lost in noise' be a fair assessment for test 2 of the Baby EM drive?

I wonder if the Yang / Fan EM Drive would produce stronger results if scaled down to Baby EM drive scale?  (longer, shallower frustum)

Offline deuteragenie

  • Member
  • Posts: 71
  • Germany
  • Liked: 22
  • Likes Given: 0
I do see a half-decent anticorrelation now, on the 2nd half.
But I won't eat my hat until I see a similar set of data with the cavity reversed.

I see here a stock market that is looking pretty good right now, with a great trend, it looks like great momentum for trading:



Oh , no sorry, that's a random walk of 10,000 coin tosses   ;)

We know that B. Mandelbrot has shown "deeply" fractal markets behave (which do not fit your coin tossing example), and why the Black-Scholes model for determining option price is broken as it relies on incorrect assumptions.  But what has that to do with demonstrating that this TS is random.  The simple averaging I suggested above shows it is not so simple...

Offline zen-in

  • Full Member
  • ****
  • Posts: 531
  • California
  • Liked: 468
  • Likes Given: 365

Since I already had this plotted, it was pretty easy.  And more interesting, especially the second half of the test.  Comments?

The only reason why it there appears to be a correlation with the RF on time is because your averages are synchronized with it.  This is called wishful thinking or biased data.  If you used a sample period that was not in synch with the RF switch when you calculated the average, they would not have the appearance of being related.   However you would still be ignoring the error limits, which are many times more than any possible signal.  This data is just background noise, nothing more. 

https://en.wikipedia.org/wiki/Bias_%28statistics%29

« Last Edit: 06/14/2015 09:32 PM by zen-in »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5845
  • USA
  • Liked: 5927
  • Likes Given: 5270
...Another thing is to do some autocorrelation analysis on the TS.  R has what you need.

I was showing that the human mind is built by Nature to see patterns, many times where there may not be any patterns.  (The face on Mars, trees on Mars, etc.)





Focusing on the Movax experiment, they should try to use oil (instead of water) to dampen the vibrations.

If calculate we must, calculate we will.  We can also do autocorrelation and power spectral density with Mathematica:

(I guess we could also try to do a cross-correlation with the on/off signal later on...)
« Last Edit: 06/14/2015 09:57 PM by Rodal »

Offline jmossman

  • Member
  • Posts: 72
  • San Jose, CA
  • Liked: 58
  • Likes Given: 173
Second test of the baby EM drive and problems with oscillation. I think they are looking for suggestions to resolve this?

https://www.youtube.com/watch?v=Y8uyIgzdzS4&feature=youtu.be

More info.

https://hackaday.io/project/5596-em-drive/log/19417-torsion-test-no-data-due-to-oscillations

My guess is the pendulum is being excited by vibrations.   Even the smallest vibration or movement of the whole apparatus will make the pendulum swing back and forth at its natural frequency.   This effect is much more noticeable than any possible thrust from the RF.   A method used by holography experimenters is to use a container filled with sand as the base, and have that container situated on a concrete floor.  The apparatus has legs that are sunk into the sand.   Any table or floor of a wood frame house will be swaying from vibrations, wind outside, etc.   This movement is coupled to the whole apparatus, making the pendulum swing.  But the effect of seismic activity, waves crashing on a distant shoreline, heavy trucks passing, construction activity, etc, etc, will still affect the measurement.   Any thrust from the RF will always be too far below the noise level to even be measured, no matter what methods are used to analyze the "data".

Another option might be to approach this in a manner similar to noise canceling microphones:  construct a second "null" apparatus (with identical design, mass distribution, etc) and then collect simultaneous data from both the "null" apparatus and the "baby EM drive" apparatus.  Unfortunately the DIY nature of their current torsion apparatus probably precludes such a technique.  If the apparatus was redesigned, running a simultaneous twin "null" for noise correlation might enable recovery of some signal-to-noise ratio.

However, I suspect there are also chaotic air currents in play, which probably won't correlate in a twin "null" configuration.  A bit more control over their testing environment is probably worthwhile.  (i.e. attempts to isolate apparatus from vibrations, air currents, etc)

I imagine Paul March could give us a few stories about noise sources in his pursuit of EM drive SNR. :)

Offline vulture4

  • Full Member
  • ****
  • Posts: 987
  • Liked: 305
  • Likes Given: 89
The possible statistical manipulations are limitless and some analyses such as power spectra will produce what appears to be systematic results from random input. In this case no systematic effect is present, and the hypothesis (a continuous force produced by the device) is not consistent with an effect that can only be detected statistically.
« Last Edit: 06/14/2015 10:02 PM by vulture4 »

Offline PaulF

  • Member
  • Posts: 52
  • Netherlands
  • Liked: 19
  • Likes Given: 9
...Another thing is to do some autocorrelation analysis on the TS.  R has what you need.

