Author Topic: EM Drive Developments - related to space flight applications - Thread 3  (Read 3131456 times)

Offline Rodal

  • Senior Member
  • *****
  • Posts: 5911
  • USA
  • Liked: 6124
  • Likes Given: 5564
...Wasn't Shawyer's equation some how based on cylinders also? ...
Notsosureofit's thrust force equation is based on a cylinder.  Notsosureofit's shows you this up-front in a clear manner and does this because there is a closed-form solution for a cylinder but not for a truncated cone.

Ditto for McCulloch: he states his assumptions and approximations used to calculate a thrust force.

Shawyer uses several approximate equations as well, for the same reason (because there is no closed-form solution for a truncated cone).  Shawyer does not clearly state what are his approximations.

According to TheTraveller's post in the last few pages,  Shawyer is using the cut-off equation for a cylinder.


...Could it be that the best force is some how slightly off resonance? ...
According to TheTraveller, Shawyer is seeking maximum Q, hence maximum resonance. Being slightly off-resonance means much smaller Q.

According to what I recall Paul March stating from his experiments there is no unique monotonous relationship between thrust force and Q.  The Brady report shows cases where a lower Q produced a higher force.

Mode Frequency(MHz) , Q   Input Power (W) Peak Thrust (μN)
TM211 1932.6            7,320 16.9                116.0
TM211 1936.7         18,100  16.7                    54.1

Same mode, same frequency, same power (practically) :

Notice that a  Q less than half as high (7320 instead of 18100) resulted in twice as high a thrust force (116 instead of 54)

In Thread 2, Paul March shows other cases where there is no monotonic relationship between Q and thrust force.

... Do we have plots of force as frequency is changed from peak resonance to slightly off?  ...
I have not seen data plotted that way.  However, we have data from NASA Eagleworks showing that lower Q sometimes produces higher force


... Could throwing in the dielectric slightly throw it off resonance?

A dielectric insert lowers the natural frequency.  However the natural frequency still has a peak with side bands, and a Q.  The  Q with a dielectric is always lower because of tan delta losses.
« Last Edit: 06/01/2015 12:05 am by Rodal »

Offline TheTraveller

Shawyer uses several approximate equations as well, for the same reason (because there is no closed-form solution for a truncated cone).  Unlike Notsosureofit, Shawyer does not clearly state what are his approximations. I find Shawyer's papers, regarding his equations and his free-body-diagrams to be very unorthodox.

According to TheTraveller's post in the last few pages,  Shawyer is using the cut-off equation for a cylinder.

SPR then numerically integrates a lot of these points along the side wall to get the effective guide wavelength. From that value, end plate spacing is then calculated to get resonance at the desired number of 1/2 waves at the external Rf frequency and excitation mode.
It Is Time For The EmDrive To Come Out Of The Shadows

Offline Mulletron

  • Full Member
  • ****
  • Posts: 1150
  • Liked: 837
  • Likes Given: 1071
Ok, so I'm being told by PM that IRT this post above,
http://forum.nasaspaceflight.com/index.php?topic=37642.msg1382564#msg1382564

these two thrusters aren't the same flight thruster. What I'm getting from TheTraveller is that the flight thruster that Shawyer is standing next to in the pic above is copper (kinda does look coppery when you zoom in), and that the one on the table is copper inside/and coated outside. I'm not sure. @TheTraveller, did I get that right?



« Last Edit: 06/01/2015 12:00 am by Mulletron »
And I can feel the change in the wind right now - Rod Stewart

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97
Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
« Last Edit: 06/01/2015 12:09 am by phaseshift »
"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97


these two thrusters aren't the same flight thruster. What I'm getting from TheTraveller is that the flight thruster that Shawyer is standing next to in the pic above is copper (kinda does look coppery when you zoom in), and that the one on the table is copper inside/and coated outside. I'm not sure. @TheTraveller, did I get that right?


It would be really nice if there was a thumbnail of each drive on the wiki - next to their experimental results.  I've gotten confused often as well as to which is which
"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline aero

  • Senior Member
  • *****
  • Posts: 3629
  • 92129
  • Liked: 1146
  • Likes Given: 360
Guess what! After struggling for months trying to understand how to model dielectric losses, publishing my control file and relaxing for a day, I came across this paper, using the right Google search terms:
http://www.satcs.co.za/TanD-Res-info.pdf
Its a lot more than I want, but the authors included background which is what I was missing.

e* = e' -je" and tan d = e" / e'

Using this, and Paul's note
Quote
The high density polyethylene discs dielectric's relative permittivity is 2.27 at 2.0 GHz with a dissipation factor of ~0.0005.

HDPE e* = 2.7 - j 0.0035

Now I might be able to code that into Meep!

