Please see theClock3 for a more functional, and better documented electromagnetic wooden clock for makers.
This project is being documented posthumously in order to preserve the lineage of my clock oriented projects so it does contain all the fusion360 files, DXF files, and other results of the project in case you are interested.
Around November of 2022, I had completed the 20 mm CNC machine project and was using that machine to create various wooden boxes and other projects. I was having fun with the laser cutting capabilities of the machine and decided I wanted to try something with wooden gears in it. After a few weeks of mulling the kinds of things I could build with wooden gears ... novelty boxes, or maybe wooden robots, cars, etc ... I hit on the notion of making a wooden clock.
So I began obsessing about clocks in general and wooden clocks in particular. One prolific online source of information is Brian Law and his many clocks, webpages and designs. Although not wooden, I quickly found Steve Peterson's series of 3D printed clocks interesting. I explored dozens of existing designs, from the most artistic, to the most rudimentary simple escapement mechanisms.
My diode laser cutter is limited to cutting about 1/4" (or less) wood.
As it turned out, I could not find a single existing design that I could make using the tools and materials I had available to me. My goal was to laser cut most of the pieces from thin, cheap, available plywood and possibly 3D print a few other parts. This is not to dis any of the many great, and complete, and sometimes free designs out there. But, I actually live in an apartment, don't have access to a table saw or similar typical woodworking tools. Most of the designs for clocks that I found online involved thicker woods, typically hardwoods, and the use of the usual woodworking tools.
Although you can find a number of clock making kits that use laser cut plywood for sale online, I could not find any free online designs for such clocks. Finally, with regards to the existing laser pre-cut kits that are available, even for pay, none of them captured what I was trying to achieve. Although beautiful and complex, the available kits are more or less uniformly intended as snap-together novelty clocks and not really designed to be permanently constructed and act as actual time pieces.
So, I determined I would just learn what was necessary and design my clock from scratch.
Probably the first requirement that I arrived at was that I wanted the clock to be electromagnetically driven. The real study and discipline of Horology, and clocks as we know them, centrally involves the evolution of gravity escapement mechanisms. But I did not want to build a clock that had to be rewound (or it's weight lifted) every day or week. I wanted something I could put on a shelf, and forget, and that would keep time fairly reliably ... that is that it would function as an actual clock in my living space.
So I elected early to have an electromagnetically driven pendulum, while maintaining the idea that it would have wooden, laser cut gears.
I thought it would be so much easier than it turned out to be! In fact, my original notion was to build a chaotic clock that, despite being chaotic, still kept accurate time, an idea I may yet revisit. The idea was to build a dual pendulum clock that could be intentionally driven into a chaotic motion, but that could be recovered to periodicity at small amplitudes, and then to have the clock randomly intersperse chaotic movements of the pendulum with regular time-keeping movements to correct and adjust for the chaos. It was to be called The Perception of Time, alluding to the fact that humans just do not experience time in a linear manner. Also, in my dreams, it would be artistic and/or have crazy gear trains like squares, triangles, or helices.
In any case, that remains a project for the possible future as I decided that, since I only wanted to spend a few weeks on the project, I had just better build a simple clock.
I had a lot to learn
As with all of my projects, I started off by listing a few requirements for the project.
- It would largely be made of thin laser cut plywood (as described above)
- It would use an electromagnetically driven pendulum
- It would have concentric hour, minute, and second hands
- It would keep reasonably accurate time, the definition of which was yet to be determined.
- It would not look like, or be constructed like a snap together kit.
- It would look like a permanent construction .. a real clock, yet ....
- It would be built in such a way that it could be disassembled and repaired or modified if necessary
- It would make use of an MPU (Microprocessor) like an Arduino or an ESP32
- It would have a custom made PCB (Printed Circuit Board)
- The PCB and electromagnetic coils would be hidden, unobtrusively powering the clock
- The magnet on the pendulum would pass between two coils rather than over one coil (discussed below)
The requirement about being able to be disassembled and and repaired or modified warrants a brief discussion.
Especially since I had no idea if the clock would even work, or what refinements might be needed, I decided to design the clock in such a way that, as much as practicable, I would be able to disassemble and re-assemble it multiple times in case I needed to modify the mechanical aspects of the clock along the way. Also, not having any idea about the lifespan of such a clock, it was important to me to be able to, at some point in the future, disassemble and repair and/or rebuild the clock in case the gears or parts of the wooden mechanism wear out.
