Direct Thrust Measurements of An EM Drive and Evaluation of Possible Side-Effects
Direct Thrust Measurements of An EM Drive and Evaluation of Possible Side-Effects
Direct Thrust Measurements of An EM Drive and Evaluation of Possible Side-Effects
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
Nomenclature
c
f
F
P
=
=
=
=
I. Introduction
ll present propulsion systems rely on the exchange of momentum and therefore require either propellant on
board (chemical, nuclear electric propulsion) or an external field/radiation pressure against which they can
push (electromagnetic tethers, solar sails)1. Only the concept of the photon rocket may be seen as a true
propellantless propulsion system as it converts onboard electric power into directed radiation that in turn produces
thrust. However, the thrust is exceptionally small (F=P/c) where megawatts of power are needed to generate milliNewtons of force. Therefore, such a propulsion system has only been studied as a concept so far. Recently, Bae2
demonstrated that this force could be drastically increased by pumping radiation between two highly reflected
mirrors. However, here two satellites are required and only a relative radiation force between the two can be
generated.
Some years ago, Shawyer3-7 claimed to have invented yet another type of propellantless propulsion system called
EMDrive that only uses onboard electrical power similar to the photon rocket, but with orders of magnitude more
1
Professor, Director of Institute and Head of Space Systems Chair, Email: [email protected], Senior
Member AIAA.
2
MSc Graduate.
1
American Institute of Aeronautics and Astronautics
Copyright 2015 by Martin Tajmar. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
thrust and without the need of another satellite. If true, this could certainly revolutionize space travel. His concept is
illustrated in Fig. 1. Microwaves (e.g. generated using a Magnetron) are guided through a waveguide into a tapered
cavity which is highly reflective on the inside boundaries. He refers to the work of Cullen8 who says that the group
velocity of a microwave depends on the diameter of the waveguide. Shaywer then interprets the tapered cavity as
two waveguides with different group velocities and therefore different radiation pressures at the ends. This is
claimed to result in a net radiation pressure that scales with the Q factor (amount of reflections inside the cavity
before the radiation is absorbed) of the cavity. As Newtons law of action and reaction must be conserved, the whole
thruster is then supposed to react against the inside radiation pressure in the opposite direction which can be used to
propel e.g. a spacecraft. The thrust direction should therefore point from the larger to the smaller diameter of the
cavity.
Fthruster
Smaller Group Velocity
Theory
Fthruster = - Fradiation
Magnetron
Fradiation
Net Radiation Force
Waveguide
Fradiation
Tapered Cavity
2
American Institute of Aeronautics and Astronautics
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
We started by designing a model optimized for a frequency of 2.45 GHz using COMSOL in order to be able to
use commercial magnetrons used in standard microwave ovens. We iterated our design several times by consulting
with R. Shawyer to be as representative as possible. Our final tapered cavity design had a top diameter of 38.5 mm,
a bottom diameter of 54.1 mm and a height of 68.6 mm as well as a side entrance for the microwaves as shown in
Fig. 2. The cavity was made out of three copper pieces where the lower and middle part as well as the side flange
were hard soldered using silver and the top part was able to adapt its position in order to optimize for a high Q
factor. A standard WR340 waveguide was then used to connect the magnetron to the EMDrive.
Q=
f0
2.44 GHz
=
= 48.8
f 2 f1 2.45 GHz 2.4 GHz
(1)
This Q factor is of course much smaller compared to the models from Shawyer, Yang and Brady (which was in
the range of 10,000 100,000). A much larger resonance peak appeared above 3 GHz, but as we did not have a
variable frequency microwave source we had to stick to Q50. As our magnetron had an output power of 700 W, we
expected a thrust of 98.2 N according to Shawyers models. This was much higher than the resolution of our
measurement equipment (< 0.1 N) and we therefore decided to go ahead with testing and explore this low Q factor
regime. After all adjustments, epoxy adhesive was used to fix the EMDrives top part on the cavity. Afterwards,
some vibration testing was done and the Q factor measurement repeated to be sure that it does not change after
extensive testing.