I'll have to assume that Paul used e* = e' - je", and not the other version I've seen, e* = e' + ie"

Oh well, can't have everything.

Now, if I happen to run across the loss tangent for copper at 2 GHz, I'll be golden.
Retired, working interesting problems

Offline TheTraveller

Ok, so I'm being told by PM that IRT this post above,
http://forum.nasaspaceflight.com/index.php?topic=37642.msg1382564#msg1382564

these two thrusters aren't the same flight thruster. What I'm getting from TheTraveller is that the flight thruster that Shawyer is standing next to in the pic above is copper (kinda does look coppery when you zoom in), and that the one on the table is copper inside/and coated outside. I'm not sure. @TheTraveller, did I get that right?

What I know for sure is Shawyer said making my Flight Thruster replication from copper was fine. I did ask about doing silver and gold platings inside as EW did.  Was told the inside should be highly polished copper with no other treatment or coatings needed.
« Last Edit: 06/01/2015 12:32 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97
@TheTraveller,

Just a thought, but if you use Google Sheets, instead of Excel, you can write your own functions in javascript instead of using 1000s of cells to perform calculations and do vlookups.

I haven't done it in years but its pretty easy. Conceptually it might be easier in some cases.

For example, instead of the 4000 cells used to calculate the resonate guide wavelength you can use something like this (though this is in ruby it will translate to javascript with ease):


def compute_resonate_guide_wavelength( small_diameter_meters, large_diameter_meters, frequency_hz, steps, jC )

    #---------------------------------------------------------------------------
    # set up some variables that are used as constants in the loop
    # jC = BesselJ Cutoff
    #---------------------------------------------------------------------------
       
    cM = 299705000.0
    total = 0
    delta = ( large_diameter_meters - small_diameter_meters) / steps
    cf = cM / frequency_hz
       
    y = ( cM * jC ) / ( Math::PI * frequency_hz )
   
    lambdaG1 = cf / Math.sqrt( 1.0 - ( y / large_diameter_meters ) ** 2.0 )
   
    #---------------------------------------------------------------------------
    # loop from large_diameter_meters to small_diameter_meters by -delta
    # add the result of each step to the total
    #---------------------------------------------------------------------------
   
    for i in 1..steps
       
        diameter = large_diameter_meters - delta * i
        lambdaG2 = cf / Math.sqrt( 1.0 - ( y / diameter ) ** 2.0 )
       
        total += delta * ( lambdaG1 + lambdaG2) / 2.0
        lambdaG1 = lambdaG2
       
    end

    rgw = total / (steps+1) / delta
   
    return rgw
   
end


Cheers, hope this might help
« Last Edit: 06/01/2015 12:50 am by phaseshift »
"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline TheTraveller

@TheTraveller,

Just a thought, but if you use Google Sheets, instead of Excel, you can write your own functions in javascript instead of using 1000s of cells to perform calculations and do vlookups.

I haven't done it in years but its pretty easy. Conceptually it might be easier in some cases.

Cheers

I did it the way I did so I could graph the guide wavelength change from end to end and be able to see if and where it hit cutoff.

I'm sure others will come up with many ways to do this. At least now it is understood how SPR calcs end plate spacing to get resonance at the desired excitation mode number of 1/2 waves and applied Rf frequency.
« Last Edit: 06/01/2015 12:53 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97
@TheTraveller,

Just a thought, but if you use Google Sheets, instead of Excel, you can write your own functions in javascript instead of using 1000s of cells to perform calculations and do vlookups.

I haven't done it in years but its pretty easy. Conceptually it might be easier in some cases.

Cheers

I did it the way I did so I could graph the guide wavelength change from end to end and be able to see if and where it hit cutoff.

A javascript callback would take care of that (I think). But yeah, I see why you did it that way
"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline rfmwguy

  • EmDrive Builder (retired)
  • Senior Member
  • *****
  • Posts: 2205
  • Liked: 2713
  • Likes Given: 1134
I think it is important for all to remember that the EM drive is a physical system, not a mathematical one. In the physical system, cutoff is not a line in the sand that you shall not cross, rather it is (probably) the center of a range where propagation drops below some relative value of db. The EM drive will do as it does over a range of frequencies, plus or minus, just some will do better than others.

Add: Its also important that the magnatron drive is a noisy source so the cavity will select its own operating frequency. It would be nice to have the maximum power transfer from source to cavity but very often "Perfect" is the enemy of "Good enough."

