Zip's DIY Rotary 6DOF VR Rig

The Idea
I should’ve started this thread earlier in the project, but it’s too late to go back now. This is still far from complete even though it is fully functional. Here is a CAD mockup of just about what I ended up with. The drawing is mainly just the frame and major components — I left out a bunch of small parts and I didn’t bother to link it all together into a full blown assembly.


After wanting a motion sim for well over a decade, I've finally reached a point where I figured I can actually pull it off in terms of space, time, money, knowledge, skills, and tooling. I actually started this after chatting with a few friends about racing this past 4th of July.

I’ve been lurking this forum a bit since then, but none of the builds were exactly what I was looking for. Every build had things that I liked and things that I didn’t (or were restricted due to budget or some other reason). Initially I had thrown around the idea of just slapping together a super cheap 2 or maybe 3 DOF (perhaps just a seat mover), but the more I read through this forum and thought about it, I decided to go full 6 DOF or nothing. I took inspiration mostly from the DOFReality H6 (first post so I can’t link external sites, but you’re probably familiar with it) and from early_m’s Compact Racing Sim - 6 DOF.

Project Requirements
Since I’m writing this after-the-fact, I’ll do my best to recap the build and walkthrough some of my decision making. Let’s start with the project requirements:

- 6 DOF
- Small footprint. Target around 3×5 ft.
- 120 V compatible
- Budget around $1000 to $1500 (definitely overshot this)
- Moderate passenger weight capacity around 200 lbs (although I’m only 125 lbs)

I’ll expand on each of these constraints below.

DOF: Dirty discussed some motion cueing principles in a video in his 6 DOF thread (https://www.xsimulator.net/community/threads/dirtys-6dof-aasd-servos.14621/page-3#post-202008) and that really swayed me away from making anything short of a full 6 DOF. My worst fear was to spend hundreds of dollhairs on a seat mover or a 3 DOF rig, only to be disappointed by the feel of it — primarily by any false motion cues.

Footprint: Another constraint in my decision process was space. I needed to keep the footprint as small as possible since it was going into an 11×13 foot room that already houses my servers, 3D printer, and various other electronics and general storage. I sorta wanted to make a clone of the DOF Reality H6. Of all the 6 DOF rigs I looked at, the H6 had among the smallest footprint at 4×5 feet. If it wasn’t for limited space and monies I would have gone for a real Stewart platform with linear actuators, but alas, gotta work with what I got.

Power: Most of the builds I’ve seen use either brushed DC motors (12V or 24V) or AC servo motors (which always seem to be the same 750 W motors). In my shopping around, I had a hard time finding suitable brushed motors with sufficient power and torque. On the other hand, all the AC servo motors that I could find either or both 1) ridiculously expensive, or 2) runs at 240V, which in the US I only have 120V.

In hindsight, maybe I could have run the 220V motors undervoltage on 120V, but I feared that could result in excessive current draw and didn’t want to take a $1000+ gamble buying six of them without knowing for sure how it would hold up.

I don’t want to babble on too much here, but choosing a motor was by far the most difficult decision because it directly impacts all other drivetrain options. The motor choice determines the gearbox selection as well as the resulting travel range and load capacity of the rig.

Budget: Sadly I’m not a millionaire, so there’s only so much I can spend on this.

Weight Capacity: I only weight 125 lbs, so most motors would be able to push me around no problem. However, I would like to have headroom for other friends to try out the rig. 200 lbs would safely cover most other participants.

Proof of Concept Model
Before I ordered anything for this, I wanted to get the software I’d need and do some early testing. I slapped together a python script for a Pi Pico to output six RC servo PWM signals. Then I sketched out and 3D printed a little 1:10 scale model platform, or at least close to scale. I didn’t take any pictures of this. After rigging up the interface, all that was left was the game-to-hardware bridge. I’ll admit I didn’t do as much research into this as I should have. Regardless, I ended up buying SimTools 3. From that point it wasn’t long before I had the model following along as I drove around in Live For Speed! The motions weren’t perfect, but it was close enough and I figured I’d dial in the mixing ratios when I can actually feel what’s happening.

I still have the RC model on my desk, just not wired up anymore, so here’s that.


