Doomtenna: Range and Crashes

In the previous post we wrote about switching our rooftop LoRa node, lovingly called "The Doomsday Device" or "Doomtenna", from Meshtastic to MeshCore.

After that software update, we are continuing our series of Doomtenna updates with the hardware side: range, crashes, solar power, batteries, and the still slightly mysterious behavior from the roof of our Hilversum office.

Two things have been a constant struggle: the range is much stranger than we expected, and until recently the node kept crashing.

Range still has us puzzled

From the roof, the antenna appears to have a nice unobstructed 360-degree view. Yet the actual range varies wildly depending on the direction. For example, in the direction of Hollandse Rading we are losing signal after only a few hundred meters, with only a bit of heather in between.

On the other hand, we have been directly communicating with nodes in Bussum (7 km) and Almere (18 km).

That contrast is the mystery part. The node can clearly reach far, but not equally well in every direction. We think the weaker directions might be caused by the air treatment unit our setup is mounted on. We recently mounted the antenna about a meter higher, but this did not seem to have as much effect as we hoped.

Other ideas we have for getting better range are replacing our 3 dBi antenna with a higher-gain antenna, or adding a second node with a directional antenna and bridging the two nodes.

Up for months, down after hours

The crashing part is the one we only recently solved. Sometimes the node was up for months, sometimes it crashed after only a few hours.

We assumed this was caused by the voltage of the solar panel slowly increasing after a drop. This is known to leave the device in a state where it does not boot properly. The only way to recover is to physically go to the device, disconnect both battery and solar panel, and reconnect them. Pressing the device's reset button does not solve the issue.

We initially ruled out the battery causing the problem, since according to all metrics the battery level never dropped below 70%.

Trying to stabilize the solar panel

Our first attempt to solve this was placing a TPS55165-Q1 between the solar panel and the device, to make sure the solar panel provides a stable 5 V of power.

This, however, did not solve the problem.

We even considered building an automated kicking machine to revive the node after a crash. Before we built something that dramatic, we decided to try replacing the battery anyway. Maybe we were too quick to dismiss the battery as the cause of the crashes.

The boring fix

We simply replaced the 1250 mAh battery with an absolute overkill of three 3000 mAh batteries in parallel.

The device has not crashed ever since.

That is not the most exciting root cause analysis, but it is a very useful outcome. Metrics said the battery level was fine, but the actual behavior of the device said otherwise. In the end, new battery setup solved the issue where better power conditioning did not.

To be continued

Since the initial draft of this post, the Doomtenna has kept running without a single crash. That means fewer trips to the roof to power-cycle it, and a more reliable contribution to the network.

With stability sorted, we can turn our attention back to improving the range. Next up: trying a higher-gain antenna. The automated kicking machine is officially cancelled.