Testing PowerTap Accuracy - Vagaries

woodgab said:

I have 3 Pro hubs and this weekend decided to see how accurate the latest one is. For each, I tried to control as best as possible, the data I got by duplicating the following:

Gear: 39 X 15
Cadence: 102
Tire Pres.: 120 PSI
Trainer mount pressure: Just higher than where slippage begins.

The same head unit was used throughout. My Cycleops fluid 2 was warmed up 10 minutes before any data was taken. The difference between the three 700c Open Pro wheels was tires. The one I was curious about came with a Maxxis tire. The two I have are Vittoria, each having seen 300+ trainer miles, with visible flat spotting.

Results after putting all power figures in Excel and deleting any cadence values other than 102 (mostly values +-3):

the lattest hub, Hub 1: 240.68 watts
Hub 2: 211.80 watts
Hub 3: 217.65 watts (recently rebuilt by Sarris)

The big difference is why I post. Does anyone else think there is something up with Hub 1, or can tires explain such a difference? As I write, I realize I should swap rubber and see what happens, but time was limited. I was shocked on the most flat spotted tire to find out that VERY little additional pressure (I don't know, 3mm deflection) into the metal roller made a resistence difference of 45.5 watts.

1) Was torque zero'ed each time you changed wheels? I find a difference between some of my hubs that the torque auto-zero will not correct for.

2) Also, I'm not an expert on this stuff but a guy on biketechreview has been doing rolling resistance tests on rollers. The impression that I'm getting is that this difference might be partially or fully explained by the tires. The resitance due to trainer roller pressure is difficult to control accurately between runs/setups.

woodgab said:

I have 3 Pro hubs and this weekend decided to see how accurate the latest one is.:...

You're doing this the hard way. Why not do a static torque test (stomp test) on each hub and compare their absolute accuracy? Here's a link to the stomp test: midweekclub.capowerFAQ.htmOpen ↗

I use a toe strap wrapped tightly around the rear brake to lock down the wheel after spinning the wheel to wake up the unit and zeroing the torque. If you have some heavy weights you can do the test by hanging weights off the right pedal for better repeatability but you can use your body weight as well by weighing yourself in cycling clothes and shoes and then balancing on the right pedal. The torque results are linear to power accuracy so you can directly compare the measured vs. theoretical torque percentages to figure out the relative accuracy of your hubs.

Good luck,
Dave

daveryanwyoming said:

You're doing this the hard way. Why not do a static torque test (stomp test) on each hub and compare their absolute accuracy? Here's a link to the stomp test: midweekclub.capowerFAQ.htmOpen ↗

I use a toe strap wrapped tightly around the rear brake to lock down the wheel after spinning the wheel to wake up the unit and zeroing the torque. If you have some heavy weights you can do the test by hanging weights off the right pedal for better repeatability but you can use your body weight as well by weighing yourself in cycling clothes and shoes and then balancing on the right pedal. The torque results are linear to power accuracy so you can directly compare the measured vs. theoretical torque percentages to figure out the relative accuracy of your hubs.

Good luck,
Dave

The static stomp test can only tell him the accuracy of the PT at zero angular velocity though. It's plausible that the source of any inaccuracy could be a relationship between the torsion/torque gradient and the angular velocity surely?

GettingFaster said:

The static stomp test can only tell him the accuracy of the PT at zero angular velocity though. It's plausible that the source of any inaccuracy could be a relationship between the torsion/torque gradient and the angular velocity surely?

It's certainly plausible that there could be a speed component to accuracy, but I'd certainly start with the static torque test. After all the PT is just a strain gauge and the accuracy of that gauge is the starting point for all power measurements. Or, put another way, if the angular velocity is the dominant error term, then what use is the static torque test which is offered as the best way to calibrate the SRM and best way to validate accuracy of the PT?
-Dave

Thanks. I'll give it a stomp. I wondered if maybe a he-man sprinter stretched out the strain gauge, or if that is even possible. I'm thinking if a strain gauge gets stretched and does not rebound, it will overstate power as a result of increased electrical resistence at zero torque. I emailed Cycleops, but haven't heard how/if these hubs generally go south.

woodgab said:

Thanks. I'll give it a stomp. I wondered if maybe a he-man sprinter stretched out the strain gauge, or if that is even possible. I'm thinking if a strain gauge gets stretched and does not rebound, it will overstate power as a result of increased electrical resistence at zero torque. I emailed Cycleops, but haven't heard how/if these hubs generally go south.

