Implement proper AGC reset for SX126x, LR11x0 & SX1276

[weebl2000]

Might be good to have this in RadioLib, based on these PRs:

• fix agc reset on SX126x, SX1276 & LR11x0 chips meshcore-dev/MeshCore#1743
• Implement 'agc' reset for SX126x & LR11x0 chip families meshtastic/firmware#9705

For SX126x & LR11x0: reset analog registers, for SX1276: sleeping should be sufficient. I based this off the SX1262 datasheet & user manuals.

relates to #521

[jgromes]

Thank you, this will be a very useful contribution!

I know the issue has been reported on SX126x - are there some reports this is also happening on LR11xx too, sicne this PR also contains reset method for those radios? And if so, is there a test that could be run on LR2021 to figure if it is happening there as well?

[weebl2000]

Thank you, this will be a very useful contribution!

I know the issue has been reported on SX126x - are there some reports this is also happening on LR11xx too, sicne this PR also contains reset method for those radios? And if so, is there a test that could be run on LR2021 to figure if it is happening there as well?

No reports on LR11x0 that his happens there too. I feel like the LR11x0 is a little more sophisticated with the AGC than the SX126x.

But I dove into the docs and LR1110 is largely done in the same way as the SX1262, so I added it for anyone that would like to try it. It could definitely help when an LR1110 is exposed to big interference say at like a cell site.

UPDATE: LR2021 is largely the same, implemented there too.

[jgromes]

Thanks, I think should be safe to merge - however, before doing that, I will try to run some tests to see if this issue actually manifests on the other radios from Semtech, because we only have it confirmed on the SX126x series so far. So if it doesn't actually manifest there it doesn't make that much sense to be including the AGC reset methods for the other radios IMO. And if it does manifest, then we will know for a fact whether AGC reset resolves it or not.

[weebl2000]

Thanks, I think should be safe to merge - however, before doing that, I will try to run some tests to see if this issue actually manifests on the other radios from Semtech, because we only have it confirmed on the SX126x series so far. So if it doesn't actually manifest there it doesn't make that much sense to be including the AGC reset methods for the other radios IMO. And if it does manifest, then we will know for a fact whether AGC reset resolves it or not.

I'm fairly certain the LR1110 is prone to the same issue, but it's just seen/reported less because it's mainly integrated in more "mobile" applications where power and form factor are most important (hence sub-ghz, GNSS and WiFi in 1 chip) . Whereas the SX1262 is a workhorse in many boards that are also used as repeaters etc.

Most people aren't running repeaters on their LR1110. The calibration reset is valid though, would be good having it in I guess for those few users who might need it.

[weebl2000]

Got my hands on a T1000E tracker with LR1110. Tried with periodic agc reset and can confirm it works fine. Not as dramatic as with Heltec v4 and SX1262 and a GC1109 LNA, but I can definitely see the AGC reset is triggered correctly. Noise floor bounces briefly, goes back to where it was fairly quickly.

[jgromes]

After trying for a few days I was finally able to reproduce the issue with stuck AGC on SX1262. It seems to be heavily dependent on the output power level of the transmitter - I was only able to get the issue to manifest with good antennas, transmitting at +22 dBm. Frequency offset does not appear that critical, I was able to get it to break with as much as 8 MHz between transmitter and receiver.

The AGC does seem to work reliably though. I will try to also reproduce this on the other radios and test that as well.

[weebl2000]

@jgromes good work. Yes, biggest issues are reported with SX1262 sitting near large cell sites. These will often have 800 - 900 MHz transmissions at 30-35 dBW EIRP. You can imagine the massive power bleeding on sidebands/harmonics if you're just a few feet away from those sources.

[jgromes]

So I tried to do this with the other radios as well, and the SX126x is the only one where this problem manifests - all others (LR1120, LR2021 and SX1272) do not have this problem and after the transmitter is turned off the RSSI returns back to the original values.

In light of that I would propose to only implement this reset for SX126x. That will remove the necessity of having this in the PhysicalLayer abstraction.

[weebl2000]

So I tried to do this with the other radios as well, and the SX126x is the only one where this problem manifests - all others (LR1120, LR2021 and SX1272) do not have this problem and after the transmitter is turned off the RSSI returns back to the original values.

In light of that I would propose to only implement this reset for SX126x. That will remove the necessity of having this in the PhysicalLayer abstraction.

done