My Most Complicated Model Yet, Finally Finished
[Please Note: The engine model described in this blog post has since been superseded by what I think is a better one, as detailed in this blog post.]
I’m way behind on blog posts, but I’m trying to catch up with things I’m pleased with. High on that list, along with my new book being out, is that after months of work and many obstacles overcome, I finally have my most complicated model to date finished and available on mechanicalgifs.com.
What you see here is a functionally complete model of a three-cylinder, four-stroke internal combustion engine. By functionally complete I mean that it models each of the critical components required for such an engine: the crank shaft, pistons and piston rods, intake and exhaust valves, spark plugs, distributor, and the timing gears to synchronize everything. If everything were made of metal to very precise tolerances, as in a real engine, this thing would actually run (but not for very long, since it lacks an oil pump or cooling system, and only if you supplied it with pre-mixed fuel and air, since it lacks a carburetor).
What makes this model particularly complicated isn’t the number of parts or the size, it’s the number of new categories of problem I had to deal with in designing and producing it.
Problem 1: The Crankshaft
I’d been thinking about making an internal combustion engine model for quite a while, but always rejected the idea because of the crankshaft. This fundamental part of the engine is a deeply three-dimensional object: it just cannot be reduced to laser cut parts in any sensible way (I tried, but it came out ugly, unreliable, and hard to put together).
Injection molding is the obvious way to make such a part, but the up-front cost of making a mold is way, WAY too high. I’ve entertained the idea of making molds and doing my own injection molding, but that would be for much simpler parts: making a crankshaft mold is far beyond my capabilities at this point.
3D Printing is the other obvious choice, but until recently the options were either filament printing, which makes horrible-looking and not very strong parts, or stereolithography printing, which makes great parts but the printers and resin needed were too expensive.
Thankfully someone recently came up with the brilliant idea of replacing all the expensive scanning UV laser parts of a typical stereolithography printer with a commodity cell phone LCD panel and some cheap 405nm LEDs. Combined with much cheaper resin, suddenly this is a realistic way of making parts for my kits.
I got one of these printers for about $450. It makes eight crankshafts at a time, taking about 6 hours to finish a batch. That’s a long time, but it runs unattended and a new batch can be started in just a few minutes, so over time it can turn out quite a few if you keep it busy.
You can see in this just-finished batch that there are a lot of support posts growing up along with the part. These support overhangs and generally keep the part from warping out of shape as it’s growing up out of the vat.
Problem 2: The Spark Plugs
It’s really not right to have an engine model without simulated spark plugs. The timing of the spark is such a fundamental element of the operation of the engine that you must have a way to indicate when the fuel-air mixture is meant to be ignited. LEDs are the obvious choice, and really it’s not such a problem to include them in a model. I was worried about the cost until I saw that I could buy perfect little LEDs for, get this, half of one cent each! And that’s including shipping.
Problem 3: The Distributor
The LEDs themselves are not a big problem, but once you have them it creates the need for a mechanism to make them fire at just the right time. That, in a real engine, is called the distributor. In modern cars it’s done with electronic circuitry, but the old way is a simple arrangement in which a contact arm, called the wiper, spins around inside a cup that has contact points arranged at regular intervals around the outer edge (one point per spark plug). As the wiper arm touches each contact point, a circuit is completed and the corresponding spark plug fires.
The points in my design are simply screw heads, so they aren’t a problem. The tricky parts are the two brass contacts (one to make a connection to the shaft, and the main one that rotates around touching the contact points). These would ordinarily be made by stamping with a punch and die set, but as with injection molding, there is a very high up front cost of making such tooling.
For my first experiments I made the contacts by crudely cutting and drilling brass sheets. Because I knew that making a punch was almost certainly not going to be practical, I kept working on the whole project only because I convinced myself that it would be possible and acceptable to sell the models with slightly nicer versions of this. Cut and drilled more cleanly, it might have been OK… But really I knew that I would have to come up with something better.
By happy coincidence I happened to be at a SciFoo conference right around the time I was worrying about these contacts, and got a tour of the fabrication lab in the Google X building. It’s filled with the most amazing 3D printers, laser cutters, water jets, etc. I asked the guy there if he had any suggestions for how to solve my problem of wanting to make a small number of nicely-cut brass sheet metal contacts. He happened to have the absolutely perfect solution: take the sheet metal, stick it down to a piece of flat scrap material with double-sided tape, and use a CNC mill with a very pointy milling bit to cut the outline of each part. He even told me the brand name of the sticky tape to use.
This seemed a bit crazy to me at first, but turns out to be a beautiful solution to my very specific problem, and quite practical since I have a CNC mill. It even forced me to learn enough GCODE (the language that controls these machines) to be able to write a Mathematica program to create optimized cutting paths for simple 2D designs. This in turn was greatly facilitated by the fact that I have thousands of lines of code I developed for optimizing quilt stitching patterns, which happens to be a very similar problem.
These are an earlier version of the contacts: what I ended up with has a shape more carefully optimized for the particular arrangement around the shaft, and has the wiper arm is now symmetrical around the arm. This illustrates one of the main advantages of making these on-demand: if I had spent thousands on a punch and die set, I wouldn’t be able to improve the design iteratively.
Not A Problem 4: The “Starter Motor”
The model includes a small electric gear-head motor that drives the model automatically, so you can just watch it running, hypnotically going through the four strokes that give four-stroke engines their name. Sourcing these motors and figuring out how to use them in a model might have been a problem, except I went through all that some time ago in connection with several other models (all of which use exactly the same motor). It’s nice being able to use this sort of experience. It’s why people, and companies, get better and better at what they do over time, as experience builds up. (The process continues until the weight of experience starts to ossify a person or company to the point that doing anything new just seems like too much work. Then it’s time to switch to something completely new, like making mechanical models instead of writing about chemistry, or whatever applies in your situation).
Problem 5: The Wiring Harness
With all this electrical stuff going on in the model, it was inevitable that there would need to be the equivalent of the wiring harness in a real engine. This is the mess of cables that ties everything together.
I want it to be the case with all my models that people can put them together using nothing but what comes in the box: no additional tools or supplies required. That meant I would have to supply all the wiring pre-assembled, with the battery, switches, and motor wires soldered together. (I tried designing it where people would make the connections using screw terminals, but it ended up looking really ugly, with lots of space wasted just connecting wires, and even then it would not have avoided all soldering.)
So I decided to go all-in and solder together all the components with pre-measured wires, so all the electrical components can be dropped in as a unit at the appropriate moment in the construction sequence. That means quite a lot of work for us making the harnesses (which contribute a good portion of the total cost of the kit). But it’s good for the end-user, and it also mirrors closely the reality in a real car factory: wiring harnesses usually arrive at the final assembly plant pre-assembled by the parts supplier.
The result of all this is really quite nice. I’ve found several people who were never all that enthusiastic about my previous models, but are quite intrigued by this one. It’s just nifty, you know? Here are some videos that show the model in action. The last one is just for fun: it shows what happens if you use a drill to spin up the model to somewhere around the idle speed of a real engine. Things do not go well for the piston rods.
If you like the look of this engine and want one for yourself, you’re in luck because they are for sale, ready to ship. Be warned, the kit will take you several hours to put together, but as advertised, no extra tools are required, no soldering, no wire stripping, and no prior knowledge of electronics or model-building is assumed.