So this weekend, I had a try at using multiple hill runs to decouple
CdA and
Crr as detailed on pg 43 of Robert Chung's
Estimating CdA with a Power Meter presentation. The conditions weren't exactly
perfect but the data that I retrieved from it look pretty decent for a first attempt, (at least to my untrained eye!). I read through this entire thread beforehand (as well as numerous other resources) and didn't see this particular protocol discussed that much, so I thought I'd share my experiences and see what you experts think!
Test conditions:
- Test course was 1.70 km with 87.6 m elevation gain; average gradient 5.2%.
- Rider: 5'7" (170cm) and 148 lbs (65 kg); weighed accurately with bike beforehand.
- Collecting data with a Quarq DZero power meter, Garmin 820 and Garmin GSC-10 with GPS off.
- Riding road bike on the hoods (i.e. typical climbing position), held the same position for all runs.
- Tight fitting winter clothing.
- Winds conditions were extremely favourable; essentially windless and well sheltered.
- Road surface conditions were typically British, slightly rough chip-seal and wet surface.
- Running 25mm GP4000s IIs with latex tubes @80 PSI.
- The first two runs each had a couple of instances of car-passing but because of the hill, the speed differential was low and they passed relatively wide (~1.5m).
As a quick summary, I did four consecutive runs, alternating high-speed/low-speed/high-speed/low-speed; the high-speed runs were @285W (~21 kph) and the low-speed runs were @200W (~15 kph). I calculated the
Virtual Elevation from these data in Excel and when I put in what I thought would be reasonable CdA and Crr values there was a clear separation of the high-speed
vs low-speed runs, but each equivalent run aligned pretty well. I adjusted the
CdA until the four runs aligned as best I could and then I adjusted the
Crr to match the known elevation gain of the climb. The resultant VE profile for the four runs looked like this:
Before tweaking
CdA and
Crr values (CdA = 0.350, Crr = 0.0060):
Click to enlarge.
After tweaking
CdA and
Crr values (CdA = 0.218, Crr = 0.0084):
Click to enlarge.
Apologies for the picture quality of the graphs; I could make the lines thicker and more visible, but then they all just obscure each other. The striking thing is that the CdA seems way too low and the Crr seems somewhat high; I guess I have a few questions that I was wondering if you guys could share some experiences on:
- First and foremost, was the speed differential in these runs (15 vs 21 kph) sufficient for this kind of testing? I could've hammered the fast runs harder but wanted to focus on maintaining a consistent and stable position.
- Were the instances of car passing likely to have ruined the data from the first two runs? I tried just analysing the second two runs (which were cleaner in this regard) and whilst the profiles aligned much more closely, the resultant CdA and Crr values were very similar to above.
- How much rolling resistance could you expect a wet road surface to add compared to a dry one?
- Does tweaking the "known" elevation gain of the hill only affect the calculated Crr? The CdA is determined by the alignment of the runs at differing speeds, right?
- I haven't accounted for drivetrain power losses downstream of the Quarq at all; how significant might this be? What might be a good way to account for this, a fixed value of power loss (e.g. subtract 10 Watts), a fixed-percentage reduction in power values (e.g. multiply by 95%) or something else?
- What other factors might have produced an artificially low CdA and high Crr?
I'm sure I can improve on this given experience and better conditions, but I thought I'd see if you guys can help me work out where my biggest sources of error are coming from.
I really appreciate all the hard work that you guys (Andy Froncioni, Robert Chung and Alex Simmons in particular!) have put in to developing these methods and especially for documenting them in such an easy to digest and replicate fashion. Thanks!