QuoteBut the first question I have is, "Why would you want to model misalignment, what is to be gained?"he wants to see how good the alignement has to be to get the modes to stabilize.
But the first question I have is, "Why would you want to model misalignment, what is to be gained?"
Quote from: aero on 10/17/2015 05:37 pmQuote from: original_mds on 10/17/2015 04:03 pmI mentioned this earlier and am not sure if it was missed or ignored. Can a meep run be done without having the end cap being perfectly axially aligned with the frustrum body? I can't help but reflect on all the geometry discussions that have occurred, but we still don't have any idea about how good the alignment has to be to get the modes to stabilize. E.g does the alignment have to be within a degree, or a thousandth of a degree? Does meep even have the resolution to be able to evaluate this type of sensitivity?Can it be done? Of course it can be done, meep source code is available, meep is a numerical algorithm running with a geometric model (in our case). Any reasonable model you care to take the time to construct in a meep control file can be run. Here are the components you have to work with in native mode. http://ab-initio.mit.edu/wiki/index.php/Meep_Reference#geometric-objectOf course if needed, you can add stuff using the C++ language and recompile. But the first question I have is, "Why would you want to model misalignment, what is to be gained?"As many of us have experienced first hand, the real world tends to rarely (never) have perfect builds. Fortunately, most systems have some margin for imperfections. For cutting edge stuff, not having a good handle on what that margin is can result in a lot of wasted time, effort, and cash.For example, a project I was working on last year spent millions on trying to trace down the source of an intermittent issue that was threatening the entire program. Similar to the EM builds, it wasn't possible to take measurements of the parameters we wanted while the device was in use, so there was a heavy reliance on FEA modeling. Our perfect models showed some potential weaker points in the design, but nothing that was the clear culprit. We went through several fix/test cycles, but the issue remained intermittent. Ultimately, it appears the issue was a very minor misalignment that led to asymetries in the forces being applied to the uni, resulting in its destruction. Tolerances were similar to Seashell's work, but we found inconsistencies in measurements when using different measurement systems. When we did additional runs with the misalignment modeled, it quickly became obvious that the design was quite sensitive to alignment. Tightening the spec and increasing the number of data points when checking alignment resulted in perfect performance ever since.In the EM drive, every model that has been discussed for attempting to establish the best modes assumes perfect alignment. IIRC, Seashell's current design is supposed to have an absurdly high Q. How sensitive is it to alignment? AFAICT, we have no idea. Given a few more simulation data points, we may find that her chosen build tolerance gives a range of possible Q spanning several orders of magnitude. Or, we may see that she can increase her build tolerance 10x with little expected impact on the results, making her build easier to accomplish.I'd like to chip in on the modeling effort, but have some outside projects on the house that need to get done before it gets too cold. I may have more time when the snow flies.
Quote from: original_mds on 10/17/2015 04:03 pmI mentioned this earlier and am not sure if it was missed or ignored. Can a meep run be done without having the end cap being perfectly axially aligned with the frustrum body? I can't help but reflect on all the geometry discussions that have occurred, but we still don't have any idea about how good the alignment has to be to get the modes to stabilize. E.g does the alignment have to be within a degree, or a thousandth of a degree? Does meep even have the resolution to be able to evaluate this type of sensitivity?Can it be done? Of course it can be done, meep source code is available, meep is a numerical algorithm running with a geometric model (in our case). Any reasonable model you care to take the time to construct in a meep control file can be run. Here are the components you have to work with in native mode. http://ab-initio.mit.edu/wiki/index.php/Meep_Reference#geometric-objectOf course if needed, you can add stuff using the C++ language and recompile. But the first question I have is, "Why would you want to model misalignment, what is to be gained?"
I mentioned this earlier and am not sure if it was missed or ignored. Can a meep run be done without having the end cap being perfectly axially aligned with the frustrum body? I can't help but reflect on all the geometry discussions that have occurred, but we still don't have any idea about how good the alignment has to be to get the modes to stabilize. E.g does the alignment have to be within a degree, or a thousandth of a degree? Does meep even have the resolution to be able to evaluate this type of sensitivity?
