NSL-69

For the much of the winter of 2018-2019, the jet stream over North Carolina was annoyingly strong. Flight simulations from central NC, typically predicted landing areas out in the Atlantic Ocean. Several flight ideas had been dreamt-up by the group over the winter with no real chance to test them out. A break finally came the morning of March 14, 2019.

Since October 2018, Paul L played with a new cheap Internet-of-Things (LoRaWAN) device out of China called a TTGO T-Beam. This $26USD device, built upon an Arduino ESP32 processor, includes WiFi/Bluetooth, GPS, LoRa transceiver, and a built-in battery pack with charger. The board also contains extra buttons and LEDs.

Would this cheap little device work as a LoRa HAB tracker? If so, it would be an inexpensive alternative for a backup tracker. Also, the 915MHz model could be operated in the US without the need of a Amateur Radio license. Code can be found here.

Several similar products exist on the market with an assortment of features like small displays. These could be readily built into portable receivers with their WiFi or Bluetooth interconnects acting as gateways to the Internet.

Paul chose a 433MHz LoRa T-Beam for testing. Arduino code was written, based upon the awesome work by Dave Akerman for Pi-in-the-Sky type LoRa trackers. This code set-up the GPS and transmitter, and at predetermined times, sent out data packets following the established HABHUB protocol. The code also read the battery voltage and processor temperature and reported it back along with the usual GPS telemetry. Debug code recorded data within the on-board flash for evaluation after the flight.

The telemetry could be received by any existing 433MHz PITS-style gateway (eg. Raspberry Pis running Dave's Lora_Gateway).

Main questions for testing this hardware:

• Would the supposed NEO-6M GPS really work at altitude?

• How well does the built-in LoRa antenna work?

• How well does the battery work? Would the violence of burst cause power hits?

• Does the device cause radio interference and how susceptible to interference is it?

When the March 14th opening came in the weather, Paul decided to perform a test flight of the T-Beam. A small "greenhouse" box, made from a dish detergent container fitted with black foam, held just three devices:

• TTGO T-Beam with its antenna pointing up.

• AP510 APRS tracker, also with its antenna pointing up.

• a rebuilt Mobius Mini camera, which previously experienced failure and needed another test.

A basic heat pipe, consisting of metal screening and copper tape, carried excess heat from the camera and delivered it to the batteries.

The payload box connected to a simple 68cm ripstop-n-tulle parachute. In case of a forest landing, a small loop was added in the bridle and secured with Polyvinyl alcohol (PVA) string. The PVA would dissolve if it got wet, releasing the payload to fall from a tree.

The entire payload weighed 360g. A Kaymont 600g balloon filled with H₂ provided roughly 1000g of neck lift. This was expected to carry the flight above 30km over the course of the two hour test.

Prior integration tests showed a working system. But during the start-up, the morning of the flight, problems with GPS interference arose. With the payload fully assembled, the T-Beam soon lost GPS lock. So the payload was quickly opened and a shield of aluminum foil was then added between the camera and the T-Beam. This seemed to do the trick and the T-Beam blinked to acknowledged good GPS lock. Launch then proceeded normally.

With the light winds, the chase team could leisurely drive to the landing zone; taking breaks to gather data on the LoRa radio's performance.

Beautiful morning -- Barring the pollen in the air, it was a great day for a chase.

The on-board video showed a calm ascent over central NC. Sadly, just 23 minutes into the flight, the Mobius camera shut down.

The T-Beam tracker's radio performed surprisingly well. Telemetry was consistently received by the "Home" LoRa gateway (fixed at launch site) up until the flight was 25km away. This gateway uses a rather poor antenna but it still heard telemetry to an SNR of -12! It might have been possible to receive the signal even longer had this station been manually operated instead of leaving it to its Automatic Frequency Control.

The chase "Car" LoRa gateway received 95% of the packets during the entire operation of the transceiver, most with a high signal strength. The chase car kept within 25km of the payload; although it sometimes 'seemed' further to the chase crew. Paul notes, "Next time this is tested, I need to get farther ahead or behind to tests distances." As the payload's small LoRa antenna was placed atop the box, it was assumed that there would be a strong reception 'shadow' directly beneath the payload. This was confirmed as the payload flew overhead and received signal strengths dropped.

