Tractor

Mon 19 January 2026 Tagged: cad, metalwork

The electric tractor is finished! I have been working on this on and off for about 6 months. It is a toy for children (and me) to drive around the garden.

Building the tractor has been fun for both me and Lucy. On many occasions she has asked "can we go and work on the tractor? right now?" and the two of us would go out to the garage and tinker with whatever was in progress at the time. Often she would get bored and go back in the house quite quickly, but that’s par for the course for a 3-year-old. With any luck some of the philosophy of creation will rub off.

Specs

The tractor is powered by a 350W brushed DC motor with a 36v Li-ion ebike battery. You may be thinking that 350 watts doesn’t sound very powerful, and you’re right. It is a toy for children to drive around the garden, being slow is a feature not a bug.

The rear axle is solid, meaning the two rear wheels always turn together.

The front axle pivots around a pin in the centre, which keeps all 4 wheels on the ground when driving over uneven terrain.

It has a cable-operated disc brake on the rear axle which is very ineffective, but slightly better than not having a brake at all.

The seat position is adjustable, so that adults can just about squeeze on to it and toddlers can just about reach the pedals.

And the tractor has very poor handling characteristics if you’re an adult, because all of your weight is over the solid rear axle and the rear tyres have much more grip than the front, so you have to lean forwards to make it steer. For a while I had one of the rear wheels free-wheeling, which makes it steer better, but means it sometimes gets stuck on hills where you can’t drive forwards, you can only wheelie to the side, which on balance is worse. I expect it’s not as bad for small children because they’re a.) not as heavy, and b.) sitting further forwards anyway.

Chassis

The core of the chassis is a plywood box which is open at the bottom.

In this picture you can see the pin that the front axle pivots on:

And you can see the pencil lines that roughly show the limits of motion of the lower edge of the axle beam.

I believe the front wheels are for a sack truck, this sort of thing:

They are very cheap, a pair of wheels with tyres and hubs and bearings (brand new) on eBay is only £12 including postage.

Steering

I tried to copy the steering arrangement from a Ferguson TE20, which originally was my reference design for the canonical tractor, although I obviously went over to the dark side with the colour scheme.

The TE20 steering wheel goes down to a gearbox quite near the driver, a shaft comes out each side, one turning clockwise and the other anticlockwise (as viewed from one fixed reference side), an arm off each shaft holds one end of a track rod, and the other end of the track rod is connected to an arm on the stub axle kingpin thing. Highlighted in red here:

To replicate this I made a steering gearbox using angle grinder gears.

(The yellow one is a test print - the final one is in black Polymaker PC-Max material with heavy wall thickness and lots of infill).

Angle grinder gears are available cheaply and are a good way to buy high-quality bevel gears in about a 3:1 ratio, if you don’t need them to be particularly heavy duty.

One issue with my steering gearbox is that there is no way to install it in the chassis because the shafts stick out the sides. This was an oversight, in CAD there is no difficulty. The solution is to assemble the parts in place inside the chassis. This is inconvenient in the extreme and if I were to do it again I would try to make it removable.

The arms that mount on the shafts are made of 3mm thick mild steel flat bar, bent around a rod, and then drilled for a mounting hole for the rod end, and drilled and cut to make a split-clamp for the steering shaft and kingpin.

This is effective and relatively easy to make, doesn’t even require any machining, I’d do this again. The only drawback is that there’s not a convenient way to key it to the shaft.

Originally I was planning to figure out a way to key it to the shaft after I had got the steering geometry sorted out, and therefore after I knew the angle that it wanted to be keyed at, but I have since realised that having these joints able to slip in the event of a crash is an "engineered failure" that prevents destroying the plastic gearbox. So I’m leaving it as it is.

I made the steering wheel myself.

It consists of 2 CNC aluminium parts, one is just a ring, and the other is the 3-pronged part for the centre of the wheel. The 3-pronged part is then bent so that the prongs sit at the right diameter, and then the 3d-printed grips are added each side, with a bolt on each prong holding the whole stack together. It is relatively flimsy, I probably wouldn’t do it this way again.

Rear axle

The rear wheels are ride-on lawnmower rear wheels. They fit a 19mm axle, which annoyingly is not a size that I was able to buy cheap pillow block bearings for. In hindsight, maybe they actually fit a 3/4" axle and it would have been easy? In any case, I don’t want to deliberately construct objects with imperial measurements, that’s just trouble for everyone.

So instead I went with a 20mm axle, and 20mm pillow block bearings, but turned the ends down to 19mm to suit the wheels.

I thought this would be easy because I was labouring under the misapprehension that a 20mm shaft would fit through the spindle bore on my mini lathe. It does not! It’s very close, but it doesn’t fit. Possibly you could bore out the spindle bore slightly so that it would fit, but I didn’t think of that at the time, and probably it is hardened.

So it can’t go through the spindle bore. And the axle is too long to support the loose end with the tailstock on my lathe.

So my solution was to support the loose end with a plastic bush in a piece of wood clamped in the vice, and kick the tail end of the lathe around to square it up until it’s not turning a taper. This actually worked very well and I ended up with a taper going from 18.94mm to 18.97mm over 150mm length, which for all I know is as parallel as I turn anything at the best of times. See this video clip.

And then the only part I can’t turn down to 19mm is the tiny bit at the headstock end which is clamped in the collet, which I filed down to match the 19mm diameter after I was finished with the rest of it.

The rear wheels have 2 flats in their bores to key them to the axle. I machined matching flats on the axle with my homemade CNC machine.

The wheels are kept from sliding off the axle by split pins, one just outside each wheel.

