Stages Power Meter Compared (my best DC Rainmaker impersonation)

[gewichtweenie]

We get it. You're an engineer. Dont know why you think that's such a special title that requires you to go on a pendantic parade in every thread to prove yourself.

Here are the 2 plausible reasons why a stages pm drift from a trainer at different cadence's:

(1) higher torque creates more slip (for those with tire-on trainer)
(2) higher torque reveals more left-right imbalance from rider ...which is picked up only by the 1-sided stages

And since bike computers only display cadence, and cyclists only talk of cadence, no need to remind everyone of grade school physics. Because the practical real world results means a stages PM might show different variance to a trainer at differences cadences and different loads, which is masked when comparing avg values across a complete ride, and becomes misleading when one assumes they "match" perfectly while turbo-training in different conditions

I have no doubt you’ve seen one meter read higher than another, but cadence isn’t the reason. It can’t be.

I’m about to explain why, but I should warn you that I design scientific instruments for a living, so I think about this subject way too much.

TL;DR: Cadence can’t explain different readings because of the way power is calculated from what these meters actually measure (torque). While there are a lot of reasons one meter might read higher than another in the short term, if the two readings converge over the course of a ride, the meters are effectively reading the same torque, and that’s what matters.

______________________

Ok; on to the gritty details:

Power is torque times angular velocity (omega). In other words, P = T ω. Omega is conventionally in units of radians per second, but that’s a straight unit conversion from RPM. (E.g., 100 RPM = 10.472 rad/s).

For two meters on the same bike, ω (cadence) is the same, because angular velocity is trivial to measure and doesn’t drift. So if the meters are giving substantially different readings, they’re necessarily measuring different torque values.

All meters do some signal processing to turn their strain gage readings into torque measurements, and the details of how they do that are part of what cause different meters to read differently at any given moment. For example, one meter may have a different thermal compensation algorithm than another, causing readings to diverge with temperature. Smoothing is a big deal in signal processing, and discretization error can affect the power displayed on your bike computer (though not as much at the meter) for short periods of time.

But the fact that two meters have measured a very similar amount of work done over the course of a ride (kiloJoules, which is essentially average power) means that they agree about torque (and they necessarily agree about cadence).

Going back to the formula that all power meters use to convert strain measurements to power: power is torque times cadence. If, at a given cadence, two meters disagree, that’s because they’re measuring a different torque. And if they’re measuring different torques beyond a certain margin of error, one or both is miscalibrated.

A power meter with a miscalibrated slope would be off by the same amount (proportionally) regardless of cadence, so you wouldn’t see of a difference there due to cadence. A power meter with a miscalibrated offset could read high at low torque and low at high torque or vice versa. But the error is torque-dependent, not cadence-dependent.

Keep in mind that most powermeters change their offset (their “zero”) multiple times over the course of a ride. Thermal compensation demands this. Meters try to do this when you’re not pedaling, ideally when stopped at a light. But many people load up their drivetrain slightly at lights. If you were stopped at a light with your left foot on the ground and your right leg resting on the pedal at 90 degrees to TDC, a Stages meter would get an accurate offset at that stop while a Qarq would get a slightly erroneous one due to the weight of your leg on the right crank. This would cause the two meters to give slightly different readings until the next zero.

The point is, there are many, many reasons why one meter would disagree with another, but cadence isn’t one of them. And if two meters agree over the course of a ride, they basically agree on torque, and since that’s what they’re measuring, you can’t really ask for more than that.

I bet you’re sorry you asked.

[youngs_modulus]

Well, because your cadence is zero in a trackstand and power is torque times angular velocity/cadence, your power output in a trackstand is zero by definition. Your Stages will read zero in a trackstand.

Powertaps can be configured to auto-zero during a ride, but Stages can only be manually zeroed, and they recommend doing so before each ride. So your Stages isn't trying to auto-zero during your trackstands. I have a Powertap, though, and I've got mine configured to auto-zero. I've stopped doing trackstands at lights for just this reason.

(Zero/auto-zero and offset calibration are the same thing. Setting the slope is a much bigger deal; many power meters don't let users set the slope).

Stages meters include accelerometers...they'll only allow you to set the offset when the left crankarm is at BDC (bottom dead center). You could try zeroing your Stages at BDC while putting some weight on that pedal, and you'd see a different offset value afterwards. Also, your meter would be wildly inaccurate until you did another zero with no load on the crankarm.

