Showroom Grigory's 6DOF #2 -- compact and no linear actuators, 220V servos

Base is complete, motors installed.




I have also connected the servo drivers and did some initial tests yesterday.

The servo drivers are controlled by step/dir signals and I plan to use Arduino Mega to generate such signals based on the target position values sent from PC through serial. Yesterday I verified that a single Mega is capable of simultaneously generating steps for all six drivers with frequency sufficient to reach the motors’ maximum speed, but that was without processing any serial communication or keeping track of individual servo positions. I should be able to update the software and add this part today. Hopefully, Arduino will still manage it.

I made a small mistake when ordering a box for the servo drivers and had quite a surprise when this monster was delivered:



Anyway, now I have plenty of space to store the electrics. Probably, this can also accommodate all the 24V stuff, including the two batteries.

I'd love to hear more about how you use the Arduino / Mega to control the servo drivers.

Well, its quite simple.

The driver is controlled by dir and step signals. Rotation direction is determined by whether dir is HIGH or LOW. Amount of rotation is determined by a number of step pulses.

The driver has a setting for the "electronic gear ratio". This defines how much the servo should rotate on a single step pulse. This is important since my motors have 10,000 encoder pulses per revolution and can rotate at 2.500 rpm. Without electronic "gears", reaching maximum speed would require issuing pulses at over 400kHz for just a single motor. I used 1:20 setting in my tests yesterday, so the servos move 20 encoder pulses per each step received.

The arduino sketch for the test was very basic. Each loop did this:

1) reverse the level of all six DIR pins
2) repeat 5000 times:

a) set all six STEP pins to HIGH
b) wait for 4 microseconds
c) set all six STEP pins to LOW
d) wait for 4 microseconds​

All changes to pin levels were done by port manipulation, not by digitalWrites, which proved to be too slow and did not allow reaching max rpm.

With my 20:1 planetary gearboxes this made all six output shafts oscillate by 180° with really high speed. I tried decreasing the wait times to 2-3 microseconds, but that caused servo driver errors, as the servos were not able to keep up. I believe that proves that the step pulse frequency is sufficient and I even have some margin to increase it.


My actual setup to control 6DOF is planned to be like this:

[simtools] >network> [my VB.Net application to calculate correct 6dof lever positions, adjust rotation axes, etc.] >serial> [arduino] > [drivers]

Upon receiving data from SimTools, my VB.Net application will calculate the six lever positions and send six integers to Arduino. Arduino will get these target positions, compare them with the current stored positions for each motor, issue a corresponding number of dir/step signals to each driver and store the updated positions based on the number of issued pulses.

I think you need to make provisions to allow canceling of current pulse stream command if new position demand arrives, as the motor will try to perform all pulses received. This means the motion will not be real time.

Just saying as I had same issue with pulse/ burst mode when trying to do same thing with clearpath servo motors. It worked marvelous, but it was not real time...

Thanks. I will keep that in mind.

I updated both the Arduino sketch and the Vb.net app and connected everything to Simtools. It seems to work well, at least at this stage.

This is one of the motors reacting to my random inputs in the Simtools output tester:



I am really impressed with the speed of this thing!

Where are the motors from? Price?

I bought both the motors and the planetary gearboxes from the local Russian reseller, but originally they are from China.

The motor model is 90ST-M04025, geaxbox -- PLF90 1:20 ratio. I do not know how much these cost in China and I understand that I hugely overpaid for getting them locally without hassle with delivery and customs.

Motor like this can be bought around 250$ as a kit (motor + servo amplifier) on aliexpress + shipping + taxes.

Nice monsters by the way.

Glad to see you got them moving. Looking forward to seeing them loaded though since unloaded doesn't really show much. With all the power you say they have I wouldn't trust those clamp on levers holding good however. But maybe they will be okay with the load spread over 6 of them. On my 3dof they wouldn't last long though.