The Design
Like any good project, this one needed a CAD model. At first I started drawing up a welded steel frame since that is my ultimate goal, however I was too eager to have a working rig so I switched to lumber for now. That will serve as a prototype or a working concept that I can use to really flesh out how I want the final design to be. Plus its way easier to make changes to if I want to add or relocate things.


Early on I thought I could get away with NEMA23-sized 180W 24V BLDC motors and the matching NMRV-030 40:1 gearboxes, however after buying one of each and bench testing it, I quickly realized how underpowered it would be. I could easily stall the gearbox output with a 3D printed crank arm and a board attached at the lowest 50 mm setting. By hanging the board and pressing down on a bathroom scale, it generated barely 60 lbs of force before stalling. That might sound like a lot but it felt very underwhelming, plus I really wanted to be able to use the longer crank settings and that would only further decrease the load capability.

After weighing out a lot of other drivetrain options, I ultimately landed on a combination of NEMA34-sized 440 W BLDC motors, BLDC-8015A motor drivers, 48V 10A AC-DC power supplies, NMRV-040 50:1 worm gearboxes, and crank arms with 50 mm, 75 mm, and 100 mm radius options. This cost around $1300 and was on track to put me way over what I initially wanted to spend on this build, but once I started having parts arrive the excitement got the best of me and the budget got somewhat ignored.

The Build
Frame
For the initial prototype I just chopped and screwed some 2×4 and 2×6 pine lumber. It’s cheap, quick, and sturdy enough that I feel comfortable climbing on the frame. Another afterthought, if I were doing this all over, I would have used construction glue (liquid nails is a good go-to) to help secure the most stressed joints in the frame. I had to come back later and add a few gussets and reinforcements to eliminate slop in the platform frame.


First up was what I’m calling the actual platform frame. This is the frame that the seat, wheel, pedals, and everything gets attached to and gets held up and moved around. It is made entirely from 2×4 except for a couple 2×6 reinforcements I added later.

The CAD model really came in handy when it came to making all the angled cuts, especially for the compound angles. Once I had all my cuts done, screwing it all together only took an hour or so. I had a little snag when it came to attaching the seat. The threaded studs on the bottom needed the be chased with a tap and one of the sliding rails had chunk broken off the roller carriage that caused it to keep jamming up until we got the pieces out. I also had to order some M8 hardware that was more suitable, but I had other screws that worked for the time being.

I only have a few pictures of this during construction and all of them were taken in my cramped and filthy shop.


With that done, it was time for the base frame. This is just a hexagon made of 2×4 and 2×6. I had to cut spacers from ½” plywood to lift the gearboxes. This was another late change. Originally I had designed it so the motors hung from the top of the gearboxes as is with most other rotary hexapod designs I’ve seen (including the H6). However, I took one of the oil plugs out to observe how much oil was inside the gearboxes. Out of the box they’re about ¼ full, maybe a bit less — and unless I fill it almost completely (meaning little space for air), the oil would always run down and away from the worm gear. Plus since the output shaft never makes a full rotation, the output gear could not be used to draw oil back up to the worm. Fearing premature wear of the gears, I flipped the design so the worm gears are always submerged in oil, but to do so I had to lift the gearboxes a bit for the motor to clear the frame.



And with that, the framing was basically complete. Next up, machine the crank arms and connecting rods.

Crank Arms
The crank arms I made from a 5/8” thick 6061 aluminum plate that was left over from my CNC router rebuild. Each crank is 40 mm wide by 140 mm long. I clamped the whole plate down on my mill and started drilling. I figured I could drill all the holes and then saw apart each crank from the plate… that almost went to plan. I thought I could use a 2 mm slitting saw in the mill to split apart the plate, that way each side of each crank was a machined face. Unfortunately, my saw was too small and the spindle would hit the part. My only other option was to bandsaw the plate apart. I can go back later to machine the sides of the cranks, but that would be just for aesthetics. They’re functional as-is so I’m rolling with it.

The only picture I got of the machining process was one showing just how close I was to making the slitting saw work.


Connecting Rods
The connecting rods are simple — 5/16” mild steel round bar, cut to 22” length, and threaded M8-1.25x22 mm on each end. No pictures of this. Threading them was a little bit of a pain. I wanted to cut to them length and just run a thread die on the end. Well, I got a die started but couldn’t hold the darn rod. It slipped and spun in my bench vise and even in my mini-lathe’s 3 jaw chuck. In order to get the job done I had to start the threads by single point cutting them on the lathe. With the threads partially formed I was finally able to run the die on without the rod slipping.

