apolloeme@gmail.com said:

I doubt if you will find a TCXO that puts out 60 Hz.  But it's only one chip
to make a divider.  The trick is to do it in software rather than hardware.

The real question is what sort of error pattern are you interested in.  The
power company has good long term stability.  How good is your TCXO?  It would
be fun to play with the numbers.  On the other hand, can you derive your 60
Hz from a GPSDO?

--
These are my opinions.  I hate spam.

Since the power line has the desired long term stability but is poor on the
short term, I wonder if a solution might be to use it as the reference for
a "power line disciplined oscillator."

You would want a filter time constant of several hours in the control loop
to smooth out the variations in the 60 Hz.

Or it might be easier to rip the guts out of the Radio Shack clock, just
keeping the display, and drive the display with an Arduino and a DS3231 RTC
chip as the reference.

On Thu, Dec 14, 2017 at 4:05 PM, Hal Murray hmurray@megapathdsl.net wrote:

--

--Jim Harman

I dont think working that way would give a stable clock in the UK. Our
frequency can vary more than the US but the number of cycles between 0800 on
one day and the next is mandated to be correct (I presume +/- 25 :-))  ). So
you would be chasing a moving target, and at no time of the day need the
frequency be 50.000Hz

Alan
G3NYK

----- Original Message -----
From: "Jim Harman" j99harman@gmail.com
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Thursday, December 14, 2017 10:08 PM
Subject: Re: [time-nuts] accurate 60 hz reference chips/ckts

That’s not a problem, that means you’re phase locked instead of just frequency locked.

This is the phase-nuts mailing list, right? :-)

Tim N3QE

Jim,

One can determine the appropriate time constant by looking at the ADEV of the two clocks [1].

For GPS and OCXO the choice of time constant is fairly obvious. But the ADEV for mains frequency isn't quite as pretty. So I'm not sure several hours would work; maybe more like days or weeks? Here's a sample ADEV for power line frequency:

http://leapsecond.com/pic/mains-adev-mdev-gnuplot-g4.png

I'd be curious if anyone has tried a mains disciplined oscillator and how well it performed. Unlike a GPSDO, absolute phase can be lost (e.g., power fail) so maybe use FLL instead of PLL. Shall we call it MFDO (mains frequency disciplined oscillator)?

Hal and I have months, even years, or mains zero crossing timestamps so if someone wants to simulate how well a MFDO might work let us know. Maybe just adapt the gpsdo simulator [2].

/tvb

[1] some examples:
www.thinksrs.com/assets/instr/PRS10/PRS10diag2LG.gif
www.thinksrs.com/assets/instr/FS740/FS740_unlockedLG.gif
www.thinksrs.com/assets/instr/FS740/FS740_lockedLG.gif
www.leapsecond.com/pages/gpsdo

[2] gpsdo simulator project
http://www.leapsecond.com/pages/gpsdo-sim/

On Thu, December 14, 2017 8:39 pm, Tom Van Baak wrote:

It appears that the appropriate design would use a local oscillator which
is stable to better than 10^-7 at 5 years and approximately 1200 days time
constant.  I would recommend using very low leakage capacitors in your RC
filter.

--
Chris Caudle

Hi

My guess is that the “best design” would likely do a sample at some specific time of day.
Just when would depend a bit on your local grid and how it is fed and loaded. There are
definitely “it’s 6 PM and everybody just got home” issues with power line phase.

Next question wold be how long to run your sample. Is a few minutes good enough? Is
an hour overkill? There would be more than a little trial and error involved. You likely
get to some SNR floor long before the 10’s of minutes point, even at the “best time of day”.

If you accept that you are getting 1x10^-4 sort of data, once a day gets you into the 1x10^-9
range. That’s not an unreasonable stability for a well warmed up OCXO for a few days to a
few weeks. A sample a day may be quite adequate ….

Of course this assumes an electronic approach. Given that it’s moving pretty slow and you
only are looking at fractions of a millisecond, one could do an electro mechanical design …...

Bob

On Thu, Dec 14, 2017 at 10:53 PM, Bob kb8tq kb8tq@n1k.org wrote:

procedures here
https://en.wikipedia.org/wiki/Utility_frequency#Time_error_correction_(TEC)

and here
https://en.wikipedia.org/wiki/Electric_clock#Accuracy

--Jim Harman

I have one of my project boxes, which monitor the main freq. Here is
graph which reflect the time difference between of RTC (based on number
of pulses from OCXO) and the "MAIN TIME" which is based on number of
zero-cross events.
The observation period is 486 hours.

On 2017-12-14 23:13, Jim Harman wrote:

--
WBW,

V.P.

I'm surprised Vlad is seeing as much as six seconds differential but maybe
I don't understand the experiment. I've done measurements of the line
frequency here in California and never seen much variation.

Jeremy

On Fri, Dec 15, 2017 at 9:02 AM Vlad time@patoka.org wrote:

--
Sent from my iPad 4.