You could make friends with someone who works in a university campus lab and have them weigh some barbell weights on one of their scales. You don't really need sub-gram accuracy since most power meters only report torque in 0.1 ft-lb or 0.1 Nm units anyway.


I suppose you could do the same thing at the post office or FedEx or UPS office, too. Ultimately, those scales need to be checked and calibrated reasonably often.

I've checked calibration on several PM's with standards traceable to NIST. Then again I work in a 16949 certified lab as well.

I used to have a rig that mounted to a ISIS BB (doing Powertap back then) but that went away when we moved in 2012 to our next facility. Used a Himmelstein rotory torque transducer. I honestly didn't find any discrepancy with the PM's I had.

I did my 650C tire testing on it too.

I realize what you are saying, but I am actually talking about accuracy, not what is OK, but as I said I come from a world where calibration is important and we do worry about these things. So what you say is fairly true.

but, that might be a start, you now have a calibrated mass, but, then you need to work with distance and time. Time probably no real issue, and then you have the distance issue... if that is calibrated to a traceable standard OK then we are on the right path. Not to overtly dis university research labs, but very few really understand calibration. Even fewer understand traceability.

Then regardless of these things we are still in the issue of the system for this to be truly validated and traceable so that any and all power meters can trace to a common standard. If one is a strain gauge do you simply measure force? This is most likely an intellectual exercise, probably no one is ever going to really do it the the NIST level I suspect but ....... what we have right now is never never land.. just because a power meter agrees with another power meter does not mean it is accurate unless that power meter is traceable to a known validated standard.

So not any argument about what you say, I am just having fun thinking about if we really go to the point of a true validated traceable standard that could be use as the basis of comparison.

I think folks would like a unified standard where an app could be used via bluetooth to at a minimum provide a generic offset or an offset by ranges (0w for 0-100w, 5w for 100w-200w, 10w for 200w-400w, 15w for 400w+).

Look, if folks want to do that to cheat riding outdoors and appear stronger, they'd be cheating themselves. Indoors Zwift races could still demand the equipment setup they want to avoid "offset cheating". I could only foresee ever needing a max of 20w anyway.

Just a random thought.

that is pretty cool stuff. It would be interesting if the companies who made the power meters used a calibration tool that was traceable... so we know that when it left the factory it has a traceability certificate, it was compared to the standard and had certain characteristics.

Static weight tests only test a small fraction of the sources of error, though. There can errors due to temperature changes, cadence errors, vibration, variations in equipment properties (e.g. the Shimano asymmetric arm issue), variations in sampling rate, etc, just to name a few.

To the OP's point a "NIST-grade" calibration would be far, far more than just hanging a weight and calling it good.

Though we're mostly just doing hobby grade stuff, and even the scientific papers just use SRM or various ergs and plausibly call it ground truth, the maturity of the industry is maybe getting to the point where some real calibration practice might be useful.

What we have now is like DC Rainmaker just correlating devices to each other and finding discrepancies. Which is fantastic work, and a real value. But it's not NIST.

I've gotten a cal cert from power2max but I really haven't looked at it. I could look when I get home tonight.

What trail said!!! or well said trail! you get my point exactly. If a company as jaetj says power 2 max does send out a certificate to what is it conforming, their own in house calibration? When I place a calibration weight on my balance in the lab, I can tell you what it was calibrated to, when and what the error is in that calibration (ie +/- xxx.xxx g) and what the confidence interval used in that it is very well documented and "traceable". If I tried to convince the lab accreditation auditor that my balance was calibrated and I do not have a calibrated weight with a calibration certificate... well that won't go very well. So there is a lot in this discussion besides just a weight, jaretj and trail I think understand what I am saying.

I am wondering out loud why the industry has not tried to deal with this? or if anyone has really gone to that level of calibration.

s5100e wrote:

If a company as jaetj says power 2 max does send out a certificate to what is it conforming, their own in house calibration?