Given a general idea of the requirements of the project, I then proceeded into a research phase to learn as much as possible about these kinds of clocks before constructing one. I downloaded and analyzed virtually any clock design that I could find, even if they did not match my construction criteria, just to get an idea of how other people had approached building wooden (or any) clocks. This included a number of complete Fusion 360 designs, but also a number of available (free) online books about the subject.
It was a wonderful journey! I spent days, and sometimes all night, reading histories and perspectives of clock making, which turn out to be pretty fundamental to science, and the development of civilization as we know it. The research phase broadened out into a review, as it were, of pretty much the entire history of science lol.
It led me to re-touch my basic understandings of physics and mathematics, going back through Newton and Calculus 101, but really took me on a journey through most of human history. From sun dials, to the Aztecs and their calendars, Pythagoras and the Greeks, I was all over the place reading about the history of how humans came up with the notion of time ... and measuring it.
Galileo famously notices the period of a pendulum is nearly independent of it's amplitude ... that is that a pendulum swings at about the same rate regardless of how wide it swings. Christiaan Huygens creates a gravity escapement mechanism to produce the world's most accurate clock of the 1600's ... accurate to a few minutes a day.
One of my favorite reads was learning more about Robert Hooke, the creation of the Royal Society, and Hooke's rivalry with Issac Newton. Did you know that Robert Hook was the first human to ever create a tone of a known frequency? LOL, he created a gear mechanism with a paper card, and by knowing the gear ratios and counting off the seconds with a clock, he could ascertain the frequency of the resulting tone.
It was approximately a G# lol, on his piano, in case you're wondering, although the tuning fork would not be invented for another 150 years, and the semi world-wide standardardization to A440 tuning until the middle 1900's - after WW1.
In any case, or rather, of course, that led me through the whole Age of Enlightment before I could relish re-learning about the greats in the discovery and understanding of Electromagnetism. Volta! Gauss! Ampere! Lorenz! Michael effing Faraday!
And then came Maxwell
All of sudden we're talking about the speed of light !!!
Another fun fact: Did you know that Michelson and Morely used a pendulum in their groundbreaking attempts to measure the speed of light? Yup. They used one to determine the local gravity in Pasadena, CA, which was part of their calculations of the speed of light.
In any case, all of this stuff was very interesting, and I got to re-learn a lot, and learn a lot of new stuff as I started designing silly wooden clocks!
I thought it would be a good idea to just build a pendulum and see if I could get it to swing with an electromagnet. I 3D printed a little white pendulum that was about 5 inches high and using some coils removed from a couple of of micro-solenoids I had laying around. I created a circuit with an Arduino and a motor controller module to send pulses of electricity to the coils.
That worked in a few hours. The pendulum had a magnet in it, and could be made to swing back and forth (fairly rapidly).
Then I designed and built a bigger pendulum, learning in the process that I'm not so bright! LOL.
I thought I would make a linear motor (I was still clinging to the chaotic pendulum idea), so that I could swing the pendulum really fast, or really slow, or even stop it in the middle of a swing. Since magnets are cheap, and coils are expensive (and fairly hard to make), and since most linear motors move the coil through stationary magnetic fields, that was my initial stab.
I designed the Pendulum in Fusion 360. I salvaged the magnet wire from the transformer of an old wall wart and hand wound the coils. I laser cut the pendulum from wood and 3D printed the plastic framework to go on some 2020 aluminum extrusions. I put two coils on the pendulum and had it swing through the fields of a series of magnets, and extended the bread boarded Arduino control circuit, thinking I could then simply control it's position in real time as one would with a linear motor.
Nope.
I got it moving, but it turns out that (a) the pendulum wants to swing at a particular rate; it did not want to move at arbitrary frequencies, (b) you basically don't want a bunch of wires and junk on the pendulum, as that decreases it's accuracy, and, perhaps more importantly, in my configuration (c) the electromagnetic coils were not nearly strong enough to stop and hold the pendulum at any given point, (d) this linear motor was at best, like a stepper motor, and the motion was jerky and not smooth.
Before moving on, I removed the magnets, one at a time, in a series of experiments, until I ended up with a single magnet and the two coils on the pendulum that I could sort of control. At that point, though I could get the pendulum to swing fairly reliably, I could not really control it's period.
Another thing I did experiment with here was using the Arduino to not only pulse the electromagnet to attract or repulse the magnets and move the pendulum, but was also to use the coil as a sensing device. Since a coil moving through a magnetic field creates a current, I could measure that current with the computer and tell, for example, when the coil passed a magnet.
Most single coil electromagnetic clocks use that current sensing capability (and a simple circuit) to detect a zero crossing for the timing for delivering the repulsive impulse to keep the pendulum moving.
