A question to any Powertap users

Depending on the length of the interval I am measuring, I either use 10s avg, 30s avg or lap avg

Herbsman wrote:

Put your display on 10 or 30 second average and keep it within a certain range of your 'ideal' power output.

Instant power is no good for road riding.

Why do you say that? It is possible to keep it steady with practice.

I find instantaneous power (non-smoothed) useful in addition to 30s rolling average.

Power, especially outdoors, is naturally quite variable - stochastic. It's normal because that's what happens.

Layered on that natural variance is an artificial variance in the displayed/recorded power numbers which is a result of the Powertap's fixed time based sampling of torque data - otherwise known as aliasing. This artificial variability does not occur with crank based power meters, which sample torque with event based sampling (i.e. full crank revolutions).

Powertap aliasing impacts the variability in reported power differently at different cadences. e.g. if you pedalled at 60rpm, 90 rpm or 120 rpm*, at which cadence there are an equal number of pedal down strokes per second, then the artificial variance is minimised, but if you pedal at cadences away from those, then the artificial variance is increased as each second of torque data contains a different number of pedal down strokes, and so affects the calculated torque average for that second.

So even though you may be perfectly smooth in power output, if you are pedalling at say 80-85 or 95-100 rpm, then the powertap will still display variable power from second to second. Of course over time these variations average out and the data is good, but it is one reason why you should also take very short duration power numbers (5 seconds or less) from a Powertap with a large grain of salt.

The normal way to address this is to apply a moving average to the displayed data. On Powertap head units, 5 seconds was a good choice for power display. I think 30 seconds is a bit long.

Also keep in mind not to fall into the trap of "chasing your power tail", where you are constantly reacting to what the display says and over compensating.


* the cadences at which this occurs is different with older model powertaps that averaged torque every 1.26 seconds, rather than every second with current models.

Because of the interaction between cadence and the 1 sec averaging that Alex has described above, you can get quite big jumps from second to second even if your output is more or less constant, maybe not 100s of watts but possibly 100 or so... You get used to it, I use both 10s and 30s averages (as well as overall average) and have all displayed simultaneously. After a while your brain sort of integrates them and/or tells you what figure is important at any given moment. The 10s average is probably most useful, you should be able to keep that more or less constant with practice. However, on anything but a perfectly flat road or a constant climb you will still vary your output quite a lot even if you are not intending to - that's why the 30s average is also very useful. On the turbo it is much easier to keep the power more constant because the resistance is exactly the same much of the time. Incidentally, if you are comparing the powertap to the (rather inaccurate) power readings you get from VR trainers such as the Tacx ones, I'm pretty sure that the "instantaneous" power readings you are getting from the latter are artificially smoothed and/or averaged over a few seconds too. With the powertap you are seeing the raw data, i.e. the averages for each successive second, which is why it appears to jump around a lot if you are turning the cranks once in one second and two times in the next second. <edit>: or more accurately, if you are pushing a pedal down twice in one second and three times in the next second.

fish156 wrote:

maryka wrote:

3s rolling avg seems to work best for me, 10s and 30s are way too long, 1s is just too annoying on the eyes.

+1
I have 3s and lap/interval power displayed.

+2