In my last entry I compared the frequencies of my two 10 MHz frequency references. One is from FEI FE-5680A Rubidium Frequency Standard (RFS) and the other one is from Trimble Thunderbolt GPS Disciplined Oscillator (GPSDO).

The frequency accuracy of RFS drifts over time. The manufacturer specifies drift of less than 2×10^-9 per year, which for 10 MHz reference is 0.02 Hz . I don’t understand the drift mechanism, but in any case the frequency offset of this RFS needs to be measured and corrected. (Why? because I could.)

The frequency accuracy of GPSDO, on the other hand, does not drift with time. This is because the guys at the U.S. Naval Observatory (USNO) continuously monitor and correct the clocks in their GPS satellites to track UTC(USNO). The UTC(USNO) in turns tracks the Coordinated Universal Time (UTC) time maintained by BIPM in Paris.

In my last entry, the way I compared the two frequencies was by comparing the phase of the two waveforms on an oscilloscope. In that short time comparison, the two frequencies were virtually identical. The two waveforms were virtually stationary with respect to each other. In order to measure the difference between these two nearly identical frequencies, I need to track the phase difference over a much longer period of time. Here is the block diagram of the circuit that helps me do this:

For convenience, both 10 MHz references were first converted to 1 pulse-per-second (PPS) signals. One PPS drives the “start” signal of a Time Interval Counter (TIC) and the other PPS drives the “stop” signal of the same TIC. The TIC measures the time interval between the “start” and the “stop” signal. The TIC is build using a digital counter clocked at 50 MHz which gives 20 ns resolution. The interval values were then sent to a PC for data-logging. And here is the result:

The top graph shows the time interval measured. The horizontal axis is the elapsed time. Each sample point correspond to 1 second. Over a period of 8.3 hours, the PPS signal of the RFS drifted by 320 ns. This drift corresponds to an error of 1.1×10^-11 or 0.00011 Hz.

The bottom graph plots the self-monitoring performance of the GPSDO. It shows that the GPSDO is functioning correctly during the duration of the experiment.

The frequency error of 0.00011 Hz is pretty impressive. This RFS was calibrated about a month ago. This hasn’t always been the case. When I first acquire the RFS, its frequency error was 0.072 Hz. It wasn’t bad, it was good, but it is impressive now 🙂

Next: improve the resolution of my counter. Short of buying Stanford Research SR620 counter, I’m planning to add a linear interpolator to my digital counter. Basically it will use the rise time of an RC circuit to measure time in-between the counter oscillator “ticks”.

Next: Understand time stablitity measurement, also known as Allan Deviation.