After a few days I decided to just go ahead and start designing the gears and things for a wooden clock.
Designing the gears and mechanics in Fusion 360 was not difficult, but was time consuming.
Somehow I landed on the notion that my clock would have four coils (two about the center of the swing and two at the extremes) -- which are really eight coils since, in my design, the magnet moves between the coils.
Another side note: I could find absolutely no scientific studies or analyses about the effect of a magnet moving between two coils. There were plenty of scholarly articles on solenoids, coils, magnets moving near a coil, etc, but nowhere could I find an analysis of the math involved when a magnet passes between two coils.
I wanted my magnet to pass between two coils, rather than over a single coil, as most magnetic clocks do, because I felt that the vector of force from a coil underneath a magnet would tend to make the pendulum bump as the impulse is applied. The force would not be orthogonal to the pendulum's axis of rotation. By having the magnet move between the two coils, the majority of the force would be put into moving the pendulum along it's line of travel, rather than up or down (assuming the pulse is applied shortly after the zero crossing) and that seemed more efficient to me.
I was also not thrilled with the technique of using the coils as sensors. I did experiment with it extensively, and it works, but it is onerous in the circuit and code to change the Arduino's pins from input to output to be able sense milliamp currents one millisecond, and then deliver 1/2 amp or more to the same coil a few milliseconds later and the results were not that great.
Because I knew I would want to know where the pendulum was in it's swing (particularly zero crossing), and I did not like the coil sensing approach, in this initial design, because I had a few on hand, I also added five hall sensors, which are electrical components that can tell when a magnet is near them, bracketting the four (pairs) coils with the idea that I could use the information, particularly the zero crossing, but also the extent of the swings, in controlling the clock..
Mind you, I still don't have any idea if, or how this thing is gonna work.
But in building the test rig above, I finally determined that I needed some kind of an actual full clock to even test these concepts. The friction from the gears and the work that needs to be done, the ratchet mechanism, the length and weight of the pendulum, all play a role in how fast the clock ticks and how much force is needed to move the pendulum.
So I just designed and built a damned clock lol, pressing through the design even if I was not sure about if, or how, it would work.
An actual full implementation allowed me to test the simple mechanics, gears, gear ratios, the ratchet mechanism, and so on, all of which worked pretty good, more or less, on the fist build, right out of the design gate, as it were.
I also made a coil winding machine, as hand winding these coils is really tedious and inexact, and I had to make eight of them!!
I started off using an Arduino on a breadboard as the computer, but quickly decided that I would rather have a more powerful and sophisticated ESP32, not only because it's speed would allow me a lot more flexibility in the implementation, but also because the ESP32 includes WiFi and a built in RTC (real time clock), whereas the Arduino doesn't. The RTC would come into play to keep the clock accurate, and with the WiFi I would be able to access NTP (Network Time Protocol) as well as provide a WebUI for remote control, monitoring, and debugging of the clock.
So I also designed a circuit for an ESP32 and built the first custom PCB for the clock. With 4 (pairs of) coils, sigh, I used two L293D H-bridge motor drivers so that I could control each (pair of) coils to either repulse or attract the pendulum magnet under MPU control.
I got it going ... it ticked and tocked .. and the second hand moved, but the clock was not even close to keeping good time
I called this version 1.0.
Without going into too much gruesome detail, I quickly learned that having so many coils was a headache and did not really help me to control the pendulum and it's timing. What I needed was a single, bigger, and more powerful coil, so I retrofitted the clock with a single bigger coil.
I also ended up embedding the coils in epoxy so that I could eliminate the plastic (on the bobbin) between the coil and the magnet to reduce the distance and increase the force. Even 1mm of distance less increased the force applied to the pendulum significantly.
As I said, I thought it would be easier. With my 4 coil design I had been planning to control the frequency of the pendulum by using the outer pair of coils to either attract, or more likely repulse, the pendulum on each swing so that I could speed it up or slow it down.
That also proved difficult to impossible, in practice, for me to accomplish.
I hit on the idea of controlling the frequency BY controlling the amplitude at about that time. If I just had a spring ...
hmmm ... what if I just have one coil and put a set of permanent magnets near the extreme of the swing that repulse the magnet in the pendulum? Then wouldn't it be the case that on wide swings the pendulum would bounce off those springs, making it go faster? And then couldn't I reduce the amplitude of the pendulum (by putting less energy into it) so that it would not bounce off the springs, and would then tick slower as a result of that?
