Very nice. I don't have the right phone or I'd try this myself. Good luck.

In fact, all the data were collected with a wired PT and the Crr calculations were done with a modified Adam Haile spreadsheet. It is just recently that I switched to a wireless PT and could use the CdaCrr app (for the 2 following points: SS23/GP4000S and CorsaSpeed/GP4000S)

I said also the size of the circle is a representation of the precision, it's a mistake, cause with a CV of 5%, it means the precision is +/-0.001 so the circle radius should be twice (circle=latex tube, square=butyl tube for at least one tire).

---

Blog | Twitter| Bike CdaCrr app

Interesting.

I have a Sony phone with ant+ and now have a wireless powertap, a match made in heaven.

I'm going to give it a shot.

Excellent. As you know, all of my early work was with a wired PT. I occasionally wonder what might have happened if I had been using a Stages at the time I did those first runs. Are you in contact with Adam, or is this an old old spreadsheet?

When I first came up with this I thought the precision was going to be poor so I didn't want to spend too much effort doing things carefully and exactly. I thought the errors in the data would be large enough to make that effort wasted. Basically, I was surprised it worked at all, thought it was kind of a cute party trick, and that getting into the right parameter neighborhood with the estimation was incredibly neat. Adam and Tom Anhalt showed me that what I thought was a limitation of the Power Tap's data collection was really my lousy experimental technique (damn them). So I had to go back and do things more carefully and systematically. Then I treated the estimation like a real statistical problem rather than a half-assed back-of-the-envelope hack, and was surprised at how reasonable the results are when careful data collection and careful calculation is done. So that's what I do now: I treat the problem like a real statistical estimation problem.

As I indicated to someone else, if you want to try, some advice:

-short loop with same asphalt
-try to keep the same trajectory on each lap
-try to ride straight and turn slowly to avoid higher rolling resistance because of steering effect
-no wind (at slow speed, wind is a precision killer)
-precision/repeatability is better by riding with low CdA and same position (so I used my TT bike in aero position)
-if you have a PT, keep coasting between 6 and 15 kmh, and accelerate at the same location of the lap
-if you have a crank based powermeter, never coast, keep pedaling slowly
-repeat, repeat, repeat to get confidence in your protocol
-....

---

Blog | Twitter| Bike CdaCrr app

It is the old old spreadsheet. I stopped using his original idea (regression method on a run with slow and fast laps), but the spreadsheet was very good for a lot of purposes and I changed it for my needs. The Crr results I showed are based on the virtual elevation method applied lap per lap, where CdA is supposed constant, and thus Crr is the unknown.

---

Blog | Twitter| Bike CdaCrr app

\

Couldn't you just buy a cheap Android phone and use it. This isn't apple, you don't need service to actually use the thing.

---

---

Save: $50 on Speed Hound Recovery Boots | $20 on Air Relax| $100 on Normatec| 15% on Most Absorbable Magnesium

Blogs: Best CHEAP Zwift / Bike Trainer Desk | Theragun G3 vs $140 Bivi Percussive Massager | Normatec Pulse 2.0 vs Normatec Pulse | Speed Hound vs Normatec | Air Relax vs Normatec | Q1 2018 Blood Test Results | | Why HED JET+ Is The BEST value wheelset

Looks great, I may have to check this out. Earth-shattering news that the SS tested very well ;)

Btw, what PSI did you use for the test, what weight and what rim?

---

---

Save: $50 on Speed Hound Recovery Boots | $20 on Air Relax| $100 on Normatec| 15% on Most Absorbable Magnesium

Blogs: Best CHEAP Zwift / Bike Trainer Desk | Theragun G3 vs $140 Bivi Percussive Massager | Normatec Pulse 2.0 vs Normatec Pulse | Speed Hound vs Normatec | Air Relax vs Normatec | Q1 2018 Blood Test Results | | Why HED JET+ Is The BEST value wheelset

I don't see the ANT+ chip inside this phone.