Classic definition of cutoff is 3db. A single cavity will not have a steep shape factor in stopbands, no brickwall is correct. Return loss will be much more transitional in the passband... IOW i'd design and tune for best S11 performance at center frequency simply to keep the signal source protected as a matter of safety and efficiency. Also the bessel function has a shallower shape factor. Its best known characteristic is flat group time delay in passband for radar/pulse applications. Really, the frustum is a poor bandpass, being so assymetrical around a center frequency....but maybe that's part of the mystery  ;)

Offline TheTraveller

I think it is important for all to remember that the EM drive is a physical system, not a mathematical one. In the physical system, cutoff is not a line in the sand that you shall not cross, rather it is (probably) the center of a range where propagation drops below some relative value of db. The EM drive will do as it does over a range of frequencies, plus or minus, just some will do better than others.

Add: Its also important that the magnatron drive is a noisy source so the cavity will select its own operating frequency. It would be nice to have the maximum power transfer from source to cavity but very often "Perfect" is the enemy of "Good enough."

Classic definition of cutoff is 3db. A single cavity will not have a steep shape factor in stopbands, no brickwall is correct. Return loss will be much more transitional in the passband... IOW i'd design and tune for best S11 performance at center frequency simply to keep the signal source protected as a matter of safety and efficiency. Also the bessel function has a shallower shape factor. Its best known characteristic is flat group time delay in passband for radar/pulse applications. Really, the frustum is a poor bandpass, being so assymetrical around a center frequency....but maybe that's part of the mystery  ;)

These high Q frustums may be difficult to deal with.

As example assuming 3.85GHZ resonance at Q = 60,000. Bandwidth at -3db points of 64kHz or 32kHz either side of ideal resonance. Then assuming we wish to operate at max 50% of that deviation, we need to hold excitation frequency to +-16kHz of the ideal and at the same time track resonate changes due to thermal expansion.

This is doable but not so easy as blasting away with a wide band magnetron into a lower Q cavity with flat end plates as against spherical end plates and a narrow band Rf generator.

Will shortly present the system I'm putting together to enable resonance tracking as the frustum warms up and alters it's length and end plate diameters.

Intend to use a very slow sweep spectrum analyser, over a tight frequency range, via an Rf coax switch that samples what is happening to the cavity via the same antenna that excites the cavity, then adjusts Rf frequency to get close, then use real time closed loop thrust feedback to centre the external Rf to the centre of the thrust curve.

Also intent to do thrust bandwidths to see how much external freq variance affects generated thrust, at the same -3db down points. So will measure both conventional bandwidth and thrust bandwidth. Then will start to get a feel for how sensitive this beast is to frequency variance versus generated thrust.

Using a wide band magnetron Rf source, there is little chance of being able to directly measure thrust bandwidth or even cavity Rf Q. So while the narrow band pathway will be slower to get there, it should yield much more interesting data.
« Last Edit: 06/01/2015 02:02 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Offline deltaMass

  • Full Member
  • ****
  • Posts: 955
  • A Brit in California
  • Liked: 671
  • Likes Given: 275
Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Offline TheTraveller

Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Cost of the high Q is your narrow band Rf generator needs to be able to track frustum resonance changes via some feedback mechanism. The Rf generator I'll be using can move in 1kHz increments, so should be OK to keep the frequency in the middle of the thrust bandwidth.

BTW I plan to work in pulsed mode. Do a pulse, measure thrust, switch the coax feed to the spectrum analyser, measure the max Q frequency, switch back the excitation antenna to the Rf amp, adjust the freq if necessary and give it another pulse. Around the loop it goes. Should be able to do this many times a second.

That way I get 2 feedback channels to help to initially keep the excitation frequency in the middle of the Q bandwidth and/or in the centre of the thrust bandwidth once it start to develop.
« Last Edit: 06/01/2015 02:32 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97
Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Speaks loudly for a tuning cylinder or RF tuner. :) 

At 1000 slices there were 4 significant digits, 50,000 slices there were 5 significant digits, 5,000,000 slices raised it to 6 significant digits - but as it's just an approximation anyway 1000 slices seems adequate considering it needs to be tuned anyway.

« Last Edit: 06/01/2015 02:25 am by phaseshift »
"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline TheTraveller

Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Speaks loudly for a tuning cylinder or RF tuner. :) 

At 1000 slices there were 4 significant digits, 50,000 slices there were 5 significant digits, 5,000,000 slices raised it to 6 significant digits - but as it's just an approximation anyway 1000 slices seems adequate considering it needs to be tuned anyway.

It needs to be continually tuned is what Shawyer said to me. ;)
« Last Edit: 06/01/2015 02:30 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Offline txdrive

  • Member
  • Posts: 48
  • Liked: 47
  • Likes Given: 14
Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Cost of the high Q is your narrow band Rf generator needs to be able to track frustum resonance changes via some feedback mechanism. The Rf generator I'll be using can move in 1kHz increments, so should be OK to keep the frequency in the middle of the thrust bandwidth.