With those done I was able to fully assemble the motion platform to the motors! I started out on the lowest 50 mm crank throw until I knew what the motors were moving where they should.


Power Supplies
Wiring up the power supply units (PSU’s) was pretty straightforward, but I can’t think of an easy way to describe it so here’s a diagram instead.


I used two separate power plugs so that the rig can be powered from two separate power circuits, but for now I’m plugging them both into the same outlet. I threw in a single 5 V PSU to power a USB hub in the rig to help with cable management.

I mounted all the PSUs in a row standing on their sides. They’re screwed to a scrap piece of 1/4” plywood that just sits on the floor under the rig. It’s not connected at all to the rest of the frame. I also screwed down the terminal strips, control boards, and the 5 V PSU to the plywood as well.


Servo Control Board
I went fully custom for the servo control interface. I recently got a handful of Pi Pico’s and have been itching to use them for something and thought this was a great opportunity. I took a few shortcuts in making the control boards, mainly to save time because I was too excited to just get the damn thing working rather than have it made perfectly.


Anyway, each board provides a USB-serial interface to control three BLDC drivers with potentiometer inputs to get the position of each servo. I was capped at three channels here because of the five ADC channels on the Pico only three are usable on the GPIO pins (one is internally wired to monitor input voltage and the other to the internal temp sensor). Obviously for a 6 DOF hexapod design, I had to make two of these boards — one for motors 1, 2, and 3, the second for motors 4, 5, and 6. In another attempt to “hurry up and make it work”, I used entirely through-hole components. I used my CNC router to mill and drill the boards, so making a single-sided board cuts the milling time in half.

Each board basically contains a Pi Pico, a FT232 USB-serial module, some discrete level-shifters, an enable bus for the kill switch, a few connectors for the potentiometers, and a few filter capacitors sprinkled about for best practice. I even included an extra screw terminal for a 5 V power supply, but the USB provides more than enough power.


The code (can a python script be called “firmware”?) running on the Pico’s is a bit, how should I put it… crude. The servo control loop takes a very naive approach to determine how fast to run the motor in whichever direction it needs to go. There’s no acceleration ramp and a very basic deceleration (if you can even call it that). Again, this was just because I wanted to hop in and start driving ASAP.

I will be revising the code to implement a smarter motion algorithm to smooth out the movements. I am already looking to make a new control board that fixes my other complaints with these. This post is already way longer than I expected, so I’ll post more info later if anyone is interested.

SimTools 3
I mentioned earlier that I got SimTools 3 for testing. I’m a little disappointed that the licensing relationship between SimTool 2 and SimTool 3 wasn’t more evident to me sooner. I thought they were separate programs entirely because of the split between this site hosting v2 and simtools.us hosting v3. Had I noticed earlier that they licenses are cross-compatible, I would have tried to get a DIY license instead of buying it outright. Oh well, too late now.

To get the rig up and running, all I had to do was modify the interface configuration for both control boards. This, in theory, should have been quick and easy. I sat on the floor in confusion for a good two hours before I finally got the motors moving the way they’re supposed to. I don’t even remember what all the problems were that I had that night. I think the “ah-ha!” moment was when I realized that SimTools was not sending out the interface’s “start” command the way that I expected it would. During testing I had written a small python program for the desktop that gave me sliders to control each of the motors — everything worked fine there… start, update, and stop commands. I copied the start command I had in SimTools and pasted it into the Pico’s code, essentially hard-coding the start command. Boom, everything immediately worked. Not sure why that was an issue. I’ll do more debugging for my next control board.

First Ride
My best friend has helped through this entire build. The grins on our faces when we took it for a spin! Well, you couldn’t really see our face because of the VR headset, but oh man! There were plenty of times when the rig would reach a maximum and stop moving, but when it wasn’t it was a blast. That’s not to say there wasn’t any tuning to be done, because it definitely needed it. I had almost no idea what I was doing with the axis assignments settings, but eventually I got it feeling alright. It wasn’t but the next day when we moved the connecting rods out to the full 100 mm radius crank arm setting. To my surprise, the motors handled it like champs! I was worried the extra torque would stall them, but nope! I even made extension blocks to add 125 mm and 150 mm mounting options to the crank arms. The 125 mm is where it is now and it’s still performing just fine. I’m not ready to try 150 mm yet because if anything fails the crank arms could potentially drive down hitting the floor. I’d have to raise the frame or be completely sure that the control boards and potentiometers won’t fail and cause a crash.

As an unintended bonus, the PSU’s I got are adjustable from 0–48 V and I’ve mostly been running them at 30 V. I did recently crank it up to 48 V because one motor was always running hotter than the others. I figured higher voltage means less current for the same power, so maybe it’ll help it run cooler. I haven’t had any long sessions yet, so I’ll check it when I do.

The picture below is the closest I have to how it looked on the first test drive. The only difference is that on the first drive the cranks were pointing inward instead of outward and the connecting rods were on the smaller 50 mm crank setting. The bystander had control of the emergency stop switch (that’s on the floor and barely in-frame) if anything went awry. The desk fan is always on because it gets hot in the VR headset after a few minutes.


I’ve done a little bit of cable management since then, but it’s still quite messy. As long as nothing is about to snag or get ripped off, I’m just leaving it since I plan on replacing the control boards anyway. It’ll get rewired then.

The Bugs
Of course no project is perfect and I fully expected many bugs and problems to fix later… and it didn’t help that I rushed through just to make it work.

Rod End Articulation
I underestimated how much travel range this rig would be capable of. Using the SimTools movement test I positioned the rig so that was lifted and rolled (or something like that) so that I could more easily reach under the rig to work on something. In doing so, I noticed one of the connecting rods was angled further than the rod end balljoint can articulate which results in the rod itself bending. This wasn’t the first time that one of the rods had bent (I’ve got a short story about another incident), but it was the first time it was due to the bearing reaching its limit. The other bending instances were due to the crank pointing downward and the rod crashing into the crank itself.


Motor Temperatures
After some racing sessions I’ve gone around and put my hand on each of the various components. I noticed that one motor in particular, and its driver, is always much warmer than all the others — almost hot at times. I verified that all PSU’s are dialed in to 30 ± 0.2 V. I don’t know, I’ll keep an eye on it, especially if I have any long gaming sessions.

Slow Movements
This is somewhat subjective since I have practically no experience with motion sims outside of this build, however I’m sure slow movements should be able to be recreated smoothly. I feel like the slow movements of my rig are usually quite jumpy or jerky. Like if there’s supposed to be a slow and steady raising of one motor, I feel it twitch up a little bit, stop, wait, then jump a little further, and repeat. Faster movements, however, feel rather smooth. I’m guessing some (or probably all) of this comes down to my crude servo control loop code, but not sure if that’s the entire problem or not. I guess I’ll find out when I get around to implementing a better motion algorithm.

LFE Vibes
After seeing the response speed of the rig, I decided to get one of those seat shakers to fill in some low frequency vibrations. I’ve never experienced one of those before, but I have some background in building audio systems with some beefy subwoofers. The LFE transducer look like it’s just a weight on a voice coil, so it’s like a subwoofer without the woofer. I screwed it to the framing that the seat is attached to and wired it to a small 50 W amplifier. This is fed from a cheap USB “sound card” (USB to 3.5 mm TRRS adapter) that I got. I was genuinely surprised how much it shakes! I threw on some bass-heavy music and damn! Reminds me of when I had subs in my ride!

I really just wanted to fill in some sort of sensations for when the rig is fairly still, such as driving in a straight line on a smooth road. There’s parts of some race tracks where the rig is basically motionless for a second or two at a time and it sometimes breaks immersion.

I think this is definitely going to help, that is, once I get the damn software to work! Every time I try to enable the Vibe interface in SimTools 3, the whole program just closes… crashed… no error. Doesn’t matter which audio device I set for the sound card. As soon as I click “turn on”, I see two of the transducer icons turn blue, then a quarter second later the window disappears.

My next step is to try out SimTools 2 and see if that crashes as well. Hopefully it’s a bug in v3 rather than an issue with my PC or something else.

Future Plans
As I said in the beginning, this build is far from complete. I mean, between fixes, modifications, upgrades, and additions, I don’t know if there ever will be a definitive finish line. In the meantime, I’m super happy with the rig as-is and only look forward to things getting better.