I was showing that the human mind is built by Nature to see patterns, many times where there may not be any patterns.  (The face on Mars, trees on Mars, etc.)

I believe the term for that, from the top of my head, is Pareidolia.

Quite the coincidence, two days ago I rewatched Mission to mars from 2000 for the first time since it was released on VHS at the time, lol.
« Last Edit: 06/14/2015 10:25 PM by PaulF »

Offline PaulF

  • Member
  • Posts: 52
  • Netherlands
  • Liked: 19
  • Likes Given: 9
...Another thing is to do some autocorrelation analysis on the TS.  R has what you need.
Focusing on the Movax experiment, they should try to use oil (instead of water) to dampen the vibrations.
If I may, maybe the best vibration dampener could be corn starch, as it becomes liquid when there is no force on it, but becomes stiffer when moved. I can see this is a boone especially for very small forces as it will dampen large force practically logarithmically compared to small forces.
« Last Edit: 06/14/2015 10:24 PM by PaulF »

Offline Flyby

  • Full Member
  • ***
  • Posts: 384
  • Belgium
  • Liked: 447
  • Likes Given: 47
That second test suffers heavily from the "black cat in a dark room" syndrome...

Me... I can't see anything....either I'm blind or there is nothing to see... 8)

Offline TMEubanks

Second test of the baby EM drive and problems with oscillation. I think they are looking for suggestions to resolve this?

https://www.youtube.com/watch?v=Y8uyIgzdzS4&feature=youtu.be

More info.

https://hackaday.io/project/5596-em-drive/log/19417-torsion-test-no-data-due-to-oscillations

My guess is the pendulum is being excited by vibrations.   Even the smallest vibration or movement of the whole apparatus will make the pendulum swing back and forth at its natural frequency.   This effect is much more noticeable than any possible thrust from the RF.   A method used by holography experimenters is to use a container filled with sand as the base, and have that container situated on a concrete floor.  The apparatus has legs that are sunk into the sand.   Any table or floor of a wood frame house will be swaying from vibrations, wind outside, etc.   This movement is coupled to the whole apparatus, making the pendulum swing.  But the effect of seismic activity, waves crashing on a distant shoreline, heavy trucks passing, construction activity, etc, etc, will still affect the measurement.   Any thrust from the RF will always be too far below the noise level to even be measured, no matter what methods are used to analyze the "data".

Another option might be to approach this in a manner similar to noise canceling microphones:  construct a second "null" apparatus (with identical design, mass distribution, etc) and then collect simultaneous data from both the "null" apparatus and the "baby EM drive" apparatus.  Unfortunately the DIY nature of their current torsion apparatus probably precludes such a technique.  If the apparatus was redesigned, running a simultaneous twin "null" for noise correlation might enable recovery of some signal-to-noise ratio.

However, I suspect there are also chaotic air currents in play, which probably won't correlate in a twin "null" configuration.  A bit more control over their testing environment is probably worthwhile.  (i.e. attempts to isolate apparatus from vibrations, air currents, etc)

I imagine Paul March could give us a few stories about noise sources in his pursuit of EM drive SNR. :)

May I suggest that the first thing someone with a new experimental setup should do is run tests with the drive off, and work on vibration isolation etc. until they understand what the null experiment is doing, and only then try things with the power on. If the drive could be run with the power on but no thrust being generated, I would do the sequence power off, power on, drive on, in that order. If you cannot understand the motions you are seeing in the null case, it makes no sense to try and understand the full-up test.

I know that goes against people's desire for results (and I could understand an initial test to see if the drive was so strong it's thrust was beyond doubt), but I strongly think that's the way to proceed.

