-
#680 by RotoSequence on 24 Dec, 2015 02:47
-
Are your 360 csv files still 100 GB? Probably simpler to mail a DVD at that size. And cheaper, too.
In the realm of burnable media, that'd be two double-sided blu-ray disks. Fortunately, USB drive storage capacity continues its inexorable march upward.
128 gigabytes, $28.91 on Amazon. -
#681 by DnA915 on 24 Dec, 2015 02:57
-
I have 20GB of memory with two 6 core Xeon with hyper-thereading (which reads as 24 cores) running OSX if you want me to run something for you. I don't have MEEP installed, so I'd need a brief intro, but I am a Computer Engineer with come CFD training, so I don't think I'd have too much trouble. Let me know if I can be of service.
- David
Wow - that machine makes my mouth water.
I don't have anything at the moment that needs that much computer but VAXHeadroom may.
The quickest way to get meep up and running is to install a virtual machine with a meep image loaded. Here is one.
http://emdrive.wiki/MEEP#Virtualization:_Virtualbox_image
If you need some realtime input when starting to run meep, I use Pidgin internet messenger. (I think it will interface with whatever chat messenger you use, but it is a free download if you need it.) Anyway, send me a PM and we can get your machine running meep quickly.
That's a pretty cool machine! Problem is for the visualizations I'm trying to do I need the data files and there's no good way to transfer 100GB of data without mailing a USB stick (or 3)
I have what I need at the moment - I'm just trying to visualize what others are digging into - I'm NOT an RF guy, I'm a programmer and pretty good at generating 3D moving pictures to help understand what's going on
And aero was great at getting me up and running meep. If you're going to install VirtualBox I figured out how to create shared directories between the virtual box running Ubuntu and my Win machine (should be similar for Mac) and have written some linux scripts to gen the 360 CSV files I needed
Why do I need to virtualize this on OSX? Should just be a bunch of math classes. I'm going to pull from Github and see what happens:
Hey, have any of you guys used Slack? Its fantastic for coordination, chat, and its free! I use it across the company I run, plus another two working groups. Would you be open to starting an EMDrive Slack for people actively working/helping on this? I can set up a group and send invites if you're interested
http://slack.com
https://github.com/stevengj/meep -
#682 by ThereIWas3 on 24 Dec, 2015 03:09
-
Meeper update: Adding a second antenna directly opposite the first (as I believe Shell's design is doing) does not change the field patterns at all. See my previous post for pretty pictures. But adding the second antenna does result in meep being able to find a resonance at 2.4942 GHz, Q=96063. Perhaps runing the simulation with higher resolution would find subtle effects that were throwing it off in the unbalanced case. I used a resolution of 100.
Next project: modelling a trombone.
-
#683 by DnA915 on 24 Dec, 2015 03:15
-
Why do I need to virtualize this on OSX? Should just be a bunch of math classes. I'm going to pull from Github and see what happens:
Hey, have any of you guys used Slack? Its fantastic for coordination, chat, and its free! I use it across the company I run, plus another two working groups. Would you be open to starting an EMDrive Slack for people actively working/helping on this? I can set up a group and send invites if you're interested
http://slack.com
https://github.com/stevengj/meep
I went ahead and created a new Slack channel. I do think this would help communication. My company has been through 3-4 different programs through the year for realtime remote and local collaboration. This program blows them all out of the water on ease of use and power. Its fantastic and free
https://emdrive.slack.com
Please PM me if you guys want to start using it and I will send out invites and make anyone who is reputable in this forum admins.
-
#684 by ThereIWas3 on 24 Dec, 2015 03:17
-
Are your 360 csv files still 100 GB? Probably simpler to mail a DVD at that size. And cheaper, too.
I do not bother generating CSV files - I make PNGs directly from the H5 files. And the PNG files are not large at all. But CSV files, being text, probably compress pretty well with any of the Lempel-Ziv algorithms.