Next, we used thermal imaging to determine the temperature profile of our EMDrive under power in air. The
magnetron itself was by far the hottest part. After a pulse of 90 s, the temperature on the microwave generator easily
reached close to 200C while the smaller diameter part of the EMDrive reached about 35C (see Fig. 3b). If we
assume that only the air inside the EMDrive cavity is heated up to an average temperature of 35C with respect to
room temperature, the upward force due to buoyancy is around 70 N and therefore close to the thrust that we want
to measure. Heat and ambient air is therefore indeed a possible error source that needs to be evaluated carefully.
However, it should be also easily identifyable as buoyancy thrust only points upwards. We also wrapped thermal
isolation (glass whool) around the EMDrive and noted only a temperature increase of 4C which was measured
more than 4 minutes after power shut down due to the low thermal conductivity through our isolation.
3
American Institute of Aeronautics and Astronautics
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
III. Measurements
A. Beam Balance Setup
We first tested our EMDrive on a beam balance setup using a sensitive Sartorius AX224 sxale with a resolution
of 0.1 mg which translates into 1 N. Since the EMDrive was much heavier than the maximum 220 g which the
balance can support, the thruster was mounted inside a large aluminum box on one side and counter weights together
with the balance on the other side using a knife-egde balance setup14 on top of a granite table to reduce vibrations as
shown in Fig. 4. The magnetron was connected with three cables to the high-voltage electronics that was powered
by a computer-controlled power supply (two from the HV transformer and one grounding cable). After installation,
the box was sealed using an aluminum sheet and tape around the box such that hot air can not easily escape the
measurement box. All other surface-edges inside the box where sealed using silicon.
4
American Institute of Aeronautics and Astronautics
Fig. 5a shows the summary of thruster upwards direction measurements comparing the different isolation
methods. The EMDrive was powered up for 15 s with a 600 s delay time before and after the impulse to ensure that
any heat generation was sufficiently removed before the next run. The measurements were repeated and signal
averaged up to 38 times. We can see in Fig. 5a that there is obviously a turn-on effect and then a steady increase
after abount 8 s of the measured weight change (thrust). In case of no countermeasure/isolation, the weight change
also further increased after turn-off. Thermal shielding significantly reduced the buoyancy effect and the addition of
magnetic shielding did not change much of the observed offsets (only the offset at the end of the power peak was
different). The implementation of all isolation methods (thermal, magnetic, air circulation block) resulted in the
cleanest measurement with an expected behavior such that the thrust appeared after turn-on, then steadily increaseed
until power turn off. It then remained there and slowly decreased as the EMDrive cooled down.
Fig. 5b shows then the comparison of upwards, downwards and vertical thruster direction measurements using
the full isolation configuration. Remarkably, we can indeed see a farily large difference between thrust directions.
The difference between upwards and downwards measurements was 229 N and therefore close to our expectation
of 2x98 N. The horizontal direction was supposed to be our zero thrust reference, and indeed it was about only 1/3
of the downwards measurement. Our observations are as follows:
The balance configuration seems to indeed measure thrust in the correct direction and magnitude as claimed by
Shawyer.
The horizontal direction was supposed to measure only thermal effects and no thrust. We observed a turn-on
effect (of the same magnitude compared to other thrust directions but with an opposite value) and then an
increase to about 100 N until the power was turned off. We then saw a behavior that was indeed expected from
a thermal side-effect: The thrust still further increased a bit (delay from thermal shielding) and then went down
to zero.
The thruster up/down direction showed a very different behavior. They increased to 620 N and 391 N
respectively and then remained constant for a much larger time compared to the horizontal direction. A different
orientation of the magnetron (horizontal versus vertical) may have caused different thermal signatures and
therefore buoyancy effects. Still, this behavior was really different and repeatable. In the much lower power
measurements from Brady et al13 on the torsion balance, we can also see that it took some time after power turnoff that the balance reading went back to zero as if the EMDrive got somehow charged and produced thrust
which rather decays contrary to a simple switch off after power is removed.
