Login required to started new threads

Login required to post replies

Help with aero data
Quote | Reply
I did my first field testing yesterday and just looking for some input to make sure I'm doing things correctly in Golden Cheetah Aeorlab. I have my testing parameters below as well as some screen shots. I've read through as many threads here as possible but now looking for some personalized feedback. If anybody has time for some feedback it would be greatly appreciated!

Course - 3 loops for a total of ~2.04km. Each loop has some sharp turns that causes a bit of coasting but no braking. Did my best to hold the same line on each loop. Has incline and decline that allows for variable power. No wind.
Bike - Speed Concept, Garmin 500 (GPS off), PowerTap P1 pedals, DuoTrap for speed.

Goal was to test my usual 2XU sleeve-less jersey to a new Castelli Stealth sleeved jersey. Also, tested road (Giro Atmos) versus aerohelmet (Giro Advantage 2).

Results (assuming I did things correctly in Aeorlab)
I did 3 tests with each setup and then ran them a second time.

2XU w/aerohelmet - test one 0.258, test two 0.2602
Steath w/road helmet - test one 0.2669, test two 0.2695
Stealth w/aero helmet - test one 0.2683, test two 0.2628

Curious if the repeats for each test should be closer in value?

I have screen shots of Aerolab below. What I'm doing to get the CdA is measuring the elevation at the beginning of the blue line by hovering over it and then sliding the CdA until the end of the blue line matches the beginning elevation. Once they match I assume the value on the CdA slider is my value. Is this correct?

I have screen shots below. The first is of the complete test (I was hitting Lap at the beginning of each test). I then broke it into separate rides in GC. The second shot is data before adjusting CdA. Third shot is after adjusting CdA slider to get the beginning and ending elevation to match. I then did that for each run.

If this is all accurate then I'm kind of disappointed that the Stealth would be slower than the sleeve-less. Also the aerohelmet doesn't appear that much better than my road helmet. Or perhaps my data is all crap and I wasted and hour!


Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
The repeats give you some insight regarding variability, but you may also have drift due to changes in posture, PM zero, changes in temperature and wind, or whatever.

It's best to only consider two options and do A-B-A-B-A-B, if you want confidence in your data... or confidence that you need to do more testing.
Quote Reply
Re: Help with aero data [rruff] [ In reply to ]
Quote | Reply
This

OP could consider saving time changing things by doing A B B A A B.

Developing aero, fit and other fun stuff at Red is Faster
Quote Reply
Re: Help with aero data [rruff] [ In reply to ]
Quote | Reply
Thanks for the feedback.

I did all the testing with in 1 hour, but it was at the end of the day and was getting close to dark by the time I was done. It didn't seem the like the temp changed but I also wasn't monitoring it. Would a couple of degrees difference make have an affect?

I think my posture was the same each time put with a few tight corners on each loop (3 loops per test) perhaps I didn't hit each one the same. I may have to find a better test sight.

Should I zero the PM before each run?

Final question. Did I come up with the CdA correctly? Matching beginning of blue line to end of blue line?
Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
Many things can make a difference. Temperature is just one.

Yes, zero your PM.

The "trick" is to have a good course and protocol, correct for all variables you can, and get enough data to average out variables you can't.

For instance if changing configurations is a pain (and often it is), try A-A-B-B-B-B-A-A. Then go out on another day and do B-B-A-A-A-A-B-B. Or if you don't want to switch at all during your ride (like if you just want to do some CdA loops while training) that can work also, but you will need many days of data.

The low points on the graph are just as important as the high points. Match VE to itself, don't worry about your barometric elevation.
Last edited by: rruff: May 31, 16 12:55
Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
I'm relatively new at this as well, but so far I've always thrown a few sanity checks in. With my sanity check being interjecting a few laps riding in the bullhorns between each trial. It should stick out like shit in milk in the data. It makes a nice control, and also gives me a break from riding in a super strict aero position.

For what it's worth my own testing of road helmet vs. aero helmet showed crystal clear wins for aero helmet. Dramatic and repeatable differences in CdA. This is an outdoor velodrome so I do have a nice repeatable course, though subject to some wind (which I average out by doing 10 laps per trial).
Quote Reply
Re: Help with aero data [trail] [ In reply to ]
Quote | Reply
Thanks All.

So based on my results I have above, are these numbers too noisy to make an accurate choice on what is the best setup? Or do I need more / better data?
Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
I'm certainly not convinced by your comparison. You need to establish your protocol first and demonstrate your margin of error. Do a bunch of laps on different days with the same configuration and position, and see what the variance is. Don't change anything, and account for all the variables you can. If there is a course you can access whenever training, that would be ideal.
Quote Reply
Re: Help with aero data [rruff] [ In reply to ]
Quote | Reply
Is there an acceptable margin of error that most folks go by? Couple of percent for the same setup?
Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
It depends entirely on how much accuracy/precision/resolution you need.
Quote Reply
Re: Help with aero data [Burhed] [ In reply to ]
Quote | Reply
With the exception of your last configuration, your results are under 1% change (the last being just over 2%).
Considering the apparent lack of rigor in the test methodology and protocol, that's pretty good IMO.

For reference, relatively expensive load cells (measure force) are spec'd to 0.5% of reading.

If you think you need better than 0.7% (in the case of the first). Remember this one thing: IT'S A BICYCLE.


...or, you could use a t-test to determine if the difference in the means (of the configurations) are statistically significant to a desired confidence interval. Unfortunately, you'll probably need a handful more data points for each test configuration. If the results are not statistically significant, then that means configuration A, while the mean is a lower CdA than configuration B, is not actually different.
Quote Reply