Sum of torque

[EOW]

No gravity here.

I think like that it's a good device:

The torus has a mass.

1/ Turn the red arm at w. Turn the grey circles at 0.999w. Grey circles have mass. The angular velocity of the orange torus is -w*cos(a) in the arm reference with a the angle of the axis from the vertical.
2/ I use the torus like a catapult. The grey circles can be cut at small parts when I want. Like that the torus can have only a torque on it, the red arm don't receive a torque. The small parts of the circles can be eject and the velocity is higher than before the friction with the torus. Need to recover the energy from the small parts of circles. Small parts are free to move in 3d space. A small part has it velocity from the rotation and the velocity up/down from the torus, it's an addition of velocities (vectors).

It's not really a friction between the torus and the circle it's more use the torus like a catapult. The goal is to keep no torque on the grey circles.

[EOW]

I add an image in 3d to look the orange disk between the grey torus.

The energy won is : 2*(-1/2*I*(w*(1-cos(a)))^2) + (1/2*I*w^2)) = 2*1/2*I*(-(-2w^2cos(a)+w^2*cos²(a))) = Iw² (2cos(a)-cos²(a))
I: the inertia of the orange torus around itself.
a: the angle of the axis of the orange torus from the vertical.

In practice the vectors of velocities are not perfectly perpendicular (i8.png), V2 cancel a little of V1 but V3 add a little of V1, the sum is near 0.

[EOW]

With the device of the video :

https://www.youtube.com/watch?v=Pjc4dIf1aWI&feature=youtu.be

I use 2 devices like this I turn one red arm clockwise and the another red arm counterclockwise and I exchange the torus ! I lost an energy.

Or I use friction between the disks

[EOW]

Cycle:

1/ I launch the red arm and the green arm at the angular velocity w. The orange disk turn around itself at w*(1-cos(a)) in the laboratory reference. The blue disk don't turn around itself.
2/ I use stems (like a parts of belt) for give the forces F3 and F4 to the blue disk. The blue disk increases its angular velocity around itself. The orange disk increases its angular velocity around itself too. The red arm and the green arm don't receive a torque.

[EOW]

And sure, I can do this with spring(s) no need the blue disk. Replace the stem by spring(s). There is a positive torque on the red arm. The orange disk increases its angular velocity around itself. The device must turn at an infinite angular velocity. The distance of the spring never change because the arm accelerates more and more. The spring is always like I drawn: vertical.