[LR11X0] pay more attention to selection of FSK preamble detector length

[lyusupov]

Brief description

Most of FSK radio protocols supported by SoftRF project are using 8-bit FSK preamble:

1. https://github.com/lyusupov/SoftRF/blob/master/software/firmware/source/SoftRF/src/protocol/radio/OGNTP.h#L24-L25
2. https://github.com/lyusupov/SoftRF/blob/master/software/firmware/source/SoftRF/src/protocol/radio/ADSL.h#L24-L25
3. https://github.com/lyusupov/SoftRF/blob/master/software/firmware/source/SoftRF/src/protocol/radio/Legacy.h#L25-L26

Current RadioLib code uses explicit 16 bit FSK preamble detector length

This causes almost 95% loss of packets while trying to receive radio packets sent by other (SX1262 and SX1276) devices ( these packets use 8-bit FSK preamble ).

This PR is filed to fix the issue of not receiving 8-bit FSK preamble packets.
The fix has been validated on Seeed SenseCAP T1000-E device - it uses Semtech LR1110 radio IC.

[jgromes]

Thank you for the contribution! Everything seems OK so this can be merged.

I think it would also make sense to add a similar configuration to SX126x, which also uses 16 bits as the default, so I would expect it to behave the same as LR11x0 in terms of the packet loss. SX127x does not change the default setting which should be 8 bits, but the configuration could also be added. Would you be willing to add those changes in another PR? If not I will add it myself.

Thanks again!

[lyusupov]

I think it would also make sense to add a similar configuration to SX126x, which also uses 16 bits as the default, so I would expect it to behave the same as LR11x0 in terms of the packet loss.

I am not sure that it is necessary....

The SoftRF uses other libraries to drive SX1276 and SX1262.radios on a regular basis

However, before I got access to the actual LR1110 device - I've emulated the logic of the LR11XX SoftRF driver by building and running on SX1262: https://github.com/lyusupov/SoftRF/blob/master/software/firmware/source/SoftRF/src/driver/radio/radiolib.cpp#L60-L61

I can say that I have not noticed any issues while running FSK modulated radio protocols with 8-bit preamble together with SX126X RadioLib driver.

It looks like the LR11X0 uses different values:

1. preamble length for Transmitter
2. preamble Detect length for Receiver

[jgromes]

It looks like the LR11X0 uses different values:

1. preamble length for Transmitter
2. preamble Detect length for Receiver

That is also the case for SX126x, the last argument sets detector length to RADIOLIB_SX126X_GFSK_PREAMBLE_DETECT_16:

RadioLib/src/modules/SX126x/SX126x.h

Line 1236 in 709fbdb

int16_t setPacketParamsFSK(uint16_t preambleLen, uint8_t crcType, uint8_t syncWordLen, uint8_t addrCmp, uint8_t whiten, uint8_t packType = RADIOLIB_SX126X_GFSK_PACKET_VARIABLE, uint8_t payloadLen = 0xFF, uint8_t preambleDetectorLen = RADIOLIB_SX126X_GFSK_PREAMBLE_DETECT_16);

So I would expect it to behave in a simlar to way tthe LR11x0 did. I will go ahead and add this, even if it seems like this is not causing a problem, there might be some other reason for that.

[lyusupov]

Yes, I agree that SX126x needs a proper setting of preambleDetectorLen as well:

Hmm, I am curious, why I have not noticed high Rx packets loss ratio with 8-bit preamble detector on SX1262 and RadioLib...
I should probably test it again, when able.

SoftRF project uses a customized BASICMAC library to drive SX1262 on a regular basis.
We select the preamble detector length as follows:

https://github.com/lyusupov/SoftRF/blob/master/software/firmware/source/libraries/arduino-basicmac/src/lmic/radio-sx126x.c#L503-L529