Ergomo Pro or Power Tap SL 2.4?

[TamMan]

To put this error in perspective, the Ergomo is specified to be accurate to about 1.5%, the PT is also spec'd at 1.5%, while the SRM Amateur is spec'd at 5%, and the pro to 2%. All of these devices are also temperature sensitive.

On that note - does anybody have any experiences with the SRM Amateur? I was wondering whether the 5% accuracy is an issue? I would assume that between the PT and the SRM Amateur, the PT would be a much better choice (apart from the financial aspect).

 

[Bruce Diesel]

If the Ergomo is extremely sensitive to bb position and torque, then I suspect that the change in temperature would have a significant impact. The different expansion coefficients of the bb shell and the bb would cause a change in the overall torque experienced by the ergomo bb. It would be interesting to test this in an airconditioned room where the ambient temperature could be changed significantly.

Also, since the ergomo uses perforated light disks to measure the torque being applied, the resolution of light is extremely high and small changes would be detected. Maybe the use of light disks and their associated resolution is actually counting against the ergomo here.

 

[cPritch67]

Could you define "a slope not significantly different from unity" a little better?

Here's what I found:

Old Chain
PT Watts = (1.0543 * Polar Watts) - 24.181W, CV = 2.4%, R^2 = 0.9762

New Chain
PT Watts = (1.0219 * Polar Watts) - 8.5696W, CV = 1.8%, R^2 = 0.9937


Thinking about this some more, if you consider that the "state" of the chain at any given time is somewhere between the 2 "extremes" I separated out, then the best fit to use overall would be the one for the full data set. The CV will be larger since there isn't any "middle" data...but the way it stands now, the fit is:

PT Watts = (Polar Watts * 0.997) - 5.7W, CV = 2.6%, R^2 = 0.9839

That looks like a pretty constant ~5-6W offset, and the CV looks reasonable considering the claimed % accuracy of the 2 devices...and that's why I've set my FTP for the PT at 5W lower than my Polar FTP.

Tom
Just so I'm clear, the older the chain causes Watts to increase (or decrease)?

Thanks
Chris

 

[Tom Anhalt]

Tom
Just so I'm clear, the older the chain causes Watts to increase (or decrease)?

Thanks
Chris

According to that data and the linear fits, the old(er) chain exhibited higher drivetrain losses than the new chain...with the difference getting smaller with higher powers. Just plug some typical power levels (200W, 250W, 300W, etc.) into the 2 equations and you'll see what I mean.

 

[acoggan]

What I'm saying is that even with moderate power levels (i.e. 200W @75 rpm) drivetrain efficiencies can vary from 91-96% depending on the chosen gear. This is shown (based on a Shimano MTB groupo) in fig. 11 and table 2 of Kyle and Berto's report in the IHPVA journal:

http://www.ihpva.org/HParchive/PDF/hp52-2001.pdf

So...isn't that going to affect your linear fits of "device vs. device?"...especially if you're analyzing a file that uses a wide range of gear choices?

I think you're misunderstanding how I generated the data that I provided: using one, and only one, gear, I pedal my bike while it is locked in the Velodyne for 3 min at each power output, tossing the first ~30 s worth of data and averaging the rest. As the data show, when comparing crank vs. hub or wheel measurements of power, the difference appears primarily in the form of a significant intercept, not a difference in slope from unity. (IOW, efficiency increases with increasing power output.)

BTW, aside from "on the trainer" data (which is a known bugaboo for the Polar system...but not necessarily for reasons of chainline) have you seen anything with "on the road" data that shows there to be a "chainline" type variation with the Polar? I'm only asking because I'd be interested in seeing that.

IIRC, it was apparent from Robert Chung's analysis of your data that even your installation shows evidence of a chainline effect.

 

[Tom Anhalt]

I think you're misunderstanding how I generated the data that I provided: using one, and only one, gear, I pedal my bike while it is locked in the Velodyne for 3 min at each power output, tossing the first ~30 s worth of data and averaging the rest. As the data show, when comparing crank vs. hub or wheel measurements of power, the difference appears primarily in the form of a significant intercept, not a difference in slope from unity. (IOW, efficiency increases with increasing power output.)

No...I fully understand how you are taking the data. What I'm asking is that given the data that shows the drivetrain efficiency at a given power level can vary significantly by just switching to a different gear, how can you confidently apply a model from a test where you only used ONE gear to analyze files that used multiple gears?

IIRC, it was apparent from Robert Chung's analysis of your data that even your installation shows evidence of a chainline effect.

Yes...but then the question becomes, given the known variation of drivetrain efficiency with gear selection, which of the 2 PMs in the "test" are exhibiting what you call the "chainline effect"?

 

[acoggan]

given the data that shows the drivetrain efficiency at a given power level can vary significantly by just switching to a different gear, how can you confidently apply a model from a test where you only used ONE gear to analyze files that used multiple gears?

That's easy: I don't.

Yes...but then the question becomes, given the known variation of drivetrain efficiency with gear selection, which of the 2 PMs in the "test" are exhibiting what you call the "chainline effect"?

You'd have to refresh my memory re. your test conditions before I could answer this question.

 

[Tom Anhalt]

That's easy: I don't.

I would submit that you actually DO...if you assume that your "one gear" analysis result (i.e. the fixed offset value) is applicable to multi-gear power files (i.e. non-single gear files taken "in the field"). Are you saying that you don't assume that when looking at a PT file?

You'd have to refresh my memory re. your test conditions before I could answer this question.

The original testing that Robert analyzed and showed a "chainline" effect was a constant wheel speed power test ON A TRAINER, which for a couple reasons is not surprising for the Polar.

Just recently, he analyzed some of my PT vs. Polar crit data and it "appeared" there may be some "chainline" effect on there...but, that's where it becomes unclear which device is contributing that appearance...with the added uncertainties involved in determining the power vs. gear data due to the differences in the recording algorithms of the 2 devices. In short, that exercise was inconclusive IMHO....

 

[acoggan]

I would submit that you actually DO...if you assume that your "one gear" analysis result (i.e. the fixed offset value) is applicable to multi-gear power files (i.e. non-single gear files taken "in the field"). Are you saying that you don't assume that when looking at a PT file?

The only time that I try to correct for power losses in the drivetrain is when attempting to extract CdA and Crr data from field test results obtained using an SRM. I do such tests using only one gear, however (and even then just use a nominal efficiency of 97.5%, as anything more complicated is, IMHO, overkill).

The original testing that Robert analyzed and showed a "chainline" effect was a constant wheel speed power test ON A TRAINER

Those are the results that I had in mind.

 

[IronmanMadison]

Hello all,
I have a CALRIG, which is a dynamic calibration rig. Other calrigs are located at the Boulder University, at the Powertap lab, at the Australian Institute of Sport, and available commercially from Vacumed. Google will also take you to a study which compares "SRM and Powertap accuracy", out of the box, and over a racing season of 11 months. After testing numerous setups, there are indeed significant descrepancies between wattages displayed and wattages generated by the CALRIG. The CALRIG is accurate to the watt. A recent test of a new Velotron at Des Moines University showed aggreement within +/- 2 watts between the Velotron and the CALRIG from 50 watts to 500 watts. A few Computrainer load units have varied up to 30 watts from the intended 300 watts. Once the "truth" wattage is known, a software adjustment can be made with several but not all powermeters. Yes, the computrainer load unit has an adjustment potentiometer located under the label, which can correct the discrepancy. Also, the CALRIG can maintain higher power outputs (beyond 1000 watts) for as long as needed.

If you really need to know it's accurate, then consider a CALRIG test. But for most riders, repeatability is the most desired attribute.