-
#2200 by SteveD on 16 Oct, 2017 21:36
-
Is 3 uN really a "significant" noise source.
-
#2201 by WarpTech on 16 Oct, 2017 22:41
-
Could you describe your rig when you are running calibration pulses? Does disconnecting all of the batteries on the rig have any effect on the noise profile during calibration?
For your calibration pulses, are you using a permanent magnet on the torsion arm, and an electromagnet nearby? Would it be possible to physically move the electromagnet after the calibration pulse? A moving permanent magnet can induce currents in a nearby electromagnet, so I was wondering if that might be one of the noise sources. Increasing the size of the air gap after the calibration pulse finishes would help test whether the electromagnet was having any effect on the noise.
Does the noise profile look different if the calibration pulse is stronger/weaker?
With everything unplugged and unpowered, I still get the noise. The calibration pulse is simply a small electromagnet outside of the enclosure. Inside the enclosure attached to the pendulum is a small aluminum arm that extends to 4cm away from the calibration coil. Attached to this arm is a small ferromagnetic screw. I use the electromagnet at a known distance and current to exert a force on the screw. Yes, I still get the noise even if the calibration coil is taken away.
I also performed a test increasing the air-gap. See below. The little spike in the middle in me moving the coil. This was one of the better runs.
The noise isn't as important at the signal to noise ratio. You did a lot of work to maximize the performance of the frustum. If you haven't done the tests, how do you know your signal is still down in the mud? The biggest issue with past data is not so much the noise as it is the "drift" to a new baseline.
Also, try putting your crash cart on a memory foam cushion to decouple vibration induced from the floor.
IMO, you should present what you have. Compared to what Prof. Tajmar presented last year on the MEGA, you've done a helluva lot more work and have a lot more data than he had to present. I think a lot of people will really appreciate all the work you put into this, and how much difficulty there is in eliminating the noise.
-
#2202 by WarpTech on 16 Oct, 2017 22:51
-
This is probably the dumbest thing I've ever said but have you tried adding the opposite of the noise signal in a feedback loop to see what happens?
If someone wants to try, here are the excel files. If there is an easy way to do that in excel, please let me know the formula I should use.
Just subtracting the blue and orange traces eliminates the slow noise. The question would be, when the Frustum is powered, does the reaction happen fast enough to cause a differential in the two sensors? But I don't think that's what Bob012345 had in mind. Connecting an Op-amp to the signal to drive the Electromagnet in opposition to the noise, could be used as a feedback loop to stabilize the arm.
Still, since the noise has a period of 30 to 60 seconds, and the Frustum can charge to full power, Max Q in what, microseconds (??), I don't see the problem. The low frequency noise would only be an issue if the thrust is due to thermal heating, not "something else", I think. What you need to determine is if the signal is significant compared to the noise.
-
#2203 by Monomorphic on 16 Oct, 2017 23:05
-
IMO, you should present what you have. Compared to what Prof. Tajmar presented last year on the MEGA, you've done a helluva lot more work and have a lot more data than he had to present. I think a lot of people will really appreciate all the work you put into this, and how much difficulty there is in eliminating the noise.
Thank you, but I have already sent the cancellation notice to Heidi Fearn. She was very gracious in her reply. There was a late addition, so there shouldn't be issues with the schedule. She said I would be invited to present at the next conference in Estes Park 2018 to give me more time with the experiment. -
#2204 by SteveD on 16 Oct, 2017 23:29
-
IMO, you should present what you have. Compared to what Prof. Tajmar presented last year on the MEGA, you've done a helluva lot more work and have a lot more data than he had to present. I think a lot of people will really appreciate all the work you put into this, and how much difficulty there is in eliminating the noise.
Thank you, but I have already sent the cancellation notice to Heidi Fearn. She was very gracious in her reply. There was a late addition, so there shouldn't be issues with the schedule. She said I would be invited to present at the next conference in Estes Park 2018 to give me more time with the experiment.
