BikeTechReview.com

For the past 4 months I've been using a new venue for my on-road Work per Lap Bootstrap Crr testing experimentation (chip-seal galore, but no cars!).

During that time, I've accumulated 57 laps (400-ish meter laps) across days with the same tires, pressure, and same position. 31 of those laps were within 1 std deviation of both the overall mean of CxA and Crr values. Based on these reduced data, the stats suggest the following for my 120 psi runs:

CxA =0.351; 95% confidence interval = 0.004; CV = 3.3%
Crr = 0.00672; 95% confidence interval = 0.00011; CV = 4.7%

for my limited testing using 90 psi in these tires I have 7 laps total with 3 usable laps:

CxA =0.357; 95% confidence interval = 0.024; CV = 5.8%
Crr = 0.00591; 95% confidence interval = 0.00037; CV = 5.5%


at 140 psi I have 7 laps total with 5 usable laps:

CxA =0.349 95% confidence interval = 0.014; CV = 4.4%
Crr = 0.00609; 95% confidence interval = 0.00026; CV = 5.0%

What's my conclusion? This field testing stuff ain't easy...and I need more usable data over many more days at the other test conditions in order to gain any confidence in being able to reliably detect differences that are interesting to me.

So I've got my next item on the field testing Crr agenda...more laps at 90 psi over many days in order to drive that 95% confidence interval down!

Giddy up!

Hello Kraig,

Am I reading your report correctly?

Crr:

90 psi 0.00591
120 psi 0.00672
140 psi 0.00609

Trend curve is different from what I would expect.

It appears that for a rough chip seal road I should not be using my usual 120 psi but rather 90 psi or 140 psi (my clincher wheels are limited to 125 psi I think. Zipps.

Cheers,

Neal

Neal wrote:


My recommendation would be to read the results in context with the reported 95% confidence intervals, and also noting the difference in reported CxA's.

So, if you are drawing conclusions based on my report, I reckon (IMO) you aren't reading it correctly.

This is why this sort of testing is rarely done (correctly)... lots of time... tedious time.

Are you controlling for temperature? Wind?

Ron Ruff wrote:


Yes, it is time consuming and if you aren't in it for the long haul, I reckon you could fool yourself into thinking something is faster, when it actually might not be.

i.e - I reckon some folks have been "fooled by randomness" so to speak when using this, or any other type of field testing (or wind tunnel testing for that matter when the data isn't placed into the appropriate context).



While not controlling for it explicitly, I am recording ambient/road/tire temperature, and wind magnitude and direction (global wind measured after the run using the eyeball method on my Kestrel 4000) -> it's a range with a 1m/s bracket...so, if over a 1-2 minute period approximately, I'll eyeball the wind speed and give it a "0-1 m/s from the SW" note...I don't get crazy with that bit of data yet.

kraig wrote:


What's the velocity range (max/min) for those laps?...oh, and what do you mean by "usable laps"?

Tom_Anhalt wrote:


Well, conceptually, I'd describe the range of speeds to be in the ballpark for what I'd expect to encounter for an Ironman triathlon distance -> i.e about 5kph to about 35 kph - it might be a touch more or less depending on the day and the tire pressure, or how well I was controlling the stubble on my chinny-chin-chin!

...I'd also like to take this opportunity, too, though, to share, that I'd also ballpark (based on my perception of the data I've personally acquired doing field testing via many different methods and techniques and venues over the past 7 years or so) the CV/repeatability across and within days for a given setup (y'know, I've never really done equipment testing using field testing methods in those seven years, BTW) as similar to the higher max speed and lower min speed venues I've tested on...y'know, I've pretty much just resigned myself to trying to document things around the broad/vague question of "how much do I trust these #'s that my power meter and the data reduction strategies the pundits put forth" for the past 7 years.

Along those lines, FWIW, since the early days of the Lim method, I've found that I have to pretty much be a caveman and throw a lot of data, and a lot of time, and a lot of effort, at the problem in order to trust things as much as I'd like to.

