-
#200 by Rodal on 14 Mar, 2016 17:34
-
It's definitely the big end that can cause all the problems! Also, the higher frequency the distortion (how rough it is) the better. This is because 2.45Ghz cannot "see" features smaller than 1/2 the wavelength. This means that frustums that are skewed or warped in large ways, perform worse than those with a rough surface of equal displacement. In my opinion gross geometric tolerances of 0.5mm (or less) should be a goal for DIYers.
This is really great work!
We learned a lot from it, not just the overall tolerance to aim for but also what is most crucial in order of importance. It is posts such as this that make it worthwhile to spend time at NSF !
I am looking forward to further comparisons comparing the geometrical distortion effect on the Q quality factor ! -
#201 by X_RaY on 14 Mar, 2016 17:44
-
Sorry for interrupting the current discussions, have to read several posts to come close to the last posts of today.
The discussion about 13µm differences of the endplates is crazy.
At first what does it mean?
https://en.wikipedia.org/wiki/Surface_roughness (definition?)
As Dr.Rodal stated the sidewalls are of interest as the endplates, there are also losses due to eddy currents.
At the moment and in general there are much bigger problems (thermal expansion, antenna matching, thermal rise/ballooning and so on) than to polish the wall and plates of the frustum up to mirrors quality. Structures much smaller than lets say 1/10 of the wavelenght are less of interrest since it decrease the Q only by a small amount.
It make no difference if the theoretical Q_u is 60000 or 60500 or even 58000 until the general problems are solved by the DIY.
To get a look to formula of unround shapes serarch for it in the pdf-file below or run simulations.
For me the biggest problem is still (pic):
-
#202 by Rodal on 14 Mar, 2016 18:04
-
Paul Kocyla in Aachen, Germany is beginning to test the 24 GHz emdrive on a rotary flotation pad. Unquantified force measurements, appears to be calibration tests:
Would appreciate somebody familiar with this testing program to clarify:
1) are the battery and the mini-EM-drive integrated together on the testing platform for the Kocyla test? DeltaMass and I had agreed that by far the best proposed test was TheTraveller's proposal to have a battery and the EM Drive on a rotary platform together (rather than having the power be fed from a stationary source to a moving EM Drive which has a big testing flaw: the center of energy-mass is outside the moving EM Drive, therefore measuring an acceleration in such a test is flawed since in space the source of power would need to be in the same spaceship as the EM Drive)
2) What is the present testing platform arrangement?
I tried to find answers by going to the hackaday website but I was quickly disillusioned:
* unlabeled plots
* grainy picture
* lack of technical clarity on the presentation of test results
very unclear as to what they describe as noise, what they describe as signal, thermal effects, etc.
I confess however that I don't have the time to read all the information they have on that website, as I just speed-read to try to gather some information.
************
* on another subject, I agree with X-Ray regarding the importance of the RF feed for all these tests.
When I have the time I would like to formally analyze the flaw in all these tests (with the exception of the proposed test by TheTraveller with the battery integrated together with the EM Drive in a self-contained package) regarding the center of energy-mass of the system. -
#203 by Monomorphic on 14 Mar, 2016 18:12
-
At the moment and in general there are much bigger problems (thermal expansion,
How much will a ~20cm copper frustum expand due to heating? If it is less than 1/2 a cm or so, then it shouldn't be too much of a concern.
Best to have a tunable frustum IMHO. And from what I've learned today, it might be best to have the big end be the side that is tuned. -
#204 by SteveD on 14 Mar, 2016 18:17
-
It's definitely the big end that can cause all the problems! Also, the higher frequency the distortion (how rough it is) the better. This is because 2.45Ghz cannot "see" features smaller than 1/2 the wavelength (~6cm). This means that frustums that are skewed or warped in large ways, perform worse than those with a rough surface of equal displacement. In my opinion gross geometric tolerances of 0.5mm (or less) should be a goal for DIYers.
