[time-nuts] Question regarding 1PPS timing accuracy and GPSDO
magnus at rubidium.dyndns.org
Tue Jan 11 18:58:53 UTC 2011
On 11/01/11 18:06, Chris Albertson wrote:
> On Tue, Jan 11, 2011 at 8:14 AM, Tijd Dingen<tijddingen at yahoo.com> wrote:
>> When a GPS devide XYZ datasheet quotes a "timing accuracy" on the 1PPS
>> output of lets say 1 us RMS, would it be useful? I.e would /you/ consider
>> using it? What if it was 100 ns RMS on the 1pps? Or 10 ns RMS? Or ... At
>> roughly what number would you consider the 1pps as a candidate for your
> Buying a better GPS unit is the most cost effective why to improve the
> total system performance. Everything else you can do has a llower
> "bang per buck" ratio. For example going from a 1us to a 10ns GPS
> cost nearly nothing but s 100X improvement
No. You won't get that kind of performance step. The reason is that when
receivers is speced to 1 us is that the receiver is so old, that it was
in the SA-days and they essentially copy-pasted from the GPS-ICD-200
spec the 1 us value. It was not based on measurements.
When measuring the stability, you also need to compare related values,
if it is 10 ns RMS, 20 ns 2-sigma (99%) or 30 ns 3-sigma (99,7%) values,
then you have the same number but may be fooled to believe that the
first spec is better.
So just blindly looking at that spec may really fool you.
Also, just using the PPS may be less optimum. Also getting the timing
error of the PPS is important for newer receivers, as otherwise will the
period time of the internal clock (typically 15-20 MHz) will limit the
resolution, so the "sawtooth error" can be reduced. The resolution of
the sawtooth correction becomes another factor.
There is even more things to do.
A good antenna and good antenna location is worth investing in.
A timing-capable receiver with T-RAIM should be a good investment. The
M12+T for instance.
More information about the time-nuts