So I retrofitted the clock with these magnetic spring thingees, switched to a single coil, rewrote most of the software, and within a week or so of the first build, I had achieved a clock that could sort of keep time.
I called this version 1.1.
I spent a lot of time messing around version 1.1, particularly mucking around with the code.
It was not pretty. There is a PID controller that adjusted the energy put into each swing based on the milliseconds error, but I also needed to add a few "heuristics" (kludges) to keep the clock running reasonably ... to prevent it from "stalling" while trying to slow down, or banging back and forth while trying to speed up.
After a week of effort, Version 1.1 was probably accurate to a few minutes a day.
Since I wanted to make a clock as a gift to someone for Christmas, and I was running out of time, I did one more quick design pass to rectify the worst problems with Clock1 and pressed on with another build:
- the wooden frame was not square and the clock did not sit flat, so I designed a 3D printed base which at least kept the clock itself from rocking back and forth
- I formalized the design of the magnetic springs so that they could be more accurately adjusted.
- I designed and built a new PCB that only used one LM293D motor controller
- I added a face with numbers to the clock.
I laser cut a whole new clock, and all the parts, 3d printed stuff, made the new circuit board, soldered it up, wound coils, sanded all the gears (pain!), painted everything, and so on, put it together, and, after putting it through a quick testing cycle, I sent it away.
I would say that it was accurate to a few minutes a day at it's best.
I did not even get a chance to test it for a week before gifting it :-(
Nor had I, in the code, gotten around to any notion of utilizing the ESP32's RTC (real time clock), much less using NTP (Network Time Protocol) to make the clock accurate over any time scales (though I had laid the groundwork in the code, WebUI, etc)
It just tried to beat at 1000ms per swing.
I called this "theClock - Version 2"
Part of it was the mechanics .. it was missing beats or sometimes grabbing two gears. There were friction problems, and, I think, misalignment causing the gears to bind at times. But also part of it was the software and heuristics, which were hamstrung with the information available to the software in order to control the pendulum, and somewhat hastily written and not really that well tested.
From these initial clocks I came up with the following list of problems:
- The use of dowels to connect the different planes of the clock did not work well. If the dowels were fitted into thick pieces of hardwood, and the holes were drilled square, then the dowels could facilitate a square frame. However, because the holes are merely in thin pieces of plywood, the dowels were not inherently square to the frame. In other words, there were alignment issues between the planes of the clock, which caused a lot of friction and binding of the clock as the hands and gears turned. This was particularly true when taken with the requirement that the clock be disassemble-able and the clock was only screwed together in the final assembly. To get it to work, I had to dry build* the clock, twist the dowels and position the planes the best I could, apply epoxy glue, and then literally hold it together with my hands as the glue set in order to achieve any degree of alignment.
- The ratchet mechanism requires the pendulum to do all of it's work while swinging in one direction, and does no work while swinging in the other direction. This made the pendulum swing (and forces involved) lopsided and caused an uneven beating of the clock. It would swing much more, and more easily, to one side than to the other.
- More importantly, the ratchet mechanism is directly coupled the pendulum to the pawls, so that the wider the pendulum swings, the further the second hand moves!! This resulted in the mechanism being unreliable, particularly to the degree that I was trying to control the frequency of the pendulum by increasing or decreasing the amplitude of the swing. I found that, at wide swings, the pawl would sometimes grab two teeth and advance the clock two seconds at a time, or arguably worse, in smaller swings it would sometimes fail to move the seconds hand at all (a condition I called a stall).
- The five Hall Sensors did not provide me with enough resolution about the position of the pendulum to implement a good PID Controller (algorithm for controlling the clock), and as a result, the swing would vary wildly while trying to speed up or slow down the clock, sometimes banging the left and right limits of travel (and grabbing two teeth!) or other times failing to even grab a tooth and move the seconds hand (stalling).
- Due to time constraints I had not been able to implement any code that took advantage of the ESP32 RTC (Real Time Clock) or the available NTP (Network Time Protocol) to make the clock more accurate.
Although I compensated for the resolution problem by putting heuristics into the code for these cases, in subsequent clocks I utilized a different, more accurate way of sensing the position of the pendulum which led to a much simpler algorithm, and smoother and more precise control of the pendulum, in turn resulting in a much more pleasing and consistent swing.
After I built theClock3, I felt bad for my old forlorn clock version 1 sitting there on the shelf, so I retrofitted it with the AS5600 angle sensor and software from theClock3 resulting in it keeping much better time and having a much nicer swing (even though it is not as accurate, or trouble free as theClock3) ..