The Keyless Works – Part 1

Today the first phase of the keyless work design was completed. The hardest part was designing an effective clutch that moved the castle wheel sufficient distances, whilst at the same time allowing adequate room for the winding pinion, having sufficient overall strength and occupying the smallest possible space.

The final result of todays work is shown below. Still to add are some gear teeth, springs, detent, click action and levers for the balance stop.

Below is the motion study, showing the clutch & lever action.

It still needs to be determined if the radial angular profile of the clutch teeth on castle wheel is too difficult to machine. Most watches just use a simple straight cut, however current thinking is that this angular cut allows the teeth to both mesh more easily when winding, and slip more easily when winding in the wrong direction. Perhaps more importantly, in this design the engaging surface area is greater, thus making a more reliable clutch less subject to wear. Radial profile clutch shown below.

render 3 radial small

In order to arrive at this lever design, several iterations took place, experimenting with shapes, pivot points and relative lengths. For example, and early iteration is shown below.

early version

Finally as a sanity check, the work so far was placed into the movement to ensure a fit (note that the large circular plate is shown only to ensure parts fit within the necessary boundaries of 42mm):

In Movement


Adding the Barrels & First Bridges

In the most recent iteration of the watch movement design, two bridges were added. These bridges will be most visible from the back of the watch. The watch will feature an  ‘exhibition back’, so it’s very important to get the aesthetics right.


The bridges were designed to follow the contours of the movement as well as accentuating key parts of the mechanism. Yet the bridges also have to be structurally strong to eliminate any flexing. Ensuring that the bridges are secured from both sides allows for a thinner bridge whilst retaining strength.

Additional bridges which will be added to the design later, hence the reason why some wheels are still floating in mid-air.

Also two barrels have been added, along with the linking pinion. The design of the barrels is not complete at this stage.

Below are some renderings of the movement so far, with the two barrels and two new bridges clearly visible.

movement2 movement1

A simple case was created to test the housing of the movement, this will be further improved later.


However there is still a lot more to do to complete the design process. Here is the current ToDo list (in order):

Winding Mechanism
Setting Mechanism
Castle and Keyless
Crown + Stem
Geneva Stop Work
Power Reserve Meter + Escape Stop
Cannon Pinions and Slip Pinion for setting
Balance cock, hairspring collet, adjuster, etc
Change Wheels to have 5 Arms
Final Bridges + positioning studs
Pivots + Jewels
Screws (+modify holes)
Dial + any dial screws
Modify case to fit movement + final touches
Asthetic Final Touches
Create Blue Prints and Design Book

Gear Train Modelling

Gear train calculations and modelling are made incredibly easy on a PC. When I read the old watchmaking books it’s clear how much easier we have things today with spreadsheets and a CAD packages. Using these can really save you a lot of time.

For my watch I added all the gear ratios into a spreadsheet, along with desired wheel tooth counts and wheel pitch diameter. The spreadsheet then calculates all the other numbers including:

  • Pinion Pitch Diameter
  • Module
  • Addendum (uses the correct formula depending upon if the gear should be cycloidal or involute in the case of the cannon pinion)
  • Timing of each wheel (starting with the assumption that the Center Wheel turns once per 12 hours)
    • Ensures the seconds wheel rotates once per minute and the Cannon Pinion once per hour)
  • The necessary balance BPH of the balance to support the desired speed of the Center Wheel
  • Center distance between each wheel and pinion
  • Various other useful calculations

Here are the calculations for my watch which I will explain more in a future blog post.

Modelling using CAD

Once these calculations have been made, the pitch diameter and center distances can be placed into a sketch in your CAD package, I’m using Fusion 360. Center distances should be set as constraints.

Now one can move around the wheels in the gear train, whilst satisfying all the constraints, as shown in the below video.

If you find yourself in an impossible situation to arrange your gears effectively, simply go back to your spreadsheet and try some new Pitch Diameters for your wheels, all the pinions sizes and center distances will be recalculated automatically within the sheet, then type these back into your CAD package.

Adding 3D Wheels and Pinions

Next you can connect 3D representations of your gears and pinions directly to the circle centers in the previous phase. This means you can still easily rearrange your gears later. To generate the cycloidal gears I use Dr. Rainer Hessmer’s excellent free program. I then import these into Fusion 360.

