CdA Variation in Initial Testing

[nikshar]

Hello! I got my aeropod recently, and I put it to test yesterday and today.

During and post calibration, I noticed the watts stayed at 0. I later realized that it is sending them through another channel, and when I added that as a new sensor in my Garmin, the watts started showing as expected (i.e 0-100 W increase during calibration, and normal watts thereafter). This channel was different from my DFPM channel. Now that everything is calibrated, I'm assuming it doesn't matter what channel I select to show the watts from? They could come via aeropod or directly via the DFPM, it shouldn't make a difference?

My main question is about the variation I noticed across the test laps.
Yesterday, I tested it on my TT bike, on an 1.1 mile out-back stretch, that took about 3 minutes per lap. The CdA across laps varied from 0.165 to 0.185 (same was true with the "analyzed CdA" in the VA software).
I then mounted it on my road bike, re-did the calibration, and did the same laps. Though the overall CdA was higher than my TT bike (as expected), the variance was about the same order.

I took my road bike to another place today, re-did the calibration, and did some more test runs. This time, it was a 3.3 mile out-back stretch, taking about 9 minutes per lap. The CdA on my Garmin was much more consistent (0.239, 0.237, 0.243, 0.237), but the analyzed CdA had more variance (0.185, 0.192, 0.201, 0.186).

I'm attaching all three files if that helps (in the order mentioned above, the names are

Velocomp_09_01_2025_1757_8_Miles, Velocomp_09_01_2025_2017_8_Miles, Velocomp_09_02_2025_1130_24_Miles

) , but could you help debug the issue here? Am I doing something wrong? All my laps were with minimal interference from cars/cyclists

[Velocomp]

Lots of comments:

1) You are doing a fine job of testing. You are doing multiple laps for each test sequence, and you are riding consistently for each test sequence.

2) The first five minutes of any ride file are recalibrating the AP. So, your first lap of each of the ride files should be looked at suspiciously. In your case the results are comparable to the other laps, so it's OK to "count" them.

3) On your test routes there are lots of hills with slopes going from -5% to +5%. You'll see that your speed and power vary consequently.

Slope/speed/power variation causes problems for highest-accuracy CdA measurement. If you check out websites of other aero sensor companies, you'll find that they are all doing their testing in level velodromes, flat outdoor courses, or underground tunnels. These slope profiles are completely known and almost unvarying, such that slope is not a variable; bike speed and power can be held reasonably constant. This makes CdA measurement much easier.

In the testing you did, slope is constantly changing and by a significant amount. Unlike other CdA measurement methods, AeroPod can account for slope change but it is definitely more challenging to get highest consistency between runs.

If you can find a test route with fewer hills that might yield more consistent results (though, frankly, the results you got were pretty good anyhow...)

4) You had the same speed sensor and DFPM for both bikes. Is this correct? Or, is it the case that you had TT bars on your bike and these were two different ride positions on the same bike?

5) When analyzing CdA, it's important to include ALL the test results in the overall measurement. It is almost impossible to keep all variables unchanged between tests--this is particularly true for wind speed, which is totally out of your control. I generally would like 4-5 tests for each sequence. You only had 3 tests for your first road bike test on Sept 1.

6) Finally, when comparing configurations, riding the same route is best practice. You have two different routes (and their profile is considerably different) from Sept 1 and Sept 2. This creates another variable to consider for comparing results.

7) When the "Analyze CdA" feature is used in Velocomp Aero (VA), additional corrections are applied to the ConnectIQ-reporterd results, giving highest-accuracy reports of CdA. Here are the results I get for your files:

Sept 1, 17:57 file (TT):

Lap 2 (calibration lap): .174
Lap 4: .180
Lap 6: .171
Lap 8: .165
Lap 10: .177
Lap 12: .180

Average laps 2-12 (TT): .175

Sept 1 (road)

Lap 2 (calibration): .193
Lap 4: .191
Lap 6: .213

Average lap 2-6 (road): .199

Sept 2 (road), different test route from sept 1 rides

Lap 2 (calibration lap): .185
Lap 4: .192
Lap 6: .201
Lap 8: .186
Lap 10: .213
Lap 12: .180

Average lap 2-10 (road): .193

What do I find from your tests?

