The (long) History of my DIY Steward Platform (Part 1)

Twelve years ago, I started a new project: building a DIY Stewart platform for flight and racing simulators.

I first came across the idea through a Hackaday article and was immediately fascinated. Not long before that, I had bought an Oculus DK2 VR headset, and the thought of combining the immersive VR experience with a motion platform sounded amazing.

I start building small scale models of different designs:

So I gathered the parts needed to recreate one of the linear actuators from this YouTube video.

It was a complete disaster.

The steel rods I had bought for the linear guides were far too soft and were already badly damaged after just a few test runs by the overly tight linear ball bearings. On top of that, despite my best efforts, I couldn’t drill the plywood frames accurately enough. The guide rods ended up ever so slightly out of parallel, which was enough to make the carriage bind and get stuck.

After a week or two, I abandoned that design. I was convinced there was no way I could manufacture the parts accurately enough without access to a CNC router.

A few weeks later, I started over with a completely different approach. Instead of linear actuators, I decided to build the actuators using rotating servo arms. A good friend of mine, who owns a bicycle shop, had given me the idea: why not use bicycle sprockets with crank arms and standard bottom brackets? For the drive, relatively inexpensive scooter motors seemed like a perfect fit, since they already came with sprockets for the same type of bicycle chain.

More scale models:

And finally a working scale model using small RC servo motors:

The small-scale model worked so well that I decided to start drawing up plans for a full-size version of the design. Eventually, I ordered six 24 V, 350 W scooter motors from Alibaba.

The build itself stretched over many months, from waiting for the motors to arrive from China to building different prototypes, taking frustration breaks, and slowly working through one challenge after another.

At the time, the plan was to preload each actuator with several bungee cords. The idea was that the bungees would support the weight of the platform and the pilot, while the motors would ideally only have to provide the additional force needed to accelerate the simulator.

From the beginning, my goal was to make the platform as modular as possible, so it could be taken apart and transported without too much effort. In the end, I settled on a base made up of three separate modules, each carrying two actuators. Here’s a prototype from November 2016:

Since I hadn’t yet figured out the final geometry, I built a series of lightweight wooden rod prototypes that I could quickly make on my lathe and using simple hand tools. These photos are from November 2014.

Wooden prototypes for the drive shafts. The bungee cords for the preload system can be seen in the background:

A first "full size" prototype with test linkage made of square timber:

A second test setup with absurdly long drive rods, to test the effects on the geometry:

First prototype for steel drive shafts (drive rod still a wooden prototype):

As I mentioned, this project took a very long time, and there were often months when I didn’t work on it at all. This photo is from June 2017. By then, it already had the final metal rods and drive shafts, the main frame made from 50 × 50 mm timber, and an old office chair serving as the pilot’s seat.

At the base, you can spot my rather sad attempt to improve stability using some very flimsy wooden braces. Early test runs quickly showed that the whole platform was a bit too top-heavy.

The solution was to give the platform a much more solid foundation by mounting it on a 1.5 × 1.5 meter MDF base plate. This solved the stability problem, but it also made the platform take up a lot more space.

The motors are powered, just as they still are today, by three Sabertooth 2x25 V2 motor drivers. The drivers were supplied with 24 V from two 12 V car batteries connected in series. Each Sabertooth is paired with a Kangaroo x2 motion controller, while motor position feedback comes from an AMT102 quadrature encoder mounted on each motor.

After several unsuccessful attempts to mount the encoders using gears and other coupling mechanisms without heavily modifying the motors. If you look closely, you can even spot one of those early attempts on the photo of the "first full-size prototype" above. 

I eventually decided to do it properly. I disassembled every motor, used my lathe to cut an internal thread into the end of each rotor shaft, and created a direct mounting point for the encoder on the motor shaft itself.

The encoders are now securely mounted to the back of the motors using custom 3D-printed brackets. It was quite a bit of work, but in the end it was absolutely worth the effort.

By this point, it had become clear that the bungee cord preload system wasn’t necessary after all. The motors, which already had built-in reduction gearboxes, were geared down even further through the bicycle chain drive between the motor sprocket and the chainring. That gave them more than enough torque.

The first full-scale tests also revealed another issue: a complete bicycle crank arm made for a much longer lever than I actually needed. The resulting platform travel was over 40 cm, which was far too much. To fix this, I moved the pivot point for the actuator arms to the middle of the crank arms, which you can also see in the photo above. That effectively doubled the available torque once again while reducing the total travel to about 20 cm, still more than enough for the application.

With that, the idea of using bungee cords was finally off the table.

The upper frame anchor points:

The final version of the metal rods:

A new home, for both me and the platform

In March 2018, I moved for work, and the Stewart platform moved with me. Here’s a shot of me putting it back together in its new home:

And here’s what was, at least for the time being, the final version of the motion platform (September 2018):

Back then, I used the platform almost exclusively with a VR headset, at the time an HTC Vive. 

It wasn’t just for X-Plane, but also for other games and simulators like Elite Dangerous and racing sims.

Hiatus

Then, at the end of 2018, I got married. And with that, although not just because of it, my priorities shifted and other projects started taking center stage. Eventually, I had to ask myself if the space the motion platform was taking up, especially since it was mostly just sitting there collecting dust, couldn’t be put to better use.

As it turned out, the platform’s modular design proved to be a huge advantage once again. It had already made moving apartments much easier, and now it paid off for a second time. The individual actuator modules were stacked neatly in my basement storage unit, the office chair that had always been meant as a temporary solution was thrown out, and the MDF base plate was tucked away behind a cabinet in the living room.


And just like that, the motion platform disappeared from my daily life for the next six years.