AeroPod, PowerPod and Newton

I'm just getting started with my PP, and want to make sure that my profile is as solid as possible before I leave it alone and get on with training.

I'm not sure whether the PP is able to second-guess the Cda and Crr that you provide to it as part of the calibration process - it's hard to see how it could - so I'm assuming that it's relying on the estimates you enter. It seems to me that the best way of telling that you've got the numbers right would be to do roll down tests (ideally from high speeds so that you know that Cda is causing the lion's share of the deceleration), and checking that the PP is reporting an average power of roughly 0.

However, when I try this (rolling downhill, turning the cranks with no power), the Isaac software clips the power output at 0 (no negative values), so it's impossible to establish the average power that the algorithm is calculating. Is there any way to work around this?

A couple of points to note:

• I've got no reason to think that PP is giving me bad data in normal use. It's just that a coast-down test like this would be solid confirmation that it's giving me good data!
• I'm not actually interested in power when coasting, and I know that (with a cadence sensor attached) PP will show zero anyway. This just seems a great way of confirming that the profile is solid.
• The reason that I'm particularly interested in Cda is that it increases in significance at higher speed. An error would therefore impact consistency of power data; a reported power of, say 250W might be almost accurate at low speed, but dramatically wrong at a higher speed. This would make it difficult to train to power.

Looking at the above question a different way; there has been a fair amount of discussion about the value of calibrating PP, and evaluating changes in Cda, by comparing the output to a trusted DFPM. A downhill coast provides any PP owner with a completely trustworthy power reading, 0W, so it's a shame it seems so hard to make use of this data point.

I've enclosed two files, in case they're interesting:

• A calibration ride
• A subsequent ride including coast down between 7:30 and 8:20. I'm 5"7, riding on the hoods with flexed arms, aero bike (carbon, Zipp 303) and Conti GP4000 clinchers. It's hard to get my power output down to zero without some extraordinarily low Cda/Crr numbers.

Thanks!

Richard

Congratulations, you're on to a really interesting aspect of PowerPod technology!

Yes, you can do coast-down measurements with PowerPod. It's a very sensitive measurement, but as long as you're careful you can get some good results.

Here's how to "geek out" on CdA measurement

1) You should do an out-and-back calibration prior to the coast down measurements

2) You should find a place with a gentle down-hill, where you can coast down for about a minute. The slope of the downhill section should be about the same, and the road roughness should be uniform and smooth (say, asphalt).

3) Winds should be calm

4) Traffic should be light

5) During the measurement, you should hold your ride position as steady as possible

6) It's important to get your profile parameters: weight, height, tire type (Crr) correct

All of these conditions are necessary so that your measurements will be reliable

I notice that your "raw" data from PP used a Crr that was too high (0.005), but that you adjusted it post-ride (0.004). This is OK.

It looks like your coast down starting at minute 7:27 meets these criteria, so here's how to use the data:

1) Set filtering to zero, so you can see the power detail

2) In your file, you can see where you stopped pedaling, then started soft pedaling at 7:27

3) When you first started pedaling at 7:27, there is a brief power "blip" where your cadence was around 50. I'm not sure if this means you accidentally applied a bit of power, so I started the measurements where cadence was steady, at around 40, at 7:35

4) There are 4 small power spikes after cadence stabilizes at 7:35. This means your default value (from PP setup) of 0.328 is already very close: note also that from 7:50 to 8:20 power is zero
5) I then played with the profile, lowering CdA a bit, and watching the power spikes from 7:36 to 7:49. You're correct that the effect of CdA is more pronounced at higher speeds, so I tried to remove the 4 power spikes from 7:36 and beyond.

6) In the Edit Profiles menu I manually adjusted CdA (in this example, to 0.29), then clicked Accept. IMPORTANT: WHEN CHANGING CDA IN THE EDIT PROFILES WINDOW, MAKE SURE THE RADIO BUTTON IS SELECTED THAT HOLDS WIND SCALING CONSTANT.
7) I then used the command to show the power output for the modified CdA:
8) Here is the power curve with the 0.29 CdA
9) So, the 0.29 CdA "killed" 3 of the 4 the power spikes that start at 7:36 and beyond. The only question is, is 0.29 too low? Here, you simply have to do some playing around. You want the CdA to be as high as possible, such that it forces downhill watts close to zero. REMEMBER: you won't get it to be exactly zero.

10) After trying this with several values of CdA, I settled on 0.312. This CdA crushed three of the 4 peaks. Average watts from the three peaks in this region is 6W
11) One way to check the reasonableness of the CdA is to compare the 6W "error" to the wattage you'd be producing on-the-flats at the same speed. That is, if riding on the flats (where CdA is important), what is your wattage, and what the effect of an 6W error?

I used analyticcycling.com to figure this out. In the highlighted section the bike speed is 10.44m/sec. So, riding on the flats in this ride position the wattage is 251W
Assuming that the CdA "error" is 6W, the PP would measure 251+6=257W. This is an error of 2.4%, well within the measurement accuracy bounds of PP (or any other DFPM).

So, the coast-down estimate of your CdA is 0.312, a bit lower than the 0.328 that comes from PP setup.

What is the overall effect of this change? Your average watts for this ride file, at CdA of .328 was 141W; at .312, 137.9W.

So, "out of the box", your PP was close.

But it is a lot of fun to do these kinds of experiments!