You could still see about presenting a poster on fabrication techniques and the expirmental setup, perhaps with preliminary data from a couple of runs. Folks may want to crib some of that for their designs. Me, I just want to see if there is a jerk action somewhere in your automated tests. Maybe setup with labview and generate a bunch of runs this weekend. -
#2205 by dustinthewind on 17 Oct, 2017 01:35
-
Could you describe your rig when you are running calibration pulses? Does disconnecting all of the batteries on the rig have any effect on the noise profile during calibration?
For your calibration pulses, are you using a permanent magnet on the torsion arm, and an electromagnet nearby? Would it be possible to physically move the electromagnet after the calibration pulse? A moving permanent magnet can induce currents in a nearby electromagnet, so I was wondering if that might be one of the noise sources. Increasing the size of the air gap after the calibration pulse finishes would help test whether the electromagnet was having any effect on the noise.
Does the noise profile look different if the calibration pulse is stronger/weaker?
With everything unplugged and unpowered, I still get the noise. The calibration pulse is simply a small electromagnet outside of the enclosure. Inside the enclosure attached to the pendulum is a small aluminum arm that extends to 4cm away from the calibration coil. Attached to this arm is a small ferromagnetic screw. I use the electromagnet at a known distance and current to exert a force on the screw. Yes, I still get the noise even if the calibration coil is taken away.
I also performed a test increasing the air-gap. See below. The little spike in the middle in me moving the coil. This was one of the better runs.
I am tempted to point you in the direction of possibly wanting your experiment to osculate at some desired frequency - possibly a very low to no frequency (low pass filter). Some possible solutions then come to mind.
The first I will suggest is a low frequency filter. Basically a capacitor and inductor/resistor circuit set up so that current flowing in the circuit filters different frequencies. Inductors/resistors resist rapid changes in current and so develop voltages. The capacitor part of the circuit is the high frequency filter while showing voltages for low frequency changes in current. (I'm probably not describing anything new to you but may be to other readers.)
Proper choice of your capacitor inductor components will allow you to filter out the low frequency signal of about 60 seconds while looking for a desired higher frequency signal if using a "High pass filter". Using a low pass filter you could filter out the 60 second signal in favor of a much longer signal of constant thrust.
http://www.learningaboutelectronics.com/Articles/High-pass-filter-calculator.php
http://www.learningaboutelectronics.com/Articles/Low-pass-filter-calculator.php
It is possible to use both and analyze low and high frequency aspects using separate filters.
There are some op amp versions of these filters with gain.
Another option using a frequency might be a lock in amplifier. https://en.wikipedia.org/wiki/Lock-in_amplifier You might be able to do this artificially using just the software. Basically a lock in amplifier observes two signals. One signal is a seed frequency that when it goes negative it flips the other observed signal. Because noise when flipped between positive and negative and integrated adds to zero as well as many undesired signals this eliminates noise. You then pulse your EM drive at the seed frequency (hopefully) causing an osculation in the thrust signal while recording the thrust signal.
This thrust signal is then inverted according to the seed signal and integrated. The seed signal can be either a sine or cosine signal or both and many use both. They are phase sensitive so you have to be in the right phase but the dual phase amplifiers eliminate this.Quote from: https://en.wikipedia.org/wiki/Lock-in_amplifierBasic principles
...
More advanced, so called two-phase lock-in-amplifiers have a second detector, doing the same calculation as before, but with an additional 90° phase shift. Thus one has two outputs: {\displaystyle X=V_{\text{sig}}\cos \theta } {\displaystyle X=V_{\text{sig}}\cos \theta } is called the "in-phase" component, and {\displaystyle Y=V_{\text{sig}}\sin \theta } {\displaystyle Y=V_{\text{sig}}\sin \theta } the "quadrature" component. These two quantities represent the signal as a vector relative to the lock-in reference oscillator. By computing the magnitude (R) of the signal vector, the phase dependency is removed:
The phase can be calculated from
Matching the seed signal to the resonant frequency of a pendulum can further amplify the signal which may/may not be desired.
As stated earlier by WarpTech if your signal to noise ratio is large enough the noise may not make a difference.
In the back of my mind I was also wondering if your radiant body heat could have some effect on the apparatus. If you are moving about the room your thermal radiation may also have an effect.
