Engineering Excitement (or: more MATLAB graphs and FFT fun...)

So, end of the degree means that some fairly cool research gets to be done to produce a thesis. My project has ended up being trying to explore the effects of a cylinders orientation with respect to RF (Radio Frequency electromagnetic radiation) wave polarisation on their RF reflectivity. A bit of a mouthful that ends up with: can I find a method to detect pipes really easily? If I can that would be really useful for the construction of pipe-detecting devices. The really cool bit was that initial research indicated that dielectric (non-metal) cylinders also exhibited properties along these lines, which is really intriguing, and could represent a massive improvement over current methods of pipe detection.

So a couple of months of reading and simulation later I found myself in the lab collecting data from a physical experiment, the results of which are below. For this test I had a plastic and a metal pipe which I rotated between vertical and horizontal in the testing apparatus. My investigation had indicated that 

Data!!!!

This graph shows the approximate power reflected from a set of sample cylinders at varying distances from a test antenna hooked up to a VNA (meaning I was actually looking at the total power reflected from the antenna).

Two points, firstly I can be fairly certain the figure above shows power reflected due to the samples as there is another set of data that was collected with no sample in the testing apparatus - this established how much power was being reflected from (and into) the antenna under no-sample conditions. This power was subtracted from the result of tests when there was a sample to give the result displayed.

'But Wait!' I hear you say, 'you have negative power over there!'. I have cheated and used a dB, logarithmic, scale to display the results. Negative readings actually indicate that less power than the no-sample test was reflected under these conditions. This is due to RF wizardry, at these points the sample has put itself somewhere where it can absorb lots of power (or lead to it not being reflected into the measurement device), rather than reflecting it (into the antenna where hopefully the measurement device can detect it). This is a combination of the sample 'coupling' with the antenna, changing its characteristics, and it being close enough to provide an effective load to the transmission line the antenna is attached to! (I did say wizardry)

Secondly, the y, distance, axis is approximate - there is some offset which I have not accurately measured yet, unfortunately Lab-Me only took down a rough measurement! Additionally it is the nature of a lot of RF work that there is a certain level of roughness to results. This is mainly as the measurements we take are affected by all RF radiation waves being thrown about, including are own, that like to interfere with each other, themselves and reflect, refract and diffract off of whatever they can. This leaves us with a non-trivial system that is always going to have an element of randomness simply because we cannot predict everything that is going to be important to the behaviour of the system. The best that can be done is to try and calibrate all of our measurements to a baseline and then not change the test environment if we can help it. This is why the measurements all look very messy - luckily this data set represents about one of 640 (from this experiment alone) I can use to help reveal how the effects I am looking for are affecting the system! Additionally I have another antenna set made up able to look at a different set of frequencies (these tests were conducted at about 5GHz).

I got really interested in the oscillations in the data so, naturally, I ran a quick Fourier transform on the data to uncover its frequency content. Note that this frequency is separate from the frequency of the RF operation, it is actually the frequency of reflective peaks over the separation distance.

MOAR DATA!!! (ignore the vertical scale)

Unfortunately for those trying to avoid confusion, the frequency plot shows a definite peak at just over 0.3333333 Peaks/cm, or a wavelength (distance between two peaks) of just under 3 cm. Why is this interesting: because the wavelength of the RF radiation used to produce this graph was just under 6cm! In other words the sample reflects the most power when the sample-antenna separation is some multiple (plus an offset) of half the RF wavelength!

It is always nice to get results that indicate more investigation is needed!