6-Channel Servo Controller
In addition to improving the positioning algorithm, I intend on making a new control board that can run all 6 servos by itself. To get around the 3-ADC limitation of the Pico, I’ve been looking at various ADC IC’s — they’re more expensive than I expected! Not like break the bank, but $3–$8 for a decent single channel ADC, which means 6x that on top of $6 for the Pico. It doesn’t sound like much, but those components add up, especially when you’re not buying in bulk. Not tracking those prices can easily turn into a board with $70 of components.

I also considered switching from potentiometers to magnets and using a rotational position hall effect sensor that can be read digitally — that would avoid the ADC all together. Again I can’t post links yet, but if I went that route I was looking at the MLX90316KDC. Those are $6 each, so at least $36 for all six. It’d still take a bit of planning to interface those with a Pico because each sensor would need its own chip select output from the Pico and I don’t think there would be enough GPIO pins for that.

I haven’t fully decided yet, but I ordered a few ADS131M06 6-channel 32-k sample-per-second 24-bit ADC (can’t link) and I’m going to try sticking with the potentiometers for now. Those were around $9 each, but a single ADC can read all 6 inputs simultaneously! The next step is to get the Pi to interface with the ADC chip. I didn’t notice until I got the ADCs that they require a clock input synchronized to the serial data clock, so maybe I’ll need a co-processor (ATmega328P?) to in act as a bridge between the Pico and the ADC. Again, I’m not sure… I’m still early on in this redesign.

FlyPT Mover
I really want to try out FlyPT Mover as it looks like it far more capable when it comes to motion algorithms. The ability to send the motion data through various low/high-pass filters before the channel mixing… that alone seems like it would help a ton in getting the feel ‘just right’. I’m holding off trying FlyPT Mover until I either A) fix my emergency stop switch so I can safely configure one interface (3 motors) at a time, or B) replace the interface with a 6-channel board.

System Monitor
I would like to implement is some sort of system health monitor. It might be a bit much, but I’d like to have data logging for current draw from mains power (and I suppose mains voltage in order to calculate total power draw), input voltage of each motor driver, motor driver temperatures, motor temperatures, and the alarm status of the drivers. The hardest part of that I think will be measuring the voltage and current of the mains inputs while keeping the high and low voltage sides properly isolated to ensure safe operation.

Cast Crank Arms
Another issue I was afraid of was the framing around the foot pedals crashing into the front actuators. This was a valid concern cause, uhh, it’s crashed a few times. I’ve already shortened the screws sticking out the crank arms in the picture, but the rods and cranks are still well within collision range. Part of my solution (ideally) would be designing a new crank arm that’s not just a flat plate. I’m thinking of doing an aluminum sand casting for these.


Original write-up was too long for the 25,000 character limit. Continued below...

Front End Framing
In addition to remaking the crank arms to avoid collisions, I also want to rebuild the framing for the foot pedals and dashboard sections. Not only due to crashing, but also the joints attaching the front-end to the rest of the platform is weaker than I anticipated. It’s not that important since I still plan on making a metal frame eventually, but for now I’d like the rig to not break itself apart.

Motor Couplers
Another one of the shortcuts taken was how I mated the motor shafts to the gearbox. The gearbox input has an 18 mm bore hole with a 5 (or 6?) mm keyway, yet unfortunately these motors have a ½” shaft with a small flat side. I know the ‘right’ way to mate them would have been to make a proper coupler that’s 18 mm on one end while the other end has a grub screw to grasp the motor shaft. That size coupler would require the motors to be mounted further away from the gearbox, meaning an adapter plate. I didn’t want to spend the extra time just yet to machine a set of 6 adapter plates and 6 couplers.

Instead I cut out some rectangles of 22ga (or maybe it was 24ga) aluminum sheet and bent it around the motor shaft. That brought the diameter up to nearly 18 mm. Then I took a 1/8”×1” steel flat stock cutoff, faced the top and bottom till it was 4 mm thick, then stuck it in my CNC router to cut out six 5×25 mm shaft keys. I stuck the aluminum round shim into the bore hole, slid the key into the keyway, then the motor shaft slid in with what I thought was a snug fit. After taking one back apart, I think the keys are too short which introduces additional backlash where the motor shaft can wiggle in the gearbox.