Here is a tip - if you insert this line into your .ctl file, the H5 files will be about one half their usual size:(set! output-single-precision? #t)
Internally, meep does all calculations in double-precision, and the default is to write out the binary H5 files the same way. By inserting this directive, you tell it to write out in single-precision.
-
#685 by Tellmeagain on 24 Dec, 2015 03:43
-
Progress update on my pendulum-based test.
Good job! Your design is a good one. Looking forward to see your result. I bet no thrust, though. Please make sure your frustum does not touch anything metal. -
#686 by cee on 24 Dec, 2015 05:01
-
Progress update on my pendulum-based test. While I cannot yet report on whether EmDrive is producing thrust or not (congrats to Shell!), I can at least report that a 50 Ohm dummy RF load is definitely not producing any
RFp-Could you explain in a bit more detail what is the force causing the pendulum oscillation of ~.3 Hz ? Also was there any noticeable heat waves from the dummy load ? It might be interesting to add channels for the Frustum side wall and end plates temps. Sensors are reasonable , examples are http://www.ebay.com/itm/151151176084?_trksid=p2057872.m2749.l2649&ssPageName=STRK%3AMEBIDX%3AIT ; feed them with a constant current source .http://www.ebay.com/itm/10pcs-J511-Current-Regulator-Diodes-TO-92-/181025851501. This is a good solid start; rumor has it that the only thing one needs to change now in order to obtain thrust is to use a frustum-shaped cavity instead of a dummy load… We shall see.
Here’s the completed test pendulum with all the components illustrated. The suspended platform is self-contained with a power source, RF generator, RF pre-amp, RF power amp and even a simple RF power meter for visual conformation on whether the forward RF power is as expected. There is also a “virtual wire” link to the base via a pair of XBee modules to control RF power on and off. The base has a high voltage power supply to energize electrostatic plates used as a source of controlled force to the pendulum. It also has a laser position sensor and a DAQ board.
Here’s a 30 s idle run of the system. (This setup turned out to be sensitive not just to my HVAC furnace, but also to the fridge being on where the fridge is ~20 ft away on a concrete garage floor!). Still, the run looks pretty stable. My own DAQ software is detecting min and max points (with some applied hysteresis) and drawing a line via computed mid-points.
Here’s the actual test run. 50 s total: 10 s idle, 10 s RF power on, 10 s idle, 10 s high voltage on, 10 s idle:
This is a zoom-in on the RF on period:
This is a zoom-in on the HV on period:
From this I would infer no force being produced during the RF on cycle, while an easily recognized force can be seen during the HV cycle. Now I just need to replace the RF dummy load with a matched frustum cavity and there should be thrust during the RF cycle, right?
I will be starting with an empty cavity at ~2.35 GHz. Assuming no thrust, will then add some dielectric to it. If there is still no thrust observed then I will need to invest in another RF amplifier to move to ~1.8 GHz range and to replicate the EW design.
Happy Holidays!
P.S. This setup is capable of computer-assisted electrostatic damping. So far this turned out to be the most difficult to implement (and, frankly, a mostly useless) but a pretty cool feature. Numerically solving pendulum differential equations on the fly to arrive at a perfect damping pulse position and duration (there must be an easier way to do this ) -
#687 by aero on 24 Dec, 2015 05:39
-
I have 20GB of memory with two 6 core Xeon with hyper-thereading (which reads as 24 cores) running OSX if you want me to run something for you. I don't have MEEP installed, so I'd need a brief intro, but I am a Computer Engineer with come CFD training, so I don't think I'd have too much trouble. Let me know if I can be of service.
- David
Wow - that machine makes my mouth water.
I don't have anything at the moment that needs that much computer but VAXHeadroom may.
The quickest way to get meep up and running is to install a virtual machine with a meep image loaded. Here is one.
http://emdrive.wiki/MEEP#Virtualization:_Virtualbox_image
If you need some realtime input when starting to run meep, I use Pidgin internet messenger. (I think it will interface with whatever chat messenger you use, but it is a free download if you need it.) Anyway, send me a PM and we can get your machine running meep quickly.
