Need help with calibration

Post Reply
brookside
Posts: 48
Joined: Tue May 05, 2015 7:03 am

Need help with calibration

Post by brookside »

I picked up a Powertap wheel and a Joule, I've been riding with them along with the Newton. The Newton is wildly off, even after doing a new Cal ride, usually 40 to 80 watts too high, particularly on the flats and going down even slight slopes. It's just a little better going uphill, the steeper and slower the better. Even if I do a Check Cal fit to DFPM it remains too high, but it's down to 20-40 watts error.

In investigating this I found something odd, and may be a clue as to problem and fix. When I choose to do a "Change profile after Ride" and select the very same profile I used during the ride the average watts drop dramatically and end up within 10 watts of the DFPM. What's up with that? If the Newton readout during the ride was what Isaac now reports with the exact same profile I'd declare victory. How can this be fixed? Today's ride file included, maybe you can extract the profile then do a "Change profile" and see what I mean.

Thanks,
Attachments
iBike_2017_02_24_1352_24_Miles.ibr
(310.96 KiB) Downloaded 312 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Comparing DFPM and Velocomp Power Data

Post by Velocomp »

This is going to be a bit of a long response, so please bear with me.

Let me begin by saying that I have great respect for PowerTap products. They have gotten really good through the years, and I like their technology very much.

When I see differences between DFPM and our power meter data, I do a lot of checking to make sure that nothing in the rider's PM profile setup looks odd. In your case, all the weights, CdA and Crr coefficients look ok, and your wind data looks solid. I think your Newton is properly calibrated.

So, if there is a difference between DFPM and Newton readings (and there is a significant difference in this ride file), which power readings are to be trusted?

We don't know the details behind the way any DFPM calculates power--that is, we don't have the underlying torque measurements, temperatures, etc. That is not a criticism; it simply means that while we may know WHAT the DFPM power number is, we don't know HOW it was calculated. So, after-the-ride it isn't easy to validate the accuracy of the DFPM readings and calibration.

With Velocomp power data, it's different. Our ride file includes all the information used to make power calculations: bike speed, wind speed, slope, CdA, Crr. Why are these parameters useful after the ride? The only reason a cyclist must pedal a bike is because wind, hills, friction and acceleration oppose the forward motion of the bike. According to Newton's 3rd law, if those quantities are known, the amount of power used to climb a hill, ride on the flats, or any other combination of riding, can be calculated.

In fact, there is a website I use, "bikecalculator.com", that allows the user to input bike speed, hill slope, wind speed, ride position, and bike/rider weight, to estimate watts. So, in the Isaac ride file I can select any ride segment, get the bike speed/wind speed/slope/CdA/Crr and weight parameters for that section of the ride, and use the website to compute the watts. I can compare the bikecalculator.com watts calculation to the Newton/PowerPod watts number and PowerTap watts.

And I can pick various points in the ride file, to isolate the performance of our sensors and the rider's profile settings.

1) I try to find places in the ride file where the rider is riding on the flats at a constant speed. On the flats, wind speed dominates (if it is truly flat and the rider is traveling at a constant speed, then rider/bike weight is not a factor). If the Newton/PP watts are wrong, this would mean a problem with the wind sensor setting (wind scaling) or the CdA.

2) I find places in the ride file where there is a hill, to check accelerometer performance and rider/bike weight settings

3) I find places where the rider is on a gentle climb or descent, to check combined wind/accelerometer performance.

I go at least 10 minutes into a ride file, so that both Newton/PP and DFPM are stabilized.

And, to get "best" results from the bikecalculator.com site, I try to find ride segments where bike speed, hill slope, and wind speed are fairly constant (the website calculations assume these parameters don't vary; if they do then the calculations are less accurate).

With all of that as a preamble, what does this ride file show?

Overall, the data shows the PT on average to be 40W LOW. In fact, when looking in detail, the PT seems to be low almost everywhere. What's going on?

In the screen shots below, I input to bikecalculator.com the parameters of the ride segment I've selected. I show the Newton measurement for that section, and the PT measurement for that same section

1) Hill climb, time 1:12:30. Slope 3.31%, bike speed 10.1 mph, headwind 3.4 mph, rider weight 186 pounds. On hills, gravity watts dominate. If bike/rider weight is correct, then Newton watts should be solid. PT: 148.9, Newton reading: 182.8, bikecalculator: 189W. Newton is correct.
Climb PT
Climb PT
ClimbPT.png (1.21 MiB) Viewed 8758 times
Climb Newton
Climb Newton
ClimbNewton.png (995.03 KiB) Viewed 8758 times
2) Flats. I picked a 1 min 42 segment around the 15 minute mark: 0.72% slope (pretty flat), bike speed 18.7 mph, wind speed -0.2 mph. In this section, wind force dominates. If wind speed and CdA are correct, then Newton watts should be solid. PT: 157.5W. Newton: 231.4W. Bikecalculator: 225W. Newton is correct.
Flats PT
Flats PT
FlatsPT.png (2.01 MiB) Viewed 8758 times
Flats Newton
Flats Newton
FlatsNewton.png (1.98 MiB) Viewed 8758 times
3) Sprint. This is where both acceleration and wind are important. I picked the second sprint, around mile 12. Bike speed 24.2 mph, wind speed -0.9 mph, slope 0.57%. PT: 312W. Newton 368W, bikecalculator 375W. Newton is correct
Sprint PT
Sprint PT
SprintPT.png (735.91 KiB) Viewed 8758 times
SprintNewton.png
SprintNewton.png (740.06 KiB) Viewed 8758 times
In all of these sections (and everywhere else where I checked at random) the Newton was nearly identical to bikecalculator, and the PT reads low. How can this be? Well, all DFPMs require calibration prior to riding, and for best results they need to be re-zeroed during the ride. I have a PT wheel and when calibrated it works just fine. But I do find that calibrating it is a somewhat mysterious process, and that I some times think it is calibrated when, in fact, it is not. This is not a criticism; I am probably doing something improperly.

As a different kind of gut-check, I used Isaac to figure out the rider/bike weight and CdA that would cause the Newton readings to most closely replicate the PT measurements.

If the total bike/rider weight is lowered to 165 pounds (from 186), and the CdA is lowered from 0.33 (road bike) to 0.23 (TT bike), then the Newton very closely copies the PT result (within 5W). These are extreme adjustments that, in my opinion, are unreasonable to make.

Another way to "bludgeon" the Newton data to fit the PT is to lower the net wind speed. This drives down watts. I found that if the average wind speed for the ride was -3.5 mph, then the overall watts were close. But then, the ride file wind looks "funny" and doesn't seem possible. And note that, with this wind measurement adjustment, on the hills the Newton still reads high (accelerometer measurements dominate).
Ride file with artificially adjusted wind speed
Ride file with artificially adjusted wind speed
Screen Shot 2017-02-25 at 7.30.01 AM.png (955.51 KiB) Viewed 8758 times
In sum, I don't find any reason to question the validity of the Newton measurements. I strongly suspect that the PT was out of calibration for this particular ride.


One final point: I did try the trick of using the "Check Calibration" correction on the Newton ride file, using the same profile. This trick certainly does bring the Newton close to the PT. Why? I have no clue. Isaac calculations are a mystery to me. However, there is no reason that the same profile information, applied to a ride, should modify the ride data. So, there appears to be a bug in Isaac.
John Hamann
brookside
Posts: 48
Joined: Tue May 05, 2015 7:03 am

Re: Need help with calibration

Post by brookside »

John, Wow! Thanks for doing all that work, I really appreciate it and it's helped me understand what's going on. First I want to assure you that although what I wrote, and what I write here, may sound confrontational or argumentative it's not intended in that way. I'm trying to solve a problem because I really like the idea of the Newton and I want it to work for me. The problem is this, I live in the northeast, I spend 5 full months a year on the indoor trainer, which is an Elite RealAxiom and not on your list of supported trainers. It gives me power readings based on speed, I want to true those up to the Newton. The Powertap wheel is lingua franca that allows that. By choice of tire and tube, and how many turns I use on the mechanism that pushes the trainer's roller into the bike wheel I've been able to equate the PTap to the RealAxiom (pretty much, the curves are a little different). So getting the Ptap to match up with the Newton means my work 5 months of the year is in concert with 7 months I'm outside with the Newton.

Let me address your technical concerns/comments in turn. Calibration: since I was using a Powertap Joule alongside the Newton the first question is if the Newton is properly calibrated to the Ptap hub. It is. The Joule does auto calibration everytime I stop peddling for 4 seconds or longer, which happens frequently, and did at least once before the first of the segments you analyzed. The Joule's power readings match the Newton's DFPM readings, therefore the Newton is calibrated correctly.

This is a general comment that will apply to all sections: the power/speed website you use isn't fine tuned for the purpose at hand. We have no idea what the Crr and Cda values are that correspond to "clincher" and "hoods". I use cyclingpowerlab.com for stuff like this. Although I can't change Crr, it's set to .400, I can change everything else including Cda, which I've set to both .0035 and .0033 for comparison, figures below reflect .0035 for hoods.

Section 1, hill. I know all of these sections of road very well, I ride them almost everyday I'm outside. This section is a gradual incline who's slope increases slightly, there are no flats or negative slope sections. The ISAAC screen shows a minimum altitude of 157 ft, which is at the start, and a max of 181 ft, which is the end, 14 ft gain. Ridewithgps shows 16 ft over that section, which we'll use. This is a grade of 2.2%, not 3.3. The power required according to cyclingpowerlab is 133 watts. Newton showed 183, Ptap 149. 40 watts of the error is in how Newton measures slope. If it had used it's own altitude measurements instead of it's calculated slope it would have been much closer. See attachment 1.

Section 2, flat. This segment is .52 miles, Newton reports an altitude gain of 12 feet (ridewithgps says 11). It looks dead flat while riding but we'll use 12 feet over .52 miles, or a slope of .4 %. Speed is 30.1 kph, wind is -.32 kph. Computed power is 155 watts. Ptap said 157, Newton 231. See attachment "Section 2".