For constant thrust at very low frequency maybe a low pass filter at 0.001 Hz might do the trick? -
#2206 by jay343 on 17 Oct, 2017 05:19
-
Here's a hunch: try changing or removing the damping oil, or sealing it with a plastic film with the smallest opening possible for the paddle strut's movement. I'm wondering if there is a drinking bird effect going on.
-
#2207 by Peter Lauwer on 17 Oct, 2017 08:28
-
Is 3 uN really a "significant" noise source.
Yes, I would say it is. With a comparable torsion balance, I had a 0.3 µN resolution. I hope to get even a little below that with my improved setup. -
#2208 by Peter Lauwer on 17 Oct, 2017 08:33
-
...
The first I will suggest is a low frequency filter. Basically a capacitor and inductor/resistor circuit set up so that current flowing in the circuit filters different frequencies.
....
Another option using a frequency might be a lock in amplifier.
...
I agree that adding a low-pass filter can be benificial. Using a lock-in amp is a bit of an overkill for a setup like this, I would say. That should not be necessary. -
#2209 by Peter Lauwer on 17 Oct, 2017 08:41
-
I'm thinking that I will back out of the advanced propulsion workshop in November. I've been working non-stop for a week now trying to collect good data and I can't seem to nail down the noise problem
...
I will still not be able to beat you, Jamie (in being the first in the new series of DIY experiments). I estimate I need another two weeks full-time work to get everything working. Which I don't have available at the moment.
We just keep on going. Like you, I learn a lot. First of all, I have to be sure which resonances I am seeing in the frustum cavity.
Cheers, Peter -
#2210 by ThatOtherGuy on 17 Oct, 2017 10:17
-
I'm thinking that I will back out of the advanced propulsion workshop in November. I've been working non-stop for a week now trying to collect good data and I can't seem to nail down the noise problem (see image below). I can easily detect forces below 3uN if the pendulum is completely still, but that is seldom the case - which is very frustrating. Just when I think I have solved the problem and get a clean run, the noise returns the next day with the same vigor. There has to be something fundamentally wrong that i'm missing, or this is the reason these experiments are performed in a vacuum! In order to minimize the noise further I would need to increase the torsional spring constant and get a laser displacement sensor with higher resolution - which requires an ADC with higher resolution. In other words, another major retooling and expense.
Jamie, just shooting in the dark, but did you consider that the noise you're observing may be somewhat picked up by the cabling ? Also, if I'm not wrong, the computer and the analyzer are sitting on the arm, I suppose either or both have an internal oscillator used for the clock, and that may generate some noise, same goes for USB signals.
Not willing to teach you how to run your business, just throwing some ideas on the table in an attempt to helpThere were emails this weekend about how they may have too many presentations on the emdrive this year. Without good clear data, I wouldn't have much to add. It's also just a hobby for me and the idea of presenting this material in front of a bunch of professional scientists is very intimidating. I know a cancellation means I probably won't be invited back, but I can't present this kind of data and expect it to be taken seriously. There is also a deadline for a final paper that I just can't commit to. I should have never accepted the offer to present to begin with. I am deeply sorry to those who recommended me and everyone counting on me.
I will continue working, as it brings me joy and satisfaction, just at a slower pace.
My Best,
Jamie
That's a pity, even if you don't have data, you still have a whole bunch of informations about building a test rig, not just that, confronting ideas with people at the conference may give you ideas and/or possibly solutions to the noise issues
Keep up the good work, Jamie !
[edit]
You also made some measurements while changing/rearranging/recabling... maybe comparing those with the current noise figures you may have some idea about what's causing that noise floor; that said... what's the "expected" thrust from your cavity ? See, if it's above the noise level, then the noise may still be dealt with, no ?
-
#2211 by flux_capacitor on 17 Oct, 2017 11:41
-
I was about to post something already pointed by Mr Peter a week ago, that I personally missed within the fast pace of recent developments on this board. As others may have also missed Mr Peter's original post (nobody commented on this document) I put it here again with more emphasis, and a few comments.
This is a 70-page long thesis on the EmDrive by Joshua Steven Pennington, PhD Student, BS in Physics (2015), BS in Mechanical Engineering (2015) and MS in Microelectronics-Photonics (2017) at the University of Arkansas (UARK), Fayetteville.