That's a pretty cool machine! Problem is for the visualizations I'm trying to do I need the data files and there's no good way to transfer 100GB of data without mailing a USB stick (or 3)
I have what I need at the moment - I'm just trying to visualize what others are digging into - I'm NOT an RF guy, I'm a programmer and pretty good at generating 3D moving pictures to help understand what's going on
And aero was great at getting me up and running meep. If you're going to install VirtualBox I figured out how to create shared directories between the virtual box running Ubuntu and my Win machine (should be similar for Mac) and have written some linux scripts to gen the 360 CSV files I needed
Why do I need to virtualize this on OSX? Should just be a bunch of math classes. I'm going to pull from Github and see what happens:
Hey, have any of you guys used Slack? Its fantastic for coordination, chat, and its free! I use it across the company I run, plus another two working groups. Would you be open to starting an EMDrive Slack for people actively working/helping on this? I can set up a group and send invites if you're interested
http://slack.com
https://github.com/stevengj/meep
Meep software is quite brittle so everything needs to be just right. Getting a consistent set of source together is the challenge but if you're good at systems administration you should be OK. Virtualize is much faster to install in my experience and the cost in overhead over bare metal is only about 3% so IMO, not worth a lot of trouble to compile. Especially if you have the resources to spare.
But Steven Johnson does do minor updates on occasion so having a working configuration file for a quick update could come in handy. -
#688 by RFPlumber on 24 Dec, 2015 05:46
-
Progress update on my pendulum-based test.
RFp-Could you explain in a bit more detail what is the force causing the pendulum oscillation of ~.3 Hz ? Also was there any noticeable heat waves from the dummy load ? It might be interesting to add channels for the Frustum side wall and end plates temps...
...
The test platform is suspended to a 3m / ~10 ft high ceiling. From T = 2 * pi * sqrt(L/g), the period of oscillation is 3.47 s = 0.29 Hz. (https://en.wikipedia.org/wiki/Pendulum). I can dampen the amplitude to a few tens of microns, the absolute value is not really important within a reasonable range as it is the mid-point of oscillation which is proportional to the external force. From pendulum geometry: m * g * sin A = F; sin A ~ tg A ~ x / L for small values of x, hence x = (L / (m * g)) * F. The platform weight is around 3 kg, which with L of 3 meters translates to roughly 10 micron of force per 1 micron of (mid-point) displacement.
The dummy load gets warm to the touch when exposed to 30W of RF power for a ~minute. However, the RF amplifier is consuming 10A at 14V to deliver those 30W, and the amplifier base plate (aluminum) gets seriously hot to the touch much faster than the load. Also, at 10A this setup drains a 1A-H battery in about 1-2 minutes. As such this setup is intended to be used in a pulsed/short-duration mode. I am thinking 10-20 seconds to span at least a few cycles of pendulum oscillation to be able to see meaningful changes to mid-point.
Xbee link has an option for 2 analog channels back from the pendulum to the base, I was contemplating to use those for capturing both forward and reflected RF power if ever needed. Could definitely be used to capture any other metric. Will bother with any of this only after seeing a mid-point change with RF on
-
#689 by DnA915 on 24 Dec, 2015 05:51
-
Meep software is quite brittle so everything needs to be just right. Getting a consistent set of source together is the challenge but if you're good at systems administration you should be OK. Virtualize is much faster to install in my experience and the cost in overhead over bare metal is only about 3% so IMO, not worth a lot of trouble to compile. Especially if you have the resources to spare.
But Steven Johnson does do minor updates on occasion so having a working configuration file for a quick update could come in handy.
I do have quite a bit of linux/unix server experience, so I will give it 20 min or so and see where it gets me. If not, I'll jump on the bandwagon. -
#690 by Flyby on 24 Dec, 2015 07:37
-
Yesterday...On my way back home, i was able to take this incredible snapshot...
Most definitely some white magic ( sorry TT
) there....
But look....there under the sled... those copper shiny things...aren't those....nah... that cant be...or can it ?
