[TC28] Desk clock

[Davidz90]

Rather unsurprisingly, I decided to build a clock. However, to make things more interesting, it will not be regulated by a pendulum. Instead, a balance wheel like in a wristwatch wil be used. This poses a lot of interesting technical problems:

-Pendulum can be made very efficient; much harder to do this with balance wheel without high quality bearings (like jewel bearings used in wristwatches)

-Pendulum is very accurate - just setting the proper length takes care of 99% of accuracy. Balance wheel, on the other hand, is very sensitive to driving force and spring characteristic.

The use of balance wheel should allow me to build a small, desk clock powered by pullback motor. However, this source of power makes things even more difficult due to limited energy and uneven spring force. Neverthless, my goals are:

1. Make it work for at least half an hour on a pullback motor (so it can be called a clock, not "egg timer" XD)

2. Make it reasonably accurate (I aim for under 1% error; trivial with pendulum, but real challenge here)

Right now, I'm testing various escapement options:

 The second design works for 5 minutes on a full rewind, 30 minutes seems possible.

[msk6003]

I tried use this part as balance spring before but failed. That hose idea looks good.

[Davidz90]

1 minute ago, msk6003 said:

I tried use this part as balance spring before but failed. That hose idea looks good.

Yes, I tried this as well! It sort of worked in this arrangement:

20241223_165650

but it's definitely inferior to the hose.

[Zerobricks]

Some brilliant ideas right there, can't wait to see what you come up with next.

[2GodBDGlory]

Wow, very cool! I'm not familiar with clock mechanisms like that, and these are very neat to watch!

[Davidz90]

9 minutes ago, 2GodBDGlory said:

Wow, very cool! I'm not familiar with clock mechanisms like that, and these are very neat to watch!

Thanks. Indeed, wristwatch mechanisms are certainly a sight to behold. Here's an example how a high-quality large scale model of one operates. Much smoother and wider swing arc than in my attempts.

To think that they date back to XVIII century...

38 minutes ago, Zerobricks said:

Some brilliant ideas right there, can't wait to see what you come up with next.

Thanks! First things first, I plan to do some measurements (and maybe computer simulations as well) to choose the best escapement type to mate with balance wheel. Eventually, I want to write a big chapter about wristwatch mechanisms for 2nd edition of my book.

[LegoTT]

WOW dude you are ALSO making a clock ?! I'm devasted now...

Not serious, I hope it will be as good as possible

[Davidz90]

Second day of trying various stuff. Managed to make it run with 40t gear instead of propeller piece, but at much smaller amplitude, faster speed and with more torque needed, so there is no big gain in efficiency.

 

[Davidz90]

I got some simulations up and running. The chronometer escapement (wheel receives a short push when passing through neutral point) is the way to go, everything else works rather poorly with high friction and is better suited for a pendulum. Wristwatch manufacturers know what they are doing, what a surprise 

so back to the drawing board.

[Davidz90]

A little more progress. The heat-formed hose proved to have poor long-term stability. It still keeps the shape, but became more stiff and seems to have more damping. Thus, a switch to the whip piece:

 

[Davidz90]

More progress today. I managed to implement a proper chronometer escapement, where the wheel receives a push when passing through middle point, in one direction. The difference from Galileo escapement may seem superfluous at first, but in fact it results in considerably more harmonic, slower motion that doesn't depend on the driving torque. Exactly what is needed for spring-driven clock.

 

[gyenesvi]

Really interesting to see those mechanisms at work and evolve!

Can you explain in short the role of the balance wheel? What I don't see is what's the role of the whip piece as opposed to the pullback motor? If I get it right, the pullback motor gives the main power to make it move for a while.

I do have an old wall clock that works with a pendulum and two weights, but it also has a spiral spring inside. If I understand correctly one of the weights serves as the main driver, the other is for making it chime every hour. So I don't know the purpose of the spiral spring, but I guess it's the same as the whip piece here?