Yes do the stomp and more importantly, make sure you zero out with each hub. I've got two hubs which regularly read 8"-lb apart. However, when I zero them, the power readings are the same.

beerco said:

Yes do the stomp and more importantly, make sure you zero out with each hub. I've got two hubs which regularly read 8"-lb apart. However, when I zero them, the power readings are the same.

Well, i haven't zero'd, but I am reporting back that the stomp test, using 45lbs revealed:

Suspect hub 1/25.4 X 172.5 X 21/36 X 45 = 161.29 vs head unit's 168

Rebuilt hub in the 19/36 = 164.56 vs 170 for head unit.

Seems pretty accurate. I thought per Cycleops instructions zeroing torque was not necessary if the watts read zero under no pedaling? Anyway, these numbers are much closer to the 2-3% range of error.

woodgab said:

Well, i haven't zero'd, but I am reporting back that the stomp test, using 45lbs revealed:

Suspect hub 1/25.4 X 172.5 X 21/36 X 45 = 161.29 vs head unit's 168

Rebuilt hub in the 19/36 = 164.56 vs 170 for head unit.

Seems pretty accurate. I thought per Cycleops instructions zeroing torque was not necessary if the watts read zero under no pedaling? Anyway, these numbers are much closer to the 2-3% range of error.

that's okay if you have the rear wheel spinning freely at a decent clip. Zero watts and zero wheelspeed says nothing about zeroed torque.

In any case, why go thru the bother you have without zeroing the torque. It takes a few seconds???

woodgab said:

I thought per Cycleops instructions zeroing torque was not necessary if the watts read zero under no pedaling?

I've seen some weird stuff on very rare occasions. Torque might drift around a bit if the hub gets very wet inside. It might drift around and pass through zero while you are looking at it and then continue to drift while for a few seconds. Getting in the habit of checking the torque zero (not necessarily zeroing out, just talking about checking it, which is part of the process) at the beginning of every ride will not prevent this but at least give you a warning when it's happening.

Also, if torque is off by, say, 0.44 (a made-up number) it may or may not get rounded-down to zero for the display and for decision making (I don't know how the firmware in the computer makes the decision). Yet, that still enough to give you a small, annoying offset to reality. If you make the effort to zero-out each time, that won't be an issue.

You may think that this is nit-picking and being anl retentive, but I've had at least one PT for 3.75 years (now I have three hubs) with very little data lost to stupid stuff like this.

I did mine for the first time today and got 99.04% of actual. But only with a 35 lb weight. While the test calls for 50 or higher. Will get one of those next.

Woodgab,

Plugged in your numbers to my spread sheet as a test and was wondering if you switched the gearring between the two as a typo. Cause I got 161.29 for the first one only if I used a 19/36 not 21/36. Making it read 104.16% of actual.

And could not get 164.56 with a 19,20 or 21/36.

Same cranks?

I want to make sure I don't have my formulas screwed up.

Thanks.

woodgab said:

Well, i haven't zero'd, but I am reporting back that the stomp test, using 45lbs revealed:

Suspect hub 1/25.4 X 172.5 X 21/36 X 45 = 161.29 vs head unit's 168

Rebuilt hub in the 19/36 = 164.56 vs 170 for head unit.

Seems pretty accurate. I thought per Cycleops instructions zeroing torque was not necessary if the watts read zero under no pedaling? Anyway, these numbers are much closer to the 2-3% range of error.

jetnjeff,

WARNING, I'M ABOUT TO GEEK OUT (its Friday):

You're right. I think I used 21/39 to get 164.56. The number comparison should have been:

Old Hub: 1/25.4 X 172.5 X 21/36 X 45 = 178.27 vs 168 indicated.
Newer Hub: 1/25.4 X 172.5 X 21/36 X 45 = 178.27 vs 170 indicated.

Sorry about that. i had to keep straight the fact that I was one down on a 12/23 cassette and at the top of an 11/21.

So, 94.2% and 95.4%. Some of the under-represented torque is probably because the crank was a couple degrees down from perfectly horizontal. Hey, what's the geometry math to figure out the error from being one degree off (cosine? tangent?..)? In any event, you would expect lower indicated values for either being high, or low, of horizontal, as the effective distance gravity is pulling the weight down from shortens (<172.5mm). No? :eek:

FYI, I used a 45lb iron bench plate over an eggbeater pedal. I am not sure at what rate being a few degrees off on the crank rotation would affect the value, but I used a smartlevel (digital) to at least be sure it was set within ~1.5 degrees on both trials.

woodgab said:

...Hey, what's the geometry math to figure out the error from being one degree off (cosine? tangent?..)? ... the effective distance gravity is pulling the weight down from shortens (<172.5mm). No? ...

