I have made several meters to measure the mains frequency to various accuracy, but not using the period measuring method. There is no guarantee that all the mains cycles are identical for several reasons. Whilst a mechanically made sine waves may all look the same, tolerance on exact magnetic field strength and windings positions will make zero crossings (or any particular voltage levels) do vary slightly. In a large grid this is even more likely when one is looking for changes of parts per million as we are here. My units used a phase locked loop to multiply the mains frequency to a convenient figure (500 kHz) but these inherently have a frequency averaging effect over a significant number of cycles. They also use ALL cycles which again means that cycle  timing errors have much less effect. I could thus measure the 50Hz to 3 decimal places at 1 reading per second, or 2 decimal places at 10 readings per second, but as the loop bandwidth (effectively an averaging time) of 1 Hz (1 second) or so gave fairly stable readings to great accuracy on a smallish generator (100 kVA). These were used by the film industry to make sure that lighting flicker from discharge lighting did not cause varying exposure of film. Modern electronic cameras don't have this problem for a number of reasons, and one can see the effect live anyway, but of course one never knows what is actually on film until it is developed! I think this variability on DD is due to lack of averaging, and cycle by cycle timing differences, but these do not make any difference to normal usage of electricity.

I have made several meters to measure the mains frequency to various accuracy, but not using the period measuring method. There is no guarantee that all the mains cycles are identical for several reasons. Whilst a mechanically made sine waves may all look the same, tolerance on exact magnetic field strength and windings positions will make zero crossings (or any particular voltage levels) do vary slightly. In a large grid this is even more likely when one is looking for changes of parts per million as we are here. My units used a phase locked loop to multiply the mains frequency to a convenient figure (500 kHz) but these inherently have a frequency averaging effect over a significant number of cycles. They also use ALL cycles which again means that cycle  timing errors have much less effect. I could thus measure the 50Hz to 3 decimal places at 1 reading per second, or 2 decimal places at 10 readings per second, but as the loop bandwidth (effectively an averaging time) of 1 Hz (1 second) or so gave fairly stable readings to great accuracy on a smallish generator (100 kVA). These were used by the film industry to make sure that lighting flicker from discharge lighting did not cause varying exposure of film. Modern electronic cameras don't have this problem for a number of reasons, and one can see the effect live anyway, but of course one never knows what is actually on film until it is developed! I think this variability on DD is due to lack of averaging, and cycle by cycle timing differences, but these do not make any difference to normal usage of electricity.