In Reply To:
"
the tests are performed over a range of yaw angles."
if this is the case, then the speed at which you ride your bike does matter, and 30mph is not going to give you the same results as 20mph "scaled up." furthermore — and maybe you can speak to this — i'd be more comfortable with some idea as to the protocol. were the results obtained at 5 degrees of yaw on the way toward 10 degrees, or on the way back down toward zero, or with some admixture of each?
I think this is where some confusion is coming in...the speed of the tunnel is chosen for a multiple of reasons, the most critical being the sensitivity of the balance being used. Higher speeds give higher forces and it's easier to "tease out" the differences between setups. In addition, 30 mph seems to be somewhat of a "standard" in the bike component testing arena. This makes it easier to compare drag values (although personally, I prefer to compare CdAs vs. the actual drag forces since it's basically wind speed and air property independent over the test ranges).
So...since the drag force is proportional to v^2, this can be used to figure out what the drag force AT A PARTICULAR YAW ANGLE will be at 20 mph. It's merely scaled by the 20^2/30^2 = .444
edit: The power, as reported in the chart, will be scaled with the cube of velocity, 20^3/30^3 = .30, or 30%
However, as you point out, when someone is travelling at 20 mph, for a given course and wind direction, they are MUCH more likely to be exposed to a higher relative wind yaw angle. So, as was pointed out above the bell curve weighting centered at zero yaw may not be the best way to evaluate the differences. In fact, it may be better to have higher yaw angles (e.g. 10-15 degrees) weighted more than zero yaw. This is apparently how Zipp evaluates wheels, both their own and competitors. Another option would be to keep the bell curve centered at zero, but "flatten" the weighting, so that higher yaw angles are weighted higher relative to zero.
I think it would be interesting to rig up an instrument that measures apparent wind angle on a bike and have a "typical" triathlete or Master's TTer ride some events and see what the true apparent wind distribution looks like. It sounds like Zipp (and perhaps Hed) have done this...but maybe only with elites?
So...to answer your questions. At a PARTICULAR YAW ANGLE, the drag force results at 30 mph DO scale down to 20 mph...it's just that at 20 mph the "most likely" yaw angle the wheel is exposed to is not necessarily going to be a bell curved centered on zero, like it would be for someone travelling at 30 mph. And if it is a bell curve, it may not be as "tall". That means that the weighting used to make up a chart like the one above should be different at 20mph vs. 30mph. Therefore, yes... I agree that it's not really valid to just take the weighted values above and scale them down to 20 mph. Make sense?
However, that doesn't mean that the test results aren't useful in making general comparisons between wheels. I just wouldn't put a whole lot of stock in small differences between wheels close together each other on the chart. Change the weightings, and they just might swap positions. Then again, like I said above, you can change the weightings all you want, but I still think the 2 endpoints (808 and R-Sys) would remain the same ;-)
Lastly, as far as the details of the protocol go...all I know is what is revealed in the link above. Judging by the data charts, it appears that the increments of yaw were at most 5 degrees, and may have been less. I don't know for a fact that they went 0-35-0 either, just that Jordan stated they did above.
http://bikeblather.blogspot.com/