• Pennington, J. S. (2017) "Optimization of Miniaturized Resonant Microwave Cavities for Use in Q-Thrusters". Theses and Dissertations. 2473.
http://scholarworks.uark.edu/etd/2473
PDF file attached to this message. Pennington's poster also attached.Quote from: Josh PenningtonAbstract
A gedankenexperiment was considered to compare a hypothetical thruster that used no reaction mass to propulsion methods currently in use. A brief discussion of previous research work done on closed resonant cavity thrust devices was conducted. Using the previous work as a template, a simulation plan was devised. Computational models of resonant microwave cavities were constructed and investigated using COMSOL software. These COMSOL simulations were verified against known analytical solutions using Matlab software as a computational tool. Multiphysics simulations were created to study the microwave heating environment of the resonant cavities. From the COMSOL study outputs, the electromagnetic field magnitude, temperature, surface resistive losses, volume resistive losses, quality factor, and energy contained in the electric field were presented and discussed. The disagreements between the computational model and real-world resonant cavities were also presented and discussed.
Key points:
• Josh Pennington's work consists in modeling cavities and antennas in COMSOL to understand electrical and thermal effects and their possible contribution to real world experimental error.
• EagleWorks helped him. That's why Pennington talks of "Q-thrusters" instead of an "EmDrive" nor the more neutral terms of RF resonant cavity thruster. Yet he cited Shawyer in his thesis as the inventor of the "EM Drive". But IMHO specifically talking of "Quantum Vacuum Plasma Thruster" introduces a bias on how such a thing might work.
• The dominant electric field shape is the resonant mode of the cavity rather than the radiation pattern of the antenna.
• His Future work includes:
- Construction of a sensitive torsion pendulum.
- Construction of a Q-thruster.
- Use coupled Q-thrusters to investigate the current theory of operation.
But what strikes me the most is:
• The spherical COMSOL simulation had a high amount of disagreement with the spherical analytical models (unlike the rectangular and cylindrical COMSOL simulations which had a very high amount of agreement to the analytical models). This disagreement exists because the COMSOL solution does not completely match the analytical solution, whatever the mesh used in COMSOL (fine or coarse). The COMSOL output can be seen to be smeared across a range of answers not matched by the spherical Bessel function. The lack of change in the data set even after manipulation leads to the belief that COMSOL's Eigenstudy does not directly compute the electric field in the geometry by directly solving Maxwell's equations, but instead relies on some non-physical overlay to create arbitrary form factors for the electric field within resonant cavities.
Although COMSOL simulations are not about thrust predictions, they are eigensolutions to the eigenvalue problem. So what exactly is this discrepancy between COMSOL and analytical solutions? Can it introduce errors in further calculus, if one only relies on COMSOL to build frustum cavities with very tight precision? I can't help but think about Roger Shawyer telling TheTraveller two years ago (in Thread #3) that SPR does not relies on COMSOL or similar simulation softwares because of problems for the specific case of the EmDrive, and that they had to resort to their own in house software instead.
Rodal then disproved TT (who was reporting Shawyer's email message about the problem of COMSOL when applied to the EmDrive, especially the controversial "cut-off rule" of a cylindrical waveguide applied to a closed cavity as a key for thrust generation) with a peer-reviewed paper validating the precision and power of COMSOL, but this paper treated precisely of rectangular cavities, which COMSOL does not have any problem with according to Pennington.
Surprisingly, Pennington does not seem to care as he concludes:Quote from: Josh PenningtonIt is not expected that the failure to accurately compare COMSOL’s spherical resonant cavity model to that represented by the analytical solution will affect the cavities that are the focus of this thesis.
Even so he will use COMSOL to design his cavities and built his test articles.
EDIT: Sorry I attached the poster twice. The PDF "Optimization of Miniaturized Resonant Microwave Cavities for Use in Q-trusters" is now correctly attached.