Quote from: original_mds on 10/17/2015 10:34 pmQuote from: aero on 10/17/2015 05:37 pmQuote from: original_mds on 10/17/2015 04:03 pmI mentioned this earlier and am not sure if it was missed or ignored. Can a meep run be done without having the end cap being perfectly axially aligned with the frustrum body? I can't help but reflect on all the geometry discussions that have occurred, but we still don't have any idea about how good the alignment has to be to get the modes to stabilize. E.g does the alignment have to be within a degree, or a thousandth of a degree? Does meep even have the resolution to be able to evaluate this type of sensitivity?Can it be done? Of course it can be done, meep source code is available, meep is a numerical algorithm running with a geometric model (in our case). Any reasonable model you care to take the time to construct in a meep control file can be run. Here are the components you have to work with in native mode. http://ab-initio.mit.edu/wiki/index.php/Meep_Reference#geometric-objectOf course if needed, you can add stuff using the C++ language and recompile. But the first question I have is, "Why would you want to model misalignment, what is to be gained?"As many of us have experienced first hand, the real world tends to rarely (never) have perfect builds. Fortunately, most systems have some margin for imperfections. For cutting edge stuff, not having a good handle on what that margin is can result in a lot of wasted time, effort, and cash.For example, a project I was working on last year spent millions on trying to trace down the source of an intermittent issue that was threatening the entire program. Similar to the EM builds, it wasn't possible to take measurements of the parameters we wanted while the device was in use, so there was a heavy reliance on FEA modeling. Our perfect models showed some potential weaker points in the design, but nothing that was the clear culprit. We went through several fix/test cycles, but the issue remained intermittent. Ultimately, it appears the issue was a very minor misalignment that led to asymetries in the forces being applied to the uni, resulting in its destruction. Tolerances were similar to Seashell's work, but we found inconsistencies in measurements when using different measurement systems. When we did additional runs with the misalignment modeled, it quickly became obvious that the design was quite sensitive to alignment. Tightening the spec and increasing the number of data points when checking alignment resulted in perfect performance ever since.In the EM drive, every model that has been discussed for attempting to establish the best modes assumes perfect alignment. IIRC, Seashell's current design is supposed to have an absurdly high Q. How sensitive is it to alignment? AFAICT, we have no idea. Given a few more simulation data points, we may find that her chosen build tolerance gives a range of possible Q spanning several orders of magnitude. Or, we may see that she can increase her build tolerance 10x with little expected impact on the results, making her build easier to accomplish.I'd like to chip in on the modeling effort, but have some outside projects on the house that need to get done before it gets too cold. I may have more time when the snow flies.Good answer. Maybe later - for now I think all of our "Citizen Scientists" are keeping the tolerances as tight as they are able. Tighter than necessary, don't know.
Quote from: aero on 10/17/2015 10:57 pmGood answer. Maybe later - for now I think all of our "Citizen Scientists" are keeping the tolerances as tight as they are able. Tighter than necessary, don't know.Nice post!For me testing the device, the more cause for error that I can take out and stay within a small budget the better I can define any results gained. I'm not sure if it's going to be good enough or over kill, but it at least it's starting with the least amount of unknowns. What meep has shown is that with the extreme tolerances (it uses something like out to 12+ decimal places) that a very high Q could be achieved, do I think I'll see those Qs, good grief no. I couldn't hand build it to those tolerances, but I can within a small budget do a frustum to good tolerances and try to negate some of the red flagged problems. Fine control non-active tuning, resonance locking with captured end plates, addressing thermal expansion issues and cavity warping. Stable microwave generation will come later as I have some tests I want to sweeping through the different cavity modes using a broadband Rf source. It could just be exciting multiple modes and the interactions of those modes are the reasons for the increased Q and exciting a sweet spot in additive mode actions is a key. Will it be be enough? The only thing that can be assured of is it will be enough to point me to the next build issues that need to be addressed.Shell
Good answer. Maybe later - for now I think all of our "Citizen Scientists" are keeping the tolerances as tight as they are able. Tighter than necessary, don't know.
Shell, Aero & Crew:The manufacturing tolerances for building these EMDrive based room-temp copper frustums does not have to be very good to get Q-factor results that are quite usable in obtaining interesting thruster performance. Our unloaded, (-7dB down from the VNA S11 amplitude reference plane assuming near optimal antenna coupling using a magnetic loop antenna), with no dielectric discs, the TE012 resonance at 2,167 MHz per our 2014 AIAA/JPC paper's copper frustum came out to be ~54,000. Considering our garage construction crew used a civil war vintage bending mill to form the copper sheet into a cone, which was then lead/tin soldered together with two half inch wide exterior flanges butted together, and pulled together using 0.050" thick by 1/2 inch wide copper hoops that I hand routered out of copper sheets, which were then lead/tin soldered to the cone, should tell you that great precision for your first frustum prototypes is not required. And since I also just used semi-flat 1/16" thick FR4 printed circuit boards with one side plated with 1.0 oz (34.8 microns thick) copper with the copper side towards the inside of the cavity, super parallel surfaces on the end caps is not required either. BTW, since the wave-length of ~2.0 GHz RF is 5.906" (0.1500m), keeping within 1/100th of a wavelength (0.0591") tolerance of your design in your first build as the telescope builders do, one should just use moderate (0.03") shop tolerances for your first prototype builds and go from there.Best, Paul March
VNA and Spec An finally working on laptop. Awaiting a parts donor magnetron to use radome for frustum testing.
