I go to Taiwan for 1 week and I miss all of this!!! Damon drops a single word analysis of the protocol.. Tom and Robert reminisce about Joaquin and Shambolic and Dan drops a wicked SNL reference.. feeling so far off the back!!
I did share some backchannel chatting with Hambini a few weeks back and he seemed reasonable in that, but ultimately his answers to my logistical questions don't hold up in my opinion. Taking an alternate approach, my questions were form the direction of how can you possibly have put a live rider in a tunnel for a 26 minute protocol x 20+ wheel combinations, and even assuming that's possible, how can you possibly claim 2.5% error when that really isn't possible once there is a human involved. He claims they did it, but from my experience this just isn't possible for a number of reasons, the big one being that humans are human.. getting a few minutes of clean and repeatable data can be hard, so getting 20+ blocks of 26 minutes feels impossible to me and there's no way to claim that level of repetition as we've shown that something even as simple as a rider getting cold over the course of a tunnel test can change their position enough affect data way more than 2.5% making repeat runs not repeatable, so claiming that level of accuracy and repeatability over weeks of half hour tunnel sessions doesn't hold up IMO.
I have to say that I love the concept of using a sort of wind tunnel duty cycle to look at a wide range of situations and flow characteristics..I also love and can appreciate the need to add more 'real world' type flow imperfections, however, it seems like the biggest benefit of doing this would be to look at the boundaries of the handling piece of the equation which as we've discussed in the 'Josh and Slowman Debate Handling' thread, is something that we know some stuff about, but really not all that much when you get down to it.
I liked ruff's question:" What are your thoughts regarding the premise of Hambini's testing protocol; that small oscillations in the flow field and transient yaw movements (more closely modeling outdoor riding conditions) can have a significant effect on the drag and stability results? "
As Robert and Tom pointed out, the current tunnel protocols have succeeded in producing CdA graphs that can be used to predict real world event times within a few %. I have no doubt that we are collectively still missing some terms from our equation, but the reality is that what we are missing must be orders of magnitude smaller than what we already know and likely have more of an effect on handling or other factors which are not captured in pure watts/velocity/time calculations. Again though, if our current models are giving us predictions accurate within a few %, then we have to believe that whatever we are missing at most represents those few % and nothing more on the drag/power side of the equation and if there is any significance it would be more in the handling equation.
Also, greenplease, rruff, trail and nealhe all had good questions about flow measurement and adaptability. Zipp started designing for higher yaw based on some models that assumed a bi-modal distribution of yaw that looked more or less like 2 bell curves meeting at zero and ending at some peak yaw value... we used to have all this on the zipp site which is now pretty thin on info.. anyway the original assumption from the late 90's early 2000's was that if peak yaw for a day was say 30degrees and you were riding a loop shaped course (not out and back) then your distribution would be a bell shaped curve from -30 to 0 and another one from 0 to +30 so you would have peak probability somewhere from say 10-20. This was somewhat validated with early GPS as we started overlaying weather data (with reduced wind velocity to account for ground effects) onto course data. By 2004 we were pretty accurately predicting TT times with this model for CSC.
As riders get faster, this peak yaw decreases, but we realized in our development that the best rim shapes would lie on the curve generated by a disc wheel.. so the goal became to mimic a disc to as high a yaw angle as possible. Then companies started doing some public testing using whirligig style probes and showed real world yaw angles were distributed much lower and closer to zero than our modeling had shown. We then compared whirligig type instruments to the Alphamantis Aerostick and realized that there was a smoothing/damping effect to wind angle with this type of probe. With the aerostick we saw that wind angle changes can and do happen in steps (I think Kraig Willet posted something like this) but the whirligig smooths these changes into curves as it has to traverse a range of angles to make the change. So there is a sort of central limit theorem effect in play with that device as at say 30-60 Hz is ends up recording a lot of transitional yaw angles that you don't see with something like the aerostick.
http://www.SILCA.cc Check out my podcast, inside stories from more than 20 years of product and tech innovation from inside the Pro Peloton and Pro Triathlon worlds!
http://www.marginalgainspodcast.cc