If resonance increases with some value of sidewall distortion any chance those genetic algorithms you were working on can optimize it? Wondering if the future here might be some kind of 3D printed sidewall together with precision endplates. Suggest also running spherical endplates. -
#205 by X_RaY on 14 Mar, 2016 18:25
-
1) https://forum.nasaspaceflight.com/index.php?topic=39772.msg1502437#msg1502437
At the moment and in general there are much bigger problems (thermal expansion,
How much will a ~20cm copper frustum expand due to heating? If it is less than 1/2 a cm or so, then it shouldn't be too much of a concern.
Best to have a tunable frustum IMHO. And from what I've learned today, it might be best to have the big end be the side that is tuned.
2) Better tune the small side, its less sensitive to the plate displacement (due to larger guide wavelength)(finetuning).
Thats one reason why I like Shells design.
-
#206 by SeeShells on 14 Mar, 2016 18:35
-
It's definitely the big end that can cause all the problems! Also, the higher frequency the distortion (how rough it is) the better. This is because 2.45Ghz cannot "see" features smaller than 1/2 the wavelength (~6cm). This means that frustums that are skewed or warped in large ways, perform worse than those with a rough surface of equal displacement. In my opinion gross geometric tolerances of 0.5mm (or less) should be a goal for DIYers.
THANKS!!! Very very nice work Monomorphic!
Shell -
#207 by SeeShells on 14 Mar, 2016 18:52
-
It's definitely the big end that can cause all the problems! Also, the higher frequency the distortion (how rough it is) the better. This is because 2.45Ghz cannot "see" features smaller than 1/2 the wavelength. This means that frustums that are skewed or warped in large ways, perform worse than those with a rough surface of equal displacement. In my opinion gross geometric tolerances of 0.5mm (or less) should be a goal for DIYers.
This is really great work!
We learned a lot from it, not just the overall tolerance to aim for but also what is most crucial in order of importance. It is posts such as this that make it worthwhile to spend time at NSF !
I am looking forward to further comparisons comparing the geometrical distortion effect on the Q quality factor !
Sometimes a picture is worth a thousand words.
-
#208 by rfmwguy on 14 Mar, 2016 19:15
-
You will like my new mechanical design on NSF-1701A then. More later, still working out the mechanicals.
At the moment and in general there are much bigger problems (thermal expansion,
How much will a ~20cm copper frustum expand due to heating? If it is less than 1/2 a cm or so, then it shouldn't be too much of a concern.
Best to have a tunable frustum IMHO. And from what I've learned today, it might be best to have the big end be the side that is tuned. -
#209 by zen-in on 14 Mar, 2016 19:36
-
At the moment and in general there are much bigger problems (thermal expansion,
How much will a ~20cm copper frustum expand due to heating? If it is less than 1/2 a cm or so, then it shouldn't be too much of a concern.
Best to have a tunable frustum IMHO. And from what I've learned today, it might be best to have the big end be the side that is tuned.
It all depends on what you are measuring. For example the Eagleworks tests from 2014 measured the displacement of the fustrum as it, in theory, pushed against the torque of the torque pendulum. That displacement ( ~18 micrometers) was actually smaller than the thermal expansion of some parts of the apparatus. To date no one knows how much of their anomalous force results were from thermal expansion.
Your margin of 5,000 micrometers is quite a bit larger. If you have a test force device on your apparatus that will provide a means of comparison. -
#210 by X_RaY on 14 Mar, 2016 20:40
-
I have implemented the expansion coefficient into my spreadsheet to get an idea of the frequency shift. I added the delta length(for endplates and sidewall length{not center heigth}) to the reference data of Frank Davies, for reference temp I used 20°C.
Using the expression
I come to the conclusion that a temperature rise of 80K (20°C -->100°C) leads to a frequency shift of ≈-2,87MHz** for the TE012 mode.
**other deformities than linear shifts are not included
-
#211 by Rodal on 14 Mar, 2016 20:54
-
A few people posting are assuming thermal expansion is uniform throughout and that EM Drive is unrestrained in the experiments, calculating the thermal expansion as the coefficient of thermal expansion times the change in temperature.
