New linear actuator design

Browsing the forum I've seen quite a lot of commercial and DIY linear actuators. Most of them use similar configurations of a ball screw and a hollow push rod with linear bearings. (Source: this thread)

The advantages are that it's relatively straightforward and easy to build. However, if the motor is big the ratio of usable stroke to total length is not very good. This could be improved by mounting the motor on the side of the aluminium extrusions and replacing the shaft coupling with a timing belt drive. This led to another idea. How about using the ball screw itself as push rod and drive the ball nut instead of the spindle.
This results in the total length (retracted) to be only slightly above the stroke plus the rotating nut assembly. The drawback is that true universal joints are required, now, because they have to transmit the torque and simple ball joints do no longer work.
My U-joints look more complex than they actually are. I just designed them in a way to use as much equal parts as possible that can be machined from aluminium sheet.

I don't have a 5-axis CNC machine and 3D machining a plate from one side only is a lot easier than taking a big block of solid metal and remove 80..90% of it as chips.
I don't really know yet if this is going to work. Maybe there will be problems with the spindle wobbling due to too much runout of the rotating nut. I've checked that buckling (Euler) is not a problem. A 20mm diameter spindle with 500mm stroke theoretically can take >50kN of push force before failing. I'll buy a spindle and try it out. If it doesn't work I can still convert into a rotating spindle and linearly moving nut configuration.

Interesting approach. Possible problems I see:
- the (open) spindle is most likely greasy and could collect dirt, which could end up in the ball screw
- since the motor is not at the lower end of the actuator, its weight adds to the mass that needs to be moved around (horizontally)

The ball nuts have wipers so bigger particles should stay out. But of course, dry slider bearings like the IGUS bushings collect less dust, that's right.
The static weight of the motors and the power required to move the motor as the actuator changes angle position doesn't matter much as it moves with a much lower speed as the push rod. But the mass/spring system of the motor together with the slightly bending ball screw could cause problems when vibrations hit a resonant frequency.
Difficult to say if that will really be a problem. I have to try it out.

On the other hand... I've seen actuators with multiple very thin push rods like in this and this video. They look so weak that I thought they would self destruct easily. But it seems they work without problems. So I'm not worried much about my design with the motor mass in the middle. Time will tell.

you could get problems when the nuth is not balanced out

@Klaus Schmidinger : I just saw your rig:

Were this the motors I have bought from you? The actuators also look nice and solid. If my design doesn't prove itself so well I can re-use the ball screws and build it like that.
I also like the idea to make the universal joints out of a fixed spindle bearing and an additional hinge. I had a similar idea but discarded it quickly because I had worries about the pole/singularity point when the angle get's close to 180°. But rotating the axis of the bearings into the XY plane avoids this problem quite cleverly.

I don't think that imbalance (COG outside rotation axis) is a big problem. I use 20mm pitch so the nut only rotates at 1500RPM max. All ball nuts I recently saw don't have external return tubes for the balls but small plastic deflection bits. There shouldn't be much imbalance. But of course, if there is excentrity (ball screw axis not equal to rotation axis) the screw will stagger and wobble .

20mm pitch is a good move IMO.
Maybe use this project for some inspiration.

@cfischer Ah, interesting. One thing that came to my mind: At the CNC router you don't have any yaw movement. This would have no effect at all since the tool spindle is spinning around the vertical axis, anyway. But this is different for a simulator. Yaw introduces an error because the ball screws are attached to the rig. The error is small, though. 18° = 1/20 revolution yaw gives 1mm push/pull at the ball screws with 20mm pitch. This could be compensated in FlyPT Mover by adding a yaw component to the position of each actuator. I don't know if you would feel a difference.

The LinuxCNC hexapod also shows the problem with the U-joints quite clearly. This is the classical U-joint configuration. The bearing axes cross each other in a plane orthogonal to the rotation axis when the joint is straight (180°).

This gives the best torque transmission and least backlash near the straight angle and works well up to 45° deflection or a little more.
For the CNC router they have choosen this configuration:

It has one bearing with a vertical axis and an additional "hinge" with a horizontal axis. This causes problems because there is a singularity when the hinge is at 180° (pushrod at 180°, exactly orthogonal to the pod plane). Then the angle of the vertical bearing is under-determined and the hinge is over-determined. This can cause excessive bending force and break the hinge.

This is much better. The rod of the actuator always goes downward so it can never reach the singularity which in this case would be horizontally.

Yes, these are the ones.

The original design of this rig was made by @Dirty. See https://www.xsimulator.net/community/threads/dirtys-2nd-6dof-aasd-m4s-controller-sfs1620.17639.

I checked @Dirty 's design, it looks really well! He also gets a fairly good stroke to overall length ratio. If I calculated correctly his should be around 630mm stroke to 890mm total length (without U-joints). Mine is 510mm stroke to 630mm total (without U-joints) or 708mm from center to center of the U-joints. Thats around 71% (Dirty) compared to 81% (mine). Not much difference. A Festo cylinder with 160mm stroke has ~240mm length. That's only 66%.

Hi @Aerosmith ,
nice to see that you calculated the stroke ratio (stroke/overall length) for your designs. I find that is a parameter not to be overlooked. I have simulated many different rig configurations in software (rig-o-meter) and found that this ratio is quite telling. Often times I see people aiming for as much stroke as possible (nothing inherently wrong with that!) but simply going with longer and ever-longer ballscrews will only get diminishing returns. Eventually the problems will outgrow the benefits.
I find it almost equally important to reduce the overall length and remove "dead space" from the design. If I was given the choice between 600mm stroke out of a 1000mm actuator vs. 550mm stroke out of a 900mm actuator, I would choose the latter. Even though on paper it has less stroke.
And I also usually consider the effective overall length (including U-Joints, measured from center-of-joint to center-of-joint).
Happy building..... and designing :-D

Cheers... Dirty

Unboxing time the spindles and bearings I ordered have arrived

Now, I have to order the aluminium tubes and machine the other parts.

I've just did a quick check of the ball spindles and nuts. I have to admit that I gambled and took the cheapest spindles I could get. I payed around $350 for 6 spindles and 12 ball nuts including shipping. I expected to get reaally crappy quality but after all it's not as bad as I expected.
With support bearings at both ends I measured around 50µm runout at the center of the spindle which should have no negative effect since the spindles don't turn in my actuators. I checked two of the nuts and one is nearly perfectly balanced. The other has ~20µm runout (mismatch of spindle axis to outer cylinder axis of the nut). This can cause some vibrations when rotating fast but should be acceptable. Axial backlash is around 20 to 30µm which would render them useless for high precision CNC applications but doesn't matter at all for simulator rigs.

Today, I've machined the first parts.

Bearing flanges for the rotating ball nuts and the yoke side-parts for the U-joints. Took quite a while to get the tight fits for the bearings right.

Some more parts machined...

Wow, I wasted a whole 4kg bar of 20mm aluminium, ruined a 8mm TC cutter and produced a bucket full of chips just to get some spacers and connecting parts of less than a pound.