Powertap Readings

Richy P

I've just bought one of these, and looking at the power and cadence readings from rides when I am pedalling at a steady cadence with no accelarations, the readings periodically jump by 10 -20 %. I have backed this up by putting the bike on the turbo trainer which has a power meter, and whilst the trainer produces steady readings, the jumps on the Powertap still occur.

Has anybody experienced this? Is it normal?

Also, on the torque settings, should the desplay read 000 or ---?

Thanks

liversedge

power is often referred to as being "stochastic" i.e. highly variable. On the road my power at any single point in time varies, although over a rolling 30sec average it evens out. Even on a CT in ergo mode power will vary, although not as much as 10-20%.

do it on a turbo in ergo mode and if you still have jumps, despite working at an even pace then you may have a problem.

Alex_Simmons/RST

There are two things that explain the high degree of variability in power readings as shown on a Powertap CPU:

- as has been described, power output when riding is actually quite variable, even when riding relatively smoothly, the power output does naturally vary with the smallest changes in pressure on the pedals, the road surface changing, the minor variations in gradient, wind and so on. That is entirely normal and is a common observation from new power meter users.

- the second element is a "feature" of the way Powertap calculates (and displays) power data. It is known as aliasing and is to do with the power calculations being based on measurements made over a fixed duration (rather than, e.g. on a complete revolution of the crank, as per SRM).

Aliasing with a PT:

The PT records a power data point every 1.26 seconds and uses the average of the hub torque readings (sampled at 61Hz) to determine the power for that 1.26 seconds.

When we pedal, we provide torque to the hub which pulses in a pseudo-sinusoidal manner, with two torque peaks per pedal revolution (one from each leg).

So at a cadence of 43, 95 or 143 rpm the number of torque pulses at the hub will be equal for each 1.26 second time period at 2, 4 and 6 pulses per 1.26 seconds respectively.

It is also a even number of pulses per 1.26 secs at cadences of 71, 119 and 167 rpm but with an odd number of pulses per 1.26 secs (3,5, & 7) - so as long as there isn't much of a leg imbalance, then the power reading from one 1.26 sec interval to another will be consistent.

The further you deviate from those cadences the likelyhood that consecutive 1.26 sample periods will contain different number or fraction of torque pulses (which will raise or lower the average torque recorded for that interval). Since power = torque x cadence, if cadence is fixed then your power reading will vary in step with the torque measurement.

So e.g. at 80 rpm (between 71 and 95 rpm) the number torque pulses will vary for each consecutive 1.26 second time period and hence the instantaneous power reading will naturally move up and down with each time period.

It makes no difference when averaged over a longer duration and the solution is to lengthen the power display rolling average interval. You would need the Pro CPU to choose longer display averaging intervals. 3 or 5 secs works well.

Richy P

Many thanks.

liversedge

Thanks Alex, do you know if this "feature" would result in any anomalies in my data that I should be aware of?

DaveyL

Thanks for the explanation Alex. I may try a different display interval to see what the effect is. I must admit my initial reactions after a couple of rides with the PT were along the lines of "Bloody hell, how are you supposed to use this for pacing!?"

Alex_Simmons/RST

liversedge wrote:

Thanks Alex, do you know if this "feature" would result in any anomalies in my data that I should be aware of?

Well about the only thing to be wary of is the peak power reported - if it is just the one data point and is not a replicable number for you, then the aliasing may have affected that one number. Apart from that, it does not affect the overall accuracy of the data.

When you view the data, you will note how the power line looks a little "choppy".

As an experiement, on some rollers or a trainer that you can ride very smoothly on, ride at difference cadences and view the power line to see how smooth or choppy it looks. try say 85rpm and 95rpm. at 95rpm it should appear smoother from point to point.

Alex_Simmons/RST

DaveyL wrote:

Thanks for the explanation Alex. I may try a different display interval to see what the effect is. I must admit my initial reactions after a couple of rides with the PT were along the lines of "Bloody hell, how are you supposed to use this for pacing!?"

Yes, one can fall into the trap of "chasing their power tail"* instead of realising that it is a feel thing and the meter is a guide to the overall effort.

* by that I mean, you look at the meter, it says you are under the desired target, so you react by putting more pressure on the pedals but since the power reading is instantaneous, it then quickly over shoots the target and you react by easing the pressure and it falls back under again!


And just to be clear - set the rolling average on the CPU display to 3 or 5 seconds. That ususally softens down the jumpiness of the displayed power. Most find 5-sec is a good choice.

BUT - don't set the sampling rate to more than 1 second if you can avoid it. Most power meters provide an option to reduce the recording frequency so the memory storage lasts longer (e.g. if you are going away and expect to ride more hours than the CPU has room to store and you don't have PC to download).

Normally recording, say every 2-secs or 5-secs instead of every 1-sec is not a problem overall (it will dampen down sprint power measurement perhaps but for most general training that's no big deal).

However on the PT CPU - if you choose 2-second recording frequency (well 2.52-sec), instead of averaging the power for the last two data readings and storing that, it actually discards one of the readings and stores the other (it only keeps every other data point). Now over the life of a ride file, that's not too big a problem but if you went to 5-sec recording, then the data will be significantly short changed as it would only keep one out of every four data points (4 x 1.26-sec = 5.04-sec).

DaveyL

Thanks for the info, Alex. I have been guilty of "chasing my power tail" as you put it on the last few rides I've done with the PT (at least when I've been on a climb and, as an experiment, have been seeing what particular wattage I can hold).

I'll change the display interval readout and see how that goes.

bahzob

I find even 3 - 5 seconds is too short. There's an excellent intro to training with power here http://www.trainingsmartonline.com/power_cycling_training.php and that recommends setting average to 30s for intervals.

Done this and find that it makes a difference in terms of interval quality.

DaveyL

Thanks for the link, bahzob. I know from previous threads that you used your PT to pace yourself up the cols in the Marmotte - did you use this same 30 s sertting for that?

bahzob

DaveyL wrote:

Thanks for the link, bahzob. I know from previous threads that you used your PT to pace yourself up the cols in the Marmotte - did you use this same 30 s sertting for that?

To be honest I dont recall. I do remember that that power reading did not vary wildly during the climbs but cant remember what setting I had it on.