Yep. They're all self-certified. 4iiii did some interesting things with the Locomotion Laboratory at the University of Colorado but the data published only went to 85rpm and 350W (and didn't show any L/R data from their dual sided unit).



What's out there has been 'good enough' to date. And still is for the majority of cyclists.

Things have changed a lot in the last few years with the popularity of indoor cycling eSports/eTeams/etc where ±1% isn't good enough. That's the difference between wearing official UCI rainbows or 20th place.... or coming first or last in the hourly races on Zwift (which holds just as much prestige as UCI stripes for some).

The next can of worms is smart trainer accuracy. Most I see and test are nowhere near as accurate as claimed. Here in Australia if the ACCC (our consumer watchdog) are alerted, I suspect there's going to be a few problems for these manufacturers.

---

Shane Miller - GPLama
YouTube | Web | Twitter | Instagram | Facebook | Strava

The cycling industry has not dealt with it because it's making a mountain out of a molehill. Any NIST traceable certificate will be for hanging a calibrated weight off the crank arm with no pedal, calculating the torque, even at different temps, and that's it. The rest is software from the vendor. Calibration labs are 17025 accredited and they are not going to test the software as it would require a dynamometer and bike fixture. How often would they use it besides little and none. It would be a poor investment. The PM vendors should have a fixture so, assuming they are doing due diligence then their self certification should be fine. All the PM vendors could create a consortium and agree on the procedure and standards used, but they are not going to do it. Its just not important enough, since a consistency is more important than absolute measurements.

The main imperative in power meter sales in recent years has been to push price points down. The cost of calibrating to certified standards simply hasn't been worth it given the vanishingly small number of people that care even a little bit about this. Maybe Zwift competition will change this, but it's been a non-starter from a commercial perspective up until now.

FWIW, when I was working on development of a power meter product back in the day, one of the biggest problems we had was that our approach allowed us to output extremely precise real-time power data at incredibly high resolution (well sub 1 second). We had to de-tune the hell out of the thing in order to come up with something that consumers could deal with, and we wound up with a design spec that basically read "make the output look like an SRM". No one wanted to look at the power they were actually producing, they wanted to see something that averaged that power over at minimum a few seconds, and ideally that number would look a hell of a lot like whatever they were already using, regardless of whether that number was any good or not.

There were a ton of (other) problems that wound up dooming this product, but we were able to produce power numbers that were incredibly precise and consistent when the device was calibrated. My guess (with some reasonable data to back it up) was that right about 1% of our customers ever calibrated the unit. We had traceable, calibrated measuring standards in the building and could have run units through before shipping if we wanted to, but no way in hell was it ever going to happen. Just takes too damn much time. Which =$.

---

Tech writer/support on this here site. FIST school instructor and certified bike fitter. Formerly at Diamondback Bikes, LeMond Fitness, FSA, TiCycles, etc.
Coaching and bike fit - http://source-e.net/ Cyclocross blog - https://crosssports.net/ BJJ instruction - https://ballardbjj.com/

Strike that, I was wrong

16949 is quality managment
17025 is lab and calibration

Just looked at the certificates that came with a few Power2max PM's. There is nothing that indicates any type of lab certification or traceable standards used.

Now the company may have a 17025 or ISO9001, I'm not sure of the European requirements.

I suppose you've seen this?

https://d-nb.info/1174145994/34

I make that the assumption that all power meters are different (bias to each other) but good ones are precise.

I rode 20 minute efforts side by side with a friend last week at subthreshold for him....he averaged 210W, I averaged 190W....we are similar height and weight, both on road bikes....I am a BIT more aero but not enough to account for a 10% difference.

I check any new power meter on an indoor trainer (old style) at various speeds (with same rear wheel and tyre) to compare to prior ones....some are close but some are substantially different.