A. Both TT and road bike configurations have very good (low) CdA values. Your road value is surprisingly low.

B. Road CdA is higher. What is the "correct" value? Tough to be sure, because your route for the two sets of test was not the same (note that wind tunnel and/or velodrome test results will also not be the same for different facilities). Probably the best comparison is the results from Sept 1, since the same route was used for both tests

C. For highest-accuracy results, If possible I would find a flatter course. Also, I would use the SAME course for all tests.

I think you have done a great job overall. Keep at it!

[nikshar]

Hi John,

Thanks a lot for the detailed feedback, much appreciated.

A few follow-up comments:

• At this stage, I'm just trying to understand the variability I could get when I ride a given course in a given position. So, when I actually test out changes, I'd have an estimate for variance and margin of error. Hence the variation in my test courses/bikes.

• My main goal is to do A/B testing. So I don't need any kind of absolute CdA, and at least at the moment, I don't particularly require intra-day CdA comparisons

• I used the same speed sensor across both bikes, but they have their own individual power meters. (I'm not really trying to compare road CdA vs TT CdA, so it should be fine I suppose)

• For course selection, I could trim the bigger loop from Sep 2 and include only the middle section. I feel like an ideal test course for out-and-back testing would be something that starts with downhill, so I can gain decent speed (35+kph) without having to accelerate hard, then start the lap and ride on a roughly flat course, followed by a slight up-hill section that can naturally make me slow down enough to turn around without braking. I feel like a completely flat course would have the challenge of muddying the results during braking and turning and getting up-to-speed again? What do you think?

• In the VA software, I see that I could use the `shift` button and select all laps and get an overall analyzed CdA. Is this just an average of the analyzed CdA of all the laps selected? I ask because in my out-back testing, every other lap is just me going past the end point, turning around, and getting to the start point again, which means I need a way to exclude it from the selection

• I actually did do another run yesterday evening, this time on a flat bike trail with a tiny rise half-way through. No cars, no bikes, but the biggest challenge was keeping the "slow down and turn" part consistent. I think I got decent results from this. Also, I changed the Crr of my tires to 0.003 (a rough figure I estimated from the info available here (https://www.bicyclerollingresistance.co ... 00-s-tr-28). With this new setting, my CdA was more in the 0.29 range (on my road bike). My live laps CdA were 0.292, 0.301, 0.306, 0.308 and analyzed CdA were 0.279, 0.291, 0.295, 0.291. Attaching the file herewith, if that helps (in which case, ignore Lap 2 - I stopped it slightly earlier than the start point)

Thanks again, for the pointers and discussion!

[Velocomp]

IMPORTANT!!!

We just noticed that in your ride file 9/1 17:57, that you are not stopping the lap at the same physical location where you start the lap.

How do we know? Your starting elevation (404' for these laps) and stopping elevation (433') are DIFFERENT.

Here is lap 2 (the other laps show the same elevation difference between starting and stopping point).

1.png (146.79 KiB) Viewed 705 times

You must start and stop each test lap at the same physical place. When using profile 4, if the start/stop point is different, the CdA measurements will be incorrect.

Your Garmin .FIT file includes GPS information, and you can merge GPS data into your AP data. Please do this for future rides. It will help us see where you are starting/stopping your test laps.

[nikshar]

Hi John,

Thanks for pointing that out. Quick question:

Is it ok if the stop point is on the other side of the road as the start point, as long as they are both at the same level/elevation? On my bike-path test route, this seems irrelevant, since it's a narrow path and I pass by the same start point on my return leg, but on my other route, which has one car lane + one bike lane on each side of the road, the stop point is on the "other" side of the road. I could choose my start point to be on a flat section of the road, so the return point would also be flat

[Velocomp]

Hi John,

Thanks for pointing that out. Quick question:

Is it ok if the stop point is on the other side of the road as the start point, as long as they are both at the same level/elevation? On my bike-path test route, this seems irrelevant, since it's a narrow path and I pass by the same start point on my return leg, but on my other route, which has one car lane + one bike lane on each side of the road, the stop point is on the "other" side of the road. I could choose my start point to be on a flat section of the road, so the return point would also be flat

Yes, that is fine.

[nikshar]

Hi John,

I did some more testing today, a new short 3-min loop, pretty much all flat. The CdA values look pretty consistent (in the attached file, my testing laps were 6, 8, 12, 14, 16, and 18), but I was checking the elevation tab, since you mentioned that being a source of error last time, and I noticed two things:
1. Despite starting and stopping at the same point, the start and stop points for my laps show an elevation difference of about 5 meters. Is that amount of difference significant? I'm guessing it's relying on elevation data from Garmin's GPS and that that is the source of error?
2. I added elevation data from my Garmin fit file (also attached here), and when I use the "map" Display in the VA app, it shows erroneous GPS locations corresponding to my laps. Like the start of a lap is shown at a different location to where I actually pressed the lap button. I don't quite understand the source of error.