BTW I plan to work in pulsed mode. Do a pulse, measure thrust, switch the coax feed to the spectrum analyser, measure the max Q frequency, switch back the excitation antenna to the Rf amp, adjust the freq if necessary and give it another pulse. Around the loop it goes. Should be able to do this many times a second.

That way I get 2 feedback channels to help to initially keep the excitation frequency in the middle of the Q bandwidth and/or inthe centre of the thrust bandwidth once it start to develop.
Well, you could just use an amplifier, some high speed clamping diodes, and a moveable feedback antenna (feeding back into the amplifier). With some passives to dampen unwanted frequencies, the cavity itself will very directly determine the frequency of oscillations.

Not that this all matters, because 1: Shawyer's formula for computing thrust has been definitely falsified by EagleWorks down to 1..2% , and 2: EagleWorks level of thrust is within the range of inaccuracy (as estimated from the disparity between the thrust and the thrust when the device is rotated 180 degrees).

More accurate experiments exist with regards to microwaves in the vacuum not exchanging momentum with anything unknown down to parts per trillion or better (an exchange in momentum implies that a photon changes it's wavelength or is absorbed, by the way, where the absorption by the unknown will only get you photon rocket level thrust, and change of wavelength would be incredibly noticeable by various electronics), so the prior experiments also weight heavily against it working.

This is merely one screw short of being normal science, one screw being having a sensible estimate of error bounds. I'm sure they'll get it right eventually. (Of course, if you are missing some crucial little thing in science, it all falls apart, hence the "confirmations" and hype).
« Last Edit: 06/01/2015 02:48 am by txdrive »

Offline phaseshift

  • Full Member
  • *
  • Posts: 104
  • Seattle, WA
  • Liked: 84
  • Likes Given: 97
Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Speaks loudly for a tuning cylinder or RF tuner. :) 

At 1000 slices there were 4 significant digits, 50,000 slices there were 5 significant digits, 5,000,000 slices raised it to 6 significant digits - but as it's just an approximation anyway 1000 slices seems adequate considering it needs to be tuned anyway.

It needs to be continually tuned is what Shawyer said to me. ;)

Exactly! So an approximate solution is adequate IMO.

Can you sample and feed at the same location? I thought the sample needed to be at center and the feed toward one end or the other?

"It doesn't have to be a brain storm, a drizzle will often do" - phaseshift

Offline TheTraveller

Just for fun I took the TheTraveler's equations for determining the end plate spacing, but instead of slicing the frustum into 1000 cylinders I sliced it into 5,000,000. 

His result: 139.22907 millimeters
My result: 139.3682 millimeters

difference of 0.139 millimeters

From what TheTraveler has written, and as I have understood it, Shawyer used a similar technique to perform approximate calculations, and as others have said "good enough to get the job done". I'll have to go with this for now until something better comes along. Thank You TheTraveler.

In order to figure out why the EM drive works we can't use this technique (probably), but to build one this seems good enough.

:)
A simplistic way to see if this makes the nut for construction is to multiply the discrepancy by Q and see if the result remains substantially less than a quarter of a wavelength. So for 2.5 GHz (lambda/4=30mm),
60,000*0.139 = 8340 mm. Well, sorry about that, but that's the price for high Q.

Speaks loudly for a tuning cylinder or RF tuner. :) 

At 1000 slices there were 4 significant digits, 50,000 slices there were 5 significant digits, 5,000,000 slices raised it to 6 significant digits - but as it's just an approximation anyway 1000 slices seems adequate considering it needs to be tuned anyway.

It needs to be continually tuned is what Shawyer said to me. ;)

Exactly! So an approximate solution is adequate IMO.

Can you sample and feed at the same location? I thought the sample needed to be at center and the feed toward one end or the other?

Prefer to sample via excitation antenna, so only one version of what is happening inside. Besides what the excitation antenna see is where the action is. So better to sense from there. Less unknowns.
It Is Time For The EmDrive To Come Out Of The Shadows

Offline TheTraveller

Not that this all matters, because 1: Shawyer's formula for computing thrust has been definitely falsified by EagleWorks down to 1..2%

EW did not use the correct excitation frequency as per the SPR resonance method, so nothing is disproven.

Really don't understand why no one asked Roger Shawyer how he calcs resonance frequency. I did, he explained the process and how we have my EM Drive Calculator that can do the job.
« Last Edit: 06/01/2015 02:51 am by TheTraveller »
It Is Time For The EmDrive To Come Out Of The Shadows

Tags:
 

Advertisement NovaTech
Advertisement Northrop Grumman
Advertisement
Advertisement Margaritaville Beach Resort South Padre Island
Advertisement Brady Kenniston
Advertisement NextSpaceflight
Advertisement Nathan Barker Photography
0