0xfeed

Here is a summary of the final tracker hardware used in the tests:

The only custom-made part of the tracker is the antenna that I built according to UKHAS payload antenna instructions. The antenna is very simple: the radiating element is made of RG174 coaxial cable center conductor and the ground plane radials are pieces of single-core wire.

The two receiver (gateway) devices I’ve built use mostly the same hardware: the Raspberry Pi and the LoRa HAT are identical to the transmitter. I have placed both receiver devices insinde plastic enclosures in order to keep them safe during car rides.

The only difference regarding LoRa signal reception are the antennas I used for testing. Because the devices are used in a chase car during a balloon flight, I opted to use magnetic mounts for the antennas so they can be placed on top of car roof. The antennas I used were a Diamond AZ510 dual-band (2 m / 70 cm) mobile antenna and a slightly taller Diamond SG7900 dual-band antenna.

After swapping antennas for a couple of times, my conclusion was that I did not notice any significant difference between the mobile antennas, so the I ended up using the shorter and lighter AZ510 for obtaining the test results below.

The maximum line-of-sight distance I could achieve easily by driving public roads was a little over 11 kilometers. While most of the telemetry packet transmissions were received correctly, I was not able to receive complete images.

The following map, generated using an excellent tool called GPS Visualizer, presents color-coded RSSI (Received Signal Strength Indicator) readings from the receiver device. Gray color indicates no reception. The unit of the RSSI values in these measurements is dBm, which means decibels of power referenced to one milliwatt: 0 dBm is 1 mW (milliwatt), -30 dBm is 1 µW (microwatt), -60 dBm is 1 nW (nanowatt) and so on. The maximum (theoretical) sensitivity of the LoRa receiver using the LoRa configuration listed above (with spreading factor 7) is -127 dBm (based on the estimate given by LoRa Modem Calculator Tool by Semtech).

The graph below represents the RSSI readings at a specific distance:

While the test was successful, the limited visibility caused by terrain and forests did not provide a good environment for testing the true limits of LoRa, so I started searching for a more suitable place for doing a second round of tests.

This time the maximum line-of-sight distance was ~16.5 kilometers, but the receiver device was able to decode packets from ~19.2 km away while I was driving. Reception was intermittent, so images were not successfully received at the maximum distance.

I had issues with the transmitter device freezing from time to time, because of a concurrency-related bug in the code I had written, so the results are not as conclusive as I would have hoped.

Again, the unit of the RSSI values in the measurements is dBm, which means decibels of power referenced to one milliwatt: 0 dBm is 1 mW (milliwatt), -30 dBm is 1 µW (microwatt), -60 dBm is 1 nW (nanowatt) and so on. The maximum (theoretical) sensitivity of the LoRa receiver using the LoRa configuration listed above (with spreading factor 8) is -130 dBm (based on the estimate given by LoRa Modem Calculator Tool by Semtech).

And the graph representing the RSSI readings at a specific distance:

While the total LoRa transmission range achieved in the tests gave some insight to the reliability of using LoRa for tracking high-altitude balloons, I was not completely convinced. Issues (that were later fixed) with the ERT software prevented me from doing proper comparison between different LoRa configurations.

To overcome the possible limitations in LoRa range, I decided to use a second tracker as a backup in case I’m not able to receive LoRa transmissions. The Trackuino backup tracker is documented in more detail in the next blog post.

There are several ways to configure Raspberry Pi to conserve power. You can:

Raspberry Pi CPU and GPU operating frequencies can also be adjusted, but - based on my measurements - lowering the frequencies had only a negligible effect on power consumption. There are also more advanced methods to reduce power consumption, e.g. by changing the voltage regulator of the Raspberry Pi board.

Batteries powering the tracker have to be chosen carefully, since the effect of low temperature to batteries is often significant: the battery voltage may drop very quickly to a level that is not enough to power the tracker. I chose to use Panasonic Eneloop Pro rechargeable NiMH batteries, since they are widely available and rated for temperatures as low as -20 C.

To supply steady 5V voltage for the Raspberry Pi micro-USB port, I used an Adafruit PowerBoost 1000 Basic voltage booster, which works with voltages as low as 1.8V.

To simulate the cold temperatures of upper atmosphere layers, I decided to place the tracker (transmitter) device to a large freezer set to minimum temperature. That way I could reach a temperature of around -24 C. While it is a lot colder, up to -40 to -50 C, at an altitude of 30 kilometers, I left the payload enclosure (containing the device) open, so that the device was exposed to the raw temperature in the freezer.

Here’s a summary of the tested equipment:

And the configuration used for ertnode tracker software:

The total running time for the tracker was 7 hours 18 minutes and 40 seconds, counting from the first received telemetry message to the last one. Over 7 hours of run-time is definitely enough for a balloon flight, so the test was a success: there was no need for further optimizations!

I also checked the power consumption of the tracker device, both with and without the USB dongle:

Here we can see that the both the LoRa radio transceiver and the WLAN dongle draw a significant amount of current: The LoRa chip consumes about 70 mA while transmitting and the WLAN dongle a total of 90 mA while being connected to an access point (without active TCP connections). It would be possible to extend the run-time even more by leaving out the WLAN dongle.