sx1262 radios were not being fully init if config.lora.use_preset is false #4167
Conversation
I was testing a virgin board (just using default channels, flash etc...) and had set only lora.region. I noticed warning messages about waiting for rx completion when those shouldn't be occurring. So I looked at the boot messages and saw: INFO | ??:??:?? 3 SX126x init result -10 We weren't treating that as fatal (which we probably should). Later when the channel was setup many of the lora registers would then get properly set but unfortunately there is some init in the last half of SX126x::begin this is never run when this function bailed out. The root cause seems to be: * config.lora.use_preset is false * therefore the second clause in RadioInterface::applyModemConfig is used * but config.lora.sf & coding_rate & spread_factor & bandwidth are zeros and nothing then provides sensible defaults. * Those zeros get passed in to sx1262::begin and it bails too soon This fix is to provide sensible defaults. Later in the boot (once channel is selected) these values are reset correctly anyways. Tested on a heltec tracker 1.1 but I suspect any sx1262 board should show it
src/mesh/RadioInterface.cpp
Outdated
| if (sf == 0) | ||
| sf = 11; | ||
| if (bw == 0) | ||
| bw = 250; // Just guess and use long fast settings |
There was a problem hiding this comment.
Maybe also carry over the ternary logic for checking if we're on 2.4G lora?
There was a problem hiding this comment.
I don't think that's super critical - because as long as these initial values are at least acceptable for the hw, later in the boot we'll be plugging in the 'real' settings based on the channel.
|
btw - I don't know if any release builds have been made that include this bug, but if they have then I think doing a release might be worth considering |
geeksville
changed the title
|
hmm @thebentern couldn't reproduce. And I went back to my board (been using it for other stuff) did factory reset and I couldn't either. From looking at the code I don't see any way for a factory reset to fail to set config.lora.use_preset true. But it definitely was on my board at somepoint. I wonder if we have some sort of odd corruption bug and I was unlucky? oh well. Changing this bug description to be non-urgent. but IMO still a good change because this !use_present clause was definitely wrong (and then much later in the boot leading to crummy rx behavior because DIO2 stuff had never been run. |
|
It looks like I had this happen on a tbeam. I have no idea how it got into this state. |
I was testing a virgin board (just using default channels, flash etc...) and had set only lora.region.
I noticed warning messages about waiting for rx completion when those shouldn't be occurring. So I looked at the boot messages and saw:
INFO | ??:??:?? 3 SX126x init result -10
We weren't treating that as fatal (which we probably should). Later when the channel was setup many of the lora registers would then get properly set but unfortunately there is some init in the last half of SX126x::begin this is never run when this function bailed out.
The root cause seems to be:
This fix is to provide sensible defaults. Later in the boot (once channel is selected) these values are reset correctly anyways.
Tested on a heltec tracker 1.1 but I suspect any sx1262 board should show it