ADR interaction with interference? (US915)

ChirpStack Network Server
1

I’m getting started with a LoRaWAN deployment; initial testing is with a Mikrotic LR9 gateway on a high-level site (HAAT is approximately 477m, with no other known 900MHz transmitters at the site) and nodes located ~15km from the gateway with a slightly-obstructed view of the gateway. Nodes are RAK5205s; US915 uses SF7-SF10. Traffic on the gateway is light; I observe traffic from 3 other nodes regularly

My test nodes - one is using the little “2dBi” antenna, the other is using a 5.5dBi Larsen ground-plane. Both are mounted at an indoor window. Chirpstack is configured with an install margin of 10dB.

Regardless of starting DR, ADR on the node with the Larsen ground-plane quickly drops to DR3 (SF7) as expected, packet reliability is good (close to 100%). ADR then drops TxPower on the node, quickly taking it down to a TxPower value of 10 (I believe that’s 30dBm - 2 * 10dBm -> 10dBm).

With node power reduced to 10dBm (~15dBm EIRP), the packet reliability is low, less than 30%, and ADR then starts tinkering with nbTrans, raising it to 3 most of the time, sometimes having an episode, lowering it to 2, then 1, then back up to 2, then 3.

Hence, most of this time, the node is sending 3x packets at DR3 (SF7).

Meanwhile, the other node, with the small “2dBi” antenna, settles at SF8 (DR2) when started at SF9 (DR1), TxPower is usually 0 (20dBm) and nbTrans is usually 3. Packet reliability is roughly the same as the node with a better antenna.

Edit: just checked carefully; the “good” antenna node gets ~55% packets, the “small” antenna node gets ~71% packets. So the “worse” node gratuitously does better.

In other words, ADR appears to adapt each node to about the same gratuitously poor packet reliability (less than 70%) regardless of antenna quality :slight_smile: I’m not sure that’s the intention of ADR.

One thing that occurs to me, we have a great deal of power/gas utility traffic on 915MHz widely across the region, using Itron / Silver Spring mesh networking. I have not had a waterfall receiver at the high-level site but can only imagine it is hearing a cacophony of short data packets across the uplink channels. While LoRa modulation is resistant to interference, I suspect that interfering packets around the same time-length as LoRa packets at SF7 may cause unpredictable SNR variations at the gateway.

This brings me to my question, which might be better for Semtech now that I think about it: does anyone have an idea how well ADR responds to this kind of interference?

Thanks -
Dana

1 Like
2

I note that my node with the small antenna, which has better packet reliability than the node with large antenna, is running SF8/DR2. Now I wonder if SF8 is more resistant to the (suspected) Silver Spring interference (the SF8 packets will be much longer than the interfering pulses).

I’ll try limiting ADR to DR2 and see what happens.

3

A quick update. After setting max DR2 and waiting overnight, I’m now seeing ~75% packet reliability on the node with a small antenna, and ~90% (or more) reliability with the ‘good’ antenna. I’ll have to wait for the next LinkADRReqs to confirm it, but it looks like both nodes are running nbTrans == 2, and it appears that ADR set the ‘good’ node power to 10, aka 10dBm.

I do wonder how the algorithm thinks it’s better to turn the power all the way down and then send every frame twice; perhaps I should increase intallation margin for ADR?