Repeatability and the San Diego Wind Tunnel

A few years back, I wrote this sucker:

http://www.biketechreview.com/aerodynamics/uncertainty.htm

whoa, that's probably the third time I've used partial differentiation in a setting outside of academia!!! :-)

To be honest, my experience as a consumer product development engineer in the bike biz (2+ years) and the golf biz (coming up on 10 years now... holy cow!) is much more caveman based than what doing partial differentiation might suggest.

Granted, the tools/toys I get to play with in the golf biz these days are a couple steps above the beer cans and bits of string I got to use in the labs I had at my disposal while in the bike biz... ;-) ...but, still, I think that even with the fancy tools I get to use these days, I find myself relying on my caveman instincts when it comes to judging "goodness" of data that so many of the high tech gadgets can spit out.

Y'know, folks will have to assess for me how well their methodology and instrumentation they have used can repeat a given measurement/setup condition within a day and across days before it gets my attention. For example, with a pedaling rider in the tunnel, I have seen things (axial force) repeat to within less than 10 grams...but I've also seen things not repeat so well. Over the years, and more than a thousand runs with pedaling riders, I've grown to know how much I can trust what the tunnel is telling me...and that knowledge drives the way I choose to test.

The same kind of familiarity with repeatability is helpful when placing "equipment only" wind tunnel test #'s into context.

Not sure where I'm going with all of this, other than, I don't think folks think about experimental uncertainty enough - especially when it comes to doing field tests with a power meter. "Subjective validation" of these data might kick in if you wind up getting the answer you were more or less looking for:



Anyway, speaking of repeatability, I was checking out some additional repeat data I have on a specialized trispoke (the same one linked to in the partial derivative link above) today. I've tested this exact wheel/tire combination in a couple tunnels (texas a&m and lswt.com). I've tested the trispoke in the san diego wind tunnel eight times since 2005 (yeah, that would be over a four year time period) at a beta=0 flow condition.

What was the standard deviation of the multiple runs over that 4 year period for the exact same wheel/tire setup? 3.8 grams of axial force at 30mph. That seems pretty good to me. What do you think? What can the other tunnels do over that same four year time period in terms of "equipment only" repeatability?

So, yeah, that std deviation tells me about how well I can trust the data coming out of the facility here in San Diego over time. My caveman instincts are comfortable with these data out of San Diego! :-)

La Vuelta: Final TT

Don't know if anyone has been following the Vuelta this year, but I have. I thought those consecutive mountaintop finish stages were top notch. Great drama, and interesting to see how the big doodz approached those final climbs.

Anyway, the final TT was pretty exciting! Great fun to see the final results of the penultimate stage of the vuelta and the limited "still" photos on the interwebs.

I was stoked to see Samuel Sanchez (who has tested at the wind tunnel here in san diego with me during the winter for the past couple of years) get a great result in the final TT, and a super finish on the final GC (y'know - I think Samuel is the one person I've spent the most time in the tunnel with over the years - Orbea/Euskaltel invest pretty well in their GC riders when it comes to TT performance):



Also, from my perspective, it is really interesting to see how final TT stage winner David Millar (with whom I've tested with here in san diego) has evolved things over the years. It's wild to see how he has been influenced by various folks over the past three years - in the end, though, he proves that "faster is faster" - glad to see him finally win a stage in a grand tour.

Here's a pic of David and I when he tested with Saunier Duval a few years back:

Amour de "Low"

I'm not sure how to spell "love" in french - it's Amor in spanish, and Amore in Italian, that much I'm sure of, and so, well, I reckon it's probably close to "amour" in french...

Anyway, after watching today's ITT in the tour, I'm pretty amazed at how much of the peloton is demonstrating behaviors symptomatic of the "disease of lowness". By that, I mean all the guys who wind up riding the tip of the saddle, only to shift themselves back on the saddle every 5 pedal strokes (I don't know what was more painful - how Contador felt during the ITT or how I felt while watching the DVR'd coverage a few minutes ago). The funny thing, for me, after seeing first hand how reach and drop interact in a wind tunnel with a wide cross section of athletes (from elites like Kristin Armstrong, Sarah Hammer, Phinney, Hincapie, Astarloza, Leipheimer, Popovych, Danielson, Marchante, Simoni, Millar, Sanchez, etc.. etc... to masters National champions like Ruth Clemence, or Alpenrose kilo record holders like BTR member Snigelmannen - way to take the record from Marty Nothstein! - to IM folks like Sindballe, Evans, Andersson, Fuhr, Ferguson, Major - to chubby, amateur, wannabe time trialists/IM'rs named Kraig) is that this disease has a cure...

The cure is simple, and it's called raising the bars in order to decrease the drop. huh? I mean, everyone knows that if you want to be aerodynamic, you have to have lots of drop, reach be damned, eh?

The favorite refrain from the "prophets of low" is: "move the saddle forward" or "get steep" isn't it:

http://www.biketechreview.com/performance/faster.htm

Move the saddle forward and drop the bars "a little", or get "steep" is the magic elixir for the sickness of being too low, according to the pundits. Well, yeah, that seems like kind of an indirect way of solving the "bars are too low" issue, eh?

The fact of the matter is that from an aerodynamic perspective there exists a relationship between reach and drop for each individual, it's not an either/or deal...and despite what the interweb forums are full of, the UCI really isn't limiting things in the "forward" department based on my experience.

I'll use myself as an example of the "disease of lowness" - the last time I tested my TT/IM position in the wind tunnel was just a week or so after my IMAZ effort last november. During that test, I baselined my position, then looked at how reach and drop interacted. At three different bar heights, it became clear that if I "tipped it" (riding the nose of the saddle, rather than sitting on the saddle square), I was less aerodynamic than if I wasn't "tipping it"...and despite lowering the bars (more than "a little") the most aerodynamic overall position came at the highest bar height I was able to achieve - this bar height was probably a couple cm higher (or more) than the position I used for IMAZ.

These tunnel data suggest that if I were to take the advice of the "prophets of steep and low", (i.e - you just need to "move the saddle forward, and maybe drop the bars a little") well, I would be less aerodynamic and, therefore, slower. Thanks for the blanket, mantra-driven advice, but I think I'll pass, and let the beta/yaw equal to and not equal to zero data speak.

So, yeah, I can't really be bothered by all the "get low shenanigans" or "get low theatrics" the pro peloton seems to be brewing up these days. The wind tunnel here in san diego is the medicine that cured me of my own personal "disease of lowness". Keep in mind that I'm not alone with the uniqueness of how my reach/drop interact. Others demonstrate this same unique trade-off (some are listed above) of reach and drop from an aerodynamic perspective.

If one takes a "forest driven" rather than a "tree-driven" approach or process to TT setups, one just might realize that there exists a real opportunity to explore how much power one can produce (or wants to/chooses to produce in the case of IM) for the duration of their intended effort as a function of different reach/drop combinations.

I mean, if one can raise their bars, extend their effective reach, be more comfortable, be more powerful, and have the same (or better aerodynamics), well then, that sounds like a pretty good deal to me.

In the end, I'm pretty much enamored with "fast" and am not burdened by "the disease of lowness" anymore.

Kind of along these lines, I'm pretty sure LANCE demonstrated today, that once again, it's not about the bike...it's really about the floppy, un-aerodynamic jewelry hanging from your neck!!! ;-)

(and yeah, I think LANCE needs to raise his bars back to where they used to be four years ago... ;-) )

The words in that picture at the top of this page

I got a good reminder of it this past week during a discussion on the BTR forum. I made some off-hand comment about aero helmets and the stage 12 giro TT. Well, it turns out that that picture at the top of this page has got it about right.

check out the thread here if you are open to the possibility of seeing things from a different perspective:

http://biketechreview.com/forum/viewtopic.php?f=1&t=2532

on another note - hey, it's pretty crazy to think that the structures guy (check out the downtube on the giant):

http://www.cyclingnews.com/photos/2009/giro09/giro0912/PIC31528576.jpg

made the aero guy:

http://www.grahamwatson.com/gw/imagedocs.nsf/2ad5fc39030e64aa86256c8600642e1a/bbf308cffdd4279e862575bd006b9e82/$FILE/4.jpg

look silly on this stage. ;-)

Wheel CFD

I received a note from Jürgen, one of the principals of the website:

http://www.wing-light.de/

this morning. He gave me permission to host the pdf document that is distributed on his site that describes the work he and his crew have done with aero wheels:

http://www.biketechreview.com/images/wheel_simulation_wing-light_090430.pdf

I asked a lot of inane "cfd groupie" questions offline, ones that I hope he'll be able to shed some light on.

Anyway, I just thought that the work he was doing was cool and figured I'd point it out to folks.

enjoy!

