Still not confident regarding accuracy

pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Still not confident regarding accuracy

Post by pb21 »

I posted the other day about how I felt the PP was overestimating my power output uphill due to a tail wind. Following positing it seemed that actually the numbers were correct, however I still was a bit surprised I was able to ride at ~450 W for 5-minutes, quite a bit higher than I thought I was capable of.

I looked again at the file and ran the ‘Analyse Route’ option and this had the effect of reducing my 5-minute by 53 W or 12% (from 444 W to 391 W), with the average gradient for that section changing from 10.5% to 9.1% (looking at mapping data the 9.1% figure is much more accurate, 115 m of altitude gain in 1.3 km). Also the 391 W value is much more believable based on other longer efforts of mine and comparing my time with other similar peoples time and their associated power on this hill; to be honest I really do not think I was riding at 450 W.

The ‘check calibration’ option for this ride suggest that the riding tilt should be -0.5%, whereas it was set to -0.4%, this doesn’t seem like much difference, plus I thought the PP was constantly ‘calibrating’ and I had been riding for over 20-minutes before this effort?

Is there a problem with my PP? It seems the PP was initially basing the power calc on a gradient which was too high. To be honest it’s a bit frustrating as I don’t have confidence that the reported numbers are accurate to within +/-10% :(

The relevant file is attached again.
Attachments
iBike_07_23_2017_1057_23_km.ibr
(243.94 KiB) Downloaded 251 times
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

I am just looking back at other ride files and am seeing this issue crop up quite often. basically the PP is significantly out in the gradient used to calculate power.

I have attached another ride file with several hill efforts.

1. Around 15 km for 1 km, the PP average gradient says 13.2% but the actual altitude change is 107 m over 930 m i.e. 11.5%.
1.png
1.png (66.73 KiB) Viewed 14821 times
2. Around 25 km for 500 m, the PP average gradient says 16% but the actual altitude change is 67 m over 460 m i.e. 14.5%.
2.png
2.png (67.66 KiB) Viewed 14821 times
3. Around 34 km for 1.3 km, the PP average gradient says 10.9% but the actual altitude change is 127 m over 136 m i.e. 9.3%.

4. Around 37 km for 1.3 km, the PP average gradient says 10.2% but the actual altitude change is 128 m over 136 m i.e. 9.3%.

The ‘check calibration’ option for this ride suggest that the riding tilt should be -0.8%, whereas it was set to -0.4%.
Attachments
iBike_07_20_2017_0927_64_km.ibr
(558.65 KiB) Downloaded 248 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

The key thing to know is that PP does NOT use slope information to calculate power.

When you climb a hill, you are fighting gravity, which is an acceleration force. You are climbing really steep hills, so you're fighting strong gravity forces.

But also remember that, whenever you ride, you are also overcoming inertial (bike acceleration) forces. Every cyclist knows that power is required to increase bike speed.

The accelerometer measures gravity (slope) AND bike acceleration forces, as a single number. This total acceleration number, which is extremely accurate, and is measured 800 times per second, is used in the power calculations.

Slope measurement is calculated by subtracting bike acceleration from total acceleration. Note that bike speed and bike acceleration measurements come from wheel rotation measurements. Speed information is updated each time the wheel makes a full rotation.

On steep hills, your bike wheels are turning very slowly. For example, at KM 26, you're climbing at a rate of 10.8KM/hr. That is only 3 m/sec. If your wheel circumference is 2096mm, that's less than 1.5 wheel turns per second. This means: the frequency and resolution of the bike acceleration measurements is poor when you are climbing steep hills.

Also remember that, when ascending steep hills, instantaneous bike acceleration is going to be highly variable as you turn the crank; for example, when you reduce the force on the pedals during a crank rotation (which happens, of course, every time you are at the top and bottom of the crank stroke), even for an instant, the bike is going to immediately decelerate (drop speed) somewhat.

The accelerometer, because of its high measurement rate, picks up these instant-by-instant changes in bike acceleration. However, the speed sensor, which gets readings only 1.5 times per second, cannot measure these fine changes in bike acceleration.

So, because slope is calculated by subtracting bike acceleration from total acceleration, the net of all this is that the slope readings are not as accurate on steep hills.

On less-steep climbs, where bike speed is faster, the wheels turn faster, and bike acceleration rates are less variable, the slope measurement is better.

In our testing we have found that PowerPod is dead-on in climbs. You are getting solid power measurements from your device.
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

John thanks for your detailed response, I think I understand what you are describing, but I have to say that I find that a bit frustrating, if unavoidable and I'm not convinced I am getting solid power measurments; basically the steeper the hill the less accurate the reading. Is there no way to factor in what seems to be an inherent issue with the way the slope is calculated, maybe apply an arbitrary correction that depends on the slope gradient?

