nRF52: STOP in Rx state -> No event?

[Bruce McKenney]

I tripped over what appears to be a change in the RADIO between nRF51 and nRF52.

For the particular case of issuing a STOP task when the radio is in Rx mode, there is no END event; observed behavior (watching the area around NRF_RADIO+0x100) is that there is no completion event at all. This of course prevents the END->DISABLE shortcut (nor its software equivalent) from working properly.

Comparing nRF51 RM (v3.0) fig. 22 and nRF52 OPS (v0.6) fig. 30, this change appears to be intentional. An analogous change was also made to STOP-ping in the Tx state, but I think that's a much more unusual case.

In case you're wondering, STOP-ping in Rx state is not so unusual: It's what one does in the case of a timeout, supposing that the other side is never going to answer us.

I was able to succeed by spinning waiting for the STATE to transition to RXIdle, but I wonder if this is what was intended.

Is this the recommended method? If so, how long should I expect this transition to take? (I don't much like spinning without a timeout.)

[Susheel Nuguru]

I am not sure 100% of the rationale behind this but it actually made some sense to me.

maybe to remove the confusion that, if we generate END event after STOP task, then it can be confused as TX/RX transaction is complete even though it wasn’t!!

The specific use case you mentioned about STOP-ing on timeout, you can do DISABLE instead of STOP task, that will achieve the same goal instead of STOP_TASK->END_EVENT->DISABLE_TASK.

[Susheel Nuguru]

I have tested the behavior on nRF51 and it is the same as nRF52. The END event is only generated when the transaction is completed and not when the RADIO is stopped or disabled. Looks like the diagram has a small error.

[Bruce McKenney]

Thanks, I'll try the DISABLE. I didn't try that before since there's no arrow for it in the diagram viz. illegal transition. (In the process of tracking this down, I accidentally tried a few "illegal transitions" and I didn't enjoy them much (:-)).)

[Susheel Nuguru]

Disable is not illegal transition. If you look at the state diagram, you can see that disable can be called from RXRU/TXRU or RXIDLE/TXIDE or RX/TX. All of them grom into RXDISABLE/TXDISABLE state.

[Bruce McKenney]

Sure enough, there it is. I managed to mis-read this (4-5 times) before.

Thank you.