So we have all those claims about accuracy of power measurement saying something like 1-2%. While the claims might be accurate what I can not find are published numbers such as:
- Torque range. Accuracy can not be the same at low and high range so please provide numbers.
- Provide temperature range for claimed accuracy.
- Hysteresis value.
- Linearity
- Zero balance
So unless claims are backed by said numbers I think one must apply some sense of reality before relying on power meter. On a positive note I think that unless something is grossly wrong properly made power meters are likely to be within 5% while estimating average power. Accuracy at low power is probably not that important since at this point we’re just simply recovering.
It gets even more complicated than the factors you have listed there.
For instance take a powertap vs and SRM:
The powertap is subject to error due to variations in the drivetrain efficiency (assuming you want to know the power the body is producing). Chain gets dirty, more error.
Something like an SRM is not.
However, the powertap’s autozero functionality is better/easier than the SRMs, so the error due to temperature fluctuations may tend to be less for the powertap seomtimes.
However, the power error due to torque error is a function of cadence for the SRM and speed for the powertap, so the powertap might be subject to more temperature induced error at high speeds than the srm!
It is important to know what things affect overall accuracy so that you can account for them, if possible.
Accuracy at low power might be very important for purposes of tracking total training load, or field testing.
So we have all those claims about accuracy of power measurement saying something like 1-2%. While the claims might be accurate what I can not find are published numbers such as:
- Torque range. Accuracy can not be the same at low and high range so please provide numbers.
- Provide temperature range for claimed accuracy.
- Hysteresis value.
- Linearity
- Zero balance
So unless claims are backed by said numbers I think one must apply some sense of reality before relying on power meter. On a positive note I think that unless something is grossly wrong properly made power meters are likely to be within 5% while estimating average power. Accuracy at low power is probably not that important since at this point we’re just simply recovering.
It gets even more complicated than the factors you have listed there. …
Sure. I just asked about very basics. Can go on and on but no need to.
So we have all those claims about accuracy of power measurement saying something like 1-2%. While the claims might be accurate what I can not find are published numbers such as:
- Torque range. Accuracy can not be the same at low and high range so please provide numbers.
- Provide temperature range for claimed accuracy.
- Hysteresis value.
- Linearity
- Zero balance
- Non-steady state use conditions
- What head unit you use to capture data (this can have a bigger influence on the accuracy of the data than the power meter itself)
- What software you use (not all report the data in the same way)
- Hardware it’s used with (e.g. chainrings)
- Sensitivity to installation
- Power meter’s firmware
So unless claims are backed by said numbers I think one must apply some sense of reality before relying on power meter. On a positive note I think that unless something is grossly wrong properly made power meters are likely to be within 5% while estimating average power. Accuracy at low power is probably not that important since at this point we’re just simply recovering.
Well you can do some things to validate the accuracy of your meter and assess the level of confidence you have with your own data, but what level of accuracy is required depends on what you are using it for.
Also, one needs to make a distinction between a meter’s variability and its accuracy. While they often go hand in hand, and if a meter is not particularly variable then accuracy is just a matter of calibration adjustment, but not all meters enable calibration adjustment (indeed very few do, but should).