Edited January 8 by gyenesvi

[Davidz90]

10 hours ago, gyenesvi said:

Can you explain in short the role of the balance wheel? What I don't see is what's the role of the whip piece as opposed to the pullback motor? If I get it right, the pullback motor gives the main power to make it move for a while.

I do have an old wall clock that works with a pendulum and two weights, but it also has a spiral spring inside. If I understand correctly one of the weights serves as the main driver, the other is for making it chime every hour. So I don't know the purpose of the spiral spring, but I guess it's the same as the whip piece here?

Sure! The role of the balance wheel is the same as the pendulum - it oscillates at some frequency, regulating the speed of the clock. The whip piece acts as a very weak spring, which is necessary to make it oscillate back and forth. Frequency is proportional to the stiffness of the spring and inversely proportional to the inertia of the wheel. Speed can be regulated by altering the inertia - in wristwatches this is the job of a set of screws on the wheel's rim. Pullback motor, as You guessed, is the main source of power that sustains the oscillations. The job of the escapement is to deliver the pullback motor's energy to the wheel in a way that doesn't disturb the oscillations too much, so that the speed is not too torque-dependent. Otherwise the clock would run slower and slower as the pullback unwinds.

Yes, in wall clocks usually one weight acts as a driver and the second one powers the chime. Not sure what the spring may be for in this arrangement. One option is so-called maintaining power - it acts as a interim power source when the weight is rewinded. It's also possible that the clock is purely spring-powered and weights are just for show. 

[Davidz90]

Another small update, experimenting with new, more compact mechanism.

 

[Berthil]

Great work so far. Looks like the whip will provide the consistency/torque needed and can provide the 1 second timing for an accurate clock, at least as much possible with a 100% LEGO clock .

[Davidz90]

Thanks! Yes, the whip is surprisingly consistent, it shows no signs of wear after a few days of testing various escapements. Time will tell how accurate the timing can be, but I don't expect miracles; my plan is to do rough tuning by tweaking the inertia of the wheel and fine tuning by slightly twisting the whip.

[Davidz90]

Yet another small update. I came up with a more robust design where all parts are mechanically linked, so it no longer depends on separate moving pieces hitting each other the right way.

 

[ord]

Fascinating and creative use of Lego. Following closely .

[Davidz90]

The new escapement design is reliable and efficient enough, runs for about 10 minutes with a single pullback motor and 1:25 gear ratio. Now let's see if it is stable enough. Thanks to loud, clean ticking sound, I can use sound recordings for measurements (incidentally, that is how technicians troubleshoot real wristwatches). They look like this:

audio1

By measuring the time between ticks (more specifically, the last, highest peak), one gets the period:

periods_1

Right now, it is slightly below 2.2 s; I'll either reduce the inertia of the wheel a little to aim for 2 s, or increase it to go for 2.5 s. At any rate, it seems rather stable, with +-0.05 s variation. A histogram confirms this:

periods_2

Finally, assuming for a moment that the mean period of 2.19 s is the ideal value, here is the error:

periods_3

One can see that the clock is never more than 0.35 seconds off during the 7 minute test run, despite the fact that pullback motor torque decreases by over 50% during that time; an impossible feat without chronometer escapement.

[aeh5040]

Wow, impressive analysis and results! Curious how the total error graph kind of looks like two line segments rather than, say, a parabola. Could be just my mental over-fitting though!

Edited January 12 by aeh5040

[Davidz90]

23 minutes ago, aeh5040 said:

Wow, impressive analysis and results! Curious how the total error graph kind of looks like two line segments rather than, say, a parabola. Could be just my mental over-fitting though!

You are right that two line segments is a somewhat surprising shape, and a parabola would be the more expected outcome; at least assuming that the period depends somewhat linearly on torque. Then again, torque characteristic of a pullback motor is not that linear either. It seems that instead of changing continously, the torque jumps between values (which corresponds to a jump in the slope of the error, giving that segmented look). To really properly analyze the torque dependence, I'd need to switch to gravity power and do multiple runs with various driving weights.