-k

Aero Torque/Watts to Spin

This aero torque topic has been on my list for a long time. I've tried many different ways of gaining insight into the deal over the years:

1) reported texas a&m wind tunnel motor measurements
2) srm power meter driving rear wheels in calm air
3) coast down of wheels in a truing stand with a fan blowing
4) srm power meter driving rear wheels with a fan blowing
5) srm power meter driving rear wheels while I stood in the bed of my pickup and my wife drove around at 30mph

more recently:

6) on a really windy day (30mph gusts) I did the srm power meter thing
7) monitoring the motor (that is used to spin the wheels) power during wheel/tire tests in the san diego wind tunnel
8) using all six force and moment component data from the wind tunnel balance during same test described in #7

I've found several independent references on the internets:

http://www.recumbents.com/WISIL/MartinDocs/Validation%20of%20a%20mathematical%20model%20for%20road%20cycling.pdf

estimated two wheels need additional 3-4 watts at 40kph 6-7 watts at 50kph

http://www.soton.ac.uk/~aijf197/Wheel%20Aerodynamics/Results%20and%20Discussion.htm

estimated one wheel to need additional 5-9 watts at 40kph

http://journals.pepublishing.com/content/g1463815454723l0/

estimated a single wheel requires more than listed above (research funded by a wheel manufacturer)

There also used to be a thesis from the same university as the latter link above that included CFD work here:

http://ir.canterbury.ac.nz/handle/10092/1800

(research funded by a wheel manufacturer)

But they have since removed the document until October of 2010. The first time I followed the thesis link in maybe December of 2008, I was able to download, save, and read the paper and the rotational watts are similar to what was reported above – though, they include wind tunnel measurements (translational/rotational) for the zipp 808 and hed three spoke. The manner in which these researchers gathered their rotational power data is clever (the motor they use to spin the wheel in the tunnel is mounted on a force platform that monitors things), and does not use the data acquired from a six component balance.

Based on my SRM tinkerings, I’ve really not been able to discern more than a few watt difference between wheels – independent of whether or not there is ambient wind present (and the numbers on the plot below include drivetrain losses):



This past time in the tunnel, I made the mental commitment to reduce the wind tunnel force and moment data in such a way that it not only reflected the translational resistive force, but also the aerodynamic torque – or “watts to spin”.

I won’t bore you with math and whatnot related to how I independently approached this topic using force/moment wind tunnel data (I'm not the sharpest tool in the shed, so I reckon it took me quite a bit longer to figure this one out compared to the big dawgs at the wheel company's - and I probably stuffed something up along the way :-0 ), but it seems as if the rotational watts can account for a chunk of the whole "fastest wheel" deal. For an estimate on the magnitude I came up with you might want to check out the insider's forum for discussion/further insight.

I’m not sure what to make of this independent finding. It’s encouraging that one is able to use force and moment data to repeatably measure this aerodynamic torque value and this fact will more than likely prompt me to investigate wheels that are different than the three spoke and the 1080 in the future. I would expect wheels like the HED 90 and 808 and shallower rim/longer spoke wheels to get rotationally bad at a relatively faster rate.

So, maybe the data here:

http://journals.pepublishing.com/content/g1463815454723l0/

Is on target?

Thread #9

Back in 2004 prior to having any real rider positioning experience in the tunnel here in San Diego, thread #9 (we're at around 2500 threads these days) on the BTR forum dealt with the topic of elbow width.

Pretty interesting to read my perspective of 5 years ago - boy, that's an eye opener for me to be sure!! This stuff isn't always as "easy" as folks would like to make it seem.

Anyway, the images in BTR thread #9 didn't make it over during the site migration I did last year, but here's an interesting frontal area look at upper arm aerodynamics as mentioned in the BTR thread:



the numbers below the images are frontal area in m^2. All sorts of things are changing between images as I reached out, huh?! It would be interesting to see what might have happened if I had controlled for elbow width and hand elevation relative to my elbow when taking those pictures.

I reckon a vertical humerus would have turned out relatively worse than it did when I took those pics 5 years ago.

So, if you look at your position and you notice that you have a vertical or near vertical humerus/upper arm, you might want to consider tweaking your reach and drop in order to explore how things change with your upper arm geometry from an aerodynamic perspective.

Pics like I took back in 2004 can be insightful...but so can a full length mirror!

-k

Frame and Rider Aerodynamic Interaction

There was a thread over on the BTR forum a couple weeks ago that sparked my memory of a data set on the topic of frame and rider interaction.

So, I did a little internets digging, and finally found an online source of what I remember seeing.