Admittedly, of course, such a simplified calculation is a good first cut at understanding the problem.
However, the actual temperature rise is nonuniform, it depends on the pattern of induction heating which depends on the actual mode shape being excited.
For such a realistic thermal expansion, the problem is thermoelastic and not just a trivial problem of unrestrained thermal expansion. For the actual thermoelastic problem (due to non-uniform temperature, one has non-uniform thermal expansion), the actual stiffness of the shell comes into play: hence the importance of the thickness to radius ratio (as well as the modulus of elasticity) in the thermoelastic problem as well.
Most important, the temperature rise is not constant through the thickness as people are assuming as a first cut.
The temperature is highest in the interior(due to the skin depth of 1 micrometer or so) and it will slowly diffuse out with time. This fact (non-uniform temperature through the thickness) plus the fact that the temperature is non-uniform thought the shell will lead to bending deformations of the shell and not just membrane deformations. Hence the thickness of the shell plays an important role also for the thermal expansion effect. In addition, there is the issue of thermal buckling (which I have addressed previously) and is also related to the thickness of the shell.
Hence the distortion of the EM Drive shell due to thermal expansion is more complicated, and in addition it is time dependent and so is the effect on Q due to shifting of the natural frequency because of distortion of the shell.
Use of thin shells (1 mm) with stiffeners, although better than just a thin shell, further complicates the understanding of the problem. Unless there is a paramount interest in reducing weight, use of thicker shells is simpler to analyze and hence easier to understand.
Use of different material for the end plates (for example ceramic at the ends) also changes the nature of the thermal distortion. If the end plates are made thicker (stiff enough) that's admittedly better.
Common sense solution: unless one can show that weight is a big issue, a clean solution is to use a thicker overall construction, (as used in commercial waveguides).
-
#212 by X_RaY on 14 Mar, 2016 21:00
-
Have to agree! As I said I did this only to get an idea of the dimension in general of the frequency shift using this linear approximation.
Its much more than the few KHz rfmwguy stated about the 13µm differences of a single endplate.
@Rfmwguy
Does you multiply it by a factor of two for both plates or do you meant the total displacement or surface roughness?
https://forum.nasaspaceflight.com/index.php?topic=39772.msg1503691#msg1503691
-
#213 by TheTraveller on 14 Mar, 2016 21:11
-
Have to agree! As I said I did this only to get an idea of the dimension in general of the frequency shift using this linear approximation.
Its much more than the few KHz TT stated about the 13µm differences of a single endplate. Does he multiply it by two for both plates or does he meant the total displacement or surface roughness?
Please review Rogers advise:QuoteThe route to high Q is to achieve very high precision in the machining of cavity components and their alignment, together with mirror finish on the conducting surface (copper, silver or gold) of at least 10X skin depth. Maintaining this quality of finish also requires a clean dry environment. This is typical flight standard for space qualified microwave equipment, and is therefore expensive to achieve.
If you aim for something like 75% of theoretical Q you will still get viable levels of thrust within a reasonable budget for a small business. -
#214 by Rodal on 14 Mar, 2016 21:13
-
Have to agree! As I said I did this only to get an idea of the dimension in general of the frequency shift using this linear approximation.
Its much more than the few KHz TT stated about the 13µm differences of a single endplate. Does he multiply it by two for both plates or does he meant the total displacement or surface roughness?
Please review Rogers advise:QuoteThe route to high Q is to achieve very high precision in the machining of cavity components and their alignment, together with mirror finish on the conducting surface (copper, silver or gold) of at least 10X skin depth. Maintaining this quality of finish also requires a clean dry environment. This is typical flight standard for space qualified microwave equipment, and is therefore expensive to achieve.
If you aim for something like 75% of theoretical Q you will still get viable levels of thrust within a reasonable budget for a small business.
