Confused by multiple CdA results

Post Reply
BillOsler
Posts: 116
Joined: Mon Apr 04, 2016 10:10 am

Confused by multiple CdA results

Post by BillOsler »

While trying to analyze some data I realized that Isaac is offering me THREE different CdA results for the same data.
I recorded a ride in Profile 4 using my Aeropod and the bike's DFPM. I edited the activity to remove everything except for 3 laps of a loop. The loop was not ideal because it required slowing for a couple of turns but I rode the entire time without touching the brakes and without changing position on my TT bike, though I did have to turn my head a bit to look for traffic when I coasted up to the turns.
CdA estimate #1 is calculated by Isaac for the entire ride: 0.244
CdA estimate #2 is from CdA Analysis. I selected the entire ride as shown on the screen shot: 0.277
CdA estimate #3 is from the HR data, 4*HRavg/1000=0.214
That's a fairly diverse group of numbers.
Why is there so much difference? And which is more credible?
Attachments
Three Different CdA Results.png
Three Different CdA Results.png (206.82 KiB) Viewed 6083 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Confused by multiple CdA results

Post by Velocomp »

Please post .ibr ride file
John Hamann
BillOsler
Posts: 116
Joined: Mon Apr 04, 2016 10:10 am

Re: Confused by multiple CdA results

Post by BillOsler »

The IBR for that portion of the ride:
Attachments
ThreeDifferentCdA.ibr
(263.85 KiB) Downloaded 202 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Confused by multiple CdA results

Post by Velocomp »

Thanks for posting the ride file.

I think there are several things going on here:

1) Routes with slowing and turning are not ideal conditions for tests. When the handlebars are turned the calculations can be thrown-off for around 30 seconds. There's no GPS data in your file so I can't tell exactly where the turns were, but the ascending CdA from minute 5 to 6 and 9 to 10 are suggestive of preceding turns.

2) Your speeds and watts are pretty low. The lower the watts, the less resolution you get

3) Your pedaling power was different on all 3 runs. Normally this would not be a big deal, but since your applied power is low I worry that this can cause variance from test to test.

4) You have quite a few places where there is 0 cadence. When pedaling is uneven (starts-stops) this can affect CdA readings.

Best practice would be:

A. Find a place where you can ride 3 minutes (or more) without stopping pedaling

B. Find a place where your test route does not have sharp turns (or round-abouts)

C. Pedal at an even power pace, and keep power constant between runs
John Hamann
BillOsler
Posts: 116
Joined: Mon Apr 04, 2016 10:10 am

Re: Confused by multiple CdA results

Post by BillOsler »

I understand that the file I submitted was not optimal for use with Aeropod. It was just a handy example of a discrepancy in the analysis of the data. I was not actually aiming to get a CdA measurement from Aeropod on that ride. In fact I'm close to abandoning Aeropod for its intended use because I have not been able to get anything resembling reproducible results despite many, many attempts. In the ride file I submitted I WAS trying to achieve best practices for analysis in Golden Cheetah using Virtual Elevation (Aerolab module) which appears to have more promise than Aeropod for after-the-fact analysis. For that analysis Aeropod is basically just an expensive anemometer because most of the data for analysis comes either from my Garmin or from a known elevation survey.
Lest you assume that the problem was caused by the variation in power, coasting et cetera I will point out that the same discrepancy occurs in rides that WERE attempts to follow the appropriate protocol for Aeropod's analysis, eg: a ride file I submitted with a previous question:
https://velocompforum.com/download/file.php?id=5784
As seen in the screen shot the same phenomenon occurs. CdA from average "heartrate", CdA from the overall ride and CdA from CdA analysis of the entire ride are all different.
Attachments
Another_Example_3_different_CdA_Results.png
Another_Example_3_different_CdA_Results.png (285.3 KiB) Viewed 6042 times
BillOsler
Posts: 116
Joined: Mon Apr 04, 2016 10:10 am

Re: Confused by multiple CdA results

Post by BillOsler »

And another example from the same day:
https://velocompforum.com/download/file.php?id=5785
It would appear that Isaac has a hard time deciding what the CdA actually is.
Addend: I realized after-the-fact that in the picture I highlighted average cadence instead of average HR in the summary data. Correcting that would not alter the net outcome, the estimated CdA from average HR does not match the other 2 estimates.
Attachments
Yet_Another_Example_3_different_CdA_Results.png
Yet_Another_Example_3_different_CdA_Results.png (284.52 KiB) Viewed 6041 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Confused by multiple CdA results

Post by Velocomp »

It is very frustrating to be given data to analyze, then to analyze it, and afterwards to be told "the file I submitted was not optimal for use with AeroPod". Please, if you are asking for assistance, provide the best information you can...

I think what is going on here is that the CdA Analysis routines in Isaac don't match the CdA measurement routines in AeroPod. AeroPod has much more up-to-date algorithms than Isaac and I am not surprised there are disagreements.

So, when you compare Isaac to AeroPod results (which is a fair thing to do) you are getting somewhat different answers.

Also, the "CdA" number reported in the ride file is a measurement taken during the calibration ride; the CdA/HR number afterwards is the result of your current test. They do not have to be the same.

There is a different way to look at CdA data in Isaac, that avoids some of the algorithm pitfalls of Isaac Analyze CdA window. Here is what to do:

1) open the ride file you want to look at. In this example, I'll use your "5784" ride file.

2) The CdA measured by AP/HR is 73.4 * 4 / 1000 = 0.293 The "baseline" CdA was 0.280, somewhat lower.