Here it is:

I initially tried to use Simtools output tester while sitting in the cockpit and moving the slider with the mouse. Not the best idea



The second attempt after adding a shaking test routine to my interface app:

I hope the levers will be fine. They have slots for the shaft keys, so I do not think they will slip.

@Grigory im really intrigued by the geometry of your platform. Are the front two servo arms reversed? How did you calculate the offset in your design? It seems to have allowed you to have a wider opening at the front with greater clearance for your legs.

Yep, the front two levers are indeed reversed. I was trying to make the rig as compact as possible and that helped to have wider movement range without having to increase the footprint.

I did not actually need any offset. All motors are evenly placed around a circle, facing outwards. All levers and rods have the same length. As the cockpit pose is determined by the distance between motor shaft and the points where the rods are attached to the moving platform, it does not really matter which way the lever is facing. If that distance is eventually correct, I mean. It’s like I have invisible linear actuators

I have a vb.net app sitting between Simtools and the servos. That app knows the limits of movement for each DOF, gets 12 bit values for each DOF from Simtools and can convert that to offsets in mm for sway, surge and heave and degrees for roll, pitch and traction loss. Then I apply 3D transforms to the coordinates of rod attachment points to understand where these points should be in the target cockpit position/orientation. Finally, I apply the method that I explained in the post for my first 6dof to get the required lever angles.

With such approach, I can really have all my lever and rod lengths different, motors placed not in a perfectly symmetrical manner — and still have correct cockpit movements.

In addition, I can place the rotation axes anywhere I want. This is, for example, the traction loss test. You can see that the rotation axis is at the pedals, not in the center of the rig:

Having resolved the RF interference problem, I finally had some time to properly test the rig.

I have not yet connected the G-seat and shakers, but pure 6DOF seems to be working quite well in Assetto Corsa and Prepar3D.

Previously, when I tried using VR on my first 6DOF rig or on the seat mover/traction loss setup, I got motion sick almost immediately. This time – not even a smallest sign of that! I’m not sure if that is because of precision and speed of movement or just because of the practice that I had playing VR on the stationary rig while waiting for the parts to arrive. In either case, I really like how it feels, especially with motion cancellation.

The motors are handling the loads just great and without any major speed loss. I am actually not sure if connecting the shakers to the seat will add anything to the feel of road texture.

Temperature also does not seem to be an issue. With my compact design, the front two motors are more loaded than the other four. The weight of the pedals, Accuforce wheel and my legs is shifted forward and its projection is always outside of the motors hexagon. But even with that after three hours of play I could easily hold my hand on these front motors. I believe this means they were not hotter than about 45°C. The motors have Class B insulation, so they can handle internal 130°C working at 40°C environment. I think I now have some margin.

I modified the software that I use to convert 6DOF values from Simtools to the cockpit pose, so that to be able to adjust the ranges of movement for each DOF. Now the software also allows to adjust the balance between linear movement and tilting in surge and sway. Combination of the settings available in Simtools and the ones that I now have in my software provide quite a lot of flexibility.




I have not yet spent much time refining the settings, but I already see that only very limited movement ranges are needed when using VR. This would be even more so when g-seat is re-connected and takes care of longer-lasting sway and surge forces. As can be seen, I limited the heave range to less than 100mm or just about 50 mm each way from neutral. Even with that, I truly regretted having a bit excessive vertical speed when touching down a Cessna.

Generally, I like how everything works now. No serious design flaws have come up and it should be possible to make things even better after more tuning.

I am wondering what the simplest solution is to show both the game and video from the webcam showing the rig in action. Geforce experience can do that, but the webcam window is way too small and I could not figure out how to make it bigger.

@SilentChill, what software are you using for your videos?

Thanks! I will try that

Here's how the rig looks in action. I certainly need to improve my driving and to replace that crappy webcam

Anyway, I just love how this 6DOF feels now with the wind and working Gseat.