pps impulse jitter

[Andy11011]

Hi) I've tested that module and I want to use it... But I would like to improve accuracy. The impulse jitter is about 2 - 7 us. The main reason I think is the jitter of gpiod_set_value() function... Also sometimes the pulse is not generated because of delay for entering interrupt... Can you give me some clue about how to get rid of these drawbacks? Maybe it is reasonable to use some Timer in pwm mode instead of gpio... I'm using AM335x SoC (Beagle Bone Black board).

[jsln]

Hi, I guess there may be many reasons that could explain differences in measured jitter, what kind of loads are you applying to your CPU while doing the measurements? Have you compiled the kernel with a low latency preemption model? RT_PREEMPT?

However, I still see 2-7us as something reasonable and I would think this is difficult to lower in the BBB unless you go with a PRU implementation.

[TNBU0]

Would a pru implementation be as simple as compiling with clpru? Do you have any hints for accomplishing this?

I am seeing ~+-7us offset on the scope for the pps between a master and client PTP device.
GPS -> Master Beagle running ptp4l&phc2sys.
Client beagle has ptp4l & phc2sys running.

I see from a discussion dating back to 2015 on sourceforge (involving you and Richard C) that using DMTimer can achieve roughly +-40ns. Did you ever accomplish this @jsln ?