Investigation of a Partially Loaded Resonant Cavity_Zeller_Kraft
G. Vector Network Analyzer Testing
1. Vector Network Analyzer Resonance Analysis
By Kurt Zeller and Brian Kraft
In order to examine the EM Drive phenomenon several design iterations were made of
an adjustable partially loaded cylindrical resonant cavity. Several dielectric materials
were introduced in varying thicknesses such as high-density Polyethylene (HDPE),
Plexiglas, and Nylon and resonance sweeps were performed using an Anritsu
MS4622B Vector Network Analyzer (VNA). These experimental analyses were
performed over the course of six months and results varied wildly between design
iterations. The primary conclusion gleaned is that resonant cavities operating near
microwave oven magnetron frequency (2.45GHz) are extremely sensitive to
dimensional tolerances, antenna quality, surface finish, and potentially ambient
conditions.
Preliminary design
The first design attempted as seen in the document labeled "BKZ1 Drawings" details an
adjustable resonant cavity with the RF source antenna placed directly into the cylindrical cavity.
After performing initial VNA frequency sweeps using our home-made antenna seen in document
labeled "Magnetron Dissection", it was readily apparent that it would not resonate. Although
analytical solutions were performed for a cylindrical resonant cavity as can be seen in
"Cylindrical Cavity Design", the antenna placement resulted in a minimum reflection coefficient
of -10dB. For perspective, RF engineers commonly cite that a reflection coefficient of less
than -20dB is a 'good match' when designing a radome or similar bandpass filter. Because the
antenna placement directly in the end of cylinder did not produce the desired resonance, it was
determined that a waveguide launcher would be necessary in order to provide an impedance
transition between the antenna and the cavity.
Secondary design
It was quickly realized that the microwave oven has an impedance matched waveguide
built in to the structure and that it could be easily cut out with a jig-saw. After cutting out two
waveguides and drilling holes in the corners to mount to the end of the cylinder, preliminary
VNA sweeps looked very promising. Note that "non conducting/conducting plate" refers to the
internal movable plate being electrically attached to the inside of the cylinder. It was later found
that this electrical contact was essential to prevent arcing from occurring during full power
magnetron testing. The first rounds of VNA sweeps can be seen at the end of this document for
7/29 and 7/30/15. Also note that the Mica insert seen below came stock from the microwave
oven in order to prevent food and vapor from entering the waveguide area. Several tests were
done with this insert but it was determined to be useless for our purposes.
2. This process was repeated countless times, however VNA results were not recorded as
diligently because efforts were more focused on thrust results, and more particularly,
eliminating arcing within the cavity. The final round of VNA sweeps before the end of Summer
testing is seen below as 9/10/15.
Later Revisions
During Winter quarter more efforts were made to eliminate arcing and new VNA
sweeps were performed using the professionally made antenna as seen below. It was
discovered that the professionally made antenna resulted in dramatically different resonances
than the homemade antenna probably due to the geometric tolerances, quality of electrical
connections, and professional design and tuning. Therefore, all of the VNA sweeps performed
over the summer should be neglected as they have resulted from a poorly created antenna.
3. Using the professionally made antenna, only a single resonance was found near where
the analytic solutions indicate it should be. Thrust measurements were performed and the
resulting pendulum deflection can be found on Kurt's Youtube page [1]. More detail concerning
this latest test can also be seen on the NASA Space Forum [2].
Conclusions
After months of VNA sweeps it can be safely said that obtaining resonance is not easy.
Using a homemade antenna will most likely produce inconsistent results and may be indicating
other phenomena at play. Every resonance found was extremely sensitive to vibrations,
meaning that tapping on the table could shift the entire log-magnitude plot by 20 dB. Every
thrust test was performed after moving the cylinder between buildings and thus the resonance
was more than likely disturbed. Although EM Pro simulations sometimes correlated with VNA
sweeps, more often this was not the case, however more computational power was needed in
order to solidify the correlation.
References
1.Kurt's Youtube page:
https://www.youtube.com/channel/UCDIeLRPy9437eZLgpyjKBGg
2. NASA Space Forum Update:
http://forum.nasaspaceflight.com/index.php?topic=39004.msg1491956#msg1491956
4. 7/29/15
Calibration was very successful (will take captures of short, matched and open calibrators next time)
Total Span: 500 MHz
Centered at: 2.46 GHz
Points: 1601 points ( 0.375 MHz step size)
First Sweep: 1 inch HDPE, non conducting movable plate, delivery waveguide from Hamilton Beach
Microwave (where the Galanz magnetron came from)
5.
6. Sweep 2: 1 inch HDPE, non conducting movable plate, delivery waveguide from Sharp Microwave (the
one with the bar across the middle, Dr. Arakaki seemed to think it was only structural not RF designed)
Sweep 3: 2nd delivery waveguide, 1 inch HDPE, metal screws instead of Nylon screws to hold dielectric
(Signal seemed much noisier, modes closer together and more of them)
7. VNA Sweeps 7/30/15
Center freq: 2.45 GHz Span: 100 MHz 1601 Data points (.0625 MHz step size)
VNA 7/30/15 Results Summary Central freq: 2.45 GHz Span: 100 MHz Step size: 62.5 KHz
Sweep Dielectric Thickness
(in)
Conducting
plate
Length
(in)
Waveguide Mica Est.
Quality
Est. S11
(-dB)
1 HDPE 1 Y 7.07 1 N 98.40 21.414
2 HDPE 1 Y 7.07 1 Y 82.02 28.55
3 Nylon .25 Y 8.15 1 N 91.12 40
Nylon .25 Y 2.292 1 N 41 .01 35
4 Nylon .25 Y 7.65 1 Y 61.5 22
Nylon .25 Y 2.986 1 Y 0 14
5 HDPE 1 N 7.15 1 Y 82.02 25
HDPE 1 N 6.80 1 Y 82.02 32
HDPE 1 N 2.164 1 Y 0 20
6 Nylon .25 N 8.015 1 Y 160 22
Nylon .25 N 3.03 1 Y 0 35
8. Sweep 1: 1 inch HDPE Dielectric, Waveguide 1 without mica insert, conducting movable plate
1st resonance
Cavity Length 7.07 inches
Sweep 2: 1 inch HDPE Dielectric, Waveguide 1 with cardboard insert, conducting movable plate (Note
reference shifted to – 3dB)
Sweep 3: .25 inch Nylon dielectric, Waveguide 1 without mica insert, conducting movable plate
14. Sweep 1: Two HDPE discs (new one closer to movable plate) (.99 in and .6185 in thick), non-conducting
movable plate, EMF strips between waveguide and cylinder with Ox Gard to seal small gaps
Screws at ~60 degrees, rod length of 5.25 inches
15. Sweep 2:
.6285 in HDPE , other conditions same as previous.
Screws at ~45 deg
1.6 inches of rod sticking out
Significant non-linearity with this resonance
16. Sweep 3:
1 in thick Plexiglas disc, all other conditions are the same as previous
Screws at 0~10 deg
Rod length: 5.8 in
17. Sweep 4:
New HDPE on top of Plexiglas
All other parameters unchanged
Screws at 0 deg
Rod length 6.0 in
18. Sweep 5:
New HDPE, then plexiglass, and finally nylon on the plate
Other parameters unchanged
Screws at 70 degrees
Rod length: 9.15 in
19. Sweep 6:
Same as previous, but with conducting movable plate (with two wires connected to the string support.)
20.
21. VNA Sweep 2/27/16
Cavity Length: 7.2 inches?
Dielectric: 1 inch thick HDPE
Professional Magnetron Antenna
Waveguide 2