1500
1250
1000
750
500
250
0
-250
-500
280
290
300
310
320
330
340
750
Sartorius Balance
Thermal+Magnetic Shield+Circulation Block
250
500
Thrust=229 N
200
150
250
100
50
0
-50
-250
280
290
300
Time [s]
300
Up
Down
Horizontal (Null)
1750
300
275
250
225
200
175
150
125
100
75
50
25
0
-25
-50
-75
350
Force [N]
No Countermeasure
Thermal Shield
Thermal+Magnetic Shield
Thermal+Magnetic Shield+Circulation Block
2000
Force [N]
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
Moreover, we also checked if the operation of the EMDrive itself does influence the Sartorius balance by
powering it up in the same setup but using less counter weight such that the balance was free. The balance reading
was stable during turn-on/off and therefore no electromagnetic influence was seen.
310
320
330
340
-100
350
Time [s]
b.) Summary of All Directions with Thermal-, MagneticShielding and Air Circulation Block
Figure 5. Summary of Tests with Balance Setup (Time Axis shows only Part).
5
American Institute of Aeronautics and Astronautics
Our weakest part in this setup was certainly the simple connection of the magnetron with three flexible silicon
isolated wires to the power supply. A current of several Ampere is flowing over those wires which can generate
significant magnetic forces (although we tried to keep the wires close together such that the magnetic effects cancel)
that may have influenced our measurements. This together with the buoyancy effect made this measurement setup
less convincing compared to a torsion balance setup.
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
The control experiment (vertical upwards direction) actually gave the biggest thrust with up to 224 N. We
could again see a turn-on effect and a steady increase during the power pulse until power turn-off. After that,
the thrust values again remained at their high offset and gradually decreased. The slope actually followed quite
well the temperature of the magnetron that rose up to around 190C and then gradually decreased at the same
rate.
The horizontal measurements for positive and negative thruster orientations rose during the pulse up to 96 N
and 145 N respectively. They showed a similar behavior compared to the vertical direction with a somehow
faster decay after power turn-off. The thrust values were now reversed (the positive thrust was smaller
compared to the negative thrust). However, considering the fact that the control experiment gave the largest
thrust, no conclusion can be derived here. The difference in thrust was 49 N which led to 24.5 N for each
direction that is about 25% of the thrust prediction according to Shawyer and our measurements with the knifeedge setup before.
We were really puzzled by this large thrust from our control experiment where we expected to measure zero. The
power signal to the magnetron consisted of a heater current (up to 5A) which was on high voltage (2000 V) with
respect to ground. We disconnected the high voltage power electronics and connected a high voltage power supply
running only the same 2000 V through the two cables without any current to check if that created any false signal
which it did not. Only when a large current was flowing through the magnetron cables, a large apparent thrust was
measured. Therefore, we believed that the anomalous signal must be due to magnetic interaction with our permanent
magnet damping.
250
400
220V
200V
150V
100V
50V
350
200
150
100
100
50
50
0
300
Force [N]
150
Temp.
Positiv
Negative
Control (Up)
1,0
250
Power Supply
Shut-Off
200
0,5
150
100
50
0,0
-50
1700
-50
1750
1800
1850
1900
1950
2000
Time [s]
200
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
Our testing was first done using magnetic eddy-current damping similar to the setup from Brady et al13. We
started with tests in ambient air (but closed vacuum chamber) as summarized in Fig. 8a. We performed a 40 s
impulse with 900 s before and afterwards during each run to allow suffient time for the magnetron to cool down. A
temperature sensor (K thermocouple) was mounted on the magnetron and the temperature was logged during the
experiments. This ensured that all tests were done with the same thermal signature. Again, all tests (usually up to 10)
are signal averaged. Our observations are as follows:
-50
0
10
20
30
40
50
Time [s]
Figure 8. Summary of Tests on Torsion Balance with Magnetic Damping (Time Axis shows only Part).
