The RF shielding effect of a berm was measured using a Multi-copter as part of my PhD program. LS of SA from LS telecom generously helped us with these measurements with their own Multi-copter measurement platform. The measurements were done with a transmitter located on the far side of the berm transmitting 9 vertically polarized frequencies in a horizontal direction, from 60m, directly at the berm. Below is a photo of the Multi-copter measuring in the vicinity of the berm.
Wessel, the Multi-copter pilot, landing the measurement vehicle
Far side of berm, opposite side than transmitter
Front side of berm, same side as transmitter
The data was then processed and compiled together from a total of 7 10 min flights at different heights and configurations. The processed data was plotted and interpolated with python on a 2D grid with an overlay of the berm. The next 3 clips shows animations of the interpolated data over different frequencies and heights.
I would just like to thank the measurement team and especially LS of SA for the great collaboration.
From the left:
Mathew Groch: Responsible for broadband loaded dipole antenna used in these measurements
Jan (crouching) from LS of SA
Nardus Mathyssen responsible for developing a pulse generator which will be used in future measurements
Wessel from LS of SA
Myself Hardie Pienaar
Brian from LS of SA
After flying in loiter mode for a while the quad copter suddenly flipped over. I immediately switched to Stabilize mode and increased throttle to try and save the situation. It stabilized itself just before hitting the ground but sadly had too much horizontal velocity when lightly snagging the grass. This caused the copter to roll and break 2 motor mounts and a landing gear. The cause of the flip was because of a bad connection to 1 of the esc’s from the APM 2.5 output.
In the beginning of 2011 my grandfather gave me his 1968 MG midget MK III. The car is still in almost perfect condition as can be seen in the photo taken in Stellenbosch. I will dedicate this part of my webpage to give updates on my MG. The bottom photo shows my grandmother and grandfather playing bowls.
Upgraded to the APM 2.5. Height and GPS accuracy are significantly better than the APM 1.0. The APM 2.5 also has a much higher telemetry data throughput due to some of the signal processing happening off of the Atmel chip. Below is a photo of the much more neat setup containing the APM 2.5.
After receiving my raspberry pi camera I thought it to be a good idea to to take some aerial photos with the quadcopter. The first few photos are from Stellenbosch University campus followed by the last few which are from my girlfriends farm near Worcester.
After getting the quadcopter to fly in loiter and auto missions with confidence it was decided to use the quad copter for what it was originally intended for. To measure RF signal propagation. This measurement was done with a transmitter on one side of a 13m high man made berm and flying a vertical path up to 50m on the far side measuring the diffraction of the continuous wave signal at 400MHz. Below is a video with the payload strapped to the quadcopter, note that the antenna (yellow block) was exchanged for a small stub antenna at the actual measurement. The measurement was done using my RS232 spectrum analyser logging onto a Raspberry Pi. The effect of the quad copter on the antenna pattern was ignored and the measured data was treated as relative. The measured data can be seen below and clearly resembles a diffraction pattern. This pattern has been verified against some prediction code of a colleague. As a first test this proved very successful as a proof of concept and was hereafter named fEMu (flying Electromagnetic Metrology unit).
It is time to revisit the RS232 spectrum analyser. Below is an image of some prototyping on a breadboard followed by a soldered working version including a SMA connection. The problem with the previous spectrum analyser was mainly in the computation of the frequency set registers which involved large 9 digit values, a hard feat for a 8 bit micro controller. This time around a 32 bit micro controller, the PIC32MX220B032, will be used. This will allow for a maximum clock frequency of 50MHz leaving the radios internal frequency settling time as the only limiting factor to the spectrum analyzer sweep time. The proposed specification for this spectrum analyzer are as follows:
Bandwidth 260MHz – 960MHz (Realisticly about 150MHz according to the input filter of the RFM22B model used)
2ms per sample
Sweep and Discrete measurement modes
Calibration mode (Calibrates the device with the help of a sweeping signal generator)
Wireless trace measurement (Two devices needed, one as a base station)
SD card for onboard storage and buffering
Connecivity – FTDI cable and rasberrypi GPIO connector header
Form factor: 1 square inch
Soldered working version without SD card
Breadboard prototyping of the 32bit spectrum analyser
After the vibration issues were sorted out it was time to test the loiter and auto modes. A video of the first successful loiter test can be seen below. The APM 1.0 seems to hold its position reasonably well given that there was a slight breeze that day, the quad-copter was also landed in loiter mode, in this video, by simply decreasing the throttle. The video was taken behind our engineering building at Stellenbosch University. We have also tested the auto mission, takeoff and landing successfully.
The new motor mounts were made by Johan Frank from Malmesbury on his own CNC machine with simple number-plate plastic. These new motor mounts are much less brittle and thicker, therefore much stronger than the original perspex parts.
After the quad copter was fully assembled, ESC’s, accelerometers and radio calibrated it was ready for its maiden flight. This flight was carried out at the back of our Engineering building, Stellenbosch, on an extremely limited piece of grass. A short video from the flight can be watched below.
After some flying the altitude hold function was tested which seemed to work. However, when testing the second time and switching it off, the throttle was in a lower than hover position, causing the quad to come down hard. This caused 3 perspex motor mounts to be snapped off. Luckily this exposed some weaknesses in the frame. The motor mounts will now be manufactured from a slightly less brittle plastic. Ignoring human error the quad copter flew perfectly the first time.
After a number of flights in stabilize mode with a very stable vehicle it was decided to test the rest of the flight modes. Our goal was to build this into a measurement platform which necessitates the use of the auto modes. Therefore, we went to an open test field near Stellenbosch where we switched the quad copter, while in a stable hover, to loiter mode. The latter action caused the quad copter to ascend at full throttle with no sign of stopping. After landing and trying again the same thing happened.
After this slightly unsuccessful test I learned that in the newer APM firmware versions vibration can be an issue and can cause the observed behavior. I decided to start eliminating all the vibration I could find. Starting with the motors I measured the vibration with a app called “vibration on my iPhone”. Spinning each motor independently while strapping the phone onto the motor arm I could monitor the intensity of the vibration and start balancing the motors in the same manner as the video I found below.
After some experimentation I realized that the motors without any nut and spinner was perfectly balanced, however, as soon as the spinners are added it would put the motor way off balance. This leaded me to balance each spinner with a piece of tape much like they balance the motors in the video. This meant I now had 4 non vibrating motors. At this stage I did not have a prop balancer and had to hope my cheap colorful gem-fan propellers were reasonably balanced.
To actually test if there were any relative improvement I enabled the IMU data logging onboard the apm. I downloaded the log after each change to relatively measure what performance difference the modification had.
It can be seen in the comparison that there was initially also some experimentation with different apm mountings such as being screwed down, double sided tape, moon gel pads and lastly, the configuration that is still being used, dampening pads below the whole electronics platform. Remember that these modifications were done before balancing the spinners, note that it almost seems as if the vibration got worse with the different platforms except the moon gel, it should also be stated that with the new dampening pads the whole layout of the vehicle was changed causing the vibration to increase dramatically. A motor which was slightly damaged in the previous crash was also replaced with more negative results. However, balancing the prop spinners and propellers showed a slight decrease in vibrations as seen in the graph below. However something weird was noticed when balancing the gemfan propellers, it seemed that they would be heavier on 1 side of the hub. It was here that I decided to invest in more expensive apc propellers which as you can see made a huge difference which in the end allowed loiter and the auto modes to work flawlessly.