Radio events clock

RE: Radio events clock

Davide — Tue, 22 Aug 2017 15:45:18 GMT

If it can be useful: going on with the tests I found that if I use the 2M PHY for doing the same test this behavior disappears.

RE: Radio events clock

Davide — Mon, 07 Aug 2017 09:43:26 GMT

Ok, no worries. If you manage to get some details of what's going on I'll be very grateful, otherwise thanks for the support.

RE: Radio events clock

Hung Bui — Mon, 07 Aug 2017 08:55:02 GMT

Hi Davide,

I tried to reproduce what you described here and saw the same thing. The value varied but never get odd value. It looks pretty strange. I will need to talk to some one in R&D. But since it's not really a bug, it won't have high priority for now. I will get back to you when I get some more information from R&D.

RE: Radio events clock

Davide — Tue, 01 Aug 2017 14:07:40 GMT

For what I see in the data I collected, it may be that the radio works mainly at 16MHz (because the interval between RADIO->EVENTS_TXREADY and RADIO->EVENTS_ADDRESS is always 627 ticks), but some time delays or some events seems to be clocked at 8 MHz (the interval between RADIO->(TX)EVENTS_END and RADIO->(RX)EVENTS_ADDRESS it is always even).

Here I copy some of the results for the the interval between RADIO->(TX)EVENTS_END and RADIO->(RX)EVENTS_ADDRESS. Keep in mind that: the values represent 16MHz ticks, and I selected the samples among all the collected ones because the last one is odd.

RE: Radio events clock

Hung Bui — Tue, 01 Aug 2017 13:46:33 GMT

Hi Davide, Sorry that I haven't got chance to test yet, but I got some information that the radio runs on 4MHz clock (not 100% sure) so the event throwing out is actually on that resolution. This could explain why you only see even value. Could you post some of the counter values you got ?

RE: Radio events clock

Davide — Fri, 28 Jul 2017 09:31:35 GMT

Yes, I get odd and even values out of the timer in the general case, but the interval between RXREADY and ADDRESS is even 99% of the times: the measurement noise, which is there, doesn't make odd values. I cannot change the address size because I'm working with Bluetooth and softdevice s140, I don't think I can change address size keeping the softdevice working.

I don't know if it can be the nature of the radio, I'm not into CMOS radio design actually...but it wouldn't surprise me if it is (symbol rate is 1Mbit, clock is 16MHz, there may be some rounding somewhere, I don't know).

RE: Radio events clock

Hung Bui — Thu, 27 Jul 2017 19:56:46 GMT

You can create a case in support portal (my page) and note my name/ link to the case there. Or u can send me a PM with the link to your file on dropbox or google drive.

As I understand , u can see both odd and even value if u test with a simple delay and gpio or other radio event. But u never get odd on that partocular READY and ADDRESS event ? Could that be the nature of the radio ? Can u change the address size and test ?

RE: Radio events clock

Davide — Thu, 27 Jul 2017 13:26:10 GMT

I'm quite sure the timer is working properly because on another RADIO event I get odd values, but this is not the event I'm interested in.

I don't have any problem in giving you the source code, but I don't know how to do it.

RE: Radio events clock

Hung Bui — Thu, 27 Jul 2017 12:53:03 GMT

Hi Davide,

I'm sorry that we are quite short in staff now because of the summer vacation period. I can't to test right now.

It would be very nice if you can provide the firmware so we can reproduce here. I would suggest to have something like:

Setting up PPI, connect a GPIOTE EVENT 1 to TIMER TASKS_START
Setting up PPI, connect a GPIOTE EVENT 2 to TIMER TASKS_CAPTURE
Start a loop, trigger GPIOTE EVENT 1, put some delay , trigger GPIOTE EVENT 2
Print out the Captured value and go back to step 3.

This will clearly show the strange behaviour that we only have even capture value.

RE: Radio events clock

Davide — Wed, 26 Jul 2017 14:04:21 GMT

That's exactly what I did with the same result. I have some noise over the measurement (due to propagation time and clock drifts on the two devices), but I (almost) never get odd number of timer ticks. With the pin toggle tecnique I see that the noise makes the transition happen only on 8MHz edges.

RE: Radio events clock

Hung Bui — Wed, 26 Jul 2017 13:56:08 GMT

How did you verify that the granularity is 8MHz not 16MHz ? I would suggest you to test using a pin toggle where you can measure using logic analyzer and compare with the Timer count you have using PPI.

RE: Radio events clock

Davide — Wed, 26 Jul 2017 12:19:36 GMT

The PPI is configured to connect the RADIO->EVENTS_READY with the TIMER->TASKS_START and then the counter value is captured connecting RADIO->EVENTS_ADDRESS with the TIMER->TASKS_CAPTURE. Then the timer itself triggers the execution of a routine which stops and clears the timer and retrieves the captured value.

RE: Radio events clock

Hung Bui — Tue, 25 Jul 2017 14:18:44 GMT

Hi Davide, I think it should be 1/16MHz accuracy. At least it's the frequency of the timer's Tick can achieve.

Could you let me know how you connect PPI between the Radio and the Timer ?