However, before changing damping, we tried to assess if air heating/buoyancy effects could still play a role as
the signal followed the temperature decay from the magnetron. Therefore, we tested the large horizontal negative
thrust direction in high vacuum by evacuating the chamber down to 410-6 mbar. As shown in Fig. 8b, we gradually
increased the voltage of our pulse that went into the high voltage power transformer from 50 V up to 220 V (usually
we operated at 230 V) exceptionally, only single measurements were recorded here. According to a magnetron
manufacturer, microwaves are starting to be generated if around 150 V are suppled to the HV transformer however
with lower power (unfortunately, we had no equipment to measure the power level at those voltages, but we noted
that if we operated below 230 V the temperature on the magnetron did not increase). Indeed we saw that only after
reaching 150 V, a thrust appeared on the balance similar in value to the one on air (no thrust for 50 and 100 V). This
thrust even increased at 200 V to 325 N. Interestingly, the thrust now also remained stable and did not immediately
7
American Institute of Aeronautics and Astronautics
return to zero after power turn-off just as we have seen it with the measurements done on air in Fig. 8a. But when
we reached 220 V, the power supply shut down due to over current protection around 5 seconds after starting the
pulse. Before the power supply failure, the thrust value was very similar to the 200 V case and it decreased to zero
shortly after power shutdown (maybe no time to charge up the EMDrive). We later found out that there was a thin
grey film around our liquid metal cups as if liquid metal had evaporated creating a shortcut. At least we could show
that the thrust we measured in vacuum had a similar shape compared to the measurements in air and that they can
not be due to any air-related side effect (at least not all of it).
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
In order to check the magnetic influence hypothesis, we completely removed the permanent magnet from the
base of our balance and replaced it with a cup of oil and a fin dipping into the oil and mounted on the balance (see
Fig. 9a). In addition, we switched the magnetron position such that it now pointed outwards and therefore as far
away as possible from our liquid metal connection (see Fig. 9b).
Figure 9. Torsion Balance Setup with Oil Fluid Damping and Magnetron on Outer Position.
Fig. 10 shows our measurements in this setup with oil fluid damping. The damping here is less effective as with
the magnetic eddy-currents, however, we can still achieve sub-N thrust resolutions. In Fig. 9a, a summary of all
thrust directions is shown. Our observations are as follows:
We could see the typical balance oscillations and that the thrust values were now greatly reduced.
Still we noted that the vertical direction (upwards) gave a thrust of around 24 N which immediately droped to
zero when the power was switched off.
The positive thrust orientation now also went positive up to a value of 18 N slightly below the vertical
direction.
The negative thrust orientation went indeed negative down to -27 N. This was the first time that we have
actually seen a real thrust reversal. The thrust orientations now coincide again with Shawyers predictions and
our earlier knife-edge measurements. Surprisingly, here also the thrust remained at an offset that slowly
degradeed. To a minor extend this was also true for the positive orientation. This might actually be a sign for a
genuine thrust produced by the EMDrive.
In Figs. 10b-c, we show the positive and negative thrust pulses together with control runs powering the
magnetron electronics only with 150 V (onset of microwave generation) compared to 230 V. No clear difference can
be seen for the positive direction but a clear difference (within the resolution of our measurement) is visible for the
negative thrust direction. The magnitude of these thrust measurements are simiar to the ones where magnetic
damping was used and are therefore 25% of the original prediction.
8
American Institute of Aeronautics and Astronautics
75
150
50
100
25
50
0
-50
-25
-100
-50
-75
2660
2680
2700
2720
2740
2760
2780
-150
2800
Time [s]
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
250
75
150
50
100
25
50
Thrust 18 N
0
0
-50
-25
100
200
-100
-50
200
150
50
100
25
50
0
-25
0
-50
Thrust -27 N
-100
-50
-150
-75
2600
250
75
2650
2700
2750
2800
2850
2900
-150
-75
2600
Time [s]
200
100
Force [N]
250
100
Force [N]
Temp.
