An attemt at a Swiss Lever Escapement

[BillM]

_SwissLeverExperiment.zip

(859.35 KiB) Downloaded 127 times

Art

I've been away from Ticker.exe for a while. I'm running v2.05B of the program. Is that still the current one?

The attached zip file _SwissLeverExperiment contains an attempt at a lever escapement. The STLs used to create the model were developed by Jaques Favre and were downloaded from:

https://www.myminifactory.com/object/3d ... ent-201013 The STLs were used to create assets re-scaled to 1/4 their original size.

I've had problems understanding the various parameters associated with springs. I need to control a chronometer model's verge-folliet escapement with a balance wheel/spring. I figured that starting with a known lever escapement I'd be able to better understand how to use springs in the Ticker program.

The model in the attached zip file runs but not as smoothly as I'd like. I'll have to experiment with the many settings to see if I can get the model to run smoother.

Once I better understand the nuances of springs then I'll start again on my chronometer model.

BillM

[ArtF]

Hi Bill:

I've downloaded your file and will analyse it to see what's up. I haven't changed the program at all for months as I'm in my summer coding phase,
(Which means I code to figure out what I find interesting to work on to enhance Ticker..).
For the last few months that's a rabbit hole called "Signed Distance field based Toolpath Simulation". Its a very interesting way to create objects and shapes from math functions, for example a 3d sphere can be defined as the simple formula P = |P| - r , so that a toolpath simulator based on them has extremely high resolution as a function has virtually infinite resolution. As a result of course I had to write toolpath functions to test it with, pocketing, contouring and such , and then of course, one needs V-Bit carving toolpaths for letters around the clocks face..and of course a tool table.. you see what I mean by rabbit holes.. I get fascinated by something and it leads to more base required to play in that area..
I have done many fixeds on the physics along the way so the next version should be more capable and solid. Zooming is now mouse position sensitive for example so you zoom to the mouse, not to center. These fixes will be in the next release..but not until I climb out of this rabbit warren..

Art

[ArtF]

Bill:

Preliminary look seems to indicate a large problem in the top pawl. If you turn on wireframe and slow the simulation, you will see the
pawls collision hull, is too tightly constrained by the slot. They overlap much of the time or are tight together, this creates singularities as
friction becomes a central component of their relationship. (Which isn't handled well internally). In my escapement you'll notice I used a much
thinner pawl as its best to rely on the reaction of a collision that has room to bounce back ( Also set the restitution to zero on such items so
they dont bounce as much, they may need more mass as well to damped the vibrations of the collisions. Physics simulations such as Bullet used in Ticker work on actually colliding not on their face, but the objects actually go into each other and then bounce back based on mass and
restitution for the final placement. I guess thats why clock escapement is a real challenge to physically model as their a very precise mechanism
in reality working on many physical principles. (Many of which Ticker is good at , some of which one much fudge alot on. :)


Art

[BillM]

Art

I know what you mean about rabbit holes: Our engineer son mentioned that he's been thinking about taking apart his Tag Heuer watch. That led me to watch a bunch of watch repair tutorials. During my searches I ran across some very educational, entertaining Google Android apps: Watchmaker Pro, Watchmaker 2 pro & Watchmaker 3. All of the apps modeled watches with Swiss Lever escapements. My search about Swiss Lever Escapements led me to the STLFinder website where I found the "printable" STLs created by Jaques Favre. The size / thickness of the parts was probably due to requiring strength because the escapement was supposed to be 3D printed.

Jaques Favre's model is quite large so when I created Assets I used a 0.25 scale factor. The day after I posted the model, I realized that I probably had the balance too close to the lever that holds the pallets. The position I used defeated the purpose of the Swiss Lever Escapement that's supposed to allow balance wheel to be free to oscillate at some frequency. When I moved the wheel a bit further away from lever the balance does move out of the lever but I haven't been able to get the right balance of forces, frequencies, etc. to get the movement to tick.

I've got a lot of experimenting to do...I never give up.

