So I've been considering this.  The only way something can fall is if it is higher than it started....
from equilibrium/stop condition, the first motion at the top actually raises the weight slightly.
gravity acts not on an entire body... but is really about the center of mass....




https://www.youtube.com/watch?v=uiyMuHuCFo4
https://www.youtube.com/watch?v=rCw5JXD18y4


*thinks* ... need some better sketches and models...
-----
Edit:
I'll have to dig way back into memory and pull out some calculus I guess...
as the top moves (rotates is the simpler model), X degrees, then the center of mass is lifted, and then falls at an angle and creates a rotation as it falls back to its original height.   This path is always longer than the height it was raised, so it must take a longer time, therefore it's accelerated by gravity for a longer amount of time than the time which it took to raise the weight in the first place.


if the angle at the top of the shaft is very slight, then less work is applied to raise the weight, though it then has a much shorter distance to fall... but the angle of incline should resemble something like a cylinder rolling down a hill accelerated by gravity.  So it's forward motion will be sin(slope) * m * g...   But; that would only work if slope < 90 degrees... if it's vertical m has no bearing.... so then is it really a cylinder on a slope?  (feather and hammer drop on moon).
https://www.youtube.com/watch?v=_mCC-68LyZM  (okay force is not acceleration)  so a heavier object gets more force applied


(distraction; not applicable: hmm a long chain falling out of a beaker goes up first... like way up over the beaker 
https://www.youtube.com/watch?v=_dQJBBklpQQ
http://www.nature.com/news/physicists-explain-gravity-defying-chain-trick-1.14523
https://www.youtube.com/watch?v=QTC3bKea2Yo (longer chains from higher heights)
)


Hmm... well due to inertia, the mass falling vertically must have the same force being applied if it's non vertical, and something like the cos(angle) vertically and sin(angle) around axis or rotation... but then to get to the bottom it must apply that force for a longer time than it took to raise the object in the first place, resulting in a net acceleration of the object greater than the initial impulse....


search for "gravity acceleration tilted axis -accelerometer -digital -sensor" didn't reveal any useful math... so maybe again it's a case I don't know what physicists would call it?


or 'acceleration deflection gravity around tilted axis -accelerometer -sensor'

i just wanted to write something like that

the easy way to emagine john device is to emagine a huge weight that is rotating on a nail
starting easy,and picking up speed it would seem like the output is huge .like he claims,
but in reality he never counected output becouse it would only stop the device.

Is there any information from Mr. Skinner that he used an elliptical path for either the upper drive system or the lower drive system?

In the video I only hear the announcer saying that the weights go round and round.

In a newspaper article about him and his device, he describes the weights as rotating 'in a circular motion'. I think some have mistakenly re-interpreted his comment regarding 'eccentrics' to mean elliptical. But eccentric simply means off-center. The circular rotation of the upper drive system is mirrored in the lower circular rotation of the unbalanced weights.

Thanks,,

no ellipse then,, not to say that an ellipse could not be used,, just that it does not look like Mr. Skinner used one.

Agreed. Also interesting is the fact that he states that the weights don't actually fall (but are simply rotated to a new COG by the angle change).

If they don't fall, they don't rise. Based on this as well, there would be no reason to depart from a straight-forward simple rotary motion. Elliptical would seem to be an unnecessary complication.

If you consider your first diagram, and even though B is not fixed to A, rotating A causes B to rotate in a horizontal circle. There could be some initial raising of the lower at startup, but once it is rotating, centripetal force will keep it rotating in a flat horizontal circle.

I dunno about elliptical... the top is actually driven by flat levers that go back and forth,

 Where is the evidence to back up this claim?

 
In Luc's video, link on page 6, post #84 I can clearly see the rotary motion of the driving arm. I see no side to side motion of the pivot point ether.

 
Confirmation that the top of the rod is circular motion is the obsevation, as shown in my video...

 
https://www.youtube.com/watch?v=u1dvSINRylk

 
that the angle of view in two views 90 degrees to each other is the same. If the rod was moving in a flat plane then the rod would move back and forth in one view and towards and away in a view 90 degrees to the first. I show this clearly on page 14, post #205 where the angle is nearly identical in the two views that are at 90 degrees to each other. The only logical explanation is the the top of the rod is moving in a circle.

 
William shows us how the device works at the point where he stops the lower weight and advances the upper weight, When he does this the lower weight rises. When he releases the upper weight the lower weigh falls and pulls the upper weight back into its rest position. In other words, when the machine is in operation loading the lower weight lifts, retards its position in relation to the upper weight and the weight is now wanting to fall to its lower position, thus outputting drive to the load.
Ron

the weights are in circle orbit but the driving top rod has best results driven in elipse,
but even in elipse you got the problem of output power ,
when you gain output the input gains  by far,
the problem is in setting the weights wright