.....if we start with an elementary proposal , intended to obtain a continuous gravity unbalance(torque difference) on the same side of the fulcrum/support , at :
www.geocities.ws/iacob_alex/Some_Drafts/text048.jpg
This design has a single problem , in the top point :we need a "self" tumbling/switching/reversing of the hanging arm , so to maintain the gravity unbalance on the same side of the fulcrum and the restarting unbalance ( torque difference) for the next full (360 *) "self" rotation(s).
Let's have a look at :
https://www.researchgate.net/figure/Conventional-Savonius-Rotor-available-on-wwwenergythiccom-8_fig1_277339136
.....but imagine a "particular" shape of Savonius (UNEQUAL arms...) attached to the yellow mass. Now , can we
test and hope for a "self" motion (rotation), due to a continuous gravity unbalance and a...Savonius "assistance" ?!
Al_ex
.....proposal , as you can see at :
www.geocities.ws/iacob_alex/Some_Drafts/text048.jpg
.....was tested for a continuous accelerated rotation for 450*(360*+90*) as self moving , due to a top starting gravity unbalance.
A full self sustaining rotation (360*) is sufficient...but storing the rotational inertia,can we "remake" the top gravity unbalance and ...the "self"rotation ?!
So easy to test this model...
All we need is to imagine a manner of "assistance" ...or a "self remake" of the top starting unbalance , so to obtain a continuous process , due gravity free fall and the in-out rotational motion storage
Al_ex
I cannot call this a true "success",
But to achieve a full 360 rotation, - the offset mass
This is a nice bar to set!
Kudos
As a possible solution, an Archer Quinn sliding offset may help
There are 2 moments of inertia, where the linear force is perpendicular
Opposite directions, one has a vertically upwards vector, the other down.
Not "up", but an angle that the force of momentum causes and outward force that
can exceed gravitational acceleration, and the mass can "rise" freely.
The other pulls down with maximum intensity. It is not 6 o'clock with your set-up,
as you can see, but slightly to the direction of rotation