This method suggests the only way to lower your drag is to lower your head, no? Would I be OK in suggesting that this method is wildly inaccurate, as it ignores airflow behind the rider, which can be just as significant in terms of drag as the front profile? This method seems to me to be extremely simplistic and ignores alot of area that creates drag, for example it doesn’t distinguish between a road or aero helmet, even if they have the same front profile.
Just wondering if someone with the some knowledge of the subject would like to chime in.
The biggest inaccuracy I see is step seven. Which is basically what you pointed out. The only significance in this exercise is calculating the number of pixels and try to reduce that number through positional changes. 99% of the time this will mean less drag. In that sense it is a very simple and effective method. I agree with you that actually calculating cda from this method is ridiculous.
This method seems to me to be extremely simplistic
It’s a really interesting exercise as part of learning about drag - seeing how you can reduce frontal area via position. I would class it as a useful thing to do alongside field testing to expand your understanding of possible avenues for drag reduction. I use some very basic measurements related to reducing FA when I’m planning field testing or tunnel sessions that have proven to relate well to lower drag.
This method suggests the only way to lower your drag is to lower your head, no? Would I be OK in suggesting that this method is wildly inaccurate, as it ignores airflow behind the rider, which can be just as significant in terms of drag as the front profile? This method seems to me to be extremely simplistic and ignores alot of area that creates drag, for example it doesn’t distinguish between a road or aero helmet, even if they have the same front profile.
Just wondering if someone with the some knowledge of the subject would like to chime in.
Yeah, this method does a decent job of estimating A but not Cd and it’s the product CdA that we care about. Some tradeoffs in A can lead to lower Cd and this method can’t account for that. For instance this method would not predict that a Superman position is as fast as it actually is nor what might happen in different yaw situations.
Still at least getting A down towards a reasonable minimum can be useful but I’d still want to back that up with field or tunnel testing.
It’s probably mildly useful. If you’re in a really bad state as far as aero, it could give you easy feedback while you’re adjusting bike geometry. If you assume your Cd isn’t going to change that much while you lop 25% off of A through positional changes, then it could be a good tool as far as a “first pass” at optimizing your setup.
It’d probably be better as a webcam app - set up your trainer in front of a wall with a constant colour, put your computer with webcam a reasonable distance away to minimize the effect of perspective, and it could automatically do it.
I think the improvement of these kinds of methods is something really good. The article is interesting and provides a recipe for riders to try. Cool stuff!
Yeah, this method does a decent job of estimating A but not Cd and it’s the product CdA that we care about. Some tradeoffs in A can lead to lower Cd and this method can’t account for that. For instance this method would not predict that a Superman position is as fast as it actually is nor what might happen in different yaw situations.
Well, one of the big advantages of the Superman position is the elimination of the arms from the frontal area. So I think it would show up well in this test.
Still at least getting A down towards a reasonable minimum can be useful but I’d still want to back that up with field or tunnel testing.
-Dave
Yup. But if you know A, then you can combine this with field testing and separately understand the 2 components of CdA. It’s interesting, for sure.
I can remember people over at biketechreview doing this like 8 years ago or more. Its a way to calculate 1/2 of cda accurately. It picks up on more than just head position. Its another tool to know about but less valuable than power meter testing.
for an individual rider the R2 might be much better though
also in the small set of 3 individuals shown, reducing A never increased CdA, so that is promising.
That says that for those riders A counts for around 60% of the variance in their CdA. I think you can get closer than that with a power meter.
I’ve often thought about putting together some software that could do this automatically via a webcam.
would be fun and probably useful too.
There’s been more than a handful of folks that have talked about doing it using one or more Kinnect cameras and the SDK for it. I’m not aware of anyone that’s actually gone from idea to reality in that area though.
It seems to me that the usefulness of this is just in something comparable to steps 1-3. It is useful to use a systematic method to measure frontal area and then try to reduce that. Going through the added steps of calculating cda is a pain in the butt, though, and you won’t get an accurate measurement of cda from frontal area alone, so the added calculations don’t add anything. So why not just stop at measuring and trying to reduce frontal area, which someone can do in a precise way on their trainer at home?
It seems to me that this usefulness of this is just in something comparable to steps 1-3. It is useful to use a systematic method to measure frontal area and then try to reduce that. Going through the added steps of calculating cda is a pain in the butt, though, and you won’t get an accurate measurement of cda from frontal area alone, so the added calculations don’t add anything. So why not just stop at measuring and trying to reduce front area, which someone can do in a precise way on their trainer at home?
I’ve often thought about putting together some software that could do this automatically via a webcam.
would be fun and probably useful too.
That would be interesting. You could take a ‘before’ image with your background (hopefully very contrasty) with only the trainer. Then pop the bike on there and hop on.
It should be able to count which pixels changed from background, thus area.
The only trick would be to trigger measurement at different positions. That would be better than just a movie with a stream of pixel counts at each frame. Perhaps you could yell something at it, then it would take a series of stills, beep to signal it is done aquiring, then average their pixel counts and display them big on the screen.
At least you won’t be wrong or dependent on excellent wind conditions, location, and perfect route duplication
That says that for those riders A counts for around 60% of the variance in their CdA. I think you can get closer than that with a power meter.
Hmmm…that plot implies a Cd of ~0.4, which is much lower than the typically quoted ~0.7 number I’ve seen for an aero position. I wonder if it would be helpful to force the fit line through zero?
Hmmm…that plot implies a Cd of ~0.4, which is much lower than the typically quoted ~0.7 number I’ve seen for an aero position. I wonder if it would be helpful to force the fit line through zero?
Helpful in which sense? It would probably improve the R^2 but I suppose whether that’s helpful depends on how you’d use the information. I agree to first order this says “reduce your A and you’ll reduce your CdA” but once you’ve picked the low-hanging fruit you need a way to reach a little higher.
Yeah…after you are done, it could spit out a graph of the ride, plotting A over time. Then you could call up the video showing your lowest A.
That way, you could also plot power and HR on the same graph. Perhaps a position yields less area, but takes more HR to maintain a set wattage, so it may not be worth it.
