RSSI variation in the uplink

[Evik]

Greetings, I’ve been conducting extensive measurements of Uplink RSSI over an extended period to analyze its fluctuations in response to climate variations. The results I’ve obtained are rather intriguing. Initially, the RSSI values exhibit a standard variation, typically ranging between 3 and 5 dBi within the first hour. However, after an hour, the variations intensify, occasionally escalating up to 20 dBm between consecutive transmissions. These two distinct behaviors alternate cyclically after each hour how it looks in the next graph:

I conducted these measurements by consistently sending messages from a Heltec node LoRa32 v2 and receiving them via a gateway over a span of two days. This gateway connects to my local chirpstack server from where I pull the RSSI data. I programmed the nodes using the “LoRaWAN_APP” library provided by Heltec, utilizing the LoRaWAN interruption example as the foundation. I rigorously tested the transmissions at intervals of both every 15 seconds and 1.5 seconds, observing similar patterns in both scenarios. Furthermore, I experimented with two different nodes to validate these observations.

I’m at a loss in identifying the root cause behind this peculiar behavior. Any insights or ideas you could offer would be immensely appreciated.

[thor]

Have you logged the SNR values aswell?

[Evik]

Yes, I have logged it but I don’t look any correlation with RSSI results. It also have several variations but It doesn’t seem to be following a pattern. This is the result:

[datnus]

Low rssi and snr at the same time.

May be due to signal reflection to materials.
Where do you put the gateway and nodes?

Or have signal interference from other source.

[Evik]

It’s possible, the gateway is not in the best place as it points in a direction where it passes over a metal roof, but you could move it and take long-term measurements to see if this changes.

But I believe it’s due to the device. In both cases—conducted on different days with varying time intervals between messages (1.5 seconds and 15 seconds)—during the period when the fluctuation happens, the messages consistently switch between low and high power almost exactly between two messages. I’ve illustrated this in the figure below, where I’ve plotted the same data in intervals of two for both the 1.5-second and 15-second scenarios.