# [ntp:questions] Jitter on \$GPGGA of Sure board

Gom gom at gom.invalid
Wed Mar 21 18:30:40 UTC 2012

```>> However if you use NTP, it will average the values, hence reducing the
>> jitter by sqrt(number of seconds of polling).
>>
>> So if you use 8 seconds, jitter of the Sure NMEA clock will be
>> 10/sqrt(8)= 3,5 ms.
>>
>> This is still waaaay jigher than the jitter of the Sure PPS : 22 ns
>
> You are comparing apples to oranges. The Sure nmea jitter is far far
> higher than that. The Sure PPS jitter is far higher than that if you use
> the RS232 level conversion. The Sure PPS jitter is far higher than that
> if you run it into your computer's interrupt routines.

NMEA jitter is measured from my PC, so it's actually correct.
Of course PPS jitter is theoretical....

Anyway on Windows, the system clock is updated every 1 ms max, so my
goal is only to get submillisec , like 100 µs,  accuracy.
With the 30 µs "jitter", I'm within my specs.

> It is like comparing the nmea sentences with fundamental timing
> limitations of the GPS (3m is about 3ns and differeential at 1 cm is
> about .03 ns.) but there is no way of getting that into your computer.
>
>
> Note that ntp does NOT take the average. it takes the median of about
> 60% of the past 8 inputs (as I understand it).
>

Yes, it does. But it's true it's from the median 60% of the past 8 inputs.

From refclock.c

/*
* Determine the offset and jitter.
*/
pp->offset = 0;
pp->jitter = 0;
for (k = i; k < j; k++) {
pp->offset += off[k];
if (k > i)
pp->jitter += SQUARE(off[k] - off[k - 1]);
}
pp->offset /= m;

```