Powertap measure very unstable: Is this right?

[aerv]

I bought a new powertap PRO+ last month and notice using on a trainer indoors using tacx (also looking my data from outside rides) that even when I keep a steady speed and cadence the values on the computer keep changing sometimes from 140 to 190 watts and cadence from 80 to 90 in less than 3 seconds.... looks like the data being sent is not accurate at all. Is this behavior normal? At the end analyzing my cadence the average is more than 10 rpm higher.
I don't have a external cadence. The csv file generated has the same numbers...Do I have a problem on this hub or is this normal?

 

[ambal]

If you have set it up properly, the readings are likely spot on. Power output isn't consistent like heart rate, its much hard to stay within a small range. Most people can keep their HR within a few % over a long interval, this is impossible to achieve with power unless your on an indoor velodrome.

 

[daveryanwyoming]

The previous post is on target, compared to metrics like HR which has a large amount of natural averaging power measurements seem really jumpy at first.

But cadence shouldn't jump around so much, especially indoors on the trainer and 50 watt jumps in power over just a couple of seconds while riding the trainer are a bit unusual.

It's possible you've got some interference from a wifi access point or other device interfering with your wireless PT Pro+ signal. I always used wired PT hubs so didn't experience this myself but a number of folks have had troubles when doing indoor training near such devices. If your signal is reliable and the data just seems too jumpy then you can go into the setup menus and increase the display averaging interval for cadence and or power measurements. Increasing the display averaging to 5, 10 or more seconds will smooth things quite a bit but won't change the instantaneous accuracy of the data recorded for download after your workouts.

-Dave

 

[aerv]

Dave,
My cadence is really wrong, too high, my power is jumping but looks like correct in average. I have my wireless access point in the same room and maybe the TACX engine also can help. I am going to test it outside as soon as the snow allows me. Thanks so much for your help. I will update with new findings outside.

 

[gregf83]

Likely the main reason for the variation in power you are seeing is due to the sampling rate (1 or 1.26 Seconds) of the Powertap. The samples are not synchronized to the crank rotation. So although the average power will be correct the individual samples will jump around. This is sometimes referred to as the precession effect. It doesn't affect SRM, Quarq or any other crank based powermeter.

The simplest solution is to increase the display averaging to 10S. It won't affect the raw data but it will settle down the display. The instantaneous power data is of no value anyways while you're riding.

 

[aerv]

Gregf,
Thanks for the explanation, I was getting some files from friends and it happens with them. You are right. Thanks a lot.

 

[ToffoIsMe]

If you aren't using the optional cadence sensor then I recommend not paying any attention to the cadence reading as it is an estimated based on your power output and speed. I have noticed that the "hub" cadence (what PT calls it when your cadence is estimated based on the hubs readings) is usually around 20 rpm lower than my actual cadence. It usually gives me an average cadence reading of around 75-80 when my actual cadence is rarely ever below 95-100. Using the optional cadence sensor will clear up your issues with cadence.

 

[Yojimbo_]

Re this precession effect - can you refer me to a location that explains this in detail? I'd like to see governing equations and that sort of thing so I can understand what's actually going on.

 

[daveryanwyoming]

In digital signal processing terms it's a form of 'aliasing' google that (or spectral aliasing, sampling theorem, etc.) and you'll get plenty of hits on the basic concept. But Alex wrote this a while back that sums up the PT situation nicely:

http://www.cyclingforums.com/forum/thread/433608/watt-readings-inconsistent#post_3581352

-Dave

 

[Alex Simmons]

Originally Posted by daveryanwyoming .

In digital signal processing terms it's a form of 'aliasing' google that (or spectral aliasing, sampling theorem, etc.) and you'll get plenty of hits on the basic concept. But Alex wrote this a while back that sums up the PT situation nicely:

http://www.cyclingforums.com/forum/thread/433608/watt-readings-inconsistent#post_3581352

-Dave

Just make a mental note that newer Powertaps now use 1-second torque data sampling, rather than 1.26 seconds I referred to in that post. Hence the effect still exists, albeit it will affect different cadences more/less than previously.

The aliasing effect could be removed if the PT sampled torque from a whole number of revolutions of the hub, rather than based on a fixed duration sample.