AGC reset doesn't appear to work with Heltec V4

[terminalvelocity23]

I've been tinkering with my repeater as of late, trying its sensitivity and noise floor in various conditions. Looks like there's an issue with AGC not being reset after getting hit by a strong signal (I've used a handheld radio to simulate a powerful interference source). The noise floor in the application drops to -120 dBm permanently, and the repeater gets "deafened" until rebooted. Building with the -D SX126X_RX_BOOSTED_GAIN=0 flag helps somewhat in the noisy urban environment, but a strong noise peak still can induce this problem.

As per code in RadioLibWrappers.cpp, "according to higher powers, just issuing RadioLib's startReceive() will reset the AGC". Apparently it isn't the case, because setting agc.reset.interval as low as 4 doesn't do nothing with this problem, and the noise floor still gets clamped at -120 dBm. Which is not surprising, according to this:
if (_noise_floor < -120) { _noise_floor = -120; // clamp to lower bound of -120dBi }

All in all, it seems to me that either this method of resetting AGC doesn't work, there's something at fault with the agc.reset procedure or something else going on.

[weebl2000]

Do you have powersaving enabled? If so it's probably the GC1109 FEM turning off. There's a PR open that fixes this. https://github.com/meshcore-dev/MeshCore/pulls/weebl2000

There's also a link there where you can build the firmware and try it for yourself.

[terminalvelocity23]

@weebl2000
No, powersaving was disabled for this.
I'll try that PR nonetheless, thank you.

[terminalvelocity23]

I ended up writing a watchdog that soft-reboots the repeater if the noise floor stays at -120 dBm for longer than 60s. It's a crutch but it works, restoring the sensitivity back to normal. Shame about losing the neighbor tables tho.

Would be nice to have a more elegant solution.

[IoTThinks]

Do you have powersaving enabled? If so it's probably the GC1109 FEM turning off. There's a PR open that fixes this. https://github.com/meshcore-dev/MeshCore/pulls/weebl2000

There's also a link there where you can build the firmware and try it for yourself.

@weebl2000 PowerSaving uses light sleep. So all GPIO state and RAM are kept unchanged during sleep.
We only need to keep GPIO state in deepsleep.

So it is likely due to other issues (if there is).

[IoTThinks]

@terminalvelocity23 My Heltec v3 also has noise floor at -120 after a day with agc.reset.interval 600.
It was based on MC 1.11.

After reboot, it will be back at -97.
Let me disable it with agc.reset.interval 0.

[terminalvelocity23]

I have feedback from somebody on my mesh that the issue can be observed on T114 as well. Nothing surprising, because RadioLib and its wrapper are the same.

[terminalvelocity23]

Do you have powersaving enabled? If so it's probably the GC1109 FEM turning off. There's a PR open that fixes this. https://github.com/meshcore-dev/MeshCore/pulls/weebl2000

There's also a link there where you can build the firmware and try it for yourself.

I've integrated your PR and still can reproduce the problem, both with powersaving on and off. The AGC gets stuck on low sensitivity, and the repeater can hear only strong signals, effectively stopping relaying the flood packets.

[weebl2000]

I've integrated your PR and still can reproduce the problem, both with powersaving on and off. The AGC gets stuck on low sensitivity, and the repeater can hear only strong signals, effectively stopping relaying the flood packets.

Can you try this commit? https://mcimages.weebl.me/?commitId=dev_plus&variant=heltec_v4_repeater

[terminalvelocity23]

Can you try this commit? https://mcimages.weebl.me/?commitId=dev_plus&variant=heltec_v4_repeater'

Sure I can. The same behavior is observed when I introduce a strong interference signal. Noise floor gets clamped at -120 dBm, I switch the interference off, and after 15 minutes uptime it still stays there. The lower flood traffic is consistent with, effectively, -30 dB of RX sensitivity (my normal noise floor is around -95 dBm).
I've rebooted and tested it afterwards for 10 minutes without overloading the AGC, saw much more flood packets being routed (like 3x).

[weebl2000]

Sure I can. The same behavior is observed when I introduce a strong interference signal. Noise floor gets clamped at -120 dBm, I switch the interference off, and after 15 minutes uptime it still stays there. The lower flood traffic is consistent with, effectively, -30 dB of RX sensitivity (my normal noise floor is around -95 dBm). I've rebooted and tested it afterwards for 10 minutes without overloading the AGC, saw much more flood packets being routed (like 3x).

Thanks. How strong is your interference signal and how are you generating it? Wondering if I can try to reproduce this myself to investigate.

[terminalvelocity23]

Thanks. How strong is your interference signal and how are you generating it? Wondering if I can try to reproduce this myself to investigate.

About -15 dBm at the antenna of my tinySA. I place it near the Yagi the repeater is connected to. This is an overkill, a -60 dBm signal from a portable radio like Quansheng UV5R+ with a matching antenna will do as well. It just heats up way more.

[weebl2000]

Thanks. How strong is your interference signal and how are you generating it? Wondering if I can try to reproduce this myself to investigate.

About -15 dBm at the antenna of my tinySA. I place it near the Yagi the repeater is connected to. This is an overkill, a -60 dBm signal from a portable radio like Quansheng UV5R+ with a matching antenna will do as well. It just heats up way more.

https://mcimages.weebl.me/?commitId=fixagcreset

Could you try this build? It calls sleep on the SX1262 when doing an AGC reset

[terminalvelocity23]

https://mcimages.weebl.me/?commitId=fixagcreset

Could you try this build? It calls sleep on the SX1262 when doing an AGC reset

Same deal so far. Gets stuck at -120 and stays there, but now it's occasionally fluctuating between -117 and -120. Noise floor level indicator doesn't appear to be rising to me.

[weebl2000]

Same deal so far. Gets stuck at -120 and stays there, but now it's occasionally fluctuating between -117 and -120. Noise floor level indicator doesn't appear to be rising to me.

That's actually promising. It suggests the AGC is working but the GC1109 FEM might be turned off now.

Can you try to build this one once more, it contains both GC1109 fix and AGC reset: https://mcimages.weebl.me/?commitId=dev_plus&variant=heltec_v4_repeater