Section 3, sprint. Here Newton gets the slope correct (1 ft gain over 106 feet, < 1%) but goes awry someplace else. Cyclingpowerlabs computes the watts required at 24.1 mph and -.9 wind to be 294. Ptap was 313, Newton 368. I think it's worth looking at the wind here. In the space of 3 seconds Newton is showing a wind variance of 2,1 mph, possible I guess, but it seems unlikely. See attachment "Section 3".

When Cda and Crr values close to watt I've programed into Newton are used in the calculations Ptap figures are much closer than Newton's. I have to believe the unit is not measuring slope and wind correctly, overstating both in all cases. Is there something that can be done to make slope and wind more accurate? I do a Wind Cal before most rides.

Thanks,
Attachments
Section 3 sprint.png
Section 3 sprint.png (113.49 KiB) Viewed 8733 times
Section 2 flat.png
Section 2 flat.png (113.34 KiB) Viewed 8733 times
Section 1 uphill.png
Section 1 uphill.png (110.28 KiB) Viewed 8733 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Need help with calibration

Post by Velocomp »

It looks like you are calculating slope data from GPS and barometric sensor data. Neither method of slope measurement is accurate for power calculations, especially over the short times of these three segments. (Many years ago there was a power meter that used a baro sensor to measure slope for its power calculations. The product failed because it was so grossly inaccurate). And if you watch slope measurements on your Garmin screen, you'll find that they are both slow and inaccurate.

In contrast, the accelerometer is taking measurements 800 times per second (measuring the total effect of hills and changes in bike speed). And this is an important point: we don't use slope measurements in our power calculation; we use total acceleration measurements.

You have a short, pretty-steep hill near the 53 minute mark, with an average slope of 7.65%. On steep hills wind speed is not an important factor (because the bike speed is low and aero watts go as the cube of the bike speed. Our experience, over many years, is that the Newton is dead-on in climbs. However, in this section Newton reads 235 and PT reads 213. So, even here the PT is low...

I have been doing CdA testing in Florida, on a road that ANYONE would say is dead flat. Except it isn't. It undulates +/- 1% over distances of about 200 yards. When I first saw the slope plots I was incredulous; this road is FLAT and I did not believe that the slope measurements could be correct. But the slope plot is repeatable, using many different meters on my bike, over many different rides and ride speeds. So I am now convinced that the slope plot is accurate and that the undulations are real.

And furthermore, both my PowerPod and my DFPM detect the surface undulations and reflect them as minor power changes.

Regarding wind: I agree that if the wind scaling in your unit is incorrect, this would be a problem. Your net ride file shows a slight head wind, but it isn't of sufficient magnitude to throw off your watts to this extent.

The only thing I cannot explain is the mysterious correction that the "Switch Profile" applies to your ride file. I'm going to do some investigation about this, but other than that I don't have any additional insight to offer at this time.
John Hamann
brookside
Posts: 48
Joined: Tue May 05, 2015 7:03 am

Re: Need help with calibration

Post by brookside »

I finally got outside to do another ride with the Ptap and Newton. Different bike this time, also recent Cal ride. The results are closer when viewed in ISAAC than the last ride we discussed. But here's an effect I noticed not just today, but with that previous ride and most others: during the ride the power numbers from Newton and the Joule are always different except for low speeds up steeper grades when I'm applying fairly equal power over time, then they're very close. Part of the during ride discrepancy is also because the Newton readings are lagging the Joule when increasing or decreasing power (decreasing power evokes the greatest discrepancy to the Newton's readings), and this is true whether the Newton is displaying its computed power reading or the Ptap hub readings. Today the Joule was doing 10 sec averaging and the Newton dynamic, but when I've set both to 5 or 10 second averaging the Newton is always several seconds behind the Joule.

On the previous ride, which had a 40 watt average difference, that difference was reduced to about 20 when I did "Analyze Route". Today the raw difference was only 10 watts which got reduced to 2 (!) after Analyze Route. Analyze Route 90%+ of the time brings the Newton's original power down, and doing that correction seems valid since my route almost always has at least the same first and last 5 miles, and many times, like today, the first and last 10 miles are on the same road. It always brings the average wind down to about 0, +/-. Before correction on almost all my rides Newton has me riding into a net headwind, and this is after doing Wind Cal's before most rides.

So here's the bottom line, or so it seems to me: to be useful during a ride, for the Newton to tell me what I'm doing at the moment, it would need to get rid of the 3 to 10 second lag in power changes (the Ptap and Joule show changes almost immediately when I change my effort) and it would need to get the wind reading more accurate, to the level eventually achieved by Analyze Route which trued up today's ride very nicely with the Ptap. It all well and good to have it be correct eventually for after the fact analysis, but what I really need to know is during a hard workout, TT, crit or road race whether I'm in the red, at FTP, below, etc., at any particular moment.

John did you ever figure out why applying the same ride profile as used in the ride afterwards in "Change Profile After Ride" gave different results. That manipulation also almost always results in lower power, more similar to Ptap numbers, than the original Newton readings.

Thanks again,
Attachments
iBike_2017_03_08_1313_43_Miles.ibr
(957.51 KiB) Downloaded 294 times
Post Reply