As the balloon passed 21km altitude, it crossed over US-70. The chase crew was parked directly below and was able to glimpse the balloon with the naked eye. An attempt was made to photograph the drifting white dot using a cellphone.

At about this same time, the chase crew noticed another change in reported telemetry. The packet counter had reached "148" and then the next packet reported "1", signaling that the T-Beam had rebooted. Earlier in the flight, the crew watched as the reported battery voltage dropped. Now, as it quickly fell below 3.2V, the received packets also continuously reported as being packet "1". The crew believed that the low voltage led to the tracker to reboot after the exertion of transmitting at 100mW. This continued for another 26 minutes, with the voltage continuing to drop. The last packet received, at 30km altitude, reported just 2.9V.

Now, running on just the APRS tracking data, the chase team made their way to the initially predicted landing zone. There were some tense moments when the HABHUB predicted landing site changed to Seymour Johnson Air Force Base. Thankfully, this track continued to slide north. Unknown to the chase crew, the heavy Kaymont balloon stayed mostly intact and was forcing the payload to fall faster than expected. At times, the balloon scraps fouled the parachute causing it to drop even faster. But thankfully, the chute was made without shroud lines and it resisted being tangled.

"I have never been so happy to see a balloon dogleg at the last minute" - Chris the flight's Public Affairs Officer

The flight drifted into residential Goldsboro, north of Seymour Johnson, and missed a large, open strip mall area. Instead it decided to land in the top of a pine tree in a front yard.

With a chance for rain the following day, the chase crew decided to leave the payload up the tree and hope that rain would melt the PVA, dropping it to the ground. But the rain showers ended up being sporadic and the payload remained dry.

On March 16th, Paul L and Tim W. visited the site. Tim brought along his trusty bow and within minutes had the payload safely down. Awesome shooting as always Tim !

A shout out to the friendly and helpful homeowner! As a middle school teacher, he had interest in what we are doing and will receive a pack of flight info and photos.

RESULTS

T-Beam Tracker: VERY PROMISING. The biggest issue with the unit was the poor 18650 battery performance. As seen in the graphs, above the T-Beam's 18650 battery died quickly. Some of this may be attributed to cold temperatures, but after review, not only were the internal temps not that bad, but the voltage curved downward prior to temperatures dropping inside the payload. Investigations after the flight showed that the 18650 battery in question was actually a fake! The battery was labeled as 6000mAh, which is way over what the legit batteries can even hold. The battery was also about 1/2 the mass of a real battery, so it is suspected that a small 200mAh battery hides inside the case.

-Would the supposed NEO-6M GPS really work at altitude?
Yes! Works just fine to at least 30km.

-How well does the built-in LoRa antenna work?
Surprisingly well. A downward facing 1/4 wave ground plane antenna would do even better.

-How well does the battery work? Would the violence of burst cause power hits?
The battery was a forgery and died too soon to know how it worked during burst.

-Does the device cause radio interference and how susceptible to interference is it?
The on-board GPS was susceptible to RFI from a nearby camera, but a simple insulated wrap of aluminum foil fixed that. The nearby AP510 saw no issues.

Transmit mode note: The T-Beam's was configured to transmit on PITS Mode=0, but with SF=8 and LDRO=0. This tweaked its transmit time to <2 seconds in length instead of the usual 'Mode=0' of 15 seconds. The packets were sent every 38 seconds with 100mW of power. If we used just the normal Mode=0, then we would probably want to step the power down to 50mW or change it to transmit once every 300 seconds. Otherwise the LoRa transmitter chip could overheat and burn-out.

Mobius Camera Shutdown: Reviewing the video, recorded bit errors increased over time and it was initially assumed that the microSD card was faulty. But repeated post-flight tests (h2testw.exe) of the card were clean. More tests of this camera are needed. Did the heat pipe fail, causing the camera to overheat? It's a shame that this workhorse camera keeps shutting down prematurely. It will be grounded until a cause can be determined.