[time-nuts] GPS noise reduction

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sat Apr 5 22:30:52 EDT 2008

> Bruce,
> Thanks for verifying that a common OCXO for the TIC clocks will work.
> I wasn't looking forward to downgrading to individual XOs. I spent
> years of lab time developing a 100 MHz TIC that doesn't suffer from
> count variations when the phase gate edges and clock edges are
> coincident. I ended up using a prescaler with a longer hold time
> than its setup time (literally a negative hold time) so whenever
> the setup time isn't met the negative hold time keeps the prescaler
> in the previous enable state. I wanted a low cost design without
> high density packages and faster TIC rates generally require high
> density parts or ECL logic, so I settled on a 100 MHz TIC. A 1/16
> 2^N scaler was added to the Shera filter so the filter time and
> loop gain can both be adjusted to 16x better resolution. Software
> sawtooth correction and smart holdover with age and temperature
> compensation were also added. Living with the remaining GPS noise
> is easy disciplining the MTI260s and the combination gives very
> satisfactory performance in the intended application.
> Thanks again,
> Richard

Actually one can achieve sub nanosecond resolution without resorting to 
ECL or fast FPGAs.

If one uses hardware correction of the PPS sawtooth error then a single 
HCMOS or ACMOS flipflop has more than sufficient resolution.
If one wants to reject outliers then one may need to use 3 flipflops, 
otherwise one can rely on the TRAIM capbilities of the GPS receiver to 
reject outliers.

If one uses an ADC to sample say a 1MHz sinewave (produced by dividing 
down the OCXO frequency and low pass filtering) on the PPS edge then 
subnanosecond resolution combined with the ability to make software 
corrections of the sawtooth error is possible.

Low pass filter the PPS signal and use it as the analog input to a 
pipeline ADC clocked by a source phase locked to the OCXO output.
Calculate the position of the midpoint of the PPS transition (or some 
related quantity) from the ADC sample sequence.
Use this as a measure of the OCXO phase relative to the PPS pulse from 
the GPS receiver.

There are also other methods such as using linear ramps to interpolate 
between TIC clock edges.


More information about the time-nuts mailing list