Offline vulture4

  • Full Member
  • ****
  • Posts: 987
  • Liked: 305
  • Likes Given: 89
If the thruster is powerful enough to cause the device to rocket out of sight in seconds, then the noise may not be a factor. If the goal is to minimize noise and record a small signal, the best strategy is to rigidly mount to device on a load cell anchored to a vibration-isolated table. A device that is free to move will be subject to motion due to noise, air currents, both already present and thermally induced, thermal recoil, power line induced magnetic fields, vibrations from air conditioners, traffic, and distant earthquakes, and a host of other factors. The investigator's statement that the device has to be free to move or there would be no thrust is unfortunately a misinterpretation of Newton's laws.
« Last Edit: 06/14/2015 11:17 PM by vulture4 »

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5845
  • USA
  • Liked: 5927
  • Likes Given: 5270
This was plotted by KevinJPluck

https://www.dropbox.com/s/t2syuuj8mtw2h04/emdrive.xlsx?dl=0

https://hackaday.io/project/5596-em-drive/log/19468-torsion-test-2-data

who wrote:

Quote from: kevinjpluck wrote 5 minutes ago
Here's some shabby analysis.  If I saw anything I would've gone deeper! 
« Last Edit: 06/14/2015 11:08 PM by Rodal »

Offline kml

Build update:

Picked up a new Bird RF power meter today at a Hamfest and started trying to tune the cavity.   No dielectrics have been installed yet.   After some tinkering I was able to get Q=1600, Pforward=8W, Preverse=200mW @1281.825MHz, TE102.    Power was measured after isolator and before feedpoint,  Q was measured using 3db bandwidth from sampling port.   The pictures show the Q measuring configuration.

Dimensions:  x=6.5", y=3.25", z=13", Feedpoint=3.375" from closest short, feed probe = .625".

I started with a feed probe of about 1.5" and incrementally trimmed while checking Q.  I stopped at .625".  One end short is an adjustable plunger covered with copper foil.   The dielectirc will likely be mounted to the fixed end plate, which currently has copper foil sandwiched between an aluminum plate and the flange.

I have a 4"x4"x0.25" Al2O3 ceramic that needs to be cut to fit.   Looks like it's time to order the A&D MC-10K balance.




Offline TMEubanks

If the thruster is powerful enough to cause the device to rocket out of sight in seconds, then the noise may not be a factor. If the goal is to minimize noise and record a small signal, the best strategy is to rigidly mount to device on a load cell anchored to a vibration-isolated table. A device that is free to move will be subject to motion due to noise, air currents, both already present and thermally induced, thermal recoil, power line induced magnetic fields, vibrations from air conditioners, traffic, and distant earthquakes, and a host of other factors. The investigator's statement that the device has to be free to move or there would be no thrust is unfortunately a misinterpretation of Newton's laws.

Yes, if the thrust > 1 g, we would not be having these discussions. Given that it is not, we have to worry about noise. Note that to measure thrust, something has to move or bend or deform - there has to be a pendulum or a spring or something with an opposing force. However, measurements at this level of accuracy can certainly be done.*

*You can go out and buy absolute gravimeters : http://www.microglacoste.com/pdf/Brochure-FG5-X.pdf that are good to ~ 10 microGals = 10^-7 m/sec^2, based on dropping a retroreflector weighing about 1 kg (thus the force accuracy is ~ 0.1 micro Newtons). This, note, is a field instrument, intended to be used in a tent.

Offline frobnicat

  • Full Member
  • ****
  • Posts: 518
  • Liked: 500
  • Likes Given: 151
...Another thing is to do some autocorrelation analysis on the TS.  R has what you need.

I was showing that the human mind is built by Nature to see patterns, many times where there may not be any patterns.  (The face on Mars, trees on Mars, etc.)

...

Focusing on the Movax experiment, they should try to use oil (instead of water) to dampen the vibrations.

If calculate we must, calculate we will.  We can also do autocorrelation and power spectral density with Mathematica:

(I guess we could also try to do a cross-correlation with the on/off signal later on...)

Actually the cross-correlation looks interesting, there is a relatively clear max (magnitude) centered around 0. Interestingly this absolute max is found at a lag of 3 frames (is it 3 minutes ?). Not rock solid statistics but lucky if it was only a coincidence in the absence of any actual correlation.

So Prunesquallor's activation periods averaged cuts seem not that much misleading, there may have a thing going on.

Attached plot :
Horizontal = lag
Vertical = sum (averaged) on all t of activation(t)*data(t+lag) with normalized data and activation +1(on)/-1(off) (0 outside)
This is using the discrete formula for cross-correlation, what I call lag is n in the formula, excitation (on/off) is f and data is g.

« Last Edit: 06/15/2015 12:19 AM by frobnicat »

Tags: