well, glad to see the speed concept is generally well regarded atleast!

But i guess it shows how important the rider is to the set-up, as to what is 'fast' or not

If you're looking for nits, you have to use a fine-tooth comb; you can't use a coarse comb.

That's a lousy analogy.

I suppose an example would be if each time you put a helmet on, it becomes set in a particular position on your head to some extent, that persists across all the laps. So you could get exactly the same CdA every lap, but it might still not be the same as the mean CdA if you were to repeat the whole exercise several times, getting off the bike, removing the helmet, and starting again each time. For this reason, when I test helmets I always do ABABAB etc, so I'm taking a fresh sample of how the helmet sits on my head every rep.

^^^^ This is a very good point that applies regardless of the test method. Wind Tunnel, ERO, Chung, whatever. There is no substitute for those ABABAB changes if you want high confidence and resolution.

Great post.

Something I have noticed in my very amateur efforts in this area is how variable the wind is, and that its stronger than you might expect when field testing. I'm waiting for that windless day just to see how windy it really is. I know this doesn't necessarily relate to controlled environments, but it does suggest that field testing benefits from being able to measure and therefore take account of environment.

---

Developing aero, fit and other fun stuff at Red is Faster

I see what you did there.

One issue here is that common devices for measuring wind speed lack sufficient sensitivity.

that has influenced my choices. Using a vane for angle for instance saves having to buy multiple differential pressure sensors. Measuring on the bike also improves resolution since you're not trying to measure air speed close to zero. It remains to be seen if I can get any really useful data from this for VE analysis, but I do have a means of measuring yaw and air density in real time.

It keeps me busy, anyway.

---

Developing aero, fit and other fun stuff at Red is Faster

https://vimeo.com/16116710

Thanks. I'm guessing it's a bit more 'polished' these days!

Back on topic, what sort of resolution and accuracy is needed for environment measurement. Also is there anything other than temp, pressure, humidity, wind speed and direction that needs to be considered?

---

Developing aero, fit and other fun stuff at Red is Faster

Something that bad suggests you have dodgy power meter data. Aero testing is a really good way to find out if your meter is up to snuff.

I don't really understand this comment.

I've been able to reliably and consistently provide aero impact data from indoor velodrome testing for years.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

Quality portable equipment exists to give you air density information at sufficient precision for relative changes.

Wind sucks when outdoors though and air movement that affects results is below the threshold of most portable measurement devices of a reasonable price. Which is why indoor tracks are far more suitable test venues. Nevertheless, outdoors it just means precision is lower, not that testing is of no value. That only matters if the difference you are seeking to tease out is less than the level of precision possible given the conditions. Collect enough data with good test protocol and you can still do good things.

I check power data quality, air density and speed measurement is made using the same wheel. Air movement is as benign as an indoor velodrome can provide.

For the three indoor Australian tracks I have tested at (DGV Sydney, DISC Melbourne, and Speedrome Perth), air density does not vary all that much through a test session even though no track here has any form of climate control. Main influence is temperature changes which are typically pretty gradual indoors, and keeping an eye on any doors/docks opening. The air pressure measurements are auto recorded with as much frequency as needed and changes are accounted for as testing progresses. I'll record at 5-min intervals typically.

If a rider can't hold their position consistently for very long, then there is something fundamentally wrong with their position. That needs attention first before worrying about aero. I suggest people be well fitted first before aero testing, and have at least some understanding of the range of biomechanically sensible/rideable positions.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

(Directed to rchung)
I'm not sure why you didn't get an answer on this, but I've tested a few people in wind tunnels and done a fair amount of field testing using a variety of methods - so I'll share what I've seen.

The sessions in the tunnel I've helped with start with repeat runs.
My experience in the tunnel (where both axial force and tunnel q are directly measured and rider position is monitored and corrected as opposed to some indirect measurements employed in field testing) has the average cxa repatability in the 0.5 to 1% range. Without referring to my database I'd say the worst is in the 2% range.

My field testing across runs is in the 1-2% range for cxa and 5-10% for crr.

I'd suggest that the typical user trying to do their own field testing won't achieve the level I've experienced

Furthermore, if one is trying to achieve statistical significance in any of this stuff using field testing, it will be a piece of work... but this shouldn't discourage anyone from trying... after all, the purpose of experimentation is insight, not numbers.

Good luck and happy testing!