Thanks. We have seen this before. What is not clear to different viewers is whether the mirror finish advise is only for the ends or also for the interior conical walls. Please further clarify Shawyer's statement. Is Shawyer asking for a mirror finish only on the end plate interior surfaces as shown by rfmwguy for example:
or also on the lateral interior round conical walls?
-
#215 by TheTraveller on 14 Mar, 2016 21:19
-
Have to agree! As I said I did this only to get an idea of the dimension in general of the frequency shift using this linear approximation.
Its much more than the few KHz TT stated about the 13µm differences of a single endplate. Does he multiply it by two for both plates or does he meant the total displacement or surface roughness?
Please review Rogers advise:QuoteThe route to high Q is to achieve very high precision in the machining of cavity components and their alignment, together with mirror finish on the conducting surface (copper, silver or gold) of at least 10X skin depth. Maintaining this quality of finish also requires a clean dry environment. This is typical flight standard for space qualified microwave equipment, and is therefore expensive to achieve.
If you aim for something like 75% of theoretical Q you will still get viable levels of thrust within a reasonable budget for a small business.
Thanks. Is Shawyer asking for a mirror finish only on the end plate interior surfaces as shown by rfmwguy for example:
or also on the lateral interior round conical walls?
My understanding, from other discussions, is that a scratch & pit free mirror like finish is required on ALL conducting interior surfaces of the frustum and any feed waveguides. -
#216 by X_RaY on 14 Mar, 2016 21:39
-
OK all surfaces sounds good
If this mechanical quality of surface roughness is required(for space applications), do you plan to Measure your Rz or whatever definition is meant in a machinist shop to be sure ALL of the dimensions of a frustum and the surface roughness is inside of these error bars of 13 micron?
Would be a lot of work and for sure expensive
On the other hand I am with you, make the experiment as good as possible to get good results. -
#217 by Monomorphic on 14 Mar, 2016 21:54
-
To accurately measure below 10um, you would need to use an optical measuring method such as laser 3D scan. A tabletop formtester like this can go down to 15um.
-
#218 by rfmwguy on 14 Mar, 2016 22:25
-
OK all surfaces sounds good
I will be sanding and planing the internal sheet metal surfaces in the same fashion...dry sand with planing wheel to 2000 grit then 9, 3 and 1 micron hand polishing. The small end was about 4.5 hours worth of work. It will be a mirror finish, just not the same thickness as the end plates.
If this mechanical quality of surface roughness is required(for space applications), do you plan to Measure your Rz or whatever definition is meant in a machinist shop to be sure ALL of the dimensions of a frustum and the surface roughness is inside of these error bars of 13 micron?
Would be a lot of work and for sure expensive
On the other hand I am with you, make the experiment as good as possible to get good results.
-
#219 by R.W. Keyes on 14 Mar, 2016 22:29
-
Back on New Year's Day, I posted a link to an article which was a popular-science explanation of this article at Science:http://science.sciencemag.org/content/351/6268/58l. Since that time, I have been busy with other projects and not had much time for EMDrive, but I now have a bit more time and resources. I have been reading about electroless deposition, and found a paper on the electroless deposition of superconductor Magnesium diboride (I found the abstract, but haven't been able to obtain the full paper) This would help solve some of the smoothness issues, but there is still the issue of imperfect 3d printing and and subsequent machining.
I have some crude, warped, and not-full-formed ideas in my head about defeating the warping and aparallelism of frustums, which has been discussed here as of late, perhaps by harnessing resonance of the frustum during forming. That is, having active RF energy help form the field, in a way which improves its Q far beyond the methods we now have. I can only roughly describe it as blowing a bubble, pumping RF into it, and the heat-loss inherent in use of an imperfect frustum keeps the bubble reforming until a near-perfect chamber is formed. No, I don't have any practical idea of how to do this, or any examples, but it is currently an idea bouncing around my head. I hope to have the idea better refined and more eloquently stated in the near future, but I thought I'd just try to drop it into some of the great minds around here ahead of time.
