Clock #3 at least runs!

Clock #3 has been plagued by various problems, mostly with the resonator/balance not being really suitable.  One of the things that was exacerbating matters was the passing spring on the detent was still (even after modifications) too stiff.  I went back to various books and found that the passing spring tended to be quite long, passing beyond the locking pallet.  So I modified the spring in that way.  Now it's much lighter.

I thought long and hard about the balance issues, and came to the conclusion that I probably had to face a right angle turn and a torsion balance.  You can see the new right-angle transmission wheel above on the impulse roller assembly.  The other half mounts to the balance through a spring, which consists of a spring, a rod, pin, wheel, and balance (at the bottom of the rod).  Here are the pieces...

... and here it is assembled.

The balance hangs in front of the lower section of the clock frame.  The rod passes through a pivot, which is formed from a steel wire.

This is perhaps not the best option, but it has the merit of making it easy to adjust the depth of the right-angle transmission.  The transmission runs fairly smoothly and noiselessly.

To first approximation, the new balance appears to have a resonant Q of around 12, whether loaded or not.  (Eyeballing 3 periods before the balance amplitude halves, then multiply by 4 according the rule of thumb suggested by Woodward.)  So the pivot and transmission is the least of my concerns, but the fact that the unloaded Q is low is definitely an issue.

The mechanism appears to run semi-reliably...

but it needs considerably more torque than I'd like.  The center wheel needs about 4.125 inch-pounds in order to maintain balance amplitude.  This is concerning because I was really hoping this to be the torque on the drive wheel -- a factor of 10 weaker!  The resonator having so much absolute loss of power is clearly a problem.

Lower on the priority list is the fact that with this balance the clock runs a bit fast.   It makes 17 "loud" ticks per minute rather than 15.  That should be easily corrected.