[youngs_modulus]

We get it. You're an engineer. Dont know why you think that's such a special title that requires you to go on a pendantic parade in every thread to prove yourself.

You misspelled "pedantic." (That's a joke, son).

I geek out on this stuff for sure, but you shouldn't take it personally. My geekiness has driven me to work with Jason at Fairwheel on his crank and handlebar stiffness testing. I wrote technical sidebars for those tests, going as far as running a finite element simulation to put an upper bound on losses due to crank flex. Super-nerdy, right? Well, those super-nerdy things are fun for me. I get excited about finding geeky answers to geeky questions, but I try to own that.

I'm not the only one, either: Dan Connel (djconnel) does awesome plots of everything from hillclimb results to frame geometry; Damon Rinard, who posts here occasionally, published solid wheel deflection test results and built a carbon beam bike in his garage. Tom Anhalt is internet-famous for his meticulous tire rolling resistance tests. Coincidentally or not, we're all engineers. Bikes attract engineers like shiny objects attract crows; we're overrepresented on boards like this.

When I write these dense technical posts, I try to make them germane to the thread and comprehensible to any non-engineers who might be interested. I'm not trying to insult you or be uncivil, so I don't understand why you seem so cranky. If you don't like my posts, please ignore them. If you choose not to, well, remember the old joke: arguing with an engineer is like mud-wrestling with a pig: after a few hours, you realize that the pig likes it.

Here are the 2 plausible reasons why a stages pm drift from a trainer at different cadence's:

(1) higher torque creates more slip (for those with tire-on trainer)
(2) higher torque reveals more left-right imbalance from rider ...which is picked up only by the 1-sided stages

Well, the OP's Stages output didn't drift from his trainer's output, so I'm not sure where you're going with this. Trainers with slipping tires aren't relevant to the OP as his trainer is chain-driven. But tires on trainers (slipping or otherwise) absorb a lot of energy: maybe 15-40 watts; that's why they get warm. So you'd expect a tire-based trainer to read substantially lower than either a crank-based meter or a non-tire trainer. But then the average power numbers wouldn't match either: you could see the difference in the meters by looking at average power. A slipping tire would just magnify the difference in average power.

Your second point could be valid for a rider with unusual physiology using a one-sided power meter. That seems like a corner case to me, but sure.

If you look at the OP's numbers, his Stages and his trainer give almost exactly the same numbers. The average power is identical, while the normalized power numbers are within 0.5% of each other. From this, we can conclude that:

• The OP's L/R balance is very close to 50/50;
• Both meters agree across of a range of intensities (average power); and
• Agreement is very good at high intensities as well (normalized power).

Your claim (as I understood it) was that workout average power doesn't tell you much about whether one power meter corresponds to another. You seem to think that's because power meter accuracy is somehow cadence-dependent. I disagreed that it was cadence-dependent and tried to explain why it really can't be. In fact, the math says that two meters that return the same average and normalized power for a given workout have comparable precision and accuracy.

I understood you to be saying that wasn't the case, and that's how our little exchange started.

And since bike computers only display cadence, and cyclists only talk of cadence, no need to remind everyone of grade school physics.

Are referring to my mention of radians per second vs RPM? I wasn't trying to patronize you; I was trying to show how an abstract formula relates to the real-world terms cyclists use.

Because the practical real world results means a stages PM might show different variance to a trainer at differences cadences and different loads, which is masked when comparing avg values across a complete ride, and becomes misleading when one assumes they "match" perfectly while turbo-training in different conditions

No, power reading correlation between two meters doesn't change as a function of cadence or rider power unless one of the meters is broken. Average power doesn't somehow mask variance, but even if it did, normalized power, which gives extra weight to intense efforts, would emphasize that variance. Both those quantities matched well in the OP's data.

You're right that one-sided meters are not going to pick up on a physiological freak whose right leg suddenly gets stronger or weaker after 113 RPM. Maybe I'm mistaken, but I suspect riders like that are pretty rare.

The OP's preliminary data (average and NP) match well, and he hasn't yet posted the raw data, so I don't understand why you keep explaining their divergence. The OP said he'd post both the raw data from this run and equivalent results comparing his Stages meter to his Powertap. I'm looking forward to seeing what those data look like.