This field testing thing is not easy (if you ask me), especially if you'd like to get at the types of information I'm interested in - ie. the 1%-ish type stuff that the internets and national federations are all a-buzz about.



well, I took a bunch more data over this long 4-day weekend (60+ laps in the pre 06:30 hours where the wind is low ) and then sliced and diced the data a bunch of different ways...it turns out that if you get enough data (and use the data reduction techniques I've described and had help with previously - i.e, the WLB method), you don't have to arbitrarily throw out laps...things seem to converge with enough data...but, for the initial post, the good laps were the laps, that for a given tire pressure, lay within 1 std deviation of the overall mean for both the CxA and the Crr across all laps/days tested. Keep in mind, that I am using the method described previously...the WLB method.

kraig wrote:


and found that (for me... and my bike.... and my tires... and my chip-seal road) there is, more than likely, less than a 2W difference at 35kph between 90 psi and 120 psi.

Having said that, with the additional runs, it was interesting to note that when I took all the laps for a give tire pressure and reduced the data such that the CxA and Crr were allowed to float (ie, let the data speak for itself) I found:

57 laps @ 120 psi

CxA = 0.3464 95% confidence interval = 0.0075
Crr = 0.00678 95% confidence interval = 0.000202


42 laps @ 90 psi

CxA = 0.3558 95% confidence interval = 0.0095
Crr = 0.00670 95% confidence interval = 0.000205

small differences in Crr from a global perspective, eh? CxA varied a bit, too - maybe 90 psi is really faster!

FWIW, I saw some really screwy 140 psi data today, but there was more wind this morning (0.5-1.5 m/s as opposed to the 0-1 m/s over the last couple days), so, not sure what to think of the 140 psi deal...I need more patience, data, saddle time in order to get a grip on 140 psi.

oh, btw, if I were to fix the CxA at 0.351 for all laps, I get the following:

120 psi:

Crr = 0.00665 95% confidence interval = 0.00018

90 psi

Crr = 0.00684 95% confidence interval = 0.00022

so, if I take the means at face value (for my bike, my tires, my chip-seal road, my data reduction methodology) ...at 40 kph or 11.1 m/s we have < 2W difference for an 84 kg rider/bike with the nod going to the higher pressure tires.

My own personal takeaway, is a simple one -> I won't be under-inflating these vittoria zaffiro tires if I am interested in going fast and not puncturing on dry roads.

Hello Kraig and All,

Thanks for the update.

Now I can keep my tires at 120 psi.

Do you have any comments on correlating your road data with Al's roller data for a 'real road' application factor?

My usual roads are usually a bit better than chip seal too.

I have a lot of respect for chip seal roads though ..... as I crashed on one once and looked like I tangled with a bobcat. (and lost)

Cheers,

Neal

The Zaffiros are an inexpensive beefy tire. If I was to take a WAG on how the type of tire would effect the results, I'd expect a high-Crr tire to show less benefit at lower pressures... ie the increase in hysteresis losses is great enough to offset the reduction in suspension (vibration losses). Actually I'd expect each tire (and rider! and road! and speed!) to experience different Crr vs pressure performance, but in general I think the better quality tires will achieve their optimum performance at lower pressures... compared to tires with higher Crr.

Ron Ruff wrote:


agreed - maybe two years ago I bought 10 700x20c flavors of them at $12 a piece, and have worn every last one of them down to the cords with only a couple flats. I've got probably two freshy's left and am wondering what my next $12 training tire will be...



When it comes to this topic, my gut says that for the types of roads I'm interested in (rumble strips/cattle gaurds don't really interest me) at the pressures I'd consider running, the tires are not lifting off the road surface...i.e, they are a pretty good suspension system.

I'd encourage you and other folks to do the type of testing I am doing (i.e, across and within days) and let's let the data speak!

Neal wrote:


I don't think I'd change the observations I made here (after conducting these tests):

www.biketechreview.com/forum/1-general-d...-testing-notes#25014

i.e - smooth roads +10% or so
rough roads +30% or so

...also, keep in mind that a 10% error in assumed/measured Crr is roughly 2-3 W at 40 kph for a good rolling set of tires.