I have looked at another ride file which included less steep hills (5 to 6%) and you are right, these do correlate better, but there is still a discrepancy on the average gradient with the slope/accelerometer showing a gradient of around 0.3% to 0.5% greater.

For instance the first main hill @18 km ascends 200 m in 3.48 km with an average gradient 5.7% but the slope/accelerometer average gradient is 6.3%; my power is based on riding up a hill 10% steeper than what it was.
123.png
123.png (183.79 KiB) Viewed 14810 times
In this case my average reported power was 331 W for ~11 minutes but in reality it was closer to 310 W which is ~6% out, maybe within the acceptable range of accuracy? So for me it seems, at the speeds I ride at, on a ~6% slope the accuracy is within ~6% and for a 11% slope the accuracy is around 14%. I will just have to factor this in when riding I guess.
Attachments
2017_06_24_1106_95_km.ibr
(763.7 KiB) Downloaded 242 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

I think you are missing the point.

You cannot, and should not, use slope numbers to assess power accuracy.
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

I'm quite possibly missing the point!

My understanding is that the reported power value is partially a factor of the slope value, which on steep hills when traveling slowly, is not as accurate as it usually would be on less steep hills travelling faster?
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

There definitely is a misunderstanding here. Let me try a different way of explaining things.

Gravitational acceleration is a vector that is perpendicular to the surface of the earth. The earth pulls down on everything, towards the center of the earth, with an acceleration of 1G. A free object at the earth's surface, with no wind or other inhibiting force, will increase its speed (acceleration) by 9.8 meters per second per second, in a direction towards the center of the earth.

The 1G gravitation wants to accelerate you ONLY towards the center of the earth. So, if you’re a cyclist riding on a perfectly level road ("level" in this sense means that your elevation above the ground is unchanging), the gravitational acceleration is totally perpendicular to your direction of travel. Thus, on level ground none of the earth's gravity will oppose your forward motion. In short, on perfectly level ground you don't have to apply any pedal force to overcome gravitational acceleration.

Things are different on hills. When you start to climb a hill (similar principles apply on a descent) the earth's gravitational acceleration remains 1G, pointed towards earth. However, because you're on a hill, you are no longer traveling "level"; your elevation is increasing over time (that's why it is called a hill!) and your direction of travel is no longer perfectly perpendicular to the gravitational acceleration vector. Thus, the gravitational acceleration you experience on the bike is no longer perpendicular to your direction of motion. On hills, the gravity vector has two components: one component of acceleration is parallel to the direction of your travel, and one component of acceleration is perpendicular to the surface you are climbing. The vector sum of these two components of acceleration remains 1G, pointed towards the center of the earth.

Therefore, on an uphill, there is a component of earth's gravity that is parallel to the direction of travel, and it will slow you down (deceleration). Don't believe it? Start riding your bike up a hill, stop pedaling, watch your speed go down (decelerate), and get ready to fall off your bike...

The magnitude of opposing gravitational acceleration on hills is mathematically related to the steepness of the hill you are climbing (get out your high school trigonometry books). But the ultimate source of the opposing acceleration force hills is the earth's gravity.

Next, note that acceleration is defined as the rate of change of velocity.

And now, think of acceleration in the “normal” sense: an object whose second-by-second speed is changing, for example a bike going faster and faster because the rider is stomping on the pedals. Every cyclist knows that accelerating a bike requires power. And note that, whenever you change the speed of your bike, you are accelerating your bike in the direction of your travel.

So in fact, whenever you are riding your bike, there are TWO sources of opposing acceleration in your direction of motion: gravity (on climbs), and changes in bike speed. These two acceleration elements together constitute the total acceleration of you and your bike.

The PowerPod uses an accelerometer that measures, precisely and very frequently, your TOTAL acceleration in the direction you are traveling.

And here is the crucial point: because PowerPod accelerometer measures total acceleration, the accelerometer does not, and cannot, distinguish between acceleration caused by gravity, and acceleration caused by changes in bike speed. Einstein got his Nobel prize, in part, because he understood that these two forms of acceleration are indistinguishable.

So, the accelerometer does NOT know the slope of the hill you are climbing. The accelerometer knows ONLY the TOTAL amount of acceleration that exists from moment to moment—and, in fact, total acceleration is what is needed to accurately compute a critical component of power.

As explained in the earlier post, using total acceleration measurements, Isaac computes the slope percentage data from the fast and accurate TOTAL accelerometer measurement, and the slow and less accurate measurement of the speed sensor. Bike speed and hill steepness affect the accuracy of slope measurements, but not the accuracy of total acceleration measurements. Thus, Isaac slope is good, but due to sensor disparities, reported slope is not quite as good as other PowerPod measurements.

In conclusion, PowerPod does NOT use slope measurement percentages to compute power. And because calculated slope calculations are not quite as accurate as other PowerPod measurements, it’s difficult to check power calculations.
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

Again thanks for taking the time for a detailed reply I now appreciate that the power value is a function of total acceleration (and speed etc.) only, and not the slope value itself.

However there still seems to be an issue for me here, and that is that the reported power for me uphill appears to be too high, maybe 20% too high.

The first hill I really noticed this on (the one originally mentioned) is basically 5-minutes long and climbs 110 m in 1.34 km with an average gradient of 8.2% (the Isaac/PP average gradient is 10.5%).
Untitled.png
Untitled.png (129.34 KiB) Viewed 14776 times
For those 5-minutes my average speed was 16.2 km/h, the total weight is 80 kg (me and bike), I have a CdA of ~0.36 m2 and the Coefficient of rolling resistance of my tyres is around 0.004.

Based on the data above and doing a ‘simple calc’ (https://www.gribble.org/cycling/power_v_speed.html) my average power for the 5-minutes should be ~350 W (believable), of which 88% is overcoming gravity, whereas the reported/PP average reported power was 445 W (unbelievable); a significant 95 W/22% difference. However if the average gradient was 10.5% then the average power, based on the simple calc, would be ~435 W (only 10 W/2% off what PP said).

Could the reason for this be due to the fact on steep hills I am out of the saddle and there is more ‘side to side’ bike movement?

Edit: in fact in the image above (and file attached) you can see @1min53s where my power drops instantly by ~100 W, likely where I sat back down on a 'flatter' section, then @3min31s where my power jumps up by ~100W around when I would have got out the saddle again on a steeper section.
Attachments
iBike_07_23_2017_1119_1_km.ibr
(61.27 KiB) Downloaded 240 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

Bike calculator apps like Gribble use inputs such as speed, slope, rider weight, etc. to compute power.

To use these apps properly you have to be very careful to feed them data consistent with their assumptions. The critical assumption in these apps is that the numbers you input are unchanging.

On the interval you showed below, there is a significant decrease, then increase in the slope of the hill during the climb. Similarly, speed drops, then increases, then drops. These changing input parameters are inconsistent with the app's critical assumption, so the app's power analysis won't be correct.

What I did is use your interval, but selected just the portion of it, where slope and speed are relatively unchanging. Basically, this is the first slope "bump" of your interval. Your reported slope is a pretty constant 13.8%, and speed is a pretty constant 14.1KM/h. I also entered the appropriate values of weight and wind speed (note: at these low bike speeds, wind speed is pretty irrelevant).

Your bike speed is low, which as posted above, can create inaccuracies for bike acceleration measurements. However, because the section I selected has almost-constant bike speed, your bike acceleration in this section is nearly zero. This means that the accelerometer's measurement is driven almost completely from hill steepness. So, PowerPod slope measurement can be pretty accurate.

I used my favorite program, bikecalculator.com (it allows for CdA adjustment) but I'm sure gribble would produce similar numbers.

The result? bikecalculator.com: 502W. PowerPod: 495W

I am very confident that, if you had a DFPM on your bike, you'd get results nearly identical to your PowerPod.

I'm beginning to wonder if the "real" problem here is that you are a much stronger cyclist than you might think! :-)

Screen Shot 2017-08-11 at 6.48.26 AM.png
Screen Shot 2017-08-11 at 6.48.26 AM.png (1.02 MiB) Viewed 14758 times
John Hamann
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

To illustrate the point about apps requiring constant values for the section analyzed, I had a bit of fun.

I selected a place with an uphill, then downhill, for equal distances.

On this section, average slope is about -0.13%, and average bike speed is 20.5 km/h There is a slight tailwind of -0.3 km/h

PP reports 153W for the section, but bikecalculator reports only 61W. Why the huge difference?

The app assumes that speed, slope and wind are unchanging. It gives the "right" answer for these assumptions.

The app has no clue that there is a big climb, followed by a big descent. It has no idea that bike speed is much greater on the downhill. And it does not know there is not any pedaling on the downhill section.

So, the app is wrong...not because it can't do math properly, but because the data was fed with data inconsistent with its assumptions.
Attachments
Screen Shot 2017-08-11 at 8.00.49 AM.png
Screen Shot 2017-08-11 at 8.00.49 AM.png (1.97 MiB) Viewed 14751 times
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

Hmm, I am still not convinced! You are using the reported slope of 13.8% to calculate my power for that section, if that is the actual gradient then my power would rightly be ~500 W.

However over that distance of ~400 m I have only ascended ~40 m, going from 30 m to 70 m above sea level, see the image below, so the average gradient is actually much nearer to 10%, and my power ~370 W.
Untitled.png
Untitled.png (290.31 KiB) Viewed 14751 times
The 40 m in 400 m corresponds with the Isaac elevation and distance numbers, and also Ordnance Survey mapping data, and also looking at my Garmin ride fie also that elevation data, I am pretty confident that this is correct?

I appreciate that simple power calcs have limited use, but for one short hill effort only with no downhill gradient, they should be

In case you want to look yourself at this hill its 'Bear Road', in Brighton UK.
slope.png
slope.png (1.33 MiB) Viewed 14749 times
BBQcyclist
Posts: 49
Joined: Mon Apr 10, 2017 8:38 pm

Re: Still not confident regarding accuracy

Post by BBQcyclist »

Therefore, to summarize - Is PP basically "ok" to help estimate general, ball park power - using its multiple "algorithms and assumptions" - especially at its price point?

Notably in STEADY STATE ENVIRONMENTS:
- i.e. unchanging route slopes, constant rider positions (either low in drops vs sitting up on bar flats or 1/2 sitting on brake hoods), constant rider weight and/or associated bike/clothing aerodynamics & weight/backpack & miscellaneous tools weights, not too many changing wind gusts / cross winds, or when pack riding - not too many different pack position changes/turbulences .... etc.... etc...

What did we really expect from a ~$200-270 USD, non-DFPM - which doesn't require changing cranksets or wheels ... and is Campy compatible?

It's ok for the price - especially if one can get past the somewhat outdated, non user friendly software, or instructions, or if one gets a good QC sample - with solid usb charging port and actual 15-20 hr battery life. An on board or head unit display of battery life left - wouldve also been a nice feature.

I'm generally happy with it - at it's pricepoint. At 350 - 500$ or more I wouldve opted for another brand.
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

BBQcyclist wrote:Therefore, to summarize - Is PP basically "ok" to help estimate general, ball park power - using its multiple "algorithms and assumptions" - especially at its price point?

Notably in STEADY STATE ENVIRONMENTS:
- i.e. unchanging route slopes, constant rider positions (either low in drops vs sitting up on bar flats or 1/2 sitting on brake hoods), constant rider weight and/or associated bike/clothing aerodynamics & weight/backpack & miscellaneous tools weights, not too many changing wind gusts / cross winds, or when pack riding - not too many different pack position changes/turbulences .... etc.... etc...
Well, actually, that is very far from what this thread is saying. PP does not need constant slopes and speeds and weights and ride positions; bike calculator apps do. Constant rider position is irrelevant for this thread, because there is essentially no effect of wind on uphill climb watts. Weight is not going to change significantly in a 1KM hill climb. Drafting is irrelevant.

In fact, PP does extremely well in dynamic environments.
John Hamann
BBQcyclist
Posts: 49
Joined: Mon Apr 10, 2017 8:38 pm

Re: Still not confident regarding accuracy

Post by BBQcyclist »

John,

I understand very well this thread being specifically re. "Slope" and power algorithm estimation calculation challenges of the PP.

Just was pointing out some of its other challenges.

As well as highlighting some of its positive points.

Trying to be objective - that's all.

O.p. Sorry.for unintentionally "hijacking" your thread. It's all yours!
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

BBQcyclist wrote:John,

I understand very well this thread being specifically re. "Slope" and power algorithm estimation calculation challenges of the PP.

Just was pointing out some of its other challenges.

As well as highlighting some of its positive points.

Trying to be objective - that's all.

O.p. Sorry.for unintentionally "hijacking" your thread. It's all yours!
Any device that measures anything has challenges. Just ask the people trying to figure out what the temperature outside has been outside through the years!
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

John,

I was out today riding and I was thinking about this problem. I work with accelerometers to measure ground vibration, so have a basic understanding of them. We use triaxial accelerometers with a manufactures sensitivity of 1,000 mV/g, however we do get these calibrated regularly to ensure we get accurate data and I know that the sensitivity is never 1,000 mV/g, on the one I normally use (Dytran 3233A) the three axis are all slightly more sensitive, maybe around 5% to 10%, i.e. a sensitivity of 1,050 mV/g (in fact the manufactures stated sensitivity for this model is +/-10%).

Could it be the case that the accelerometer in my PP is more sensitive than it 'should' be and therefore there is more voltage than there 'should' be and the PP thinks there is more power being generated than there is?

Incidentally I appear to have the same issues on today's ride, that is the reported power appears to correlate well with the reported slope value, not the actual gradient and my power uphill is not correct and the steeper the hill the more off it is.
ronpei
Posts: 120
Joined: Sat Jun 08, 2013 7:47 am

Re: Still not confident regarding accuracy

Post by ronpei »

it is more than likely a mass market accelerometer like the ones in phones and PP would not be 100% true.. however the one thing that people get head over heals about is accuracy... it is very difficult to prove that any power meter is accurate since power is a derived measure and not a direct measurement. So unlike a weigh scale where you can take a calibrated mass place it on the scale and directly calibrate the scale, that is a direct measure, however since all power meters must take several measures to mathematically calculate power if any of these measures are in accurate then the result is inaccurate. So called direct force power meters still have to measure speed and distance as well as angular force (another multi factor measure) to calculate power. If any of these drift for any reason then so does the result. The more important issue is precision, or how well it measures the same thing day after day, and for anyone who is training with power that is far more important. This is the only way you can determine if your training is benefiting you. If the precision is poor then there is a possibility that your decision regarding training could be flawed. So whether the accuracy going up hill is true is nearly impossible to determine, all one can say is that if 2 meters on the same bike during the same ride are relatively the same then they compare well, it is impossible for normal people to determine if they are accurate. Even calibrating a DFPM is only a close approximation and is very dependent on what you use to calibrate and how accurate that is. So it becomes a circular argument. Your comparison of an online power "estimating" tool is the same as counting cricket chirps to tell temperature, it might work but all assumptions must be met. So the bottom line is that all power meters are challenged to be accurate, and the precision is a cost factor, to get better sensors cost money so if you want more precision your gotta pay for it. I think the PP is a good device and the cost/ benefit is 100% in line. I know there are a number of assumptions that may not be met 100% on every ride but at the end of the day it works and seems to provide the measure at a fair price to performance ratio. So it may be true that your PP is not 100% accurate going up hill, but the real question should be is it precise?!
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

The problem isn't so much the inaccuracy, although the 35% discrepancy in the 400 m section posted above isn't exactly great!

The problem is consistency and it seems as though the reported value is more and more inaccurate as the slope increases and gravity is the dominant factor (~90%). Then the acceleration is more and more dominant, so the error becomes more and more dominant.

400 W on a 5% hill at x km/h isn't the same 400 W on a 10% hill at y km/h.

I'm also fairly sure riding out the saddle causes the power value to increase even though my power hasn't changed, possibly because my bike is then accelerateing side to side.
ronpei
Posts: 120
Joined: Sat Jun 08, 2013 7:47 am

Re: Still not confident regarding accuracy

Post by ronpei »

"The problem is consistency and it seems as though the reported value is more and more inaccurate as the slope increases and gravity is the dominant factor (~90%). Then the acceleration is more and more dominant, so the error becomes more and more dominant. "
if there is an error due to the sensor which gets multiplied then the error may be consistent but multiplicative, still I bet it is consistent and reproducible, if not that is a huge problem, otherwise it is consistent, which is what precise means, if you measure the same thing 2x and get the same result it may be inaccurate but it is still precise.
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

Sorry for the delay in responding; I have been traveling.

Yesterday afternoon I spent a fair amount of time investigating this situation. Thanks to Google Earth I've become a bit of an expert about the elevation profile of Bear Rd... :-)

I also found, separately, a topographic map of the area!

And I looked at the ride file, in detail.

After looking at all this information, it appears to me that, for the 400 meter section of ride pb21 has posted (first part of Bear Rd climb), the PP accelerometer slope reading is higher by 2.5% compared to the GE slope of 10.7%. The PP measured 480W for this section; bikecalculator.com calculates 400W, a 20% difference.
first section of Bear Rd climb
first section of Bear Rd climb
Screen Shot 2017-08-12 at 2.36.07 PM.png (1.28 MiB) Viewed 14661 times
I note, however, that in the second part of the hill climb, which according to GE is even steeper (14.3% on GE), PP reads 14.8% for the same section--very close, with watts within 2.6% of each other (549 vs 535).
GE elevation profile of second section of Bear Rd climb
GE elevation profile of second section of Bear Rd climb
Screen Shot 2017-08-12 at 2.20.11 PM.png (2.65 MiB) Viewed 14661 times
power for second climb
power for second climb
Screen Shot 2017-08-12 at 3.02.59 PM.png (1.19 MiB) Viewed 14661 times
I have 4 thoughts:

1) I am intrigued by the possibility that side-to-side motion is contributing to the additional accelerometer reading. This should be a simple thing to test--ride up the same road, without standing up! Most assuredly you won't achieve the same level of power, but what we're looking for here is possible differences in slope (accelerometer) measurement

2) Ronpei raises the interesting question of accelerometer calibration. We don't use the cheap accelerometers used in smartphone applications, but it is possible that this particular accelerometer is out of calibration. We will know more if the testing in 1) is done.

3) If side-to-side motion is causing the increased accelerometer reading, this begs a question: could the PP reading be BETTER than we might think? Specifically, cyclists know that side-to-side/out of saddle riding is inefficient and requires more effort than riding in the saddle. Additional energy is expended doing all that wiggling around. If the accelerometer is capturing some of that wiggling energy, then that might explain the slope difference, and the extra watts.

4) It is unreasonable to conclude that there is a 20% discrepancy (not 35% as suggested; see image below) in the 400m section (482 vs 400). We are using GPS/topo map information, and baro sensor readings for determining the "correct" slope. I'm personally not comfortable with either measurement for stating definitively the slope of this section.

Finally, I think ronpei's point is well-taken: if you're riding in a consistent manner, on the same roads, PP readings will be similarly consistent.

For me this has been a very interesting thread and I'm glad this conversation is going on!
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

John

I think the Google Earth numbers are wrong for that section. Based on Ordnance Survey elevation data for that section noted above the elevation gain is ~23 m (from 96 m to 119 m), this also corresponds with the PP barometer, my Garmin barometer and my perception riding it! This section is ~200 m so the average gradient is ~11% not the 14.8% reported by PP (or the 14.3% based on GE numbers). On the basis of OS survey data, and the PP and Garmin barometers I am very confident that the gradient is ~11 % and my power should be much closer to 390 W and not 550 W.
1.png
1.png (666.98 KiB) Viewed 14641 times
2.png
2.png (457.23 KiB) Viewed 14641 times
I will head out later and ride the hill twice, once sitting the whole way, once standing the whole way. I will also try and maintain the same speed for both efforts...

Edit: it's also worth pointing out that my FTP is ~300 W (based on 10/25 mile TTs and specific indoor testing). I would expect 5-minute power to be ~360 W based on that value being ~120% of FTP.
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

So I have just been out and I did a couple of 'tests' and to be honest the results aren't very encouraging. The first was a short 500 m hill with an average gradient of 5.2%. I rode it four times, seated, standing, seated and standing, each time I tried to ride it at a similar speed, this occurred at 5.5 to 9 km in the ride. PP/Issac results below:

Seated 1:

Speed: 18.4 km/h
Reported Slope: 5.9%
Reported Power: 305 W
Expected Power: 290 W

Standing 1:

Speed: 16.1 km/h
Reported Slope: 8.5%
Reported Power: 362 W
Expected Power: 249 W

Seated 2:

Speed: 18.4 km/h
Reported Slope: 4.9%
Reported Power: 260 W
Expected Power: 263 W

Standing 1:

Speed: 17.2 km/h
Reported Slope: 9.4%
Reported Power: 427 W
Expected Power: 256 W
3.png
3.png (106.23 KiB) Viewed 14623 times
I should note that I tried to do the seated efforts very smooth and the standing efforts more exaggerated than normal, particularly the second standing effort. Nevertheless it seems fairly clear that riding out of the saddle increases the acceleration the PP measures and therefore power is massively exaggerated.

At the end of the ride I rode the original Bear Road hill seated the whole way. The actual gradient of this hill is ~8.2% for 1.34 km. Originally I rode this standing and seated and this had a reported slope of 10.5%. Today I rode it seated and it had a reported slope of 9%. Its worth noting that whilst I was seated this is a tough hill and I was holding the bars tight and there would have been some movement of the bars.

So there is at least one issue and that any acceleration of the PP is assumed to also be movement of the bike and person riding the bike, in addition this is likey compounded by an inherent variation in accelerometer sensitivity not taken into account.

You will also see that on the long hill the reported power from today is much 'smoother' with less of the jumps in power where I either sat down or got out out of the saddle on the original ride.
4.png
4.png (106.22 KiB) Viewed 14623 times
Attachments
2017_08_13_1033_75_km.ibr
(575.01 KiB) Downloaded 148 times
Last edited by pb21 on Sun Aug 13, 2017 10:04 am, edited 1 time in total.
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

Thank you for your new PP data!

1) Your Bear Rd., climb 1, shows PP 11.1% grade (vs GE 10.7%)

2) Your Bear Rd., climb 2, shows PP 14.4% grade (vs GE 14.3%)

So, it would appear that, when you're seated, PP is getting the slope readings "right". It also shows that the second climb is, in fact, steeper than the first...

It would also appear that, for your style of out-of-saddle riding, PP slope (accelerometer) readings are higher when you are out of the saddle.

When you're out of the saddle your bike frame is flexing, and your body and bike are tossing all around...There is no doubt that this requires extra power.

The interesting questions are: how much extra power are you expending when you are out of the saddle? And, is this extra power being "captured" by the accelerometer?

I honestly don't know how to answer these questions with high confidence. A crank-based, both-leg DFPM might be able to measure the amount of power going in to the drivetrain (I'm not sure how the strain gauges respond when they're being knocked-around side to side). Most definitely, a one-leg PM or hub-based PM would NOT do the job.

I think it would be very interesting to see your PowerStroke data, which captures side-to-side and front-to-back accelerations. As thanks for the extra work you've done, I've attached a PS key below. Hopefully you will post some new rides...

We most certainly will take to heart your very interesting information, and use it to guide us in future product development activities.

In the mean time, ronpei's point remains valid: if you're doing your rides in a consistent manner, your PP will report data in a consistent manner.
Attachments
APowerStroke_ab7e1136.enc.zip
(266 Bytes) Downloaded 175 times
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

John

Thanks for the PowerStroke key, very generous of you, I will download it and will definitely post the results.

I was thinking that the extra acceleration could be taken into account if it can be differentiated, otherwise the calculated power is presumably based on the total acceleration and the total weight. If I held the accelerometer in my hand and shook it as I cycled up the hill the measured acceleration level would be massive and only a small amount associated with my actual movement of 85 kg, the rest associated with the movement of my hand.
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

I've just been out for a quick test ride which involved riding up a nearby hill (elevation gain of 39 m over 600 m, 6.5%) 5 times. The hill has a fairly constant gradient and I tried to ride it at the same speed (14.5 km/h) the entire way up for each effort.

1. Seated (14.6 km/h 258 W based on gradient vs 285 W based on PP - 10% difference)
2. Standing (14.2 km/h 250 W vs 337 W - 35% difference)
3. Seated (14.5 km/h 256 W vs 278 W - 8% difference)
4. Standing (14.0 km/h 246 W vs 324 W - 31% difference)
5. Seated trying to be very smooth (this was ridden slower in a lower gear, 13.2 km/h 231 W vs 240 W - 4% difference)

For both the seated and standing efforts the reported gradient, 7% and 9% respectively, is higher than the actual gradient, also the reported average power is also higher for the standing efforts by ~50 W despite been ridden at the same speed. I'll let you look at the data yourself though!

Looking at the PS data none of the efforts look particularly 'good', but I suspect the standing efforts are off the scale?! I'm also not fully clued up on this.

Is this movement of the bike the cause of higher power values than otherwise anticipated, looks like it is, but am I really 'wasting' 50 W when standing out of the saddle on this hill?

I need to do some more efforts, particularly of Bear Hill...
Attachments
2017_08_13_1655_8_km_HiDef.ibr
(228.34 KiB) Downloaded 161 times
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Out-of-saddle measurements

Post by Velocomp »

The data is really interesting, and helpful to understanding better what is going on:

1) In the saddle, your PowerStroke looks good: 3 cm side to side, 1 cm front-to-back, 8 wasted watts
In-the-saddle PowerStroke
In-the-saddle PowerStroke
Screen Shot 2017-08-13 at 2.51.25 PM.png (209.4 KiB) Viewed 14585 times
2) When out of the saddle you have a huge 14 cm side-to-side sway and 6 cm front-to-back sway; honestly, the largest sway I've ever seen in a ride file, by at least a factor of 3.
Hill 3
Hill 3
Screen Shot 2017-08-13 at 2.24.53 PM.png (362.01 KiB) Viewed 14585 times
The impact of such a large sway on accelerometer readings is something we have not previously studied, but obviously it is having a material effect. Such a large sway may also be having an adverse effect on your riding efficiency. PowerStroke is estimating 175W of wasted watts (I don't think it is really that much, but it is definitely something to look at)