Positiv
Negative
Control (Up)
Force [N]
125
2650
2700
2750
2800
2850
2900
Time [s]
Summary of Tests on Torsion Balance with Oil Fluid Damping (Time Axis shows only Part).
In order to check if the lower thrust may be due to an even lower Q factor at the end of our extensive test
campaign, we performed another resonance measurement. Indeed we measured that our Q factor was reduced to
only 20.3 probably due to the fact that our inner surfaces were now much more oxidized compared to the start of
our test campaign after a visual inspection. This reduces our theoretical thrust to 41 N which is only a third less
of what we have measured in our last runs and is therefore well within the expected range.
IV. Conclusion
We have built and tested an EMDrive using a commercial standard magnetron with a resonance frequency of
2.44 GHz and 700 W of power in setups similar to the ones used in the past in order to assess possible side effects
and their claimed thrust values. Our thruster had a considerably smaller Q factor (around 50 for the first tests and 20
at the end) compared to others (10,000 100,000), however our test facilities had a higher sensitivity as well.
Our first tests were done with a knife-edge balance configuration and we assessed different isolation scenarios in
order to see any thermal or electromagnetic influence. As expected, we noticed a large thermal effect that could be
significantly reduced by thermal isolation and by blocking any air circulation inside our measurement box. We
indeed found thrusts that changed with the orientation of the thruster and magnitudes in line with the theoretical
predictions for our low Q factor. After turning off the power, the thrust values in the order of several hundred N
remained and slowly degraded after power shut-off. Considering that the EMDrive and especially the magnetron
mounted on it can get hot, such a setup does not seem to be able to adequately measure precise thrusts.
We continued with testing on a torsion balance inside a vacuum chamber. Here we also found thrusts but quickly
realized that there was a strong interaction with our magnetic damping system. Still we used this setup to test an
EMDrive for the first time in high vacuum down to 410-6 mbar observing similar thrusts (although at somewhat
lower power levels) ruling out any air influence in this configuration. After changing the position of the magnetron
(outer position) and replacing the magnetic damping with oil fluid damping, surprisingly we could still observe
thrusts that are indeed reversing with thruster orientation but with control runs in vertical direction producing similar
thrusts compared to the positive direction. However, negative thrusts were only observed with firing the thruster
indeed in a negative direction. Running the magnetron also in this direction at lower voltages produced similar
positive values as the vertical control experiment. The thrusts observed with the oil-damped torsion balance were
close to the original prediction taking our small Q factor into account (around +/- 20 N for 700 W of microwave
power still an order of magnitude more effective than pure radiation thrust). We also observed that the thrust
appeared not to go down to zero immediately after power is switched-off but rather noted a gradual decrease as if
the EMDrive was charged up and slowly reduced its thrust effect.
The nature of the thrusts observed is still unclear. Additional tests need to be carried out to study the magnetic
interaction of the power feeding lines used for the liquid metal contacts. Our test campaign can not confirm or refute
the claims of the EMDrive but intends to independently assess possible side-effects in the measurements methods
used so far. Nevertheless, we do observe thrusts close to the magnitude of the actual predictions after eliminating
many possible error sources that should warrant further investigation into the phenomena. Next steps include better
magnetic shielding, further vacuum tests and improved EMDrive models with higher Q factors and electronics that
allow tuning for optimal operation. As a worst case we may find how to effectively shield thrust balances from
magnetic fields.
9
American Institute of Aeronautics and Astronautics
Acknowledgement
We would like to thank Roger Shawyer for his assistance, attocube (Dr. Martin Zech) for their generous support
and supply of the FPS interferometric sensor for our test campaign and Prof. Plettemeier from TU Dresden who
assisted our Q factor measurements.
References
Downloaded by CSIR - NATIONAL AEROSPACE LABS on July 28, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4083
1Tajmar,
10
American Institute of Aeronautics and Astronautics