A three steps experiment...

[TinselKoala]

I replied to your comment on the video page, but I haven't gotten to the PMs here yet.

Real data doesn't "go against my beliefs" at all... if it is reliable and actually tests an hypothesis. The video wasn't a "real" experiment, it was an illustration of an experimental method. The numbers from that video aren't reliable, as the controls weren't tight enough. There is bearing wobble, the release is done by hand, the heights of the weight aren't exactly equal from trial to trial, there aren't enough trials in each condition, etc.

Mond's original design included some latches that were designed to hold the magnets in position at some point during their travels, and I don't have that system in this video. But there is another video of this same apparatus where I use an idea given to me by JK (of the rotating plasma in the cup) as an alternative latch system that works pretty well. And it has even been plagiarized! The ripped-off version of the latch-testing vid is appearing in the related videos list on the right side of this vid's page.

Unfortunately this apparatus is not available to me at present, having been left behind with a lot of other neat stuff in the other undisclosed laboratory location. Maybe.... some benefactor will send it along to me here, eventually, and I'll be able to retest it.

I recall that getting the magnet tubes smooth enough for reasonable action was a little difficult, I had to chuck the wheel in a big milling machine and carefully ream-hone the rough holes to a smooth internal finish with a fluted chucking reamer and, finally, small cylinder hones. Lots of WD-40, my favorite lube for machining acrylic, was used up in the process. (It smells better than kerosene, which is also a good lube for acrylic machining.)

ETA: Here's the ripped-off version of the latch-testing video. I can't read the Cyrillic writing so I could be wrong, but I don't see anything that looks like it could be a credit to "tinselkoala" in there. Can someone translate?
http://www.youtube.com/watch?v=AHyRGv7gseo

And here is my original:
http://www.youtube.com/watch?v=WcPuKv9Z-XE

ETA: A further refinement that should be done is to re-do the control baseline runs after the experimental manipulations. That is, one should do "A-B-A" rather than simply "A-B". If the second baseline runs aren't the same as the first baseline runs, this means that some other "third variable" has changed over the course of the experiment, like maybe  the bearings working loose. Checking the baseline after the experimental condition allows the experimenter to make sure that this hasn't happened. Even better is "A-B-A-B". With 20 trials in each condition, this can easily eat up a whole afternoon. Tedious, but not difficult.

[mondrasek]

FWIW, CLaNZeR also did some playing around with the "mondrasek wheel."  You can find his work here:  http://www.overunity.org.uk/cmps_index.php?s=62b4ff7ded5a272c8a5aaad8ea997fb2&pageid=mondrasek

I don't advocate anyone spending time working on that exact design just because I did and/or TK and CLaNZeR were kind enough to build models themselves.  That exact design was based on a false premise and could never work as presented, AFAIK.  The various builds showed even more complications that further detract from even being able to achieve unity.

M.