[time-nuts] Line Frequency

Tom Van Baak tvb at LeapSecond.com
Mon Feb 10 18:39:56 UTC 2014


> My desired accuracy is based on what I have seen displayed on line frequency monitoring sites on the 'net.
> Numbers like 60.053 etc. That says (roughly) 1E-5 accuracy.

Ok, I'm with you. I understand your 1e-5 number now. Using that display format will be fine.

Also realize there's more to frequency measurements than digits. The other factor is the time constant, or the interval over which the frequency is measured, or the refresh rate. For example, is the display an average over the past second, 10 seconds, or minute or hour, etc.

> Now it may be that they show more precision than is warranted. Or they are averaging over a number of cycles. 

No. Perhaps you're thinking that "averaging" is an automatic path to greater precision. In this case it isn't. That's the amazing thing about this field of time & frequency -- there are many examples where the longer you average the *worse* the precision. And there are examples where more or less averaging makes *no* difference in precision. It takes people by surprise. This is why ADEV plots are so revealing: they clearly show over which time spans that averaging helps vs. averaging makes no difference vs. averaging actually hurts.

Here's the MDEV of mains (100 days, US western grid): http://leapsecond.com/pages/mains/mains-picpet-22.gif

For averaging from 1 second to 20,000 seconds your mains precision is either worse or no better with more averaging. Only after 5 hours of averaging do you start to get more precision. You reach the 1e-5 (10 ppm) level of stability after about 2 or 3 days. You get 1 ppm accuracy after about 2 weeks.

/tvb




More information about the Time-nuts_lists.febo.com mailing list