Question 6DOF Linear vs Rotary

Hello, I plan on building a 6DOF system in the near future and I am trying to evaluate which motion system to go with.
The criteria that I am trying to evaluate are:
- Cost
- Performance
- Space

Regarding the cost I kind of see they are nearly equivalent, they both require 6 motors + drive. The rotary one requires either worm gear or (better) low backlash planetary gear while linear one will require some hardware, alu profiles, ball screw and nuts. Does somebody have a better price estimation that might differ from what I found?

Regarding space it's straightforward that the rotary one can be made to use less space and this is a nice plus for me.

The main open question I have is performance, which system is superior in terms of feel and precision. Can a rotary system be precise as a linear one? (assume 5-8 arcmin backlash planetary gear 50:1) Can the rotary system produce the same g forces and movements?

If you have any resources that summarize better this topics please let me know.

p.s. I would like a rig that does racing well but can also handle flight

This might be too much information but, Im my experience you dont really need precision for motion sims. Not like what most people think when the word precision is used, like cnc.

You do want smooth and often quite motion for flight, go read posts from username dirty on his journey building linear actuators. And you want large range of motion.

For racing you need good stiffness to handle the speed without a bunch of vibrations from a floppy rig. This will give false cues. People dont talk about stiffness much. IMHO its paramount.

I like to think of a simulator as a means to get data into the drivers body - correctly (minimize false cues).

My next sim iteration will have a rotary stewart platform on top of a continuous yaw dof. This will be racing centric and I will draw inspiration from the guys at kinemaniacs using direct drive motors and having hardware hold the weight of the sim. This will give great high frequency data. But I will also take inspiration from full motion dynamics senior design project from 10 years ago. The key to their success was to get the upper triangle where the motors connect, to be as small as possible. This gives more stiffness, but also way more range of motion - especially in pitch and roll. It is also quite compact.

I'm trying to plant the seed out there that we should aspire to give 1 to 1 pitch roll and yaw to the driver and not add surge or sway to pitch or roll. Also the center of rotation should be at the center of gravity of the driver. This is not how its normally done, so take it with a grain of salt. The point of 1 to 1 pitch and roll is to feel the climb up or down a hill, or to feel the upset of the car on a large curb or bank. Most simulators ignore this data. So for driving, one would measure the hills and banks and curbs of what tracks or stages they want to simulate. My goal is ±30° for pitch and nearly that for roll. Continuous yaw is way harder but allows the driver to get used to a proper turn and catch slides when breaking traction in the most data rich and intuitive way possible. Its hard to build though...


And of course supplement the low frequency g force data with a g belt, g seat, and g helmet (which is more important than the chassis if your looking for a starting point).

i thought long and hard about this

linear actuators are very neat and look the part
but
they are expensive
complicated
noisy - from the ballscrews
and often have poor performance in mm/s
they are more suited to flight sims where long travel is needed

rotary actuators have some advantages - but do not look so cool and are suited to driving sims
cheap
fast acceleration
easy to build
sometimes difficult to get the right power/speed gearbox combination

at the end of the day its your build and you have to do it your way

Not too much information, I think it's great information and we should really start to build a wiki because finding this kind of concepts in the forum is pretty hard for me. By the way with precision I exactly meant that stiffness concept and ability to move exactly of the data is showing. I think right now I am leaning much more towards rotary due to various factors but I really want to understand if I am missing out on something.

Regarding hardware hold the weight you mean the pneumatic piston as damper or holder? I have seen also a rig using the same concept but at a smaller scale.

Having a top triangle small will make it more stiff but also reduce movement precision, you will need high precision hardware (motor, gearbox) to achieve good movement. There is a nice platform from the user motion4sim which uses bent rods to achieve smaller top platform and it looks great plus it's really compact.

Why the center of rotation should be at the center of gravity of the driver? When driving the car this is rarely the case because the rotation center often ends up being the front wheels or am I missing something? This is why the 5DOF actuators + slider rig is good in the end.

Why do you say expensive and poor speed performance for linear actuators? From my calculations rotary is more expensive or the same. Assuming that we use the same motor + drive rotary need a good planetary gear reduction which I cannot find for less than 220 euro while linear requires less than 150 euro on top of that. Not considering the worm gear one which maybe is cheaper but generally has plenty of backlash (not sure if it matters much)

Regarding noise, did somebody in the community build low noise linear actuators? Could not tell from the videos because everybody is blasting car engine noise in the background ^^'

I use worm drive gearboxes with metal gears - silent and no backlash

look at the sim by knaufinator

I use brushed DC motors and he uses servo motors - its a great build, very powerful

are you designing for flight or driving ?

