Enigma brings up a valid observation.
A suggestion, Ray: That particular aspect is easily addressed in re-design by completely enclosing the spinning elements so that no external aerodynamic forces are potentialized. Since both extension arms are static there are any number of ways to do this, from individual plastic spheres to entirely exclosing part or all of the apparatus. For that matter, a million thoughts on how to achieve isolation from aerodynamic influence is now crossing my mind. I would choose whatever least expensive option for now, as any option should be equally effective.
Your suggested variation which incorporates varying surface coefficients is interesting.
I also looked at your link with regards to compensating for friction, but at this point, my opinion is that it is not necessary and actually might complicate matters.
Just throwing thoughts out now:
First of all, I consider the effect worthy of further study, because the device uses and expends energy, also, that the "effect" is less than the energy used to create it. This tells me that the device is not asking "something for nothing"
Of course, this doesn't mean that the effect is separate from normal causes or holds any special promise, it means that the effect should be investigated, in my opinion.
Ray, I have closely examined the concept behind the extension arms with the motors and off-centered balsa wood, and apart from the possible aerodynamic effects noted by Enigma, I have also noted another prevalent set of dynamics going-on: very rapid cyclic compression and extension forces placed on the support arms(frequency roughly equivalent to the motor armature)
When the balsa wood is rotating away from the support arm the support arm experiences an extension, however slight. When the balsa wood is rotating towards the support arm, the support arm compresses. In fact, the effect is not linear, and so the support arm will also slightly bend left and right as well as in and out during this off-centered rotation of the balsa wood.
In addition to the above dynamics, and as noted in your web site, a constant torque tension is being placed on each arm, due to the fact that the end of the arms support the motor housings, which produce opposite rotational torque with respect to the motor armature and vectored on the center of system mass.
On top of all this, the center of system mass dynamically alters by virtue of the rotating balsa wood, and this alteration is non-linear. That interests me.
Ray, in summary, I would propose pursuing complete isolation of aerodynamic influence as presented by Enigma. That seems a healthy first design variation.
If the effect continues to occur afterwards, then we can address the more problematic issue of designing a test which eliminates friction-walking as a potential.