[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [questions] GPS+PPS vs NTP server, why a huge offset ?



On 20/06/2022 15:54, chris wrote:
Still doesn't explain how the pps state change gets into the
system at hardware level. Devil in the detail, as usual.
It sounds like a polled system, not an interrupt driven one,

Interrupt system are I think pretty much always polled at the hardware level.

which can never be as accurate as the latter. Assuming a poll
rate of say, 1mS, not unreasonable for a modern cpu, Then if
the pps state change is not seen at one poll, but is seen at
the next, then the uncertainty can never be better than that

The uncertainty is the clock period in the capture hardware, because the hardware counts and returns the the number of its clock cycles, so you know very precisely how long it was from the the PPS to when you polled. Actually I think something like that is pretty standard for most microcontrollers.

I believe there are Ethernet cards that do this to support, I think, Precision Time Protocol, which I suspect is what high frequency traders use.

1mS, since it cannot be known when during that 1 mS the state
transition occurred. Best results for ntp really need a true
real time os, but it's  far more common to use general purpose
os's because of the out of the box functionality., but not
ideal.
--
This is questions@xxxxxxxxxxxxx
Subscribe: questions+subscribe@xxxxxxxxxxxxx
Unsubscribe: questions+unsubscribe@xxxxxxxxxxxxx