[Catalin]

Assistant tool for calibration.....I made this, not pretty, but usable!:-!

Looks like a smart setup - I would guess a 20mm shirt button with 2 holes for the wires and one 'cut' for access to the calibration part ... :-!

 

[Catalin]

Unless you have a super camera the best accuracy you can get with the video method is 1/25=0.04s. Having said that it's good that you're not seeing any drift. When it's that close I generally run a couple of tests to see where the "tick" falls, it could be anywhere in that 0.04s range.

Actually more like 1/30s if the camera and the LCD are not that great, but if the LCD is bright and you have a very strong light on the watch itself you can go to the limit from the LCD, which is 1/60 - the camera still stays at 30 fps but since the exposure time is (much) shorter than 1/30 you tend to get only one LCD frame (or something like 80% of it and 20% of the next/previous).

The other thing is that you can also look after a number of frames and do some average, and with a very careful analysis and some care on internet sync you can sometimes get to around 0.005s accuracy (which I believe can be seen in the first 2-months results on my 8F33 (not my 8F56), where the standard deviation was suggesting an error in that range or better).

 

[Catalin]

The quartz literature all mention aging. But the aging stabilizes to a constant rate after certain period.
...

I believe that I have seen very often references that the 32 kHz aging becomes very small after a few years, but certainly on higher frequency high-accuracy OCXO the manufacturers provide very clear aging numbers for like 10+ years!