[time-nuts] Logging the grid frequency....

Tom Van Baak tvb at LeapSecond.com
Sat Feb 23 03:15:58 EST 2013

> So it is not correct to measure one point to a gnat's nose
> hair and call it "the grid frequency."


Yes, and this is true for any source of frequency. That's why when we specify stability the averaging interval is critical; the x-axis of a log-log ADEV plot.

One might look at every cycle to measure stuff like jitter, not so much to measure frequency over a tau of 0.016 seconds. The nice thing about ADEV is that a single plot can convey frequency stability for all intervals from as short as a single cycle to as long as days or months or more. Now with a 10 MHz standard we don't normally start an ADEV plot at a single (100 ns) cycle. But for 60 Hz it's perfectly natural to do so.

A histogram of period is another way to show variations in cycle time. This gives more information than a single ADEV point. But to show variations as a function of averaging time, a whole set of separate histograms, or overlaid histograms, are required.

> It might be more accurate to put a flywheel on a synchronous
> motor and measure its speed, because the time constant of that
> system is a whole lot closer to that of the real grid frequency.

I too was suspicious of digital or PLL or filtered methods of monitoring 60 Hz phase. To validate the digital methods I compared against an old synchronous wall clock. In the following animated GIF, a photo was taken exactly every 900 seconds (15 minutes):

It turns out the zero-crossing microprocessor digital time-stamping method exactly agreed with the old mechanical synchronous motor/inertia method. Satisfied with this result, I do all my mains phase/frequency logging using the digital time-stamp method (picPET).


More information about the time-nuts mailing list