Axis vs. Actuator-Driven Design

When we started this project, we wanted to build a full 6-DOF motion simulator, allowing for rotation and translation in every direction.  In the interest of keeping costs low and simplifying the initial prototype, we started with a 2-DOF version, which only allows for pitch and roll rotation.  We used linear actuators in the design so that we could use the same actuator design in the full 6-DOF version down the road.

Linear actuator design

As I design the second-revision 2-DOF frame, it is becoming increasingly clear that linear actuators may not be the best option.  Here are the problems:

  1. As these actuators now stand, they cost > $1k each to build.
  2. The actuators are designed to both lift and rotate a load, so they are significantly overpowered for this 2-DOF design, which only needs rotation.
  3. The actuators only allow for +/-45 degrees of rotation.  An axis-driven design would easily allow for at least +/-90 degrees of rotation, even without slip-rings.  That would be a wild ride!
  4. The linear actuators are fairly complex assemblies, time-consuming to build.
  5. The overall assembly is more bulky because of how the actuators must be mounted.

Don’t get me wrong, I’m glad that we have the actuator design!  We’ll definitely use it on more capable motion platforms down the road!

But I’m fairly confident that an axis-driven design would help solve all of these issues.  By spinning each rotation axis directly, instead of lifting the edge of the axis with a linear actuator, we increase the allowable rotation angles significantly (although we may have to increase the frame size to allow for the necessary space).  Without linear rails and idler pulley assembles, the transmission becomes more simple.  We can size the motors appropriately without having to lift a load.  Driving the rotation axis directly will require less space.  And finally, my initial estimates suggest that I can shave between $400 and $600 off the cost of each axis with the more simple design.


A potential downside is that we’ve never attempted a design of this type, so there will probably be some hiccups along the way.  Backlash might be a big problem with the cheap gearboxes we are using.  Without some kind of shaft coupler, high peak torques could damage the gearbox.  And so on.  We can solve problems as they arrive, I’m sure.

The verdict?  An axis-driven design is at *least worth exploring seriously.  I’m hammering out details of the design now.

Follow & Like for Updates

New 2-DOF Frame Design

The first revision 2-DOF frame has a few issues that we are working to address in our next revision:

1) Rigidity isn’t great: accelerating quickly causes the frame to bounce.
2) The frame is large and unwieldy.
3) Sometimes the frame squeaks during motion – the mounting feet rub when weight shifts around.
4) It’s difficult to climb into the simulator, because the chair sits so high.

We hope to fix these problems in our upcoming V2 frame prototype.  First, we think that some of our rigidity issues stem from how the outer yoke (roll axis) connects to the base frame.  Right now, we use two opposing angular contact bearings on a 1.5″ shaft.  The idea is sound in principle, but because we compress them together around 3″ square stock, there is a lot of room for deflection.

Here is the proposed solution:

Roll yoke bearing pack
Roll yoke bearing pack exploded view

The new bearing design utilizes a fairly large (~ 6″ OD) needle roller thrust bearing.  The assembly is held together using a 1″ shaft and a single angular contact bearing in the center to handle radial loads.  Overall, we think that this will handle the cantilevered load much more effectively than the V1 prototype does.

The main frame is also being changed.  Instead of having a large base platform with two verticals, the new version will mount to a single standalone pillar, designed to be rigid enough on its own (i.e. no angled mounting plates like we have on the V1 prototype).  The “pillar” can be bolted directly to a concrete floor, or it can be bolted to a frame of some sort, which we’ll design later.  See below:

Pillar frame design
Pillar frame design – wireframe

The interior ribs provide increased torsional rigidity.  And the pillar itself is sized so that its area moment of inertia is *significantly higher than the combined result of the two 3″ verticals used in the current design (more than 2x higher, in fact).  The new yoke bearing pack mounts at the top of the pillar, as shown.  I’m a little concerned about the rigidity of the stock at the bearing mounting point – that may need more attention.

And finally… the roll axis yoke needs to be redesigned.  It is partially finished, but still needs some work.

V2 Frame Concept – still in development

The concept is shown above, with a human model for scale.  The pillar is shorter than the original verticals on the V1 prototype, and it bolts directly to the floor. The outer yoke (no arms yet) also has internal ribs to improve torsional stability.