I agree that the lever & pin on the balance is way to beefy. I'm going to go back and redesign things.

BillM

[BillM]

Art

The link to the model is simply a project file named A4. On my computer, it seems to tick fairly consistently as long as gravity is set to 980m/s.

I've rebuilt the escapement model and experimented with several parameters: masses, restitution, angular damping.
For example:
The balance now has mass 1.0, 0 angular damping, restitution of .9.
grounded spring (60Hz, 0.1 damping & 0 angle)

https://drive.google.com/file/d/12s-inv ... sp=sharing

The position/ spacing of each component is critical to the operation. If I set gravity much greater than 980m/w the model grows rapidly unstable. When things go unstable the display of wire frame becomes out-of-sync with the image of the Balance wheel.

I'm trying to make sense of the parameters. At one time you mentioned that any rpm that gets set is 10 times less than set. Does that also hold for the value of springs? Is 60Hz really 60Hz or does it run at 6Hz? If running at 6Hz this would be equivalent to 21,600 beats per hour.

I still may play around with the parameters but this time I'm happy that the escapement runs. I left it running for ~ 20minutes.

BillM

[ArtF]

Bill:

Amazingly good. I found I had to shrink 3 shafts, likely a bug on my end.. and then when I resaved and ran it, it ran perfectly.
Spring freq appears good at 60, and I think that's 60 per minute really, not hertz, Ill have to tag that for review. In some cases the
units are a bit arbitrary I think internally. But the way this ticks is truly a show of a well simulated escapement.

I note that to shrink the shafts, I had to take the minimum overrides set at 1.5 and make them -3 to allow it to run. Ill tag a
look at the saving of shaft offsets as well to see why their not sticky.

Well Done, nice Sim..

Art

[BillM]

Art

I can't take credit for the actual model. I figured if I could take a working 3D printed model then I had a better chance at getting a model running under Ticker rather than designing one from scratch.

My first attempt was from STLs created by Favre and accessible through the stlfinder.com site.

This second model was derived from STLs created by Grain and accessible at

https://drive.google.com/drive/folders/ ... QslBqhhdwh

I looked at Grain's video of the completed model (https://www.youtube.com/watch?v=x9wBv-9FwJI) and realized the importance of the banking pins. Grain's STL files are significantly larger than those from Favre. The Favre & Grain models did not self-start and required a tug on the balance to start the printed model. Within Ticker, "tugging" (Sim snap on) the balance was too unpredictable to start the Ticker version of the mode so I chose to use the escape wheel to start the Ticker version of the model.

I used the Favre bracket for the initial spacing the balance, pallet fork, escape wheel and banking pins. I guess due to the way hulls are created I had to tweak the positions of the escape wheel & banking pins.

I noticed several times that when I re-loaded the model that the Minimum shaft override for the link's shafts had to be readjusted to enable the shafts (banking pins) to interfere with the pallet fork. My solution was to move the link closer to the pallet fork.

Masses are another difference between the real world and the Ticker model. I had to make the mass of the pallet fork and escape wheel significantly smaller than the Balance wheel.

Tweaking the restitution values and angular damping values was an exercise of trial and error.

I had hoped that I'd get the model to run with a true vale for gravity of 9800 mm/sec^2. Perhaps my computer is simply not up to the task.

BillM

[ArtF]

Bill:

Its likely my code that isn't up to the task. The normal units for the simulators base code is kg and meters. As I'm using mm's and grams many of the assumptions in the code for friction and such I haven't managed to track down. In reality the engine is mainly used in tweak and make work type of settings. Bullet is popular in video games and such for the way one can tweak realism to look as if its reality. I'm quite impressed really that such a simulation actually ticks away.. there's not much that's as sensitive as an escapement. Ticker has bullet very tightly integrated into the models themselves so the behavior overall will always take some tweaking to work as everything affects everything else I've found and unrealistic settings of mass and restitution are the norm..

I will find that pesky shaft length override bug, I think its related to the settings options for shaft length extension when placing a model.

Art