Quote from: rfmwguy on 10/18/2015 03:27 pmVNA and Spec An finally working on laptop. Awaiting a parts donor magnetron to use radome for frustum testing.Hi, good to see that some measurements were done. I have some questions about.What are the conditions for this measurements?Spec.: µW-Source, Antenna/probe ?VNA: S11 or S21? Antenna/probe? Can't identify some of the numbers, JPEG quality is to bad. Shows the left y-axis of the VNA measurements the loss in dB? Have you a better pic available or a data export file (Touchstone file or something else)?At the moment I cant see strong resonance around 2.4 GHz.Nevertheless good that you post all of your results!!
Quote from: X_RaY on 10/18/2015 07:13 pmQuote from: rfmwguy on 10/18/2015 03:27 pmVNA and Spec An finally working on laptop. Awaiting a parts donor magnetron to use radome for frustum testing.Hi, good to see that some measurements were done. I have some questions about.What are the conditions for this measurements?Spec.: µW-Source, Antenna/probe ?VNA: S11 or S21? Antenna/probe? Can't identify some of the numbers, JPEG quality is to bad. Shows the left y-axis of the VNA measurements the loss in dB? Have you a better pic available or a data export file (Touchstone file or something else)?At the moment I cant see strong resonance around 2.4 GHz.Nevertheless good that you post all of your results!! These are just scans without the frustum attached. The vna will be S11 and the spec an will paint a shot when the frustum fires up. The spec an module was plug and play, the vna from miniradiosolutions.com was not. Lots of extra drivers and jre needed...real hassle but finally got it sorted out. Am awaiting another mag and will pull off radome and stick an sma connector on it. I'll use this on the empty frustum for the RL sweep.
In the air its hard for me to say, but for me I think something in the range of one to ten newtons would be needed to really get me interested (flying off the table levels of force).
Quote from: Tetrakis on 10/19/2015 03:55 am In the air its hard for me to say, but for me I think something in the range of one to ten newtons would be needed to really get me interested (flying off the table levels of force).To be clear, we get about 6-7 Newtons of force per Kw in jets. Obviously you don't mean 1-10 per Kw (Unless you do...) but it does put into perspective how tall of an order that is. Maybe lower your expectations a bit?
Quote from: Dortex on 10/19/2015 03:59 amQuote from: Tetrakis on 10/19/2015 03:55 am In the air its hard for me to say, but for me I think something in the range of one to ten newtons would be needed to really get me interested (flying off the table levels of force).To be clear, we get about 6-7 Newtons of force per Kw in jets. Obviously you don't mean 1-10 per Kw (Unless you do...) but it does put into perspective how tall of an order that is. Maybe lower your expectations a bit?Perhaps I do have excessive expectations, but a jet is essentially an efficient "thermal effect" I arrived at those numbers because differences in pressure/density are the sources of confounding factors in these experiments, and at about one atmosphere thermally induced pressure differences can probably put out a maximum of one to ten newtons of force at any realistic power level (in the ballpark of a vacuum on one side of the test article). In other words, the noise floor is pretty high and difficult to characterize. In a microtorr vacuum the maximum expected force from surrounding gases is millions of times lower, concomitantly lowering the expected height of the noise floor. As has been pointed out before, this is a sci-fi-esque energy to momentum drive being discussed. A pretty big signal is going to be needed to get more than niche interest. How "big" of a signal do you think is needed?
How "big" of a signal do you think is needed?
Seems BBC Horizon are working on a Roger Shawyer / EMDrive documentary. Claimed NASA is also involved.http://envisionation.co.uk/index.php/roger-shawyer-emdriveClick on the full Shawyer youtube interview and then scroll down to the comments. Nick Breeze has done 4 Shawyer / EMDrive YouTube interviews.
BBC would not make a document on just "some" technology.
Quote from: Chrochne on 10/19/2015 06:56 am BBC would not make a document on just "some" technology. Compare Horizon episodes from 30 years ago (they're on youtube) to an episode you get today. Today, all you'll get is a lot of inspiring music and a contentless narrative about how 'the world may never be the same again'. The information density in a typical Horizon episode is so low because they need to fill the air time with shots of scientists looking wistfully at the skies.This will hurt, rather than hinder the efforts to get to the bottom of this issue.