We’ll have this design fleshed out in more details soon.  If it looks good, we’ll get the metal parts laser-cut and start welding the new assembly together!

Follow & Like for Updates

2-DOF Prototype Build

I’ve already talked about the 2-DOF prototype, but since wasn’t active until lately, we have no in-process photos or videos from the build.  So, here’s a review!  This isn’t a complete or exhaustive review – it’s just a collection of pictures and videos that I happened to have on-hand.

A huge part of this design involved linear actuator development.  You can find details on that process in my previous post.  Unlike the actuator design, the 2-DOF frame is still in its first revision, so we have plenty of work to do to get it ready for prime-time. 🙂

Below, we machine part of the profile used for the linear actuator.

Below, we are planing the outer yoke where the bearings will attach.

Machining the outer yoke.

Early assembly test-fit.  The attached actuator isn’t finished, but it’s mounted for test purposes.

Roll axis assembly test

Testing the roll axis with a functioning actuator!

Looking good, despite the terrible mess in my garage!

Frame and actuators together for the first time!

Control Box!  We are using a Productivity 2000 PLC for control for now, but will eventually upgrade to a custom-designed control board for more advanced control.

Control box – components placed, but not yet wired.
Wiring the control box.
Hobbes is always helping!

Full assembly together and moved to my basement:

Full assembly.
Follow & Like for Updates

Actuator Designs – 2016 to Now

Like I mentioned in a previous post, we’ve been at this project for more than a year now.  Early-on, we explored different motion control options and decided that we needed a good actuator design to enable the kind of motion needed for complete immersion.  To-date, most of our development efforts have been focused on that need.

Early exploration of kinematics for the platform. Helped us decide on the required strength of the servos.

First, the Final Version

Here’s the “final” version, below.  We’ll make modifications to improve manufacturability and improve the general appearance, but this is close to what we’ll be using for the simulator designs.

In the video, the position target is being changed randomly every half second.  A PC app of our design is setting the position, and a PLC is handling the actual motion control.

Version One – Ballscrew

Our first attempt (completed just before Christmas last year!) used a ballscrew and a smaller servo motor, compared to the final version.  We ultimately decided that the ballscrew was too noisy and that it had too much trouble moving as quickly as we wanted it to move, so we switched to a belt-drive system for the next version.

Version 1 actuator used a ballscrew and a smaller servo motor. We decided it was too noisy, and had too much trouble moving as quickly as we wanted.

Version Two – Timing Belt

This version used a timing belt and a linear shaft / bearing for motion.  It had no trouble moving quickly, but it wasn’t nearly as compact as the version 1 design.

Version 2, U-Channel belt assembly.
Version 2 Actuator – fast and quiet, but had idler pulley alignment issues.

This design had belt tracking issues when the belt was under load, because the idler pulley (at the top of the picture) was only supported on one end, and by the aluminum frame at that.

Attempting to fix belt tracking issues with a new tensioner assembly.

We built a tensioner assembly to support the other end, and it worked.  But it was unwieldy and imprecise – a bad design overall.  So we decided to move to rectangular aluminum stock for the main body, instead of U-channel.

Version Three – Rectangular Body

In an attempt to fix belt and pulley alignment issues, we switched to rectangular aluminum stock.  This made it easy to support the idler on two sides, as shown below:

Revision 3 actuator. Allowed idler to be supported from two sides.

The idler was attached to steel plates on each side, and the belt was tensioned by pulling the idler upward using a CAM assembly (the circular plate shown in the image was rotated about a slightly off-center point, pulling it tight).

This design worked, but the it was more bulky than we liked.  And the linear bearing and shaft design itself didn’t do much to constrain shaft rotation under load, which again caused belt tracking issues.  It isn’t shown, but when fully extended, this actuator was *very tall.  That made it hard to use.

Version Four – Linear Guide, Modified Tensioner

So, we switched from the linear bearing to a linear rail and guide system (as shown in the first video).  This keeps the entire assembly nice and compact.  The idler is supported on both sides, but the tensioning actually takes place on the drive pulley instead of the idler.  This made the tensioning process and the idler assembly itself much more simple.

We also added limit switches to prevent crashes in case of bad control inputs.

This is the version shown in the video at the top of the post, and the one used on the 2-DOF prototype.

Follow & Like for Updates