jakapoor
commented
6 years ago Thanks for this detailed information Russell. I'll respond to the specifics below, but first I want to make a few general comments / suggestions.
1) Some of this information will be redundant with what you've included above, but we need a centralized location to compare future empirical tests of the performance of the system. Can you upload to GitHub a document (or documents) detailing a) your goals for this empirical test (i.e. what - specifically - did you want to test? b) the setup (i.e. what hardware did you use, where, and exactly when, who was present, any other relevant notes), c) the parameters of the test(s) (i.e. what settings were you using, what versions of what code, specifically referencing where they are in GtHub), d) what specific tests you conducted, and e) the results (I mean data here, not only general descriptions) along with your interpretation of what those results mean. This is absolutely essential in order for us to actually benefit from these tests: a poorly documented test essentially never happened.
OK, now to the specifics:
The first tests were performed with the coherent receiver placed in the middle of a baseball field. Preliminary angle of arrival tests with 2 antenna elements resulted in unstable sporadic angles. It was found shortly after that there was an RF signal in the environment transmitting at 431 MHz (we were using 433 MHz). It is debatable whether this would be an issue with the small receiving bandwidth being used (~1 KHz).
Explain the hardware setup. How high was the rig, were you on mains power? Also, even preliminary tests should be clearly documented. How sporadic were these two-antenna-element angles? What DF algorithm were you using? Remember to include the result data on the repository. How do you know there was significant noise? How did you take these measurements?
I was not able to test the MUSIC/root MUSIC algorithms sufficiently outside because the power supply only lasted one hour (with a signal generator and low-battery laptop connected). Further tests were made indoors in Duffield.
The low battery capacity is another reason to plan out exactly what you are going to do out there. Let's come up with a plan together before your next test. Please make sure to document your indoor tests fully as well.
For the root MUSIC/MUSIC algorithm tests indoors, the resulting angle took ~5 seconds to stabilize once the transmitter was put in a fixed position (see improvement 1).
Thoughts about why this might be the case? This is definitely not something we can tolerate in a wildlife system, where animals are in place only momentarily. We'll need to understand why that's happening. Could it be related to the snapshot length required for MUSIC?
433 MHz transmissions resulted to NULL values from the root MUSIC algorithm; however, 866 MHz transmissions never resulted in a NULL value (see improvement 2).
Can you explain this? What do you think might be happening? We ultimately need to be able to work with 433 MHz and down in frequency. Transmission properties (attenuation due to vegetation) at 866 MHz aren't compatible with a wildlife system.
Indoor multi-path effects (resulting in a clearly unexpected angle) resulted when the transmitter was pointed more than 30 degrees away from the line of sight. Towards the end of testing last night, stable angles within 10 degrees of the actual angle of arrival (measured by protractor/string) were observed. These last observations are premature though, and further testing must be done to see if these were fluke values.
First, how do you know this was the effect of multi-path interference, versus some other issue? Also, how can a non-directional transmitter be "pointed" in any particular direction? Do you mean that the transmitter was located 30 degrees with respect to the array? Please provide the data for this test. Why are these observations premature? Remember that data collection should be done systematically; you need to have a plan for exactly what information you want, how you will record it, and what you will do with it when you're done. This all comes back to the part of the project proposal that was to be devoted to creation of standardized test protocols.
Regarding improvement # 1: Is there a numeric output from the GUI Compass? How will that information be logged for later analysis? We want to avoid any error due to inaccuracies associated with reading from a user interface. Saving a float value seems like a good solution.
Regarding improvement # 2: Don't forget to provide specific information about the setup of the testing mast, detailed enough that someone could recreate your setup without speaking to you. You provide good detail here, but make sure it gets saved systematically somewhere. We can talk about the best format for this. I think it's a good idea to test whether the setup is interfering with itself, but we need to approach this systematically.
Regarding improvement # 3: This has to be an error. Does the issue remain if you re-order the units? I can't see a reason why the signals would be getting progressively weaker as you go down the array; each SDR has its own antenna with its own preamp. I'd definitely bring this up with Igor.
russellmsilva
The first tests were performed with the coherent receiver placed in the middle of a baseball field. Preliminary angle of arrival tests with 2 antenna elements resulted in unstable sporadic angles. It was found shortly after that there was an RF signal in the environment transmitting at 431 MHz (we were using 433 MHz). It is debatable whether this would be an issue with the small receiving bandwidth being used (~1 KHz).
I was not able to test the MUSIC/root MUSIC algorithms sufficiently outside because the power supply only lasted one hour (with a signal generator and low-battery laptop connected). Further tests were made indoors in Duffield. For the root MUSIC/MUSIC algorithm tests indoors, the resulting angle took ~5 seconds to stabilize once the transmitter was put in a fixed position (see improvement 1). 433 MHz transmissions resulted to NULL values from the root MUSIC algorithm; however, 866 MHz transmissions never resulted in a NULL value (see improvement 2).
866 MHz transmission tests: Indoor multi-path effects (resulting in a clearly unexpected angle) resulted when the transmitter was pointed more than 30 degrees away from the line of sight. Towards the end of testing last night, stable angles within 10 degrees of the actual angle of arrival (measured by protractor/string) were observed. These last observations are premature though, and further testing must be done to see if these were fluke values.
Improvements to be made:
The GUI Compass may have an instability issue with reported issues here and here. Therefore, the following tests will be done using the console or a GUI float sink (displaying the float value result of MUSIC/root MUSIC on the GNU Radio GUI) in addition to the GUI Compass to see if there is disparity between the end AoA results.
The testing mast contained a platform for the coherent receiver, Raspberry Pi, and usb hub. This platform was placed within the near field of antenna array elements. The platform was specifically placed 2 feet from the antenna array (near field is one wavelength - ~2.3 feet for 433 MHz). This platform will be placed further from the near field (an additional 1.5 feet lower) to reduce EM interference with electronic devices.
No matter the frequency, gain, or bandwidth setting, RTL SDR 1 receives the greatest sinusoidal amplitude, RTL SDR 2 receives a lower sinusoidal amplitude, RTL SDR 3 receives an even lower sinusoidal amplitude, and RTL SDR 4 receives the lowest sinusoidal amplitude. This effect occurs when sending sinusoidal waves to the coherent receiver from the CC1310 and the signal generator. Please see this video which displays the sinusoidal waves received from a CC1310 transmitter. Since this was a high power disparity, the RF gain of each receiver had to be increased from 35 to 50 dB in order for RTL SDR 4 to receive enough power for testing. This phenomenon may adversely effect beamforming (as implemented in Whiting's system). Since RTL SDR 1 always receives a signal of much greater power than RTL SDR 4, the MUSIC/root-MUSIC algorithms may be adversely effected as well.
Please feel free to shoot me any suggestions or ideas. A next set of tests will be performed after the above improvements have been made or reconciled.