...and I don't have a lot of weight on my front wheel when I do these on-road tests.

kraig wrote:


Interesting...I looked back at the range of speeds within a lap from my pressure vs. Crr testing on my halfpipe course and it was more like 10-55kph. Seeing as how I would think that the energy input to the tires is not only going to be a function of the surface roughness, pressure, rider weight, etc..., but also the velocity, I wonder if there may be a speed component that determines if one sees a "breakpoint pressure" in the data? Also, Ron makes a good point about the tires...in my case I was using Vittoria Open Corsa KS w/latex tubes...obviously a much more "compliant" setup across all pressures, but also possibly a better case for "seeing" the differences since at lower pressures things aren't being "swamped out" by the energy being transmitted through the stiff casings of a tire like the Zaffiro...just a thought.

A couple of other quick comments which may be apropos to the discussion:

- I find the results for VE or Adam's "ShortTrack" testing to not be nearly as repeatable when using my road position as compared to my TT bike/position. I think I must "move around" a bit more accidentally than with my "locked in" position on the TT rig.

- I also find the "Short Track" results to be more consistent on flat loops as opposed to halfpipe type out and back courses. I'm not sure exactly why...just something I've observed. It's a lot easier to get a "clean line" on a flat loop for some reason and the y-intercept tends to be much more repeatable. That makes this method and venue great IMO for determining "on road" Crr.

- On the other hand, I find the VE method on a halfpipe (and being consistent across the laps in effort) to allow for much better resolution of CdA differences for various setups. As Robert has shown, when evaluated as an AR series, I've seen the SD of the CdA prediction as low as .0006 m^2.

BTW, is your WLB script coded up the same as Adam's "ShortTrackAeroTesting" spreadsheet?

I guess these things are possible. My gut says that the excitation frequencies are in the 100's of hz range though, and the sack of water on the saddle probably has a natural frequency of 1's to 10's of hz...and the differences in tires/psi just seem a bit small to me in the grand scheme of things. I've got some additional ideas on protocols (thanks BH!) and mechanisms at work, and will continue to dig a bit, simply because it's interesting to me.


I don't really seem to see much different day to day repeatability with my road vs TT bike. In fact, I was just looking at data I took a couple years ago for my road bike, and the CxA was within about 1.5% of the average of the values I've been reporting recently.



I've got 100's of laps on all sorts of different venues like you describe, and I think wind is the biggie that affects variability within a run and across runs.



I've done some laps like this and I wasn't comfortable with the assumptions - I'd prefer, in a perfect world to let the data speak for itself with as few assumptions as possible. However, FWIW, on three separate occasions at two different venues using the WLB method, I've seen within run CV's a touch smaller than what you describe (I have no idea what AR series are...I'm just a caveman that defaults to the law of large #'s when in doubt). FWIW, I didn't really give those data I took with the extremely low CV's any more "credibility" than other days/runs. If I'm going to believe something, it has to rise above the wind noise floor that I routinely experience around here despite which venue I use (wind noise is in the 0-1.5 m/s range).



I think its the same concept. I posted some code awhile back and Adam gave me some pointers. I think the key differences are:

1) how we handle acceleration is probably different - I don't know what differencing scheme he uses -> pretty sure mine is second order
2) I base everything on the distance domain, and he probably does it in the time domain
3) I estimate my error by bootstrapping non-overlapping lap CxA and Crr values

48 laps @ 140 psi

CxA = 0.3468 95% CI = 0.0076
Crr = 0.000654 95% CI = 0.000183



140 psi:

Crr = 0.00633 95% confidence interval = 0.00014

Again, it seems like small differences to me, considering that 10% difference (~0.0006 in Crr) is maybe 2-3 watts at 40 kph.

kraig wrote:


I'm a little hesitant to post these data I took last weekend, simply because they are only based on a single day where I did 21 laps with a given configuration...but, hey, this thread is pretty much just me thinking out loud again and goin' with the flow...so, here is my data at 60 psi:

21 laps at 60 psi letting the data speak for itself:

CxA = 0.343 95% confidence interval = 0.0069
Crr = 0.00768 95% confidence interval = 0.00017

Fixing CxA at 0.351 with 21 laps on a single day:

Crr = 0.00747 95% confidence interval = 0.00017

Based on these additional data, I won't be running 60 psi anytime soon, it seems! Although, dang, it sure felt smooth and fast while pedaling over all those high frequency, low magnitude chip-seal pieces of gravel!