Real-time hardware capabilities of 5340 and 52833

RE: Real-time hardware capabilities of 5340 and 52833

Pret — Fri, 23 Apr 2021 16:59:15 GMT

Thank you very much, Einar. Very helpful indeed

RE: Real-time hardware capabilities of 5340 and 52833

Einar Thorsrud — Thu, 22 Apr 2021 15:46:53 GMT

Hi Alex,

[quote user="Pret"]

Just to make sure I get your points as intended, when you say CPU do you mean an external CPU to the nRF52 device to configure the nRF52 device for the operation (counter mode as you mentioned)? If so, would the 5340's application core be useful for that?

Same about the PWM case, when you say using CPU - would the 5340 app core serve this?

[/quote]

When I say CPU I refer to the CPU core(s) in the nRF. The key here is that the programable peripheral interconnect (PPI) can be used to connect HW peripherals within the nRF so that events in one peripheral can trigger tasks in the other. See Peripheral interface and PPI for more information. There are some minor differences between the nRF52 and nRF53 here, but the fundamentals remain the same. In a nutshell you can think of PPI as a very simple variant of the programmable interconnect you have in FPGAs, but without the programmable logic.

And when I referred to the PWM peripheral, that is a dedicated peripheral that generates a PWM signal without need for CPU intervention other than for chancing configuration. See WPM chapter in PS (or for 53 here).

[quote user="Pret"]I'm just trying to get the idea whether our application can be served by a single-core nRF chip bearing in mind the regular BT stack operation and that PWM or time-between-events counting features, or this is not something that can be handled in hardware due to BT protocol real-time workload and we need either a separate controller on board or nRF5340 utilizing its app core..[/quote]

I understand. I still do not know enough about your use case to say if the nRF52840 will do, but if you do not have strict timing requirements for when you change PWM duty cycle or when you actually process the times you have measured using GPIOTE and PPI (if you do as suggested), but if that has a bit of slack, then using a single core device might work (and often does). If not, then the nRF53 could be a good option.

RE: Real-time hardware capabilities of 5340 and 52833

Pret — Thu, 22 Apr 2021 15:23:33 GMT

Thank you Elnar

Same about the PWM case, when you say using CPU - would the 5340 app core serve this?

I'm just trying to get the idea whether our application can be served by a single-core nRF chip bearing in mind the regular BT stack operation and that PWM or time-between-events counting features, or this is not something that can be handled in hardware due to BT protocol real-time workload and we need either a separate controller on board or nRF5340 utilizing its app core..

RE: Real-time hardware capabilities of 5340 and 52833

Einar Thorsrud — Thu, 22 Apr 2021 12:54:14 GMT

Hi Alex,

I do not know enough about your use case to say which approach you should use or if it is feasible on the single core nRF52 device. There are some general concepts that are relevant, though.

For counting time between external events you can use GPIOTE and TIMER in counter mode connected via PPI/DPPI. That way this can be achieved in HW, only needing the CPU for configuration (and using the data/count for something useful).

Regarding PWM, if you need a continuous PWM signal you can use either the PWM peripheral (or a TIMER in timer mode connected to a GPIOTE via PPI), depending on your needs. If you use the PWM peripheral, it can "play out" a sequence from DMA, so you can change the duty cycle without CPU intervention for a predefined sequence. If you need this dynamically then you must use the CPU.

Einar