[Delorre]

I've done the same comparison over and over again (Stages vs Elite Drivo), and although very consitent, I have a +/- 20w diffrence between both. Stages reading 20w lower. And that 20w difference remains, be it for avg power over 1 hour or more (let's say 200w) as for an 20' FTP test I did (325w on the Elite, 308w on Stages). As I ride with power from my Drive on screen of the pc and the power of my Stages on the Garmin, I have a permanent view on the difference, and even at 150w (or 450w) I see the same +/- 20w. I tend to belive the Drive more, but if there is a difference, it would be more logic that the Drivo reads lower as there may be drivetrain power loss between the crank and the Drivo. But no, it's the other way around. And yes, I zero-reset my Stages before every ride. Is there any logic to this? If it has to do with L/R balance, the difference would grow with power increase, with is not the case here

[youngs_modulus]

Oh, that's interesting! I agree that if the difference were due to L/R balance, you'd see it increase with your effort. The fact that it's a static ~20W difference strongly suggests that zeroing the Drivo might solve your problem.

I did some homework on the Drivo and found some interesting details. It seems that rather than using resistance-based strain gages, they're using an optical occlusion system. If I understand how they're measuring torsional strain, their method would be hard to use in a mobile system but should be pretty solid as a trainer-based system.

I almost immediately found a thread in their support forums started by someone in almost exactly the same boat as you: his Stages differs by a constant 25-30 watts from his Drivo, except his Stages reads higher than the Drivo, not lower. Here's a link:

http://forum.elite-it.com/viewtopic.php?f=20&t=212

In that thread, someone from Elite walks him through the process of resetting the trainer's zero. Have you tried that? If not, would you mind letting us know what happens when you do?

[Gearjunkie]

@ Youngs Modulus: I'm no engineer so I have no idea if your posts are accurate but it seems to me that you do a very good job of explaining the science to plebs like me. You seem authoritative and persuasive, plus nice. Which is nice

So thanks for posting, I for one find your posts interesting and valuable.

One question, DC Rainmaker seems to suggest that his L/R balance shifts with power and that left only power meters (such as Stages) are therefore less accurate (and less precise?), at least for him. How does that fit with your understanding?

Thanks again (in advance).
GJ

[glepore]

If you test and train with the same pm AND you don't blindly set levels as a % of ftp ( see the 4dp thread in training as to why this doesn't work for everyone) then this doesn't matter. Test ( or measure using software of choice) reported power at training intervals. Set workouts accordingly. The Stages is precise, the *2 algorithm may not lead to accurate total power at different intensity if you are like Ray but testing at various durations fixes that.
Most of these issues are theoretical not practical.


Sent from my iPhone using Tapatalk

[TobinHatesYou]

In this thread:
Op shares data and 8 guys immediately hijack thread to go on rants unrelated to the dataset

He mentioned people who dump on Stages, so I figured I'd say why I don't really care for mine...and it has nothing to do with data or accuracy. If battery life were better and the Gen2 enclosure were improved further, I would definitely think more highly of my unit. The left-only nature does not bother me so much at this price point, and Stages support has been very good. Unfortunately while Stages broke some barriers to entry, the race to the bottom that they started has kind of squeezed them out. 4iiii's unit is cheaper, albeit without temperature compensation. The left-only pedal options are more portable and offer an easier path to dual-leg later.

[Delorre]

Oh, that's interesting! I agree that if the difference were due to L/R balance, you'd see it increase with your effort. The fact that it's a static ~20W difference strongly suggests that zeroing the Drivo might solve your problem.

I did some homework on the Drivo and found some interesting details. It seems that rather than using resistance-based strain gages, they're using an optical occlusion system. If I understand how they're measuring torsional strain, their method would be hard to use in a mobile system but should be pretty solid as a trainer-based system.

I almost immediately found a thread in their support forums started by someone in almost exactly the same boat as you: his Stages differs by a constant 25-30 watts from his Drivo, except his Stages reads higher than the Drivo, not lower. Here's a link:

http://forum.elite-it.com/viewtopic.php?f=20&t=212

In that thread, someone from Elite walks him through the process of resetting the trainer's zero. Have you tried that? If not, would you mind letting us know what happens when you do?

Thx for the info!! I had to mess around a little with the app and the BT connection : With the Drivo , if anything ant+ is connected, no way to connect the smartphone app anymore. You first have to connect the smartphone , before connecting the Edge/ Zwift or other. Long story short : I was able to do the calibration, but after the first time, almost no change at all. I redid it 3 times, and now, from a very short test (I'm in my 2 weeks after season "off"), things seems more in line with each other, but I will let you know after a proper ride, were my Garmin records the Stages and Zwift records the Drivo power readings

[Marin]

rants unrelated to the dataset

[savechief]

Finally, some data. Sorry for the delay, there was an issue getting the files imported with a Mac running Safari:

https://analyze.dcrainmaker.com/#/publi ... 8d57d6e08f

My Drivo is set via the Elite app to 10 second smoothing. I also used 10 second smoothing within Ray's DC Rainmaker Analyzer tool. Once you click on the link, you can use your mouse to zoom in.

