On 2014-02-23, Rob <[email protected]> wrote: > Chris Adams <[email protected]> wrote: >>>I further studied the matter and found that on the CentOS 6.5 system >>>where I first checked the kernel is at 2.6.32 and the pps_parport module >>>is not yet included. >> >> When I ran this on a CentOS system, I didn't use the in-kernel PPS. I >> used the shmpps daemon from EPEL instead (user-space daemon, so not >> quite as accurate). > > Sure, on my own system I use an old gpsd in which I included a user-space > PPS daemon which works quite well, but for this purpose we can use all > accuracy we can get. Timestamping in the IRQ handler is probably a lot > better than in a user process.
On my system with a parallel port pps irq device, I get offset standard dev of around 2 micro sec. Testing with parallel port out pin the total delay between the timestamp on the putting the pin up, and the irq timestamp is about 1usec. (using the microsecond clock, I cannot time more accurately than that. Maybe sometime I will use the nanosecond timestamping). looking at the offsets, they are usually around 1-2 us with some excrusions up to 10usec (presumably interrupt conflicts). No idea what you were getting on your user process. _______________________________________________ questions mailing list [email protected] http://lists.ntp.org/listinfo/questions