I agree that certifiable standards would be good but I don't see that coming any time soon.

thanks for the paper link I think in the past I had read that. Well I work in industry and understand cost of quality (part of my title is QA/QC Manager) but at the same time I hear people throwing around accuracy all the time with respect to bicycle power meters and well I think so far this thread has confirmed my feeling, there is no standard and if there was no one would use it. As for the cost, we are talking traceability, the gold standard is locked away somewhere, but at the industrial level there is a jig that is simple to use and can be calibrated against the standard on a routine basis. It is no different than lab balances, and various calibrated weights, often a good lab will have a working set that they periodically compare the the certified standards, that are periodically sent back to a calibration lab for recertification. It can be done. I think the industry is just not interested and there is not the demand, but as GP Lama said, maybe something like e-racing may change that, and a standard will be developed and if the trainer is calibrated to a standard then so can a regular power meter.

Thanks for the discussion folks, I have learned a few things.

I think some was on point here. I’m not a machine and I suspect I don’t put out an even X power during 360 pedal circle. I’m sure it varies quite a bit. What is reported to me via the head unit must involve some type of averaging software of that data stream.

If you calibrate/ certify a power meter with a machine that puts out a fix/constant power over the 360 degrees, you will miss that piece of the puzzle. A power might have an accurate data stream but shit software for averaging and you would miss that completely with a certification. I don’t suspect you will have access to the data stream.

Just the other day I was trying to measure some air pressures created in an air stream coming out of an axial fan for the CFD guys. They said we want to know if the pressure is X at this point. The data stream I got was a series of values ( from a certified meter ) that bounced all over the place. Air loves to swirl. I can take the data stream and do all kinds of different manipulation to it either give them a number that matches theirs or is way off. The data stream is accurate, the value presented has human thought in it.

My engineers would have requested a 9, 17 or 33 point pitot tube grid and then threw out any they didn't like :)

(I hope you get the 9, 17, 33 channel problem)

Ha ha... yeah I offered up an averaging array to remove me from the equation but they balked at that and wanted specific pressures at specific points.

Their CFD model had a nice laminar flow and as you probably know an axial fan provides nothing like that.

It’s going to get even more messy I suspect since this was with just a single fan running... future testing will be with 4 ...

I'd give them a kiel probe and say "here you go, find whatever you want"

I found about 10w going from a known good TT helmet to just a different one. Exact same position. What tires were they running versus you. Same wheelset? Same clothing? Same helmet?

So, yeah......it happens and can make that much difference.

I like what you are saying, but I guess in your air example there is close to true random variation? in a pedal stroke circle I doubt it is random variation. Also no matter what if the measuring item is calibrated to some form of standard regardless if there is variation then we have something to pin our hopes on? But I suspect from what you have said you probably have way more knowledge than I in such things and why I started the discussion as a learning tool. In the lab we have measurement uncertainty. It is the culmination of all of the individual measurement errors that comprised the final value. It is not fun (for most people) nor easy to do but it is a known technique to get some idea of what a number means when you read it on a lab report. if the result is 0.5 +/- 0.3 units then you know the lab is getting into the bottom end of where they can actually produce a result. Also in this same protocol there is a change in the uncertainty of a value as you get into the "better" range of the measurement system. So like in a lab (my thinking) a power meter is a system, it has a number of uncertainties in each aspect, these can be known and then evaluated to determine the system(power meter) uncertainty.

Just to be clear, suppose Stages had a NIST traceable calibration for their left-side only PMs prior to leaving the factory. Would you accept that?

They clearly state if their products are 1 or 2 sided so I would. That doesn't mean it's the best product, but if they certify what they have is within the tolerances they publish, then they are doing what they say. With that said I still know it's not the best product for my application.

I could use a turbine or a coriolis meter to measure fluid flow, (volume or massflow). Both could have calibration certification but one is clearly better than the other.

I as well would accept that, as long as we know constraints of the system then so be it as jaetj said and I agree.