-
#2212 by Eusa on 17 Oct, 2017 13:46
-
There is either a significant noise source, or a small noise source exciting a very poorly damped (and complex) mode of oscillation. Try a second damper to eliminate oscillations on axes through the damper, or failing that make the damper paddle much longer with the same total damping. Sorry to be broken record, but if you are desperate you might as well try it. No garauntee it will work.
With LIGO they have regular intentional harmonic oscillation set for masses with mirrors. Continuous acceleration dampens other noise to be better recognizable... -
#2213 by PotomacNeuron on 17 Oct, 2017 14:03
-
Jamie, just shooting in the dark, but did you consider that the noise you're observing may be somewhat picked up by the cabling ? Also, if I'm not wrong, the computer and the analyzer are sitting on the arm, I suppose either or both have an internal oscillator used for the clock, and that may generate some noise, same goes for USB signals.
Not willing to teach you how to run your business, just throwing some ideas on the table in an attempt to help
Agree. Though using an on-board computer can make it easier to automate the experiment, it also potentially introduces noise. When I did my experiment, I designed an optically activated solid stat switch that does not have internal oscillator. Everything is static. Maybe it is worth to move the computer to the framework, from where it uses laser or flashlight to control the experiment equipped with optically activated (or RF controlled) solid stat switches. -
#2214 by PotomacNeuron on 17 Oct, 2017 14:08
-
Jamie, just shooting in the dark, but did you consider that the noise you're observing may be somewhat picked up by the cabling ? Also, if I'm not wrong, the computer and the analyzer are sitting on the arm, I suppose either or both have an internal oscillator used for the clock, and that may generate some noise, same goes for USB signals.
Not willing to teach you how to run your business, just throwing some ideas on the table in an attempt to help
Agree. Though using an on-board computer can make it easier to automate the experiment, it also potentially introduces noise. When I did my experiment, I designed an optically activated solid stat switch that does not have internal oscillator. Everything is static. Maybe it is worth to move the computer to the framework, from where it uses laser or flashlight to control the experiment equipped with optically activated (or RF controlled) solid stat switches.
Though I do statistical analysis to make a living, I do not suggest filtering or other kind of data processing techniques in resolving Monomorphic's noise problem. It is not a fundamental noise. It has a source that can be removed. -
#2215 by Monomorphic on 17 Oct, 2017 14:17
-
Something is causing the pendulum to move in unexpected ways. It has too many degrees of freedom in my opinion and needs to be constrained to the one we are measuring.
The plan is to remove the piano wire and replace it with flexure bearings. I have to be careful that I use flexure bearings that do not exceed by much the current 0.00326 lb-in per degree torsion spring constant but can still support ~25lbs. That means I can use either one E-10 c-flex bearing (0.0036 lb-in per degree, 36.5lbs axial load capacity) or two D-10 bearings (0.0018 lb-in per degree, 22.7 lbs axial load capacity). I will also be replacing the rectangular damping fluid reservoir with something circular.
I ordered the three flexure bearings today, as they are not very expensive, but installing them is going to require a couple of weeks. -
#2216 by PotomacNeuron on 17 Oct, 2017 14:42
-
Something is causing the pendulum to move in unexpected ways. It has too many degrees of freedom in my opinion and needs to be constrained to the one we are measuring.
The plan is to remove the piano wire and replace it with flexure bearings. I have to be careful that I use flexure bearings that do not exceed by much the current 0.00326 lb-in per degree torsion spring constant but can still support ~25lbs. That means I can use either one E-10 c-flex bearing (0.0036 lb-in per degree, 36.5lbs axial load capacity) or two D-10 bearings (0.0018 lb-in per degree, 22.7 lbs axial load capacity). I will also be replacing the rectangular damping fluid reservoir with something circular.
I ordered the three flexure bearings today, as they are not very expensive, but installing them is going to require a couple of weeks.
I do not agree. The noise problem is caused by a noise source, not by the degree of freedom of movement. Just find the noise source, remove it and the experiment is good to go.
Also flexure bearing has its own problems. It will droop and transform mass center shift into thrust-like force.
-
#2217 by Rodal on 17 Oct, 2017 15:13
-
...