3) When using PP for slope measurement comparisons, it is extremely important to pick locations where the slope is not changing, and where bike speed is constant. You have some good data available that satisfies both conditions. You can see a clear break (reduction) in watts near the top of the hill. This is where slope goes down slightly. However, for the prior 430 meters, slope is nearly constant, and you've done a great job of holding constant bike speed (no bike acceleration effect). So, this section of the hill is perfect for slope measurement comparisons. (The data in hill 1 is OK, but nearly perfect for hills 3 and 5). Of course, the out-of saddle data on hills 2 and 4 is not used... :-)
Out-of-saddle PowerStroke
Out-of-saddle PowerStroke
Screen Shot 2017-08-13 at 2.09.13 PM.png (320.91 KiB) Viewed 14585 times
For your three, In-the-saddle efforts (hills 1,3 and 5), PP measures:

Hill 1: 30 m climb, a 7.57% slope (note there is some bumpiness in watts near the top of the climb).

Hill 3: 30 m climbed, over a distance of 430 m, or a 7.39%. Very smooth watts

Hill 5: 30 m climb, 7.38% slope. Very smooth watts

Hill 1 has some variation in watts near the end, but hill 3 and 5 are very, very smooth. So, for PP, I've used 30 m rise (from baro sensor), 430 m length (from speed sensor), and 7.385% slope (from accelerometer, hills 3 and 5)

Using the baro sensor and speed sensor data, you climbed 30 m over 430 m, which yields a slope of 7.0%, somewhat lower than the 7.385% slope measured by the accelerometer.

But here's the thing: if the distance climbed were measured to be 31.7 m, instead of 30 m, PP and baro slopes would be identical.

For baro elevation readings to be precise, the pressure inside PP case needs to be identical to the outside pressure. Under conditions of rapidly changing pressure (which happens on any steep climb), it takes about 5 seconds for internal and external pressures to equalize; consequently, the baro sensor elevation will lag behind actual elevation changes.

You climb 30 m in about 107 seconds. Each 5 seconds you climb about 1.4 meters. So, what may be happening here is that the baro elevation/slope readings are not updating fast enough to keep up with the accelerometer slope readings.

I think we agree that your out-of-saddle riding is causing high accelerometer readings. And my view is that the in-saddle accelerometer, slope and power readings are correct.
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

I was wondering that as the PP can detect ‘side to side’ and ‘front to back motion’, could the acceleration associated with this movement be easily subtracted from the acceleration value used to calculate power and slope? :geek:
oxenhoper
Posts: 31
Joined: Sun Mar 26, 2017 10:23 am

Re: Still not confident regarding accuracy

Post by oxenhoper »

This is a great informative post.. enjoyed reading it.

As a TTer i personally quite like the side to side front and back analysis as it lets me know how smooth i've been on an event
Velocomp
Velocomp CEO
Posts: 7793
Joined: Fri Jan 11, 2008 8:43 am

Re: Still not confident regarding accuracy

Post by Velocomp »

pb21 wrote:I was wondering that as the PP can detect ‘side to side’ and ‘front to back motion’, could the acceleration associated with this movement be easily subtracted from the acceleration value used to calculate power and slope? :geek:
Yes, I have been thinking about this! The thing is, we would need to get some DFPM measurements as well. My intuition tells me that the sway motion is causing more work when out of saddle.

If you could borrow a DFPM, then do some climbing with it, I think we could make some improvements!
John Hamann
pb21
Posts: 32
Joined: Sat May 06, 2017 12:14 pm

Re: Still not confident regarding accuracy

Post by pb21 »

This evening I got out on a ride which included the Bear Road hill. I was a bit slower this time, but the result is essentially the same as previously. I have attached two ibr files, one for the whole ride and one for the Bear Road section only (which occurs at 24 km in the ride).

Interestingly I was looking around the software and found the 'Tools>Power Meter Comparison' feature and that within this you can select an equation of motion to compare it to. Selecting 'speed, wind & elev' option gives the following chart (it seems the elevation used to calculate power is based on the barometer data rather than the calculated slope). Usefully this shows the overestimation at the beginning when I am out of the saddle, the middle flatter section where the two are much more closely aligned when I am riding in the saddle before the final steeper section where they diverge again.
5.png
5.png (35.11 KiB) Viewed 14483 times
I also looked at this for the whole ride and generally the two align quite nicely, although the equation of motion tends to be always be a bit lower, particularly on hills of all gradients and whether I rode them sitting or standing, although the difference is always higher when standing.
Attachments
2017_08_16_1830_1_km_HiDef.ibr
(322.97 KiB) Downloaded 147 times
2017_08_16_1738_36_km_HiDef.ibr
(777.67 KiB) Downloaded 151 times
Post Reply