(you might have to be signed in to google to access that link above, but check out page 34 and 35 if you'd like to verify for yourself)

In the mid 80's and early 90's, it seems, Chester Kyle did some tunnel tests at a couple different venues that looked at bike and bike+rider test runs. I think the idea was to gain insight into whether or not the floppy human pedaling on the bike made certain frames perform better or worse than when evaluated solo. There's not much background on the numbers I dug up out of the book entitled "High Tech Cycling" by Burke, so this could be a dog's breakfast published data set so to speak...

...but anyway, I made this plot based on what is publicly available:



so, what does everyone think is the frame(s) that, according to this data, are the least likely to exhibit appreciable frame/rider interaction?

Position Evolution Update

I resurrected my simple TurboCad program this past weekend, and went crazy trying to compare the positions I used in 2004 and 2008.




My personal observations regarding these two positions are that the one on the right was way more comfortable. I'd probably attribute this to a couple of things:

1) actually sitting on the saddle with my sit bones
2) not having to support my body weight with my arms

#2 above is also probably what allows me to relax and narrow up my shoulders via narrower elbow pads, and also maintain a better head position without neck pain - I can ride in the position on the right for 112 miles - not sure about the one on the left, though.

It seems as if when I am actually sitting on the saddle as the old-skool europeans imagined, things are more balanced, relaxed, comfortable, and in my case, much faster.

Just something to consider when you are paying for a bike fit, or setting yourself up on the bike for this years goal events.

Field Testing and Wind Tunnel Testing

Over the past several years, I've had the opportunity to do some wind tunnel testing with a variety of folks. On several occasions I've also had the opportunity to try and correlate the wind tunnel data to power meter based field tests.

Here's the field testing methodologies that I've tried in the past:

-flat, regression/Lim method
-indoor velodrome, caveman method
-dip/half-pipe chung method
-dip/half-pipe work per lap bootstrap method
-short laps-flat-ish work per lap bootstrap method
-long laps-flat-ish work per lap bootstrap method

It's been my experience with the data from the San Diego wind tunnel that the indoor velodrome has produced the best results when it comes to the academic exercise of matching wind tunnel data to indirect estimations of CxA based on power meter data. In windy conditons, it seems just a bit too easy to "post hoc adjust data" using stationary wind probes without high resolution data logging capabilities, in order to make things turn out the way one wants them to turn out.

The other methods (i.e, not indoor velodrome data) seem a bit cumbersome/time consuming and un-predictable/unrepeatable, in my experience (I had some pretty good calm Lim method data taken over many successive days that on average correlated very well with lswt.com data, but this process took something like a week of early morning test sessions, IIRC - man, I don't have the patience for that ;-) ).

It's my current perception (which is, of course, subject to change based on new reliable information) that many folks who distribute information all over the internets are using math models, or field testing data reduction techniques, that don't quite capture exactly what is going on - which I feel can lead to inaccurate results from an aerodynamic body axis coordinate system perspective.

So, basically, just as I mentioned in this forum thread:

"often times, it's helpful to pursue multiple, independent lines of inquiry when attacking a problem. Field testing is but one way to gain insight, and hopefully, this methodology (referring to the work per lap bootstrap/chung approach) doesn't steer folks unknowingly down a wrong path. That would be a bummer!"

Cool Qualitative Flow Gadget

For anyone who's got some free time this holiday season, check this sucker out:

2-d Flow simulation

yeah, it's 2-d and qualitative, and there are some obvious drawbacks to that, but it was kind of fun to play around with it. Pretty neat web application that can take your own black/white bitmap image and output some intriguing flow visualization videos.

So, I took the wind tunnel side view image from the 2004 vs 2008 post below and used these settings (which doesn't really match real reynolds number, but does qualitatively match some tuft analysis I've done in the tunnel):



to create this interesting flow visualization:




So, have a go with this online tool and let's see your results! Feel free to share here or over in the forum.


Happy Holidays, everyone!

Natural Hydro/Aero - Dynamics

What's your guess on the drag coefficent of this critter:

boxfish side view


boxfish front view

and this critter:

formula one

for reference, the drag coefficient of a raindrop is around 0.04, and the drag coefficient for a square, flat plate is about 1.2.

yeah, that's a pretty lame question I've proposed, but it always surprises me when folks talk about F1 cars and their relevance to low drag geometries -> F1 is primarily about downforce, not low drag.

12/24 Update:

Here's some great pics of various cars in a wind tunnel using smoke flow viz -> text is in german, but the pics work just fine:

smoke flow of cars

I think there's a shot of an F1 car in there somewhere, with it's Cd of ~1.2. Oh, and that square cross sectioned, angular box fish above, I think it has a reported Cd of 0.06 and was the inspiration for a mercedes concept car:

mercedes boxfish concept car