---

=================
Kraig Willett
http://www.biketechreview.com - check out our reduced report pricing
=================

Nice public service post :) The expense of wind tunnel testing makes people avoid repeats that they feel are unnecessary. Unfortunately, when your test articles produce a lot of run-to-run variance, it means that a prudent scientist would repeat the test point until statistical significance is achieved. Often, the single run, with perhaps one confirmation run is assumed to represent the mean value, when in fact it could be anywhere in the distribution. Sometimes this variation is larger than the differences between test articles.

Pope's "Low Speed Wind Tunnel Testing" has a nice section on employing statistical analysis to estimate how many independent runs are required before one can say they have converged to a mean that is within a chosen confidence interval of the "true mean," but I've only seen tests employ this under the most unstable conditions.

As you mention, the companies testing may only care enough to test until a particular run repeats in their favor. Hasty generalizations abound when money or ego are on the line.

Jim, thanks for taking the time to give us this level of insight and information. I have a related question I'm hoping you can answer.

I've seen the claims of drag reduction from aero road bars, but I'm having a hard time believing the figures. Have you done any testing on the track that reveals any significant drag reduction compared to regular round road bars?

I understand narrow would/should be lower drag, but just can't wrap my head around the magnitude of the claimed reduction from the shape.

---

_______________________________________________

Oh, dear -- yes, it is huge! 30% reduction, sometimes. Most of it comes from the bars allowing you to position the rider in a more aero configuration.



Have a look at this video, then:
https://www.youtube.com/...h?v=ftq8jTQ8ANE#t=16

(Thanks for the original video link, Alex Simmons!)

---

AndyF
bike geek

You mean like this?

http://www.tririg.com/...omega_whitepaper.pdf

That works for me.

I understand that people would feel that there's more value in testing more configurations in a given amount of time, but my take on testing is that if you don't have confidence in your data, you may as well not gather it. It's something of an all or nothing proposition in that holes in logic created by not understanding the permutations of combined changes leave enough room for someone to introduce their own bias into the data analysis. Granted, it's not always a complicated matter, but when refining at levels near to tunnel repeatability, how one handles the test and the data is the difference between making improvements or spinning in circles.

"Know fewer things, but know them better". Sounds like a good plan!

---

AndyF
bike geek

---

=================
Kraig Willett
http://www.biketechreview.com - check out our reduced report pricing
=================

Not sure what you're trying to point out. My response to Andrew Coggan was directly referring to the reported statistics in the paper he linked. The existence of systematic/deterministic and stochastic sources of error are nothing new or up for debate; the former has more to do with sloppy test methodology or modelling, or imperfect knowledge of the system, and the latter is inherent to the nature of turbulence, sensors, and certain dynamic systems.

Discussing systematic errors is a bit too messy for a casual forum since you can't truly assume linearity in aerodynamic testing. In practice, this means that you have to be very deliberate about how you design your tests.

I agree; rchung's comment is germane.

---

=================
Kraig Willett
http://www.biketechreview.com - check out our reduced report pricing
=================

I work(ed) in finance and ran performance testing for very time sensitive software algorithms. Similar things going on.

A machine has an OS. OS is doing stuff what the OS is doing change.

Network switches have firmware. Firmware responds different to different loads. Can even slow down or speed up a little based on the temperature of the room.

Servers have memory. CPU has a pipeline. And caches. So so so so so so so many things going on. Literally way too many things to ever deal with all of them.

Things that helped my performance testing that can maybe help bikes?

1) Do many independent tests. Easier in code than bike.. but take a pike of bike parts.. set it up, test, tear it down. Repeat 20 times for each bike. Throw out the best and worst 5. Average the middle 10. You can hopefully find some kind of steady state.

2) Change 1 variable at a time. Try the above with different saddles. Different wheels. Different foam people. Do it many many different times.


The reasons why this isn't done is I suspect due to cost. I had a lab with a ton of servers, I could test 100k combinations for 20 reps each over the weekend, and draw data from it. How much would 100k wind tunnel runs cost? It would give you better results. It would be expensive.

Hey Andy - i think he was asking about aero road drop bars. And I believe you're referring to aerobars on a TT bike.

I'd be interested in the original question. If that's what he was referring to.

---

"One Line Robert"

Yeah, I was referring to aero road bars. Bonus points for data when on the hoods, drops, and ghost aero

---

_______________________________________________

No idea who produced this (maybe tour magazine?) but it's all I could find, thanks jackmott.


http://www.aeroweenie.com/...a/tour-road-bars.png