And then the axle also has carriers for the sprocket and the brake disc.

I originally mounted both of these by making an M6 tapped cross-drilled hole in the axle, and making a hub with a 6mm cross-drilled hole through it, and then bolting the hub to the axle. This provides both axial location and torque transmission so I thought it was a simple and effective solution.

You can see the head of the bolt fixing the sprocket carrier to the axle in this pic:

Unfortunately fixing a hub the axle with a single bolt through a hole is not adequate because the shear force is very large. Eventually the bolt holding the sprocket on snapped. I replaced it with a new bolt and it snapped again, so then I welded the carrier to the axle.

It hasn’t snapped yet.

Countershaft

Originally I thought it would work if the motor drove the rear axle with just a chain and sprockets, but I had missed out a factor of Pi in my calculation. Having the motor drive the rear axle requires about a 30:1 reduction, which is far too extreme, would require a rear sprocket with about 300 teeth. so I added a countershaft to gain back the factor of ~3.

You can see the countershaft in this pic:

It is supported by a couple of small pillow block bearings on a big piece of steel box section.

The shaft is simply an M12 bolt. The large sprocket is bolted to a big piece of aluminium, which is keyed to the bolt with a hexagon machined into the centre, which the bolt head is hammered into. And the small sprocket has a couple of flats on it, rather like the rear wheels, and I filed matching flats onto the threads at the end of the bolt.

Reverse

The motor controller that I got doesn’t have a way to reverse the motor, so I implemented reverse by putting an "on-off-on" DPDT switch in the wires that go to the motor, so that in one position positive and negative connect to the motor one way, in the middle position it’s disconnected, and in the third position the polarity is reversed. This works fine but you need to make sure your switch can handle the current. In my case it’s not too hard because 350W at 36V is only 10 amps.

You can see the switch sticking up from underneath the chassis here:

(It has a black rubbery cover).

I later added a lever so that it is easier to switch:

The lever has no actual bearings, it just rides in holes drilled in the wood. This is more than adequate for the kind of speeds and loads that a gear lever experiences, which are basically zero. The holes just need to constrain its location. A bonus is that the natural friction in the holes prevents the lever from rattling around.

I did add a gate to indicate the selection, and constrain the movement to prevent damaging the switch:

The text is done with multicolour printing on the Bambu X1 Carbon.

Brake

I think the brake is from a mini moto. It doesn’t work very well and I haven’t done a very good job of fitting it.

You can pretty much see how it works in this pic:

The pedal on the left-hand side (right-hand side in pic) pulls on the cable, which then pulls on the lever on the caliper, which presses the pads against the disc.

Like the gear lever, the shaft is not on any sort of bearing, it just pivots in holes drilled in the wood. I think this is also fine for the brake, because it doesn’t get much use, although the force on the brake pedal is much higher than on the gear lever. If the pivots do get worn out then they can easily be drilled out and bushed.

Bonnet

I put off making the bonnet for a good while because I originally wanted to do something like a Ferguson TE20 bonnet:

But I couldn’t work out how to do all the curves. My best plan was to break up the design into large flat surfaces and small curved surfaces, and make the large flat parts out of plywood and 3d print the small curved parts and somehow join them together and body-fill over the crimes.

But then Lucy acquired this toy tractor:

And I saw that there is no need for the complex compound curves. Just a single curve will do.

So I tried to form the curved part with "kerf bending", but:

It instantly snapped instead of bending. I think I had the grain in the outer layer of plywood running in the wrong direction. It might have bent nicely if it was the other way.

In for a penny, in for a pound, I carried on:

And actually that was starting to look like I might get away with it. So I filled the gaps with glue to make it hold its shape.

And after a few rounds of body-filling and sanding, I was actually really pleased with the result.

Throttle conditioner

The throttle response is very dissatisfying. It starts off from a standstill with quite a violent kick and then immediately tapers off into having hardly any power at all.

So the plan was to make an Arduino project that would take the throttle position as input and output a "conditioned" throttle position as output, which would ramp up gradually as you initially apply the throttle.

The throttle conditioner is in the yellow box here:

While it did work, it also somehow prevented the throttle from ever reaching 100%. The throttle position is transmitted as an analogue signal, and I think the analogue output of the Arduino topped out at a lower voltage than 100% throttle. I didn’t care to fix it, and I realised that this was an unnecessary complexity, so I just removed the electronics.

I even had a magnet and sensor to detect when it was changed between forwards and reverse so that it would instantly cut the throttle when you change direction to avoid damage.

But this is not the way. A machine should obey the will of its operator, not its constructor, and it is incumbent on the operator not to operate in a way that damages the machine.

Painting

I despise painting. It is one of those jobs that takes way longer than it feels like it ought to.

Step 1: disassemble the completed tractor.

__IMG_6410_

Step 2: brush paint white "knot-block" primer onto the chassis.

Step 3: spray grey primer on everything.

Step 4: draw the rest of the owl.

Welding

It’s a long time since I did a lot of welding, and my welds on this tractor are rather poor.

I have experimented with the welder today and discovered that my prior mental model about how the welder works was totally wrong. I had thought that turning up the voltage would make it "hotter", and turning up the wire speed would make it "build up more material". As if it’s a 3d printer extruder and voltage sets temperature and wire speed sets extrusion feed rate. But that’s not how it works at all! In fact turning up the wire speed makes it hotter, and turning down the voltage makes it build up more material. I don’t really understand the physics of why that is the case, but at least I know how to control the machine now.

If you like my blog, please consider subscribing to the RSS feed or the mailing list: