Many thanks Hans!
... and Tinu out of this thread. 

My model only ever intended to show what happens in the system with various arrangements where no actual work is being done and where the movement is unrestricted.

But in my model i found that if you restrict the movement, you need even less power to keep the flywheel rotating.  Like 2 N only.

Wow 44 pages and still going strong I may as well throw my twisted opinions into the mix.
I see some odd things happening in Milkovic's machine that we should maybe consider more carefully concerning the force of gravity, centripetal/centrifugal forces and relative velocities. Imagine yourself freefalling downward at terminal velocity, we could say you are weightless and as such the force of gravity would be zero because you are no longer "accelerating", the force of gravity is an accelerating force. When standing still on earth the force of gravity is maximum as there is no acceleration involved, so the force of gravity is dependent on relative velocity in the vertical plane of motion. In the picture below, on the lower right is a half circle representing the motion of a pendulum, in mid-sectors 1 and 3 we could say the pendulum will undergo maximum accelerations both negative and positive in the vertical plane and in sector 2 minimum accelerations in the vertical plane thus gravity must act with maximum force at points 1, 2 and 3 as represented by "G" (red line) in the graph, the letter "C" represents centripetal force pulling downward/outward on point (B) of the balanced lever. We can see in the graph that maximum forces (G-C)peak at point 2 moving the lever end down at B down.What is odd is that in the upper diagram in sectors 4 and 5 the lever end "B" is rising increasing the potential of the pendulum system this potential released near the bottom of the pendulum swing near point 2. The increased potential must be paid for but from which force?, the centripital forces act in a predictable manner through the arc but the force of gravity would seem to peak just as this potential is paid for at point 2. If we could imagine holding lever end "B" and pulling upward where sectors 1-2 and 2-3 meet we can visualize that the pendulum would stay in motion, but which force provides this timed pull upward? It would seem only the mass "A" can and the force would be the force of gravity on a mass with little vertical velocity but great mass and inertia. The mass "A" must be periodically reset to the top position to lift "B" so we come back to the same question again, is it force "G" or force "C" that lifts lever end "A" through the motion of the pendulum pulling on "B"?. Judging from the quite erratic motion of the mass "A" in the video and the predictable centripital force graph line I would guess the de-acceleration of the pendulum mass on the vertical plane leading to an increase in gravitational forces has some role to play here.
This simple setup is very misleading, there are some very complex timing and force issues involved that I can't quite put my finger on yet.
Best of luck

@eugene-g
The problem with your setup is friction, friction reduces motion, motion is produced by force---so there must be an equal force wanting to keep something in motion.

I have also made ?Two-stage Mechanical Oscillator? someway (but not exactly) similar to Milkovic's, to test his "12 times overunity" claims, but my measurements in one very limited scenario, that I was able to measure, showed coefficient of efficiency equals to 0.9! That's far below 12!

But I was unable to do all tests or to construct the device by the best way...

I am an amateur (Java programmer) and I do not do or construct mechanical devices...
I have past one exam in Theoretical Mechanics on college so my knowledge of physics is limited. But Physics was my hobby during elementary and high school.

I have used my bicycle to replicate ?Two-stage Mechanical Oscillator?.
Pedals were balance beam. One pedal had pendulum made of wood and bob and other pedal had a larger (and heavier) bob (to simulate longer and heavier end of balance beam with hammer-head)...PS: The "problem" with this construction is that the both ends of balance beam are the same length...

I have made just one scenario (measurement) test by filing the system with energy JUST ONCE at the beginning.

System was fed with energy by lifting the pendulum just once so I have measured input work: A = m*g*h.
My premise was that output work is equal to mechanical-heat energy made by impact of the large bob at the other side of balance beam.
And that output (mechanical-heat energy) is equal to Sum of potential (E=m*g*h) energies  of the large bob. To mesure that sum I had to count the number of oscillations of large bob and quickly to read its achieved heights on the ruler fixed behind such "oscillator".
I have repeated the same scenario (measurement) test for about 10 times and used average measurements.

So I had coefficient of efficiency equals to 0.54 (54%)in this particular test with this particular device.
I made theoretical approximation for constant oscillation mode, based on previous test, of maximum coefficient of efficiency with my model, and it was 0.9 (90%)

I think I have also noticed things about energy transmissions like Mr. Hans von Lieven speaks about but I didn't put much attention, tests or thought about it...I did this tests more than 5 years ago...(PS: at Dec. 17. 2002.)

I don't know why this design wasn't used. If it produces 12 times the energy used then it should produce 12 times the power used to fire a coil to repel a swinging magnet. The posted picture is just one of many ways this could be utilized if it indeed does work as it is said to work. I think if it was timed right, it should put out a continuous flow.
 It would only take one pulse to get two revolutions. It's like a two stroke motor but with two compressions instead of one. It should be very easy to build.

shruggedatlas, "I believe the principle of the Milkovic design is that it produces 12 times the energy used, but only if there is no load.  If you try to put a load on, it immediately reverts to underunity mode."

 That doesn't make sense. As long as the load didn't equal more then 11 times the swinging force, it still should still work and be considered over unity.

So what is your conclusion ?

Are you saying WM2D does not work ?

Also you have designed  something different, as your pendulum is not swinging,
but indeed your results with the left upper example are quite interesting !

Maybe you are using the centrifugal forces to get maximum force at the lower position of the moved
weight on the disc ?
Maybe this will be a method to use centrifugal forces in such a system ?

I think WM2D does works.

As you said, i had the conclusion that using centrifugal forces will allow us to apply a even smaller force to get maximum force at lower position, and the smaller force will be enough to keep the flywheel rotating.  I know will only give 1 pull per revolution, but that's not what i think we are testing here.

All 3 systems shows that the system is giving (the way i did) something like 17 times the input power, but i guess is because of the way the weights are arranged.

Because of that, Hans suggested that the "well balanced system" was what allows me to lift the 50kg weight, and even some bird fart will, but i did the other 2 system to prove he's worng.

About some other theories about restricting the system will make the pendulum loose power i will say that you (my dear Stefan) are right, because you pointed some post ago, that restricting the system is even better.

I found, the less the system moves, the less energy requiered to keep the flywheel to rotate.

I think Hans is right about energy transfer, but i think this system is a way to gain some energy from gravity trough centrifugal pull of pendulum instead of the torque gain (like torque wrench tools).

I will not get into controversy about all the physics or mechanics (sorry my english ) on the system, why not or why yes.

What i cannot test with w2dm is the generator part.  Using sproket system to rotate really fast some generator, instead of the 50 Kg weight to see if i can produce more than the 5.6 N i have in my system.