I'd please like some input on some misfit in my testing. I've attached a screenshot from GCNotio, showing some areas of misfit between measured and virtual elevation around the smaller turnaround on my halfpipe course. This part of the course is least well sheltered by hedgerows, and on this particular day there as a light crosswind according to weather station data (~7kph). I can rationalise this misfit as being due to poorly captured headwind/tailwind components along this stretch, since they approximately seem to cancel. The effect seems to be worse on runs when the windspeed is higher

I'm using the out and back of this course for Notio to calculate the calibration factor, which is repeatable around ~1.38. I have it mounted using their supplied mount underneath the extensions, but due to the narrow gap between them (and shape of my extensions) offset to one side rather than centrally. Has anyone had good results mounting a notio on the end of a boom, which should lessen the asymmetry?

Any other advice also gratefully received!

Yeah, that's a lot of testing, and it's probably not ideal or even necessary. I do much less on-bike data collection, but much more post hoc at home analysis. The advantage is I can do the post hoc analysis while sitting with either a hot coffee or a cold beer.

I believe you can get an average wind for each direction and see if they cancel out. This will give you more confidence in your calibration function.

If I had to bet, I would say those are errors in altitude measurement. Which is VE ? green ?

For the mount, I 3d printed a little piece that replaces the gopro connector and allows me to rotate it slightly to get straight. Would that fix it ?

Marc, normally green is actual elevation in GC isn't it?

Not sure I agree it looks like elevation errors- the altitude data looks solid to me or at least repeatable)

I agree with Marc to have a look at the wind values in both directions makes sense.

The fit seems good traveling in one direction but not the other - Is it possible the CRR on different sides of the road are different? Is this first time on this course you see this pattern?

ryinc is correct: measured elevation in green and VE in blue

Over these 3 laps, my average wind is 0.1 kph so the calibration factor seems plausible from that respect (and for the other runs I have, bounces around 0 plus minus 0.5 kph).

I will double check that I have the axis of the pitot aligned with the bike. I assume that, once calibrated, a small amount of misalignment (either in the y or z planes) shouldn't matter?


The road surface looks at feels the same on either side, so I'd be surprised if that could account for the magnitude of the discrepancy?

 

Ok, if green is actual I changed my mind :-)

Can you share the other GC charts that shows speed, wind, altitude, CDA,......all stacked ?

Shoot me your email and I'll send you the .stl you asked for.

Hopefully the attached cover the variables that might be relevant? There's a ~90 degree part way up the small climb on this turnaround, which I go through on the TT bars. I freewheel the last portion of the climb in order too slow down enough to make a U turn, and then again before the 90 degree corner on the way back down. I've edited the file to remove the repeated power values when freewheeling, but this hasn't made much difference to the fit of the VE profile.

The notio calculated CdA values are much lower than I get when fitting the VE. I'm also not sure it gives a negative value when I turn, but wasn't planning on using these data so not worried unless it helps diagnose my VE issues.

These values are perfect. Bottom one has a time scale in minutes, all the others use distance.

Distance scale makes it easier to relate back to VE

Email sent with part

One thing I've wanted to do again for a while but had too much memory taken up on the Notio was to see how consistent I could read out really really long laps with lots of variation in terrain within each lap.

To honor the final 40k TT of Le Tour I went out to the local bike themed cafe' and rode the really hilly 7mi road race loop until I hit 25mi for a thumping workout. I Let the Notio run the whole time. I was very consistent about where I got out of aero to go uphill (like 9mph) and where I sat up for a hairpin turn, or braked. All three 7ish mile loop laps came out the identical CdA. Like, on the last decimal place. So, kinda cool.

I'd be curious to see the charts like Mitochondria posted above. It's usually a good indication of how the session was : size of rolling hills, wind, variability of cda...

I'll dig it up and get screenshots later.

Here's the route. It's the 25mi of laps around the rectangle. It's always up or down. So the CdA is always wrong given the accel/decel of the hills. I think I remember all the laps came to exactly .236 CdA. That's with me climbing hills out of aero for about 4min per lap and then coasting over the one hill for a drink from my BTS bottle once per lap. The .236 also includes the hard braking for the hairpin turn each lap and re-accel.

https://www.strava.com/activities/7515127709


Speed dist for the whole ride was:
14% under 13mph (this much per lap out of aero)
32% under 18mph
14% over 32mph
7% over 35mph
2% over 40mph

I spent 18 months with a team to develop an on road testing protocol to measure rolling resistance using an early model AeroLab sensor. Myself, a professor from UNLV and the guys at AeroLab worked together. In the process I learned a ton. The biggest take away is that in order to get good result you must take extreme measures to use these sensors correctly. I think AeroLab has the right model.

If anyone is interested in learning more about our protocol and what we did to get it right I’m happy to help.

