LoRa node encounters failure in txfsk()

utkarshshah007 commented 5 years ago

After running for about an hour using a slight modification of ttn-otaa (just sending unheard join requests, the TTN gateway I'm using is disconnected), my LoRa end node hits:

FAILURE /.../Arduino/libraries/IBM_LMIC_framework/src/lmic/radio.c:429

This corresponds to an early assert in the txfsk() method. However, this node should be running in Lora mode, not FSK mode.

I assume this means that getSf(LMIC.rps) == FSK must be true. I've set my downlink datarate to SF9, as suggested in the ttn-otaa example.

Why might my end node be in FSK mode? How do I change this?

utkarshshah007 commented 5 years ago

I was able to discover the source of this issue.

For US-915, the datarate for joining is set in lmic.c as follows:

if( LMIC.datarate != DR_SF8C ) {
        LMIC.txChnl = 64+(LMIC.txChnl&7);
        setDrJoin(DRCHG_SET, DR_SF8C);
    } else {
        LMIC.txChnl = os_getRndU1() & 0x3F;
        s1_t dr = DR_SF7 - ++LMIC.txCnt;
        if( dr < DR_SF10 ) {
            dr = DR_SF10;
            failed = 1; // All DR exhausted - signal failed
        }
        setDrJoin(DRCHG_SET, dr);
    }

When LMIC.txCnt reaches 132, dr becomes -129. However, dr is an 8-bit signed integer, which means that it’s minimum value is -128. Therefore, it overflows and becomes +127.

utkarshshah007 commented 5 years ago

Here's a PR that solves this issue: #197

matthijskooijman / arduino-lmic

LoRa node encounters failure in txfsk() #193