Multiply the cosine of the error angle relative to horizontal by the crank length to get the shortened lever arm.

-Dave

woodgab said:

Some of the under-represented torque is probably because the crank was a couple degrees down from perfectly horizontal. Hey, what's the geometry math to figure out the error from being one degree off (cosine? tangent?..)?

If you were hanging a 45 pound weight off the pedal, then I'm guessing you were also using a trainer to hold the bike. In that case, rather than clamping the brake and trying to get the crank horizontal (which, as you've discovered, is hard), simply rotate the rear wheel backward *very slowly* until the torque reaches maximum.

For what it's worth - I just did a quick and very dirty stomp test now on my new SL hub. Like the rest of you dwellers of the temperate climates, I've got my winter trainer setup as a fixture in the spare room so it wasn't a struggle getting this going.

I stepped on my bathroom scale, got my weight( :eek: ), dumped it into a spreadsheet to figure out torques for a variety of gears, tried them out. Locked up the rear wheel. Right pedal closest to the front wheel, left pedal overlapping with the rear wheel. Put both feet on the right pedal. One hand squeezing the right brake handle, the other against my body. I totally eyeballed the crank angle at 90 degrees. (So shoot me.) I did not do multiple tries at each gear. (So shoot me again.)

I tried 4 different gears and all were within 0.04-2.00% of theoretical.

BUT

I adjusted my body lean until I got the max torque indication. I don't see that as wrong though. It's not like I was "gaming the system" by doing that. I'm just maximizing how much weight is on the pedal. If anything, it probably helped the consistency of the measurement.

Had I gone to the trouble of getting my bubble level out of the basement to check the crank angle, maybe I could have gotten better results.

Does it prove anything? I guess it suggests that this hub is not grossly off the mark. That's about it. I guess it depends upon how much my right arm weighs and how much of it wasn't applied to the pedal because I was grabbing the brake lever.

That sounds better

woodgab said:

jetnjeff,

WARNING, I'M ABOUT TO GEEK OUT (its Friday):

You're right. I think I used 21/39 to get 164.56. The number comparison should have been:

Old Hub: 1/25.4 X 172.5 X 21/36 X 45 = 178.27 vs 168 indicated.
Newer Hub: 1/25.4 X 172.5 X 21/36 X 45 = 178.27 vs 170 indicated.

Sorry about that. i had to keep straight the fact that I was one down on a 12/23 cassette and at the top of an 11/21.

So, 94.2% and 95.4%. Some of the under-represented torque is probably because the crank was a couple degrees down from perfectly horizontal. Hey, what's the geometry math to figure out the error from being one degree off (cosine? tangent?..)? In any event, you would expect lower indicated values for either being high, or low, of horizontal, as the effective distance gravity is pulling the weight down from shortens (<172.5mm). No? :eek:

FYI, I used a 45lb iron bench plate over an eggbeater pedal. I am not sure at what rate being a few degrees off on the crank rotation would affect the value, but I used a smartlevel (digital) to at least be sure it was set within ~1.5 degrees on both trials.

Nice tip about rotating the wheel back until max.

Still wounder why they recomend 50 lbs or more and how much that affects accuracy?

RChung said:

If you were hanging a 45 pound weight off the pedal, then I'm guessing you were also using a trainer to hold the bike. In that case, rather than clamping the brake and trying to get the crank horizontal (which, as you've discovered, is hard), simply rotate the rear wheel backward *very slowly* until the torque reaches maximum.

jetnjeff said:

Still wounder why they recomend 50 lbs or more and how much that affects accuracy?

There's nothing magic about 50 pounds. The higher weight is about precision, not accuracy. It's slightly better to use fewer heavy weights than a whole bunch of lighter weights because they're easier to handle, and you should try to verify that the weight is accurate (cheap plates can vary a little). I use two 20 pound plates.

You'll also get better precision if you use a smaller gear (like, 39x25 rather than 53x12). Since I don't have to balance on the pedal, it's really not that hard to check two or three different gears to make sure that there's linearity across the cogs.