These are some 12-hour plots of time-interval measurements. The measurements are made with a HP5334A universal counter. Channel A is connected to the Motorola Oncore M12+t saw-tooth correction board, while channel B is fed with a 1PPS pulse coming from a FE5680A atomic rubidium standard.

This plot is taken with sawtooth correction turned off

And with sawtooth correction turned on

This data is then high-pass filtered to remove the rubidium standard's drift, so we can see the noise in the PPS pulse.

Sawtooth correction turned off.

The noise without sawtooth correction is about 9.13ns(rms).

Sawtooth correction turned on.

The noise with sawtooth correction is only 2.67ns(rms). This is about 10dB reduction in noise.