Such a shame the picture is so blurry and with a lot of background noise....
Merry Xmas to all !
-
#691 by TheTraveller on 24 Dec, 2015 11:40
-
Email received from Roger Shawyer to all EmDrive builders:Quote
Hi Phil
Thanks for the greetings, I hope you have a really great Christmas and that your health keeps on improving throughout the New Year.
The BBC spent a whole day filming at the research labs of the company we are working with, so we should get a couple of minutes of programme time! I am told Horizon should be broadcast around Feb next year, with a Discovery version to follow for an American audience.
Please pass on my good wishes to all your fellow EmDrive builders.
Best regards
Roger -
#692 by Justin Space on 24 Dec, 2015 12:06
-
BBC Horizon is very high level and heavily respected so it should be decent. That silly blog writing, armving woman with Wired, who called this all "crap" and this site "tiny" (when these EM Drive threads are into the millions of reads and the team here said it had 1100 people A SECOND visiting during the SpaceX mission this week) will be pulling her hair out such a high level programme is covering it!
-
#693 by ThereIWas3 on 24 Dec, 2015 12:24
-
So is there public interest in just setting up an AWS (Amazon cloud computing) box for MEEP simulations to be done? To be honest it's not the kind of thing that person rigs usually are useful for, because you aren't going to be doing much while MEEP is running. Moreover I've heard that MEEP only multithreads well on Debian.
You might find that Digial Ocean is easier and cheaper. You have a choice of Distros, and even the smallest configuration has a 20GB SSD and a Terabyte of internet transfer included. Setup time for a new instance is under 60 seconds.Meep software is quite brittle so everything needs to be just right. Getting a consistent set of source together is the challenge but if you're good at systems administration you should be OK. Virtualize is much faster to install in my experience and the cost in overhead over bare metal is only about 3% so IMO, not worth a lot of trouble to compile. Especially if you have the resources to spare.
I found everything I needed already present in the Ubuntu repository, so a few clicks in the Installer and everything was ready to go. Check whatever the OSX equivalent is before going to a lot of trouble.
I run meep overnight while I sleep, or when busy doing something else.
-
#694 by Rodal on 24 Dec, 2015 14:42
-
My goodness we are so lucky to have at out beckoned call these machines with gigabytes of ram and terabytes of storage. The very first computer system I did engineering work on was the Philco Ford 1000 with 32k of iron core memory and a asynchronous clock.OK - I have pushed meep as far as I can on my computer (Win 8, 8GB RAM, 1TB HD, 3Ghz quad core i7). Taking aero's CE3 file for SeeShells' device, I have updated the resolution from 250 to 275. This runs with 1 thread (in VirtualBox Ubuntu) using 85% of 6GB of RAM (maxing my computer RAM without thrashing the HD).
Have you tried using mpirun and multiple threads? I found using 3 out of 4 cores on my 6GB 2.1 GHz machine I cut resonance analysis run time about in half.
http://ed-thelen.org/comp-hist/BRL64-0224.jpg
Times have changed, software has changed and it's allowing us to help answer a very interesting problem.
I have a question that's bothering me. I have a calculated Q somewhere around 20-80 k in a TE013 mode and it can show mode movement or it can not according to meep. How long is that mode stable for? What can cause it to decay? How does the decay resemble the build during a increase of Q? Is it a mirrored operation? Does it depend on the power supply of a switching dirty maggie or will it just sit there with a high energy mode. Do ghost modes in a cavity influence the decays and growths in the frustum?
It seems to me that the rise of the mode and Q is governed and shaped more by the RF injection and then the cavity influences the decays more than the RF feed.
Dr. Rodal mentioned this several threads back and its stayed with me causing me to wonder why would a Q buildup in mode mirror the decay I can see where the frustum would be a huge factor in this. To me it's a question that hasn't been addressed, maybe it has and I missed it.
Would love to hear thoughts on this.
Good day away from building and just kind of nesting and thinking.
Shell
Shell, the knowledge base at present cannot answer your questions.., unless all observed thrust is explained by thermal effects and or systemic error.