Yes I have seen already all the 6DOF builds possible, especially the one from knaufinator and smitty and they are amazing indeed. There is also one from motion4sim which I think is even nicer but they all use planetary gears and this combo comes at 250 + 220 euro per actuator. Do you have some links to the hardware you are using?

I am designing for driving and space flight, at the moment I have no huge interest in commercial flight so I do not need long motion I guess. I know I cannot build a system that is perfect for every use case but I am a big fan of optimization and modularity so I would like a system that is good enough for everything.

For example I could build a rotary system that supports different arm lengths so I can have high torque/speed movements for driving and longer smoother movements for flight in the future.

Do you have documentation about your build?

EDIT: nevermind, I have already seen you build actually but do you have a link to the motors/gearboxes you used?

i have 3 sims

Sim 1 uses 2 x 400W motors per actuator (800W) from parvalux and a 25:1 gearbox

Sim2 uses 6 x 535w motors from parvalux with pairs of gearboxes 30:1 50: 60:1 ( on a driving sim not all actuators work as hard as others)

Sim3 - the final one uses linear ball screw actuators 6kw per motor iirc and approx 1.6 of travel - (project on hold, it can lift 1.8 tons and simulate 2g of accelration !)

Yeah the piston holds the weight on that sim, challengingly with as little damping as possible. So the motors only move inertia. I was thinking of trying a constant force spring like in the old car windows.

I agree having a smaller triangle will make the angle changes bigger and thus a small motor motion will produce a bigger change to the sim angle, but I think its still well withing the accuracy of any good enough gearbox. On my design I'm hoping for something like 25" triangle lengths.

I place a higher standard on gearboxes for stiffness sake. In motion4sims thread in the first driving video you can hear and see the wobble after abrupt changes in angle. This is full of motion that shouldn't be there and if you compared an accelerometer/gyroscope output of the sim vs the incoming data it would be pretty ugly.
There is also a large moment from the location where the force is applied to the center of mass on that rig. So the parts are much more stressed than they should be. Which is a segway to your next question.

My reasoning comes from the physics concept superposition.
https://en.wikipedia.org/wiki/Superposition_principle
It basically states that any linear system can be broken up into components. This is what we do with our sims. We break up a complicated vehicle motion into three translations and the three rotations. If the center of rotation occurs where the driver will experience the least amount of translation(center of mass) then you can isolate that degree of freedom from the data stream.

Let me give an example. If you drive on flat ground through a gradual turn then you are only rotating in yaw at that moment. The center of rotation for that turn is very far away from the sim, could be hundreds of feet for a gradual turn. Lets say you were rotating at a steady 18°/sec but you are taking the turn at a high speed. Then lets compare that scenario to a turn at the same 18°/sec but your in a parking lot with very little speed. The yaw will be the same for both scenarios, and we apply a high force in sway for the fast turn and a low force in sway for the parking lot turn. This is why the g belt, g helmet, g seat are so important.

The same thing applies to hills and banks.

Lets look at pitch and specifically the motion4sim rig. His center of rotation is under the platform(very common). The farther any part of the driver is from the center of rotation the more translation is included in the motion. Take it to an extreme and its practically all translation. As a sim builder you want the ability to give any data from the game to the driver. If your rotations are full of translation too you cant separate this data and present it to the driver very well. Pegleg sims (3dof - pitch roll heave) don't have anything to provide surge and sway so they send this motion to the pitch and roll axis and they keep the angles small. This works well to provide short and sharp surge and sway because the center of rotation is on the floor and they feel almost only translation. This is a fantastic method to deliver only high frequency surge, sway, and heave because its cheap! But cant deliver any meaningful rotation data to the driver. A stewart platform already has the ability to deliver high frequency surge sway and heave. So now make sure you can also deliver at least pitch and roll. Its fantastic to notice elevation changes on a race track, and in rally - my opinion is that pitch is the greatest contribution to the information delivered by a chassis. Everything from setting the suspension for good weight distribution in a turn, to landing a jump that was nose diving, to loosing traction cresting hills, to feeling the steep decent robbing you of brake power. Most people ignore pitch, roll, and yaw data.

One might think ideally the center of rotation should be where you sense rotations the most (the head), but that is actually not true. Because if you rotate from the head then the feet will very much notice the translations during car driving speeds and that will break immersion/trust in the data. You really need to minimize the distance between the center of rotation and any mass of the driver. Once could play a bit with biasing the center closer to the head vs the feet but only so much.