1) My paper on cut-off solutions (attached below) had a comparison with an exact solution for a truncated conical cavity, and not for a rectangular cavity. Actually, I did not perform any comparisons for rectangular cavities. The message you are quoting is one of many I wrote addressing this issue, there are stronger reasons to challenge TT's statements about the appropriateness of numerical solutions than the message quoted.
Rodal then disproved TT (who was reporting Shawyer's email message about the problem of COMSOL when applied to the EmDrive, especially the controversial "cut-off rule" of a cylindrical waveguide applied to a closed cavity as a key for thrust generation) with a peer-reviewed paper validating the precision and power of COMSOL, but this paper treated precisely of rectangular cavities, which COMSOL does not have any problem with according to Pennington.
...
2) Finite Element Methods should not be discussed as "black boxes", preferably in academic graduate research one should write the numerical routines which presumes having access to change the code.
3) COMSOL uses the Galerkin solution method, which is one the weakest methods of numerical solution. Particularly if one suspects a numerical code, one should use more than one code for comparison, particularly in academic research. For example, one should use excellent codes like ABAQUS, and ADINA that have a strong theoretical basis. Most importantly academic research should address address convergence of the numerical scheme.
4) In this case (the EM Drive) the pertinent comparison between exact solutions and numerical methods should be for a truncated conical cavity (and not for spherical or rectangular cavities). A comparison with a cylindrical cavity is also pertinent, as most EM Drive designs have been close to cylindrical. A comparison with a spherical geometry is the least pertinent (as readily apparent from an understanding of the resonant modes in a perfectly symmetric spherical cavity as compared to an asymmetric resonant cavity like the EM Drive).
-
#2218 by flux_capacitor on 17 Oct, 2017 16:06
-
You are right, I was quoting one of your messages where you were referring specifically to a peer-reviewed paper on rectangular cavities, that you cited as a demonstration of COMSOL's efficiency: "Push-Pull Phenomenon of a Dielectric Particle in a Rectangular Waveguide". Other posts of yours including your paper you attached in your message above, are more precise on that matter....
1) My paper on cut-off solutions (attached below) had a comparison with an exact solution for a truncated conical cavity, and not for a rectangular cavity. Actually, I did not perform any comparisons for rectangular cavities.
Rodal then disproved TT (who was reporting Shawyer's email message about the problem of COMSOL when applied to the EmDrive, especially the controversial "cut-off rule" of a cylindrical waveguide applied to a closed cavity as a key for thrust generation) with a peer-reviewed paper validating the precision and power of COMSOL, but this paper treated precisely of rectangular cavities, which COMSOL does not have any problem with according to Pennington.
...The message you are quoting is one of many I wrote addressing this issue, there are stronger reasons to challenge TT's statements about the appropriateness of numerical solutions than the message quoted.
2) Finite Element Methods should not be discussed as "black boxes", preferably in academic graduate research one should write the numerical routines which presumes having access to change the code.
3) COMSOL uses the Galerkin solution method, which is one the weakest methods of numerical solution. Particularly if one suspects a numerical code, one should use more than one code for comparison, particularly in academic research. For example, one should use excellent codes like ABAQUS, and ADINA that have a strong theoretical basis. Most importantly academic research should address address convergence of the numerical scheme.
4) In this case (the EM Drive) the pertinent comparison between exact solutions and numerical methods should be for a truncated conical cavity (and not for spherical or rectangular cavities). A comparison with a cylindrical cavity is also pertinent, as most EM Drive designs have been close to cylindrical. A comparison with a spherical geometry is the least pertinent (as readily apparent from an understanding of the resonant modes in a perfectly symmetric spherical cavity as compared to an asymmetric resonant cavity like the EM Drive).
If I understand all the above correctly, you do agree with Pennington that the discrepancy between the results of COMSOL for spherical cavities vs spherical analytical models is pointless in the case of the EmDrive, because a truncated cone is closer to a cylinder than to a sphere (for that matter the cylindrical COMSOL simulations had a very high amount of agreement to the analytical models). -
#2219 by SteveD on 17 Oct, 2017 16:12
-
Try moving the device to the other side of the lab. The noise may be caused by vibration from sewer or other piping running near mono's basement.