---

Blog | Twitter| Bike CdaCrr app

116 Psi (8 bars).
Crr is a mass independent coefficient.
Kinlin rim (width 19mm) for the wired PT, and Kinlin rim (width 24mm) for the wireless PT but Crr is rim width independent? Tom A. will confirm.

---

Blog | Twitter| Bike CdaCrr app

Can't you just get a dongle?

---

---

Save: $50 on Speed Hound Recovery Boots | $20 on Air Relax| $100 on Normatec| 15% on Most Absorbable Magnesium

Blogs: Best CHEAP Zwift / Bike Trainer Desk | Theragun G3 vs $140 Bivi Percussive Massager | Normatec Pulse 2.0 vs Normatec Pulse | Speed Hound vs Normatec | Air Relax vs Normatec | Q1 2018 Blood Test Results | | Why HED JET+ Is The BEST value wheelset

Right but you are going to use a different PSI in different width rims. In addition, based on how wide the tire is you are going to have adjust PSI as well. Ideally testing the same PSI on different size tires is going to lead to a better PSI fit for some tires than others. Not an easy way to do it, but ideally you would test say your tires in 5 psi increments and show the lowest RR point. 116 psi might be the optimal pressure for a 20mm but might be two much for 23mm or 25mm

---

---

Save: $50 on Speed Hound Recovery Boots | $20 on Air Relax| $100 on Normatec| 15% on Most Absorbable Magnesium

Blogs: Best CHEAP Zwift / Bike Trainer Desk | Theragun G3 vs $140 Bivi Percussive Massager | Normatec Pulse 2.0 vs Normatec Pulse | Speed Hound vs Normatec | Air Relax vs Normatec | Q1 2018 Blood Test Results | | Why HED JET+ Is The BEST value wheelset

Yes, it may work if the phone has OTG. I wonder which is the cheapest ANT+ phone on the market.

---

Blog | Twitter| Bike CdaCrr app

I've had the chance to look over the app and ride my new to PT wheel and gather some data.

This is essentially automating the Chung VE method of testing? Sort of like aerolab in an app?

I have scouted out a perfect half pipe stretch of road that I can test on in prep for doing some VE testing but I think I'll give a try first for kicks. My main goal is to test out just how much difference in CdA I'll have with my S1 in road position riding the drops VS forward seat post, lowered stem, and clip on aero bars.

A few comments:

1. Those are some low Crr values.

2. The lack of any apparent temperature dependence is interesting (but consistent with what I have found).

3. Also interesting is the fact that your between-day (I assume?) CV isn't much smaller than what I have previously reported http://www.trainingandracingwithapowermeter.com/...esults-part.html?m=1). This suggests that limiting your speed to try to emphasize Crr over CdA doesn't really help very much.

Yes, it automates (Chung) VE method. You will find exactly the same results with Aerolab, but with the app directly on your phone just after the field test. As an example, the app produces comparison between altitude (from phone barometer) and virtual elevation. For the three laps of the run given below, there was a constant headwind (~1-2 km/h) in the first half and the results were perfect (CV~0.3%). Very rare, generally you will have higher variability and will have to repeat a lot.




We may suppose there is something like a CdA difference between 0.01 and 0.03 for the two configurations tested. You should see easily this difference with VE method. More difficult with two aero helmets or two skinsuits. Good luck in your tests.

---

Blog | Twitter| Bike CdaCrr app

Thanks for your comments.

1) I was surprised too. But year after year, session after session (3 different PowerTap used), checking multiple times the algorithm, I see no reason why the Crr values are so low (except a CdA estimate too high, which it is not the case here). I can post data.

3) The CV is relative to Crr variability between laps of a single run (one point on the graphic). As you said in your blog, precision of Crr field testing is ~10 times less than roller tests and riding at small speed does not improve this. But separating Crr tests (at low speed) from CdA ones (at high speed) may have practical reasons: a long loop for aero test at high speed to have also a training benefit, a short loop for Crr testing just after to recover during few minutes.