The end result would look something like this. Note I also inserted the lever escapement made in my previous post.

gear arrangement.png

Placing 3D models of Cycloidal Gears onto the 2D sketch

Note: I’ll show how to animate this arrangement in a future post.

Note: this design has a center seconds wheel (aka ‘sweeping seconds’) and an off-center center wheel. I’ll discuss the reasons in a future post.

During this process, having scaled what is on the screen to actual size, I realised that some parts would be just too hard to construct. So what you see above is the result of several iterations. This is another way CAD saves a lot of time!


As I plan to skeletonise the watch, the ascetics are very important. It is now possible to create accurate renderings and visualise the final design with ease. Below are renderings of the above gear train.


Further modifications can now be made as the creator sees fit.

Next the dial-side works, barrels and keyless works need to be modelled.


Escapement Design

Wanting to ensure the wristwatch has very accurate time keeping and yet is still reliable I felt the choice of escapement design was either a Daniels coaxial escapement or a modern lever escapement. Most other escapements were ruled out either because they are not self-starting, can become stuck if a shock to the watch is received or they do not have accurate enough time keeping.

The Swiss Lever escapement was chosen mainly due to there being more literature and support available as compared to the coaxial alternative. The Swiss Lever is also superior to the English Lever in several ways. Mark V. Headrick gives excellent discussion on optimum escapement design in his ebook, so this won’t be repeated here. As recommended by Headrick, I chose to model and simulate this escapement movement first to find an optimal design. Headrick does not explain the software or offer instruction on how to perform the simulation. I used Fusion 360 to model and simulate, finding the results to be perfectly acceptable.

The first design attempt (shown below) was a double roller. I used the same roller vs. escape wheel scale as illustrated in Headrick’s ebook. Although this worked fine in the model, the initial attempt showed the whole escapement was going to be too big for my wristwatch. Whilst I could simply reduce everything by a common factor, the escape wheel would become very difficult to manufacture.

Double Roller

Double Roller Escapement (this design was too big)

However simulation proved that the double roller in the ebook worked and the all the appropriate angles of draw, impulse and unlocking functioned as expected – perhaps another way of interpreting this is that my simulation environment worked as expected. As can be seen in the above image, the roller is almost as large as the escape wheel, and the palette stem is undesirably long.

At the same time as looking to reduce the size of the escapement, I decided to try modelling a single roller. In Watchmaking there is a chapter on escapements and Daniels talks about using a single roller with a guard pin. This design was favourable to me over the double roller as there are fewer components to manufacture. I read some of the theory about how to calculate various angles and lengths for the safety action, but in the end I decided to simply try a few different ideas within a simulation.

The final product, after reducing the size of the palette and using a single roller with a safety action, was the design below.

Single Roller

Final Design – Smaller, Single Roller Escapement

Here is the same design in different states of movement (bankings not shown):

Addenda were added to the teeth of the escape wheel as shown above in order to increase the total angle through which impulse is supplied. The simulation exactly matches the various stages of movement as suggested by Daniels  (shown below for convenience).

Daniels Lever Angle Components

Actions at various lever angles (note action is symmetrical for a total of 22 degrees)


The safety action was also tweaked to ensure no friction or binding would occur under normal operation. Scenarios were generated and tested to ensure it was impossible for the escapement to become stuck, for example if it received a sudden shock in any given position.

I did not attempt to reduce the size of the escapement further as the parts would become difficult to machine accurately. It may still be possible to reduce the size of the palette stem and roller, whilst keeping other parts the same size, but the results were quite satisfactory and will now easily fit into the movement being designed.

The final dimensions of the newly designed escapement are shown below. It remains to be seen how easy it will be to machine such small components to the tolerances required.


Dimensions of the Final Design (in mm)

Making a Workbench

I’ve never tried carpentry before – however having a good large standing workbench is essential for lathe work and other machining jobs. I figured that by pre-planning enough and applying a good amount of precision during the cutting and assembly, the results should be reasonable.

Laminate flooring was used as the work surface, which later turned out to be a good choice as it’s incredibly durable.

The bench is very stable, I put supports between the legs, but it really wasn’t needed and didn’t make noticeable difference.

I’m not sure on the terminology, but the wood which the legs are screwed onto is a little soft. I felt that further wobble could be eliminated by using a harder and thicker wood, however the results were perfectly acceptable.