4-post 3-DOF rig issue with leg lifting

My rig is a 4-post 3-DOF rig based on the Departed Reality 500mm version, driven by Motion4Sim and FlyPtMover 3.5. The cockpit is a Track Racer TR160.
The riding position is set at 250mm - 50mm = 200 mm.
The goal was to get lots of heave - for flying GA - and have workable simulations of acceleration, yaw, roll, sway.
The goal has been met as all sensations are pretty believable. And yes, you can do yaw because a pilot is not at the COR of the plane. On yaw he is thrown out of the plane to the back ( GA ) or to the front ( Larger planes ). This can be simulated with roll and pitch.
It is a kind of an off-beat rig in the sense that you almost never see a 4-post rig with actuators having such a large range.
That being said, I see no reason why it should not work, and it does, up to a point.

I get the maths behind the fact that 3 points are always in a plane, and 4 not necessarily so. You have to perform calculations to get 4 points in a plane, but with reasonable tolerances, it should not be impossible.
The dimensions of my actuator corner-plane are 1090 x 665 mm. These dimension have been put into FlyPT.

The problem I am having is that on more extreme pitch and rolls, one of the front actuators ( where there is less weight ) comes off the floor, and not a bit. There is still range on the actuator, but it's just in the wrong spatial position.
This also make my rig twist, not a good sensation.
This problem is not present on GA aircraft, as there you constantly have the sensation of being in a floating tub, but when you get to planes like the PC-21 it's there very quickly.

Any thoughts would be more than welcome!

*** Edit : The best thoughts turned out to be my own thoughts as the problem existed between keyboard and chair. The cause of the problem was a limit on the output value I had set way way way before the platform ever was properly used, to limit the amount of movement in the upward direction, with a crop output function in FlyPT.
Unfortunately, this limit was only 20mm. You do not see this reflected in the 3D viewer, that is why it did not occur to me sooner. The value I physically measured on the actuators was alway this maximum of - 50 driving position upto 20 max = 70mm.
This round value got me thinking and sure enough a cropping filter on the output value was defined with an absurdly low max value.

Unfortunately, this is a symptom of a 3dof using 4 actuators

A 3dof only needs 3 actuators - can you fit the 2 rear ones together in the middle?

As long as the front ones are far enough apart, it will remain stable

Is there a video of the issue for us to see?

Thanks for the answer but I kind of refuse to accept this explanation. Even with basic high school math I can rotate a rectangle in space and project from its corners a line-segment onto a fixed plane, where the line segment would be orthogonal to the rectangle but not the plane. The length of those projections determines the values that a driver has to output. Now, in reality you could be off by a couple of mm, and that is acceptable, the flexing of the frame would absorb that.
As for moving actuators closer together, I don't think that FlyPTMover allows for non-equally spaced actuators in a 4-post rig. In other words there are only two dimension you can define.
Theoretically I could make it a 3-post rig, but technically that would be quite challenging to do and cause tipping issues. I don't want to consider that option until I really have no other choice.

Move the 2 rear together and move them the same ie as one actuator

or just loose one actuator

There has to be a problem with the maths in FlyPT (You may be able to adjust the dimension in FlyPT to get it to work)

BUT at the end of the day - a 3dof only needs 3 actuators not 4

have you tried using simtools only without using FlyPT ?

I think, combining only the upper joint of the rear actuators instead of running them fully in parallel makes more sense.

This is the configuration of the D-Mover shown in the ad at the top right of every page in this forum. It not only avoids the rig being mechanically over-constraint. Additionally it adds the forth degree of freedom for free: yaw, well, not much but at least a little bit.

This looks indeed like the ultimate solution for such a rig. It would require a lot of extra hardware, but looks like a better option than 3-post.

Edit : now that I have seen it in action, it does not allow for much roll or pitch, since it is semi-constrained in the front bottom pivots.

Hacking the dimension in FlyPT would be a valid option to try. In practice, if it works, it will get a lot worse before it gets better

I have not tried any other motion simulation software besides FlyPT.
I am still enjoying the fact the combination of all my work actually functions, including motion compensation via OXRMC.

Well if you only have 3 actuators I think it wouldn't be worth the money and time. For general aviation, 3DOF does the job fairly well. You don't have much side-forces in an airplane. Even badly coordinated turns (slip/skip) can be simulated quite well with roll (and no sway). Only cross-wind landings where you hit the runway in a crab-angle would require yaw and sway. I believe it can be dispensed with.

But if you already have 4 actuators it doesn't cost much, so why not make the best of it.

My current FlyPT profile actually simulates all the forces pretty well. Sway into Roll, Surge into Pitch, and Yaw into a Pitch/Roll combo, Roll into inverse roll, etc....
I did edit my original reply to the DMover design to say that I think the DMover setup is too constrained in roll and pitch.
For example, for rolling a plane, you need to inverse roll the cockpit. In reality a plane rolling to the left will push against you at the right, even from the very onset of the motion, so if you roll the cockpit in reverse, you can have continuous roll simulation. But to do so, you need quite the range because planes can roll at very different speeds and the roll DOF is also needed for sway and yaw. So having just a touch of actual platform roll is not enough.

Thanks again for the all the help, I have edited my original post to show the answer. I did learn some things along the way, mainly that properly testing your rig with static positions beats using dynamic wave signals, which hide the occasional lifted foot. I will leave you with a picture of the platform before the fix.