---

Jon Thornham
Co-Founder FLO Cycling

Yup. Good estimates of Crr are harder than getting "equivalently" good estimates of CdA. There's a statistical estimation reason why this is so but the bottom line is that you have to be very meticulous in protocol and procedures in order to do this.

We worked on this before my work with AeroLab and you helped a lot. The AeroLab work made post processing much faster so we could crunch more data.

---

Jon Thornham
Co-Founder FLO Cycling

I'd be curious to hear about the protocol.

It's long and quite detailed but here is the basic outline.

Variables Tracked Before Every Test

1. Bike and rider weight.
2. Tire pressure.
3. Tire casing width.
4. Tire temperature.
5. Road surface temperature.
6. Air temperature.
7. Air density.
8. Wind speed.
9. Wind direction.
10. Road surface roughness approximation.
11. Rider

While on the bike and running the test we collected the following.

1. Bike speed and location via GPS.
2. Relative wind speed.
3. Wind direction.
4. Vibration data.
5. Power.
6. Air temperature.
7. Air density.

After the data was collected the AeroLab team post processed the data. There is magic in their work on that end. Both process, the capture and analysis are not for the faint of heart.


What's unique about testing Crr vs CdA is that you have to remain 100% still when riding outside of your legs. Chris at AeroLab was so good at the end of our testing that he could tell if we turned our heads to look around a corner and could also tell which one of the three riders was riding during the test. We all learned to ride without moving. The reason you have to stay so still is because you have to keep CdA constant to get a good Crr value.


In my opinion this type of testing is very difficult to do. CdA is more user friendly but still comes with a lot of variables that need to be controlled. Simply giving someone a sensor to "get their CdA" is not really possible. This is why the AeroLab guys have a great model.


I do believe their is a more user friendly Crr option and we at FLO are working on that now. We're trying to make the testing for users fool proof. It's a tough project but fun.


Ride safe,

---

Jon Thornham
Co-Founder FLO Cycling

What did the riding protocol look like ?
IE were you doing laps around a velodrome with speed varying from lap to lap or varying within a lap or out and back or......

We rode outdoors on a defined course that we set. We altered speed over the course for good regression.

---

Jon Thornham
Co-Founder FLO Cycling

Always the same course ?

It occurs to me that there are different applications for estimating rolling resistance, For example, I test tires on the rollers to choose the tire with the lowest rolling resistance, so the last testing I did was to confirm that the new VeloFlex Records I bought really were significantly faster than the Corsa Speeds.

I have to assume you are interested in wheel-tire interactions, like how rim width, or hooked vs hookless design affects Crr. Would you be willing to discuss what you are doing with the Crr information?

There are many applications for testing rolling resistance and I believe different tests have different purposes.

1. If I simply want to test tires for the Crr there is a large case for using rollers or a machine built for this type of testing. You see the data for each tire. I can argue that rollers and most machines have flaws but you are controlling variables and finding numbers consistently.
2. We were looking for data to prove rim with a wider internal rim width were faster. We want to do this on road since that is where people ride. The road brings in a number of characteristics not seen on rollers. We did prove this.
3. My long term belief is that an athlete will be able to test for optimal pressure, which we are working on now. This will be user, bike, wheel, tire, and road specific. Our goal it to make this test fool proof. What you won't get is the Crr value but it really doesn't matter if you've picked a good tire to start with. For example, the Continental GP 5000 is very aero, has great rolling resistance and is durable. This makes a great tire. Now for the athlete, the goal is to find the best pressure. That's where we are focusing our work. I argue that if Crr is known the athlete does need to find that value and today there isn't a method to do this reliably.

I hope that makes sense.

---

Jon Thornham
Co-Founder FLO Cycling

Can you not do this with the Chung method today ?


https://www.youtube.com/watch?v=rT9J4ywWo9g

Ryan, a fellow STer here did this back in May. There has been some interesting development since.

In theory yes. I started with the Chung Method with Robert’s guidance. In practice it’s extremely difficult to get right. Ryan and I spoke yesterday which was a great conversation.

Focusing on vibration changes the target from measuring Crr to searching for an impedance break point. While challenging to find if we solve it the user experience will be a 10x improvement.

---

Jon Thornham
Co-Founder FLO Cycling

That video was super impressive when i first saw it, and again now. Thanks Marc and Jon - really interesting to see where this stuff goes, and looking forward to when these nifty features are available in GC. Appreciate all the knowledge you and others share on these forums.

Endurance Innovation podcast has one with John Buckley of Streamlines and Velosense
John Buckley, co-founder of Velosense and Streamlines, is on to talk about his aero and position sensors and to chip in on our favourite debate: the utility of real-time cycling CdA.
Was an interesting listen.
https://endurance-innovation-podcast.simplecast.com/