Idea - timing

[Nickleaton]

There is an online lighting detector network. It has one interesting feature. It uses a GPS in two ways.
First you get the location of the detector. Second it uses an interupt from the GPS that it will generate every second. From two interupts and a clock counter, it get get the number of computer clock ticks in a second. Millionths of a second is possible. Very accurately and that handles temperature, voltage differences over time. A GPS is of course highly stable over time.

It might be an idea to add this in to get a very accurate reference clock that doesn't depend on sound cards etc.

Just an idea.

[xyzzy42]

This is basically issue #24. Existing NTP clients can use a local GPS device to synchronize the computer's clock.

But a GPS isn't really necessary. With a network connection NTP can sync the rate better than a good quartz watch, much less any mechanical watch.

[Nickleaton]

Thanks. I was thinking about building a pendulum clock, similar to the Shortt-Synchronome.
It's easy to get an arduino to generate the external signal to the external clock.

However I was looking at doing some analysis with an external timegrapher. From having a fixed width obstacle generate a pulse, you can get a speed of the pendulum measurement. If you sync an arduino to a GPS, with a 1 second interrupt, and count the clock cycles in 1 second that is going to vary, most likely as a result of temperature changes. The syncing means you get temperature compensation of the measurement. It's this part that generated the question

The second is with a triangular shaped attachment to the pendulum, if you measure the pulse width it also tells you if the pendulum is longer and hence been affected by temperature.

Now the Shortt clock is 1/6 s per year! That is sufficiently accurate that you get variations caused by the sun and the moon changing g, and hence the period of the pendulum.