John,

this is a great description!! You said that the weights and CRR should be correct before doing this. Weights are easy. Any similar recommendation for how to use coast downs and the PowerPod to adjust the CRR?

Maybe an iterative process where low speed portions of the coast down is used for adjusting CRR and high speed portion for CdA?

Also, noticed you used 'edit profile' and 'switch profile after-the-ride'. Could you use 'Tweak CdA, Crr, Cm' instead?

Thanks,

Morten

You set the Crr to 0.004 in Edit Profiles.

It's beyond the scope of PP FW to measure CdA and Crr simultaneously.

Switch Profile After the Ride does a better job of CdA calculations.

Hi John,

Thanks for the comprehensive answer.

The aspect of this method that worries me is that if I can decrease Cda to the point where all values of power are showing 0W, my Cda will be too low. The correct value would inevitably show some readings above zero, balanced by others below zero. I need to work out the average value of calculated power throughout the coast down without ignoring the negative values. I thought that maybe "braking" would be a way of doing this - if braking balanced energy, then the average would be zero. Unfortunately, the calculation doesn't seem to treat negative power readings as braking until some threshold is reached, so this doesn't work either.

I think that, unless Isaac is enhanced to include a mode where negative power can be displayed, or where the threshold for braking is removed, there's probably nothing I can do other than adopt one of the "standard" models for Cda.

Richard

Your data shows that CdA should only be a very tiny bit lower--from .328 to .318.

You'll notice that I did NOT include 100% of your coast down data (the part where cadence was above 40) and I did NOT force all power spikes to zero. Using coast down data for CdA is mostly science, but a bit art, too.

You most certainly can continue to use the "standard" value of CdA

I'm really enjoying getting to grips with the PP, and gaining new insights into how to use it. Although I might seem to be trying to pick holes in it, I'm actually just trying to come to a more complete understanding of its capabilities, so that I can best utilise the insights it provides. Although I've not yet been able to compare its output with a known good power reading, I've been impressed at how consistently it's output matches my perceived effort. The only exception is on faster sections, where it's showing me that I tend to confuse speed with power, and don't work anywhere near as hard when I feel that I'm going fast enough!

Anyway, in my search for the perfect Cda, I did another, much longer, more controlled roll-down today, which resulted in power readings of zero throughout, except for a couple of small spikes. I tried applying a larger Cda afterwards in Isaac, but although it changed the size of the spikes, I couldn't budge the flat-line sections. It was at this point I realized that I had completely misunderstood how PP and Isaac interact! I had assumed that PP stored all the data it was using for the power calculations, so post-processing in Isaac would always be able to correct a poor profile. However, it's now clear that this isn't the case; the PP stores the calculated power plus some additional context (wind speed, etc), and Isaac can only scale the PP's calculated power rather than recalculate it from scratch.

Bottom line - once PP has calculated a power output of zero, there's nothing that Isaac can do to recover that data, even if you tell it that, actually, you had the Cda of the back end of a bus. It's rather like blowing highlights on a digital photograph; once they're gone, they're gone. It therefore follows that the way to try to evaluate Cda using coast downs is to deliberately use a profile with a slightly high Cda on the test run, because then there's some positive power data that Isaac can work with as you try lowering it. Once it's dialled in, a coast down with the perfect profile will not read zero, but will read slightly positive for something approaching half of the time.

One question arising from all this - given that PP has more data available for its computations than Isaac, does it follow that an out-and-back calibration on the PP is more effective than using Isaac to perform a calibration after the fact? I suspect not, in that when I use Isaac to check the calibration of a calibration ride, it comes up with exactly the same profile. But I just thought I'd ask, as I prefer to just include an out-and-back section within a normal ride to check the calibration, but I'd use the PP-controlled calibration instead if it were more trustworthy.

One other question. If you use the Cda analysis dialog, without any DFPM data, what is the Cda graph showing? It looks interesting, and maybe it would provide useful input when reviewing coast downs, but I'm not really sure what I'm looking at.

Cheers,

Richard

It's great to see you thinking through all this...I suppose it could be likened to a mystery novel!

How Isaac works is really complicated, and while your thoughts are consistent with what you can see, what you can't see is quite different. I'll leave it at that...

Isaac should be able to very closely replicate calibration results. It has access to a HUGE store of data that is not visible.

I don't know about the CdA graph; Travis, our Isaac magician, doesn't tell even me all the tricks he has!

Velocomp wrote: ↑Sun Aug 20, 2017 10:04 am
1) Set filtering to zero, so you can see the power detail

How do I do this?

Regards,
Chris Maness

In Isaac, in the top right of the main screen

When I do a coast down, why am I not able to bring up the power values up from zero using the "change profile after the ride", but I am able to do this with "tweak CdA etc.." function? In the discussion above, I see that the preference is given to the latter because it is supposed to be more accurate. I am also seeing braking unless I raise CdA with "change profile after the ride" to an unreasonable value. This is also better fixed with the tweak function. I will post the raw untweaked files. This is on my heavy Surly with two full rear panniers. Weight was updated just before I rolled out.
Regards,
Chris Maness

Update: I see that the AP re-calibrates after every profile change, and it looks like there was not enough time for it to recalibrate in the short ride posted above. I rode to my second job, a long 15mile climb to the college. Difference between DFPM and AP total work is now less than 1%. A lot of work to get it that close

Regards,
Chris Maness