[time-nuts] Correcting jitter on the 1 PPSsignalfromaGPSreceiver.
Tom Van Baak
tvb at LeapSecond.com
Sun Sep 14 14:06:07 EDT 2014
> They say that the system clock is 12.504 MHz and that they use both the
> rising and falling edges. That is about 40 nS between quantization time
> slots. The PPS can only appear on a 40 nS edge. I should be seeing
> 40 nS jumps in the waveforms. I do see ~40 nS jumps but they are less
"The PPS can only appear on a 40 ns edge" -- of the imprecise, unstable, uncompensated, unshielded, crystal oscillator on the Resolution-T board. It may be 12.504 MHz, but it certainly isn't (and not intended to be) 12.504000000 MHz. There's also instability in your 'scope or counter. Thus you will see significant jitter, drift, and wander in the observed 40 ns edges. All this is normal and expected any time you "beat" two oscillators against each other.
The net result is that the jumps are very evenly (and not Gaussian) distributed anywhere from 0 ns to 40 ns. Again, look at the raw data, plots, and histograms that I provided. Especially the "zebra" plots which show just how varied the sawtooth error is over the span of minutes, hours, and days:
http://leapsecond.com/pages/MG1613S/tic-72-hour.gif (3600x1800 pixels!)
> The waveform timing is clearly quantized but I am seeing ~2 nS jumps.
> My scope is a Rigol DS2202 which samples at 1 GSP in 2 channel mode.
> Could this 2 nS quantization be a result of the scope? Perhaps I should
> get my 400 MHz analog scope out (Tek 2465B) and repeat the measurements?
Off-list you mentioned you don't have a ns or sub-ns time interval counter, so yes, I guess you should try the analog 'scope. I'm wondering now if your GPS receiver more stable than your scope's timebase.
More information about the time-nuts