Full Interval Session


First 5 Minute Interval


Second 5 Minute Interval


Third 5 Minute Interval

[Delorre]

Oh, that's interesting! I agree that if the difference were due to L/R balance, you'd see it increase with your effort. The fact that it's a static ~20W difference strongly suggests that zeroing the Drivo might solve your problem.

I did some homework on the Drivo and found some interesting details. It seems that rather than using resistance-based strain gages, they're using an optical occlusion system. If I understand how they're measuring torsional strain, their method would be hard to use in a mobile system but should be pretty solid as a trainer-based system.

I almost immediately found a thread in their support forums started by someone in almost exactly the same boat as you: his Stages differs by a constant 25-30 watts from his Drivo, except his Stages reads higher than the Drivo, not lower. Here's a link:

http://forum.elite-it.com/viewtopic.php?f=20&t=212

In that thread, someone from Elite walks him through the process of resetting the trainer's zero. Have you tried that? If not, would you mind letting us know what happens when you do?

Thx for the info!! I had to mess around a little with the app and the BT connection : With the Drivo , if anything ant+ is connected, no way to connect the smartphone app anymore. You first have to connect the smartphone , before connecting the Edge/ Zwift or other. Long story short : I was able to do the calibration, but after the first time, almost no change at all. I redid it 3 times, and now, from a very short test (I'm in my 2 weeks after season "off"), things seems more in line with each other, but I will let you know after a proper ride, were my Garmin records the Stages and Zwift records the Drivo power readings

Did an easy 100min ride on Zwift, and without going into details, I still have 13w difference in AVG power, Stages reading lower. Have to say, avg looks less bad than real time comparison, where I saw more something like the previous 20w delta. So be it, It will not keep me awake at night, and I trust the Drivo more. Only 'issue', is that I have 2 ftp's to keep in mind...

[TobinHatesYou]

Did an easy 100min ride on Zwift, and without going into details, I still have 13w difference in AVG power, Stages reading lower. Have to say, avg looks less bad than real time comparison, where I saw more something like the previous 20w delta. So be it, It will not keep me awake at night, and I trust the Drivo more. Only 'issue', is that I have 2 ftp's to keep in mind...

At 250W average, my Stages typically reads a few watts higher than my Hammer, but again that correlates to what my Vector 2s have been telling me about my L/R balance.

[savechief]

Regarding other Stages issues that have been brought up besides the accuracy of doubling left-leg power:

1) Housing/water ingress: I have a Gen 2 stages, and I live in Southern California, so rarely have to deal with rain. As a result, I have not experienced this first-hand. I will admit that the o-ring is a bit fiddly, and likes to fall out if you're trying to replace the battery with the power meter vertical.

2) Battery life: I know that some people have had their battery drain very quickly. While I have not calculated how long my battery has typically lasted, I have not been disappointed in the battery life, which is maybe why I've never felt the need to check actual performance against manufacturer's claims.

The problem that I did have that was easily solved was inadvertent battery contact, which would result in the power meter not coming on at all. Tweaking the two metal tabs fully solved this issue.

[bilwit]

Everyone always complains about one-sided power meters (particularly Stages) but the fact is that it has a few Grand Tours under its belt so it's more than good enough for all-world riders/teams and regardless of "true" accuracy, the results are normalized to that individual PM so the only real issue with training with something like this is how its results translate to other PMs you might be using in other setups, in which case it would be better to keep separate FTP for each one (admittedly, would be a hassle). I guess for Zwift superstars this might be an issue since you would want the maximum amount of watts accounted for because it directly relates to everyone else's readings.

Finally, some data. Sorry for the delay, there was an issue getting the files imported with a Mac running Safari:

https://analyze.dcrainmaker.com/#/publi ... 8d57d6e08f

My Drivo is set via the Elite app to 10 second smoothing. I also used 10 second smoothing within Ray's DC Rainmaker Analyzer tool. Once you click on the link, you can use your mouse to zoom in.

...

well.. are you going to punch in the numbers for us to get the accuracy%? (expected-actual)/(expected)*100? You could also do an accuracy plot as a function of power and again with time. Just off hand it looks plenty accurate to me.