The center of rotation around the front wheels is what is commonly referred to traction loss and is a calculated degree of freedom. It tries to deliver the information to the driver that they have crossed over the traction budget in a turn and only moves when the slip occurs. This is a bit messy of a dof because of false cues and travel limits. In my holistic view of data to the driver it is to be avoided. But I can respect that people who wont go all in (continuous yaw is a bitch) will get enjoyment from it.

This is all of course my opinion. I hope to someday build the things I obsess over and then get lots of feedback from lots of actual drivers.



PS: unfortunately a lot of sims with large travel like the 401cr from force dynamics DO NOT deliver 1 to 1 yaw in their profiles. They want to give you motion cuing that tips you to use gravity as a tool for low frequency acceleration (longer accelerations than a bump). This gives lots of false cues in driving because the changes in acceleration happen too quickly. It does work in airplanes (usually) because changes in acceleration are slower and people can't tell if they can't see a change in angle under a certain speed. For me and for training, motion with false cues is INVALID. So it is my goal to build a motion simulator that minimizes false cues as a first principal.

Anyway the false cues is why there is a bad reputation for motion sims and the whole "less is more" movement. If machines don't deliver false cues then large motion ranges can work.

A lot of information to digest in this post haha, we should synthesise it into a wiki for general consumption.
Regarding the constant force spring idea I think it is really interesting but I am not sure I get the topic regarding the center of mass. I agree that the motion4sim center of mass is really bad and puts a lot of strain on the motors when doing strong pitch/roll motions but I don't understand why you say we cannot deliver the motion we want to the driver if the center of mass is not at the driver. Isn't it enough to configure the delta in the sim software and the motion adapts to deliver good roll motion even with a center of mass 400mm higher then the platform? Of course the more the center of mass is away from the center of the platform the more angle we loose due to the fact that the platform has to circle around the driver to reduce any sway/surge/heave.
My understanding was more that the best would be to fidn the best compromise between the center of mass of the entire rig+driver and the driver to get the best of both worlds where the motion range is maximized and the strain on the actuators is minimized.

As a matter of fact, I jumped into sim racing just recently, still waiting for my equipment:
- Moza R5 + clutch + ebrake + hshifter (I know is entry level but it's my first gear and I did not want to buy a trustmaster)
- Sim-Labs GT1 EVO (this thing is amazing, it is barely more expensive that DIY 400€ vs 300+€)
- OMP TRS-E XL seat (I'm a big boy)

While researching equipment I was astonished at the amounts this gear costs and most of the time for no reason.. a stupid load cell ebrake can cost more than 250€ and a 6 DOF stewart platform rig 15k € while making this tools yourself you can spend a fraction of the cost with the same or 90% outcome..

I think as a first step I will build either a 4+1 DOF slider rig by simply adding some sfx-150 actuators to the sim-labs frame and work more on my 6DOF project in the future..

This is why I decided to create a project called OpenSimGear where I develop (hopefully with the help of the community) high quality, open source, well thought out sim gear that anybody can build by themselves for much less.

I am currently working on the website, wiki, and tools and want to start soon to design the first gear in the ecosystem, for example the pedals or the handbrake which are pretty easy hardware and electronics wise. I wanted to design a ultra low cost version made for 3d printing and a more premium version (even if with optimized hardware) made for plasma or CNC cutting and the electronics should be interchangeable.

Regarding the tools I would like to create multiple design helpers for motion sim kinematics and requirements. Like entering the platform type, motor specs and calculating maximum achievable G forces/angles etc. I have seen some utilities here in the forum (excel and apps) but I want to build web based tools accessible to any platform and all in one place.

Have a lot of ideas but not much time.. maybe you can contribute in the motion sim docu/wiki!

It is true that the center of rotation can be changed virtually but it takes a lot of range to do this. Its easy to check out how much using flypt mover as a design tool to guess check geometry vs range. In my geometries and ranges 400mm offset would be far too restrictive. I have found about a 100mm offset to be as far as I can offset before my ranges are compromised. I don't yet fully understand how performance is affected, detrimental to some degree because it is after all more travel the motors have to move to get the sim to rotate with an offset.

I agree with this. Look at it this way, rotate about the center of mass of the loaded rig (seat+driver) to get the most out of your motor. Rotate about the center of mass of the driver to get the best data to the driver. Ideally the centers are quite close anyway.

If I could I would encourage you to go with low frequency tools before the chassis. Things like g belts, g seat, and g helmet will give you way more information than high frequency motion from a rig. You can start with the belts and do 4 direct drive motors. These motors can give you high frequency and low frequency surge (mostly just braking) as well as some sway and heave without feeling wrong.








