RCP model latency constraints

[carnalfromst]

Hello,

I am currently looking at the timing constraints in the RCP dual chip implementation.
Is there any timing constraint to receive any feedback from the coprocessor after sending a command or any other timing constraints to handle a cmd/evt.

I would be very concerned if the timing constraints are in the range of a couple of 100us whereas I would be happy if it is in the range of a couple of 100ms.

I would expect that all real time traffic/management are all handled in the Coprocessor and the only timing constraints I may have would be on procedure timeout which I guess are at least above a couple of 100ms.

Could you please provide me some information to understand how much the system is relaxed timing wise in RCP model ?

Regards

[bukepo]

100ms looks a lot of latency. I'm not sure what's the cause of the latency. If it's caused by the transport between the host and the RCP, I think it would significant affect the performance. For example, if it takes 100ms for a received frame reaches host, is it acceptable to your use cases? On the other hand, does it make sense for the host to wait for 100ms to know whether a frame is successfully transmitted or not?

[carnalfromst]

Hello,

Thanks for your reply.
My request might have been confusing. I do not experiment any issue as of today. I am only wondering if there are any timing requirements in the RCP model for OpenThread regarding the transmission over the interface.

What I wanted to say is that this would be very difficult to commit 100us between a command sent from the Host to the Coprocessor and the response from the coprocessor whereas it will be quite easy to commit 100ms.
As well, it would be very difficult to commit that a frame sent from the Host would reach the radio in less than 100us whereas it should be quite easy within 100ms.

So the question is more to know whereas there are any requirement on the interface speed and on the time for a device to process the command/packet to have OpenThread working fine in RCP model.

Regards.