Which means, if there is thrust clearly above systemic error and thermal effects, it will require something more than the current interpretation of the available knowledge base to answer your questions...
Right now you are our best hope for some clean data, which then may begin to lead somewhere.., that answers to your questions can be found.
I found it very interesting that your frustum walls were cool and the endplates warm....
On the idea that there may be some GR effect involved... Any alteration in Gravity, which is GR.., should be easy to test by attaching an accelerometer to both end plates and turning the thing on. If any distortion in spacetime (gravitation) occurs, it will show up on one or both accelerometers, even if the thing were bolted solid to the floor. Don't you have a raspberry something? that would work for that?, at least as a start. Or would it require something new?
<<found it very interesting that your frustum walls were cool and the endplates warm....>>
One simple interpretation of this, is that the experiment by Shell excited a Transverse Magnetic (TM) mode, as such a mode would result mostly in induction heating of the end plates (due to the transverse magnetic field induction heating the end plates). Purely a thermal effect resulting from classical electromagnetism, as one would expect from a microwave excitation of a cavity. Nothing esoteric.
This was confirmed by Paul March who provided a) thermal camera images and b) finite-element calculations showing the temperature distribution in the frustum of a cone for the transverse magnetic mode.
In more detail:
1) NASA's Finite Element analysis (using COMSOL) -in NASA's experimental case involving the mode shape for the frustum of a cone that is similar to TM212 in a cylindrical cavity
2) NASA's infrared thermal camera -that confirmed the mode shape for the frustum of a cone that is similar to TM212 in a cylindrical cavity
3) See the complete thermal analysis report by NASA, as a PDF attachment if you click this: http://forum.nasaspaceflight.com/index.php?action=dlattach;topic=36313.0;attach=634723
Conclusions:
COMSOL Predictions of Surface thermal losses shows very good agreement with Thermal IR data
• Surface heat distribution follows surface magnetic field distribution and not electric field distribution
• Thermal IR data is consistent with predictions using a single resonant mode (TM212)
PS: Induction heating of the small end in NASA's analysis is affected by the fact that the small end in NASA's experiment contains a polymer dielectric, while in Shell's experiment there is no polymer dielectric at the small end. Analysis without a polymer dielectric (that I conducted with an exact solution, using Mathematica) shows induction heating of both ends. -
#695 by VAXHeadroom on 24 Dec, 2015 15:25
-
Are your 360 csv files still 100 GB? Probably simpler to mail a DVD at that size. And cheaper, too.
I do not bother generating CSV files - I make PNGs directly from the H5 files. And the PNG files are not large at all. But CSV files, being text, probably compress pretty well with any of the Lempel-Ziv algorithms.
Here is a tip - if you insert this line into your .ctl file, the H5 files will be about one half their usual size:(set! output-single-precision? #t)
Internally, meep does all calculations in double-precision, and the default is to write out the binary H5 files the same way. By inserting this directive, you tell it to write out in single-precision.
I'm not making PNG files, I'm making POV-Ray animations, and for that I need POV-Ray include files and to gen those I need the CSV files and for that I need the H5 files. *whew* Having the H5 files means I can create any x/y/z slice I want for the CSVs and then suck those into POV-Ray...
The single-precision is PLENTY for what I doing - thanks for that tip!! That will drastically reduce my HD space - and also the time it takes to calculate/output the cycle for which I'm generating output.
The CSV files are about 2MB each. For that last animation I posted there are 2 horizontal slices so over 700 files which then turn into 224 POV-Ray files (7M each), so *only* about 3G total
-
#696 by OnlyMe on 24 Dec, 2015 15:30
-
My goodness we are so lucky to have at out beckoned call these machines with gigabytes of ram and terabytes of storage. The very first computer system I did engineering work on was the Philco Ford 1000 with 32k of iron core memory and a asynchronous clock.OK - I have pushed meep as far as I can on my computer (Win 8, 8GB RAM, 1TB HD, 3Ghz quad core i7). Taking aero's CE3 file for SeeShells' device, I have updated the resolution from 250 to 275. This runs with 1 thread (in VirtualBox Ubuntu) using 85% of 6GB of RAM (maxing my computer RAM without thrashing the HD).