Any thoughts? Thanks

[nikshar]

Hi John,

I have some more results from today's test, and there are a few questions I'm hoping to get your input on.

I did a lot of runs today, I tested four different configurations, I'll call them A, B, C, D. Wrt that, the relevant lap numbers in my file are:

• A: 10, 13, 14

• B: 16, 18, 20, 22, 24

• A: 26, 28, 30

• C: 32, 34, 36, 38, 40

• D: 44, 46, 48

• C: 50

I have the following questions:

• Once again, I see a noticeable difference in the start and end elevation values (both, within and across laps), even though every lap started and ended at the same point. Is this affecting my results?

• When I use the analyzed-cda feature and group the runs (shift-select), I get the following results: 0.232 (A), 0.237 (B), 0.237 (A-2nd group), 0.238 (C), 0.234 (D). It seems like these numbers are close enough to make out meaningful differences (for instance, my stack-height went down 4cm between A and C), but also that the 2nd group of A was significantly different to the first (0.232 vs 0.237) - can you help debug?

• Within the above groups, if I look at the analyzed-cda for the individual run, I end up getting a pretty big spread (variance). If I take all the analyzed-cda values across all configurations, and plot them, I don't see any correlation at all, so I'm not able to get any meaningful value out of these results
  
  Would appreciate your thoughts, thanks

[nikshar]

^ In addition to that, I just noticed that the power graph shown in the .ibr file fetched from the device is different from when I import power into VA from my bike computer's .fit file. The latter has much less fluctuation. I wonder if this is contributing to some issues?
Here are two screenshots comparing that for Lap 14 (I've noticed this behavior across all laps)

DevicePower.png (163.53 KiB) Viewed 216 times

FitFilePower.png (172.05 KiB) Viewed 216 times

[Velocomp]

Hi John,

I did some more testing today, a new short 3-min loop, pretty much all flat. The CdA values look pretty consistent (in the attached file, my testing laps were 6, 8, 12, 14, 16, and 18), but I was checking the elevation tab, since you mentioned that being a source of error last time, and I noticed two things:
1. Despite starting and stopping at the same point, the start and stop points for my laps show an elevation difference of about 5 meters. Is that amount of difference significant? I'm guessing it's relying on elevation data from Garmin's GPS and that that is the source of error?
2. I added elevation data from my Garmin fit file (also attached here), and when I use the "map" Display in the VA app, it shows erroneous GPS locations corresponding to my laps. Like the start of a lap is shown at a different location to where I actually pressed the lap button. I don't quite understand the source of error.

Any thoughts? Thanks

Sorry for the delay in responding...

Elevation measurements by barometric pressure sensor can be imprecise, due to variability of local weather, and GPS measurement of elevation is even less precise.

We do not use GPS or baro sensor data for CdA measurement.

Because you are using your eyes to identify the stop/start point (which should be the same...) don't worry if GPS data is not as good!

[Velocomp]

^ In addition to that, I just noticed that the power graph shown in the .ibr file fetched from the device is different from when I import power into VA from my bike computer's .fit file. The latter has much less fluctuation. I wonder if this is contributing to some issues?
Here are two screenshots comparing that for Lap 14 (I've noticed this behavior across all laps)
DevicePower.pngFitFilePower.png

I'm looking now at your 40 mile ride data.

The likely difference between the .FIT and AP data is your Garmin has some kind of smoothing factor set for power sensor data. The data the Garmin receives is identical to the power data received by AP. AP records raw values; smoothing is done in VA.

[HDPWR]

The smoothing of the power data can be configured in the AP profile, if I remember correctly, by default it is dynamic. If it is set to 3s or 5s, then Garmin will receive the smoothed data, won't it?

[lorduintah]

And Garmin can be configured to smooth Power data, too.

Tom

[Velocomp]

The smoothing of the power data can be configured in the AP profile, if I remember correctly, by default it is dynamic. If it is set to 3s or 5s, then Garmin will receive the smoothed data, won't it?

Good catch! Yes, power transmitted by AP to Garmin will be affected by smoothing setting in profile. However, the data recorded in the AP file is not smoothed.