All that to say, my shortcut is sloppy crap and I need to make proper couplers!

More Sound
For years I’ve been using a set of Logitech speakers that have been great. I had to move my gaming PC across the house to run the sim rig and I left the speakers setup at my desktop station. The only speakers in the sim rig are the ones in the VR headset (I’m using an Oculus Rift S). It’s so-so. It works, but I can definitely hear the motors (mostly the backlash clacking) over the headset. Smooth motions are quiet, but the motors are definitely audible during fast or bumpy movements. The crappy motor couplings that I’m using right now are probably a big source of the noise. The rig really needs a set of external speakers to help overpower the clunking and provide the low-end with a subwoofer.

Wind
Maybe it’s just the design of the Rift S, but anyone with a VR headset knows that wearing it for a while gets HOT. Like not going to burn your face off, but just in general sweaty hot. I’ve never had a chance to try out any headsets other than the Rift S and the Oculus DK2 that I had back in the day, so if it’s just poor headset cooling due to design and other headsets fair better here, then I’d gladly consider switching (cost permitting of course). Ultimately I’d love to upgrade to a Valve Index since I also like to play other VR games like shooters and whatnot and the knuckles controllers look awesome! But I digress, back on topic.

I currently have a cheap desk fan constantly blowing on the driver but would really like to upgrade that to a 2-channel wind simulator. From what I’ve seen it should be as easy as taking another microcontroller (running basically the same serial data parser code) and regulating two PWM outputs to control the fans. It wouldn’t take long to slap out a PCB housing another Pi (since I already have parser code from the control boards), a 12–24 V power input, and a couple n-MOSFET (or maybe NPN BJT) transistors to switch the fans.

Since I’ll be making another interface board for this— and since the Pi has dozens of GPIO pins — I’d like to support other peripherals on the same board, or maybe on version 2 of the board. Since I’m exclusively running VR, I don’t need any dashboard devices, but I like the idea of adding a monitor to the rig so that VR-incompatible people can still take it for a spin. Given that, maybe I’ll add extra outputs for shift LEDs, or a 7-segment RPM gauge or something. I don’t know yet. But wind fans are on the list!

Handbrake and Sequential Shifter
This is becoming a more desired addition. I’m looking to come up with a handbrake and sequential shifter (two separate units) that I can either 3D print or machine (or a combination of those).

I’m pretty sure I’ve seen code out there already to turn an ATmega32u4 (the microcontroller in the Arduino Leonardo) into a USB game controller. I could take that and wire up a couple momentary switches for a sequential shifter. I’d probably have a separate USB connection for the handbrake, just so they’re fully independent — feels cleaner that way.

The handbrake could be sensed with a potentiometer, hall effect sensor, or maybe a rubber pad pushing on a load cell. I think most of the handbrakes I’ve seen people use actually have a real master cylinder and they’re reading the hydraulic pressure on the back. I could go that route too, I’d just have to do some research to find a suitable pressure sensor and master cylinder.

Wheel and Pedal Upgrades
The Logitech G27 has treated me well for ages, but now that this is getting a bit more serious (compared to sitting in an office chair at an office desk), my friend has been thinking about picking us up some Fanatec hardware, starting with a new wheel and pedals. I’m not mad at the G27, but it’s getting old hearing the gears clack away over certain feedbacks like rumble strips and whatnot — that’s really my only complaint with the wheel. The pedals could be a bit stiffer, especially since it’s finally bolted down, and I’d also like to invert them so the pedals ‘hang down’ like most in cars. I stiffened the pedal springs a long time ago, as well as added a cork pad at the back to dampen the clack when the pedal gets slammed down.

I’m also curious about the feel of one of those load-cell brake pedals that I’ve seen. Up until my buddy mentioned looking at Fanatec gear, I had been envisioning a DIY set of pedals, potentially with a load-cell brake.

Wrapping Up
This post has gotten WAY longer than I anticipated! I think that covers most of my thoughts, efforts, hurdles, and progress in this build thus far. I don’t usually post on forums much and I thought about making a blog post about this instead, but I’ve seen so many other rigs on here that I figured I had to share.

It’s been a fun build and a fun ride! Looking forward to making it even better!