[time-nuts] Re: Has Line-Frequency Timekeeping Become Less Reliable in the U.S.?

[Bill Dailey, MD, MSEng, MSMI]

i decided to try to fingerprint this with my data.  i thought you may find
it helpful in troubleshooting zero crossings.  my device is FNET/GridEye
FDR from UTenn.  i just archive the data locally.

The externally supplied 31-day "Electric Clock Drift" graph (time-nuts
post, 2026-07-19, Francis Grosz, Louisiana — power-line cycle counter vs
Thunderbolt GPSDO) partially aligns with the Clinton reconstruction: best
fit places its window at 2026-06-16 19:00 → 2026-07-17 19:00 UTC (± ~12 h),
same sign convention, r = 0.54 (integral), r = 0.48 (daily slopes). Three
major features reproduce — the Jul 6–7 collapse (−17.1 vs −17.6 s/day in
the two records), the Jul 12 decline, and the Jul 13 recovery. The residual
is dominated by negative-only excursions unique to the Louisiana record,
all beginning ~14:00 local time, consistent with afternoon
(thunderstorm-season) cycle loss in a zero-crossing counter rather than
grid behavior. The two records agree on the character and scale of EI time
drift but do not overlay point-for-point at digitization resolution.


On Sun, Jul 19, 2026 at 1:54 AM Francis Grosz via time-nuts <
time-nuts at lists.febo.com> wrote:

> Following up on Hal Murray's post, I have some data.  With all of the
> discussions about AI and Data Centers connecting to the grid and looking at
> the kind of load they represent, I became interested in what will happen to
> power quality and reliability.  For several months I have been collecting
> data from my power line.  This data consists of frequency over a short term
> (1 second); maximum, minimum, and average frequency over a 20 minute block;
> and long term frequency (drift of a power line clock vs. an accurate
> clock); RMS voltage (at 5 second intervals); AC waveform; and power line
> spectrum over 30 Hz to 1 kHz.
>
> The long term frequency measurement is made by comparing a nominal 1 pulse
> per minute from the power line (1 pulse per 3600 cycles of the line) to an
> accurate 1 PPM from a Thunderbolt GPSDO and graphing that over time, one
> measurement every minute so 1440 data points a day.  A 31 day graph of the
> Eastern Interconnect time variation as seen in Louisiana is here:
>
>
> [image: image.png]
> There is no significance to the value of the difference, only the change is
> significant.  The total variance is a bit over 35 seconds.  How significant
> this is depends on the use case.
>
> One additional note:  Last year I attended the IEEE Power & Energy Society
> General Meeting in Austin to learn what was happening in this area.  There
> was a lot of very interesting information at that meeting.  Unfortunately I
> am not attending this year's meeting which begins Monday, 7/20/26.
> However, many of the sessions will be streamed on IEEE TV.  I am not sure
> if an IEEE account is needed to access them, but I don't believe so.  If
> you are interested you can find the sessions at:
>
>    https://ieeetv.ieee.org/2026pes-general_meeting
>
> The Opening Plenary Panel Session begins at 9:30 AM EDT Monday.
>
>                         Francis Grosz
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com



-- 
Doc

Bill Dailey
KXØO