Variations between PT unit and CP

kopride said:

When I downloaded the PT data onto CP, I got a very slight variation between the average watts recorded on the head unit in interval function and then the average watts for the interval calculated by CP. (NP was also higher but I understand that there is an algorithim applied). It was only about 2 or 3 watts, but I was wondering whether CP actually recalculates the raw data? it would seen to me that average is average and there should be no difference. I posted the actual data in JS back to cycling thread.

Also, how are you CP users posting graphs on this forum? I just cut and pasted the ride summary, but the graph wouldn't cut and paste.

The PT CPU tallies a running average as you ride. Think of it as plus and minus values accumulating real time during the ride. That's a bit different than having all the data and averaging it all at once. Sure you could calculate a very accurate running average but it would take more microprocessor resources than a simple tally based approach which is good enough for real time estimates but can result in the sort of differences you see between real time and uploaded data.

I suspect the difference is also related to the slight differences in the way time is captured in the CPU and in the uploaded data. It's typical for interval or total ride time to differ slightly between the CPU and computer and it's discussed in the PT manual and on their website FAQs IIRC.

I post graphs by taking a screen shot of my monitor with the graph displayed. Basically you grab a screenshot by hitting shift-printscreen which places the screenshot on your clipboard and then you need to paste it into an appropriate document. I usually paste it into a blank photoshop document and save it as a JPEG after cropping out stuff like the Windows taskbar. There are other ways, but that's usually what I do...

-Dave
P.S. If you see big differences between CPU AP and AP in WKO+ it's often because "include zeros in power averaging" is disabled. I finished a hilly crit once and my AP was wayyyyy too good to be true. Turned out zeros were discarded and a third of the course was a fast coasting descent which really skewed the AP. Sure 'nuff it showed up correctly in WKO+.

daveryanwyoming said:

The PT CPU tallies a running average as you ride. Think of it as plus and minus values accumulating real time during the ride.
I suspect the difference is also related to the slight differences in the way time is captured in the CPU and in the uploaded data. It's typical for interval or total ride time to differ slightly between the CPU and computer and it's discussed in the PT manual and on their website FAQs IIRC.

P.S. If you see big differences between CPU AP and AP in WKO+ it's often because "include zeros in power averaging" is disabled. I finished a hilly crit once and my AP was wayyyyy too good to be true. Turned out zeros were discarded and a third of the course was a fast coasting descent which really skewed the AP. Sure 'nuff it showed up correctly in WKO+.

just for yucks, I also uploaded the data in the power agent software and it showed the same averages as the CPU. Again, it is only a 2 and 3 watt difference from the wko average, but it is interesting. The two different programs either crunch average differently, or power agent drops zeros as you suspected.

kopride said:

just for yucks, I also uploaded the data in the power agent software and it showed the same averages as the CPU. Again, it is only a 2 and 3 watt difference from the wko average, but it is interesting. The two different programs either crunch average differently, or power agent drops zeros as you suspected.

Hmmm. Check the CPU set up for the non-zero average power. Otherwise, as Dave said, it's simply because the CPU uses maths shortcuts to calculate a running average to save processor resources. CPU typically under reports slightly (1-2 W). If it is reading higher, then that's more likely to be non-zero power.

I've not seen the recrorded time difference happen much (at all?) with my PTs but all mine are wired, maybe that's a "feature" of wireless.

But I have definitely noticed it with my (wired) SRM. In my 20-min efforts it "loses" about 10 seconds worth of time. In other words the total time as calculated by the interval start and ending time stamp was 11 seconds longer than the total of the number of data points x recording interval (1-sec).

WKO+ doesn't massage data the way, for instance SRMwin does (unless you ask it to). WKO+ sucks out and displays what ever is in the meter. I'm not sure if Poweragent does anything similar like SRMwin.

Alex Simmons said:

Hmmm. Check the CPU set up for the non-zero average power. Otherwise, as Dave said, it's simply because the CPU uses maths shortcuts to calculate a running average to save processor resources. CPU typically under reports slightly (1-2 W). If it is reading higher, then that's more likely to be non-zero power.

I've not seen the recrorded time difference happen much (at all?) with my PTs but all mine are wired, maybe that's a "feature" of wireless.

But I have definitely noticed it with my (wired) SRM. In my 20-min efforts it "loses" about 10 seconds worth of time. In other words the total time as calculated by the interval start and ending time stamp was 11 seconds longer than the total of the number of data points x recording interval (1-sec).

WKO+ doesn't massage data the way, for instance SRMwin does (unless you ask it to). WKO+ sucks out and displays what ever is in the meter. I'm not sure if Poweragent does anything similar like SRMwin.

I do have a 2.4 and I will check that zero feature when I go out for my ride tommorow. Thanks. Again, it is nothing that I am going to lose sleep over. The NP feature of CP is interesting and helps b/c I do contend with lights and traffic on these longer outdoor intervals. I am only working off a trial download for now, but I am interested in eventually transferring all my power agent files over and we will see how consistent the variation is. If I can see a pattern or explanation, I will post it.

Btw, your recovery is amazing. When I ran x country in hs, I often thought that I literally was doing something that my disabled father never did in his whole life, i.e. run normally, and it helped me get through the times I was feeling a bit sorry for myself. Now, when I start to feel hassled by getting my riding gear together and clearing some time to ride, I think that my obstacles are pretty petty compared to what some folks contend with. My pop never saw himself as disabled and always wanted to be measured against everyone else, it was a mindset that always served him well. I have a funny feeling you're going to move up that FTP board and start knocking us off one by one.

Won't they always be different since they're measured at different places? Kind of like a car that has 400 HP at the crank but less than 350 at the wheels...You have to account for the losses, correct?

upstateSC-rider said:

Won't they always be different since they're measured at different places? Kind of like a car that has 400 HP at the crank but less than 350 at the wheels...You have to account for the losses, correct?

You would be correct if we were talking about taking measurements at different points along the drivetrain (say an SRM measuring power at the crank vs a PT measuring power at the hub vs a Computrainer measuring power at the tyre-roller interface) but we aren't.

He's talking about the difference between what the Powertap's on bike CPU unit shows as the average power, and what the software shows you the average to be after you have downloaded the data. The data is from the same source (the PT hub) just the former has some rounding errors.