2) The previous graph with only the Supersonic tires (20mm front/23mm rear) and the two lines indicating a 1.4% variation per degree for Crr, it just shows that more data is needed: in the 5-15° range:

---

Blog | Twitter| Bike CdaCrr app

1. Have you done any roller testing to compare with your field results? I did all of my tests using an SRM, so the absolute values I have reported are dependent on the accuracy of the chain drive friction losses I have assumed. However, the difference or offset between road and rollers for my data are very similar to what Tom A. has found, and he uses a PT. I therefore don't think my numbers could be too far off.

3. That's a good idea. Field testing is obviously quite time consuming, so being able to have it serve as training as well would make it less onerous.

2. I still don't see any real trend in that data?

1) I have no roller. GP4000SII Crr on perfect asphalt is near 0.005 according to Silca recent results (https://silca.cc/...stance-and-impedance). May be my PowerTaps give lower power values than real ones near zero... Or something else like auto-zero behaviour (I remember Adam Haile said to not use this feature for field testing). Or still a big issue in my protocol. The only fact that gives me confidence in my data is that I got 0.004-0.006 values on some other roads, so in my (limited) range of tests, there is big variation of Crr (*2) according to the quality of the asphalt. I implemented recently into the app, a first attempt to measure road vibration (standard deviation of the vertical acceleration component given by phone sensors) and to log the average of this metric for each Crr lap. We will see if it adds useful information (correlation or not with the quality of asphalt and/or tires pressure).

2) Me too.

---

Blog | Twitter| Bike CdaCrr app

Road surface is definitely a wild card, but OTOH the road I used was relatively lightly traveled and had been repaved just a year or two before I started using it. Still, perhaps that plus other effects add up to account for the apparent difference.

About point 1), I recently read this (in German, but the picture is interesting) about Crr variation with speed, and as roller tests are generally done at high speed (~45 km/h for Tom tests), I suspect that with the protocol I used (VE riding at low speed between 7-12 km/h with given constant CdA to estimate Crr), we only get the first coefficient Cr1 of a Crr law which will be described as:

Crr=Cr1+Cr2*BikeSpeed

I missed may be a thread here on ST about Crr speed dependency... Is it something well known ?

---

Blog | Twitter| Bike CdaCrr app

I have not seen speed related variation in my roller tests. However, there has been a lot of discussion of "impedance" coming from surface roughness on real world roads:

https://silca.cc/...y-and-previous-works


Some believe this impedance increases when tires are inflated beyond a certain "breakpoint" pressure.They also suggest that the impedance increases (more than proportionately) with higher speeds. I guess you can think of the "bounciness" of going over rough road increasing as you go faster. For my part, I am skeptical about the whole impedance proposition. But even if you accept the impedance proposition, it's conceivable that the bouncing could actually diminish at higher speeds. Ever drive over dirt-road washboards at 50mph? It's a lot smoother than it is at 25mph.

I have spent the last few weekends trying to field test the impedance & pressure breakpoint concept, doing hundreds of laps at different tire pressures. Per a suggestion from Tom, I have tried to test at fairly high speeds (20-30mph).

All I have determined so far is that the Chung method will not work for this. We're looking for Crr differences of maybe 3 or 4 watts. But at 30mph, even subtle differences in your aerodynamics-- an elbow or a knee a little further out, different cadence, a slight puff of a breeze -- completely overwhelm differences of that magnitude.

So I retrieved my old Alex Simmons spreadsheet and will have a go at regression testing instead.

---

My latest book: "Out of the Melting Pot, Into the Fire" is on sale on Amazon and at other online and local booksellers

What you say is confirmed by the graph where Crr decreases above 40 km/h as if bouncing effect diminishes.


Indeed, not easy. The guys, if I understand correctly, measure Crr of a recumbent bike (and 3 tires instead of 2) where CdA is much more constant than a rider on a bike. But there is still the influence of wind.

---

Blog | Twitter| Bike CdaCrr app