Have you tried using mpirun and multiple threads? I found using 3 out of 4 cores on my 6GB 2.1 GHz machine I cut resonance analysis run time about in half.
http://ed-thelen.org/comp-hist/BRL64-0224.jpg
Times have changed, software has changed and it's allowing us to help answer a very interesting problem.
I have a question that's bothering me. I have a calculated Q somewhere around 20-80 k in a TE013 mode and it can show mode movement or it can not according to meep. How long is that mode stable for? What can cause it to decay? How does the decay resemble the build during a increase of Q? Is it a mirrored operation? Does it depend on the power supply of a switching dirty maggie or will it just sit there with a high energy mode. Do ghost modes in a cavity influence the decays and growths in the frustum?
It seems to me that the rise of the mode and Q is governed and shaped more by the RF injection and then the cavity influences the decays more than the RF feed.
Dr. Rodal mentioned this several threads back and its stayed with me causing me to wonder why would a Q buildup in mode mirror the decay I can see where the frustum would be a huge factor in this. To me it's a question that hasn't been addressed, maybe it has and I missed it.
Would love to hear thoughts on this.
Good day away from building and just kind of nesting and thinking.
Shell
Shell, the knowledge base at present cannot answer your questions.., unless all observed thrust is explained by thermal effects and or systemic error.
Which means, if there is thrust clearly above systemic error and thermal effects, it will require something more than the current interpretation of the available knowledge base to answer your questions...
Right now you are our best hope for some clean data, which then may begin to lead somewhere.., that answers to your questions can be found.
I found it very interesting that your frustum walls were cool and the endplates warm....
On the idea that there may be some GR effect involved... Any alteration in Gravity, which is GR.., should be easy to test by attaching an accelerometer to both end plates and turning the thing on. If any distortion in spacetime (gravitation) occurs, it will show up on one or both accelerometers, even if the thing were bolted solid to the floor. Don't you have a raspberry something? that would work for that?, at least as a start. Or would it require something new?
<<found it very interesting that your frustum walls were cool and the endplates warm....>>
One simple interpretation of this, is that the experiment by Shell excited a Transverse Magnetic (TM) mode, as such a mode would result mostly in induction heating of the end plates (due to the transverse magnetic field induction heating the end plates). Purely a thermal effect resulting from classical electromagnetism, as one would expect from a microwave excitation of a cavity. Nothing esoteric.
This was confirmed by Paul March who provided a) thermal camera images and b) finite-element calculations showing the temperature distribution in the frustum of a cone for the transverse magnetic mode.
In more detail:
1) NASA's Finite Element analysis (using COMSOL) -in NASA's experimental case involving the mode shape for the frustum of a cone that is similar to TM212 in a cylindrical cavity
2) NASA's infrared thermal camera -that confirmed the mode shape for the frustum of a cone that is similar to TM212 in a cylindrical cavity
3) See the complete thermal analysis report by NASA, as a PDF attachment if you click this: http://forum.nasaspaceflight.com/index.php?action=dlattach;topic=36313.0;attach=634723
Conclusions:
COMSOL Predictions of Surface thermal losses shows very good agreement with Thermal IR data
• Surface heat distribution follows surface magnetic field distribution and not electric field distribution
• Thermal IR data is consistent with predictions using a single resonant mode (TM212)
PS: Induction heating of the small end in NASA's analysis is affected by the fact that the small end in NASA's experiment contains a polymer dielectric, while in Shell's experiment there is no polymer dielectric at the small end. Analysis without a polymer dielectric (that I conducted with an exact solution, using Mathematica) shows induction heating of both ends.
Thanks for the link. Coming into the discussion around the 5th of November, I missed a lot.
