I love mechanical watches and clocks. I collect the former, and
accumulate the latter.
One problem with basic spring-powered, pendulum-regulated
is that they are not isochronous: they run faster when the spring is
newly-wound-up than when it's almost run down. This is because
escape wheel (which is driven, indirectly, by the spring) gives the
pendulum a tiny nudge via the escapement, at each swing, to keep
going. This tiny nudge is enough to make the pendulum swing a
little faster than its "natural" rate. The nudge is stronger when
clock is first wound, and weaker as it runs down, so the clock runs
faster and then slower by a very small amount. As an example, I
have an old station clock with an 8-day movement which gains
5 min in the three days after its wound, and then loses about 8min
over the next 4 days. Most of the gain is in the first day, and I've
set the subsequent loss to offset this, so the clock is never more
than 5min out.
Now, this 8-day movement will run the clock for 8 days or more
(hence its name), but the clock will run most accurately if it is
wound up less than fully at the start of the week, and then run for
only 7 days before winding again. In this way, you avoid the "very
fast" bit when the clock
has been wound up as far as possible, and also the "very slow" bit
the 8th day when it is almost unwound.
The problem is, though, that it's very difficult to wind the clock up
just enough to keep it going for 7 days. The only way you can tell
when it's wound is when the spring "goes tight" (ie, the coils start
touch eachother), and by this point you've given it the full 8 days
winding, and it will run very fast for the first day.
You could, of course, count the number of turns of the winding
and find that a full winding was 16 turns of the key, so 14 turns
would be ideal. However, if you're slightly out (eg, if it unwinds
13.5 turns in a week), then you'll 'accumulate' 0.5 turns per week,
and you'll soon be overwinding the clock.
A modest proposal.
The winding key should, in fact, be a miniature torque-wrench. In
this way, you'd be able to tell when the mainspring was wound
sufficiently to keep the clock running for 7 days, but not
to the point of over-driving the clock on the first day.
That's it, really.
You can stop reading now, because I am about to stop writing.
And don't talk to me about bloody fusées, that's all I can say.