3) In the stats window toggle between View DFPM checked, and unchecked. With CdA set at 0.28, AP watts are lower than DFPM watts.
AP watts, CdA = .280
AP watts, CdA = .280
Screen Shot 2019-05-21 at 2.39.41 PM.png (56.02 KiB) Viewed 5990 times
DFPM watts
DFPM watts
Screen Shot 2019-05-21 at 2.36.41 PM.png (55.55 KiB) Viewed 5990 times
3) Since AP watts are low with CdA = .280, this means the actual CdA during the test was higher than 0.280. Let's see how close the AP measurement of 0.293 is...

4) Use the Isaac command Analyze/Tweak CdA...you will see this window:
Analyze CdA window, CdA = 0.2796
Analyze CdA window, CdA = 0.2796
Screen Shot 2019-05-21 at 2.46.27 PM.png (147.64 KiB) Viewed 5990 times
5) Now, change CdA to 0.293 and click Accept. Note that AP watts rise to 189.6
Analyze CdA, CdA = 0.293
Analyze CdA, CdA = 0.293
Screen Shot 2019-05-21 at 2.47.44 PM.png (148.1 KiB) Viewed 5990 times
AP watts, CdA = 0.293
AP watts, CdA = 0.293
Screen Shot 2019-05-21 at 2.55.58 PM.png (56.36 KiB) Viewed 5990 times
6) This is still 2W low; if you change CdA to 0.296, you get total agreement. (Note that the .003 difference is within the round off error of HR)
AP watts, CdA = 0.296
AP watts, CdA = 0.296
Screen Shot 2019-05-21 at 2.50.08 PM.png (149.59 KiB) Viewed 5990 times
What the above means is:

1) HR/CdA data from AP is correct

2) Isaac CdA Analysis is not as good as actual data (no surprise there)

3) You can check HR/CdA data with the Tweak CdA command

There are issues with Isaac CdA Analysis recalculations, but in our view, AP data is good.
Attachments
AP watts, CdA = 0.293
AP watts, CdA = 0.293
Screen Shot 2019-05-21 at 2.48.52 PM.png (45.36 KiB) Viewed 5990 times
John Hamann
GooseRider
Posts: 58
Joined: Wed May 10, 2017 10:52 am

Re: Confused by multiple CdA results

Post by GooseRider »

Maybe it is already available, but I think a best practices white paper would be awesome. That should include the test protocol for most consistent data as well as how to treat/analyze the data afterwards including how best to look at the data in Isaac

Morten
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Confused by multiple CdA results

Post by Velocomp »

GooseRider wrote: Tue May 21, 2019 3:14 pm Maybe it is already available, but I think a best practices white paper would be awesome. That should include the test protocol for most consistent data as well as how to treat/analyze the data afterwards including how best to look at the data in Isaac

Morten
This is a good idea. I will work on it.
John Hamann
GooseRider
Posts: 58
Joined: Wed May 10, 2017 10:52 am

Re: Confused by multiple CdA results

Post by GooseRider »

Awesome, thx!!
BillOsler
Posts: 116
Joined: Mon Apr 04, 2016 10:10 am

Re: Confused by multiple CdA results

Post by BillOsler »

Velocomp wrote: Tue May 21, 2019 2:53 pm It is very frustrating to be given data to analyze, then to analyze it, and afterwards to be told "the file I submitted was not optimal for use with AeroPod". Please, if you are asking for assistance, provide the best information you can...
I apologize for any frustration caused. I was surprised by the request for a ride file because i thought I was posing a straightforward question that could be answered without reference to a specific set of data. That said, I am not convinced that it is possible for me to collect what you consider an optimum set of data given the geographic and climatologic situation in which I find myself. Long straight stretches of relatively level road with light traffic and little wind with no gusts are a scarce commodity, and even a velodrome would be sub-optimal because the laps would be shorter than what you prefer.
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Confused by multiple CdA results

Post by Velocomp »

Here are some guidelines:

1) Try to ride at a constant level of power, preferably 150W or more. This is because higher power produces measurements with better resolution and consistency

2) If you're doing multiple tests with the same conditions, try to hold your power constant between runs. In the "ThreeDifferentCdA" file presented, power increased by over 40% between the first run and the 3rd run (and all runs were below 150W).

3) Tests should be 6 minutes or more in length.

3) If you're using Isaac for analysis, the "CdA Analysis" window will provide useful RELATIVE CdA information, but not, necessarily, absolute numbers. This is a bug in Isaac that we will address, once we get updated Isaac for Mac working (Mac 64 bit version has not been simple).

4) If you want to get absolute CdA data using Isaac, the best way to do it is to use the HR/CdA data from AP ride file, in combination with the "Tweak CdA" command in Isaac. When CdA is "perfect" (you adjust CdA in the Tweak CdA command), opposing watts will equal applied watts. The HR/CdA data should be very close to perfect. See an earlier post on this thread for an example of this kind of analysis.

5) Also, you can use the Isaac command "Power Meter Comparison" to graphically display power data (both AP and DFPM) and CdA data. This is a good way to visually check your CdA analysis.

6) Remember that the CdA number shown in the stats window is that measured during the Cal Ride. This number won't necessarily be the same as the CdA number measured in your tests.

We did a lot of work to improve AP algorithms to get best CdA results. Those changes have not yet been transferred over to Isaac, so it's not surprising there might be differences in the Analyze CdA window.
John Hamann
Post Reply