"Attempt 2", because Attempt 1 was finished last summer, and being exhibited in two German Makerfaires. But it was unreliable, so a bit of fudging was needed so it worked with the minut and hour hand at least. 

It tells the time

Once one tediuosly has set the hands and time, it does work.

Improvements done

I have now tweaked and redesigned the mechanisms that holds the other gears in place, while one gear is being moved.

The software can now set the clock chip, via the Serial interface. It can also set where the hands are now, (ie how much catchup there has to be after power loss) Incidentally, the Serial interface is already full of debugging commands, that allow advance by one tooth, move to next hour and such like. The position of the servos is defined in EEPROM tables, these are adjusted this way, too. These are all for rebuild/maintenance only.

Still outstanding

Basically get rid of the Serial hookup to "start" the clock. Two points:

Some easy way to set the hand position. I have an idea or two how to do this with a single optical gate and some "encoder-like" addition in the gear wheels. 

As an interim design a button or two will be added to manually position the hands at 12 o'clock after power outage (assuming the clock chip battery has kept time).

The button(s) will also set the chip time at some point (design details outstanding - the challenge is the "feedback" on what time you are setting)

Lastly, the weekday and day gears need to be exchanged for ones with integral toothcount to the step size to get all hands to be spot-on. (see earlier project notes)