My comments on the warm ends and cool side wall was more about the cool walls and her ceramics. Not so much surprise... I think her frustum walls are 1/4" copper while the ends are thinner and bonded to ceramic disks. (Assuming the epoxy was JB Weld.., from the pic, but I think she just called it magic...paste?)
Ihavehad a ceramic coffee cup that when heated in a microwave would get so hot you could not touch it by anything other than the handle, while another ceramic cup would just be warm to touch. Coffee in both at or above boiling... Coffee in the hot cup and the cup itself, would cool down very quickly.., while the other would remain hot longer.
When Shell gets the frustum back in working order and starts generating data, it would be important to know just what the thermal conductivity of her ceramic end plates are, when evaluating thermal data. -
#697 by Rodal on 24 Dec, 2015 15:53
-
...
Thanks for the link. Coming into the discussion around the 5th of November, I missed a lot.
My comments on the warm ends and cool side wall was more about the cool walls and her ceramics. Not so much surprise... I think her frustum walls are 1/4" copper while the ends are thinner and bonded to ceramic disks. (Assuming the epoxy was JB Weld.., from the pic, but I think she just called it magic...paste?)
Ihavehad a ceramic coffee cup that when heated in a microwave would get so hot you could not touch it by anything other than the handle, while another ceramic cup would just be warm to touch. Coffee in both at or above boiling... Coffee in the hot cup and the cup itself, would cool down very quickly.., while the other would remain hot longer.
When Shell gets the frustum back in working order and starts generating data, it would be important to know just what the thermal conductivity of her ceramic end plates are, when evaluating thermal data.
do we know what particular ceramic material Shell used?
Aluminum nitride and silicon carbide thermal conductivity is not that bad (about 1/2 that of copper), they are used in applications that emit heat. Zirconia's thermal conductivity is significantly lower than metals (about 1/20 that of copper), hence used for kiln walls, for example.
Also what matters for the EM Drive is the thermal diffusivity, rather than the thermal conductivity alone, because of the transient nature of the problem, hence the mass density of the ceramic is also important. -
#698 by OnlyMe on 24 Dec, 2015 16:07
-
...
Thanks for the link. Coming into the discussion around the 5th of November, I missed a lot.
My comments on the warm ends and cool side wall was more about the cool walls and her ceramics. Not so much surprise... I think her frustum walls are 1/4" copper while the ends are thinner and bonded to ceramic disks. (Assuming the epoxy was JB Weld.., from the pic, but I think she just called it magic...paste?)
Ihavehad a ceramic coffee cup that when heated in a microwave would get so hot you could not touch it by anything other than the handle, while another ceramic cup would just be warm to touch. Coffee in both at or above boiling... Coffee in the hot cup and the cup itself, would cool down very quickly.., while the other would remain hot longer.
When Shell gets the frustum back in working order and starts generating data, it would be important to know just what the thermal conductivity of her ceramic end plates are, when evaluating thermal data.
do we know what particular ceramic material Shell used?
Aluminum nitride and silicon carbide thermal conductivity is not that bad (about 1/2 that of copper), they are used in applications that emit heat. Zirconia's thermal conductivity is significantly lower than metals (about 1/20 that of copper), hence used for kiln walls, for example.
Also what matters for the EM Drive is the thermal diffusivity, rather than the thermal conductivity alone, because of the transient nature of the problem, hence the mass density of the ceramic is also important.
I don't remember her mentioning specifics, but then I only really read through thread 5 rather quickly. Gave up on the shear volume of the past threads, so what I know from the past is sketchy and based on very limited searches.
From the pictures it does look maybe 3/4 to an inch thick??? -
#699 by OnlyMe on 24 Dec, 2015 16:15
-
Remember I earlier mentioned a casting material used in assembling forges. The purpose was three fold. It could be used to provide a ridge base for spherical end plates, add a ridged backing to thin side walls, used to custom cast either at home (though the cure time takes time).., and it had very good thermal resistance.
It might make using a thinner frustum wall & endplate material workable without an increased danger of thermal distortion... But being thermally insulating it might also increase the total heat build up in the metal?
) there....