Then move to a g seat like in my signature. I'm getting 170lbs of max force into each side of my body with tons of fidelity for nuance. Plus the fantastic surge dof in the seat base pushing my low back into the seat when accelerating forward. Its amazing.

Then if you really want the most bad ass data to the driver, a g helmet. Use something like a bigscreen beyond so the helmet doesn't interfere with the view through the headset. Perfect low frequency surge sway and heave to the head (only cantilevered body part in the car) right to the center of gravity of the head. Plus you can put shakers on the helmet for more excellent tactile interface.

Once major bonus of these better tools is that they are lower power and thus can be cheaper. And you don't have to bother with motion compensation for vr. Chassis motion is flashy and fun but these tools will give better data. To each their own.

This is a valiant passion and I commend you. A lot of what you want to create is sitting in the faq of this site painstakingly curated largely by @noorbeast over many years. Make sure you're familiar with that resource in your efforts.

Had to laugh at this. I think most people on this site fall into that category. I could contribute to the wiki but I think most of my direction is towards getting the absolute best data into the driver and that wont be most peoples main objective. To some degree people need to play with the cheap junk to see if it fires up some interest to go farther. Most will stop with the junk and that's fine.

I have seen many rigs that are way off, everybody is putting the set at the center of the platform and the entire front part of the rig is dangling in the front (even with monitors on top).. On many rigs I have seen I would estimate the center of mass to be outside of the top platform by more than 100/200mm

That is a good idea, the belts should be easy and "cheap" at least, the other two not that easy haha. I currently will put the rig in my office, not much space in my micro workshop. I have currently many projects in the pipeline that will make any mod easier later:
- CNC router (building the control cabinet)
- Lathe stand and conversion to CNC
- Manual mill/drill to motorized feed and DRO

Yes I plan on collecting a lot of knowledge from here and then synthetizing facts and real numbers in the open source wiki. I find it really hard to find info in the forum here even if it is curated.

Well the whole project is aimed to exploring the best tech for sim racing and documenting what is available right now plus providing possible DIY solutions for who wants the best experience. If I ever make it with it, because it will be a good amount of work and I am really good at getting interested in million other things before finishing the thing I am doing haha.
Anyway the wiki will be open source and anybody will be able to contribute, like everything else in the project.

I think we would make great neighbors. You're not in Colorado are you?

Hehe, no I am in Germany but I think I have seen you have a CNC mill with ATC in one of the videos/build logs so I definitely agree!

Due to space constraints I am building only small tools.. I have a MPCNC that needs to be assembled and a Mini lathe that needs some love: changed the bearings to angular contact bearings, have all the HW for the 2 axis CNC adaption and I am building a metal stand with concrete top to bolt it down properly.. Oh and after I do the conversion I am removing the cross slide and mounting a quick release tool post on a solid block of steel that I still need to machine... Now that I wrote all this stuff I don't even know why I got into sim racing, was a big mistake

Whenever I will have to move from this apartment it's not gonna be fun, I hoarded pretty heavy stuff in the basement

So many people on this board seem to be in Europe. Germany and England seem like it comes up a lot. Maybe you can find others and try before you build. This would teach you so much super fast. I have spent a lot of time building the normal things people make just to get an understanding of the baseline.

I can relate to the overwhelming list of projects. Just wait until you're doing construction to expand your capabilities with a family that takes priority and a job that takes your time/engery. The project list drags out to infinity.

That is exactly my situation, two small kids and a job (thankfully, wont do much in sim racing without one apparently haha), I only have one chance, make my kids interested in the basement and I am working on it!

Regarding trying, I actually never raced a real race car although I did push mine a little some times when I still had a M235i so I should be able to see if the feeling makes sense I guess. I know that in Germany it is pretty popular, the problem is that many have probably a 4 DOF+ standard rig so I am not sure if it's worth trying. I have seen I can book a sim rig hour in a couple of places around here.

For now I will slowly get into the hobby, I already spent 1.5k just to start with no motion and I went with the cheapest decent stuff I could find.. I will try to summarize all the thoughts regarding the rotation and cues for simulation because I think is pretty valuable info. And then make a plan on optimizing the money/immersion ratio over time

Respect.

Just for your info I kick started the project I was talking about, this is the website https://www.opensimgear.org!

It's mostly empty at the moment but I am working on it slowly, if you are interested you can keep an eye on it . Any non technical help is needed and very appreciated: content, style, UI/UX, ideas