Clock or anemometer?


I don't think this is your page.  What is yous, I'd like to have a  look at the code / thanks!

On 4/5/2020 7:03 AM, Bepi wrote:
Hi Mark, I am replying privately because I don't think this subject would interest everybody.
I don't think there is anything missing in the quoted lines and the program worked as it is when I tested it but it's not my work and I am not using it for the data you have seen, I am presently not using an arduino timer but one which is equipped with a TCXO clock.
Did you try the elementary code, also published on the web page, first? The 4 μs resolution one. The one you quote a few lines of is the better than 100 ns resolution, normally not needed, certainly not to test the system for the first time.
I should review my web page, I wrote it as a reminder for myself and it's not well commented  but I am available for all sort of discussions, in particular during these lockdown days.
By the way were are you located? If you don't see it on this mail my real address is pepicima (at) and I am available on skype at pepicima.
Take care, p.-

On Apr 5, 2020, at 09:52, markotime <markotime@...> wrote:

In your code the following is confusing:

      TCCR1B = _BV(CS10) | // start Timer 1, no prescaler

                          _BV(ICES1); // Input Capture Edge Select (1=Rising, 0=Falling)

is there a term missing? thanks /mark



I had a quick look at the, relatively speaking, HF power spectrum of the period fluctuations vs frequency in Hz for two 136 min intervals around a windy period, top graph h 70, and a calm period, bottom graph h 125, for the same data shown at my first post above. 

No surprises, broad and HF, the way it looks by eye.
The .1 Hz spike visible in the low noise case (bottom graph) corresponds to the .1Hz impulsing frequency. It can be bearably recognized in the time domain, below is a sample of raw period measurements on a 30 μs scale from around h 125, always the same session.

With magnetic impulsing one can do better at reducing the effect of impulsing on period but it's a slow adjustment process and I don't predict any performance improvements at reducing even more the escapement error.



During the last couple of months I left the clock alone, gathering data undisturbed. The largest disturbance by standard deviation in my case comes from the wind, presumably through the building wall vibrations. Something not everybody knows is that foundations, all foundations, in the old city of Venice consist in marble blocs, at tide level, resting on a bed of 10 to 20 m long pine piles driven in the lagoon soft mud down to the clay layer. Remarkably nothing moves much, the submerged wood doesn't rot, and very heavy buildings, like mine, have been sitting there undisturbed for many centuries. What does it mean for a clock regularity? I don't know but certainly doesn't prevent the wind from inducing vibrations in the building structure. Canal waves influence the period too but, according to my anecdotal experience, are less of an influence.

The following picture shows period against time in hours for the last almost two months, every dot represents the average over the 5 period impulsing cycle.

The city anemometer in m/s for the same period is shown in the following image:


I found another generic observation of wind noise by Robert Belleville, HSN 2006-3-11, thanks Bob for pointing this out to me:

Even for the long recording shown above there are not many extended periods where the wind is below 4 m/s, a 4 days one ( apr 9 to 12 ) has been analyzed for temperature and pressure fluctuations.

The raw period data (integrated ove the 5 period cycle) is shown here, the overall standard deviation is around 2 μs (10 s average)

The comparison with temperature and pressure is in the following picture. Data have been averaged over 10 minutes in this case.

The period data show an obvious hourly correlation (same data, same scales as above, binned by hour of the day) but not coherent with the temperature variations:

I am not particularly interested in the pressure/temperature effects, i am still trying to make sense of the intrinsic clock noise, I am only showing what I have seen for anybody to compare. Since I am otherwise rather busy right now i just collect long term data for potential future use. The pendulum is the one of a synchronome in its original case, the impulsing is magnetic, cycling every 5 periods instead of 15, the bob is a 7.5 kg lead one (lead might have to do with some temperature response but I never checked this hypothesis).