power meter drift issues?

hey all,

unfortunately i've fallen into the power meter accuracy snake pit.

recently i've been seeing what looks to me like power meter drift. you
can see a pretty good example of this from my last trainerroad workout
here (done on a pre-2017 wahoo kickr snap):

https://analyze.dcrainmaker.com/#/public/ade655fb-936b-462b-6037-ef06a2775681

one power meter is a set of garmin vector 3 pedals with a claimed
accuracy of +/- 1.0%. (i've followed the service advisory notes for the
vector and verified the spindle torque. after installing the vectors i
reset the install angles and i also did a static torque test, as
described on the official garmin support pages, using a rice lake
calibrated 20 kg weight). the second power meter is a powertap g3 hub
with a claimed accuracy of +/- 1.5%. both power meters were calibrated
(zero'd) before the workout, which was done indoors at a consistent
temperature of around 70 F.

at the start of the ride, the powertap numbers are a few watts below the
vector numbers, which makes sense given some drivetrain power loss. but
by the end of the ride the power meters have reversed, with the powertap
giving numbers about 10% above the vector numbers, which seems wrong.
(aside from the drift, the difference between the two power values seems
out of spec for claimed accuracy of the devices, especially when you
consider that there should be some drivetrain power losses).

has anyone else seen something like this before? to me it seems like
one of the power meters is drifting out of spec? any ideas what else
might be going on or how to root cause the problem (short of buying
a tacx neo or a third power meter ;)?

cheers,
ed
  • First thing to note is that depending on how tightly you torque in the pedals, they may initially take a few hard efforts to settle in. Until that happens, you may see the zero offset shift from one ride to the next. Best practice is to make sure the crank faces are clean and square, torque the pedals into the cranks at between 25 and 30 foot-pounds (34-40 N-m), and then do a couple of 10-30 second sprints at a high relative effort. Then do a zero offset (or repeat that process, if you did one immediately after installation).

    After that, if you're doing a test like this and you see a discrepancy between two or more power meters after several intervals:
    (a) make sure neither is set to auto-zero.
    (b) Do a zero offset on ONE of the power meters.
    (c) Do several more intervals and note if the discrepancy changed.
    (d) Repeat steps b and c for the remaining power meters.

    If you re-zero all of the power meters during this test, and the discrepancy does not go away, then do the hanging weight test again immediately while all the gear is at operating temperature.
  • thanks for the reply.

    the last time i installed the vector pedals was about three months ago, and i normally ride at least three times a week, so i'm assuming they've settled in by now. i also made sure to put a few hundred miles on the powertap before i started paying attention to the numbers it's generating.

    i do have auto-zero enabled for both power meters. for the vectors, i assume they won't be able auto-zero themselves during a workouts unless i stop and unclip (which i normally don't do). but i do sometimes backpedal during recovery intervals, and i know the powertap can auto-zero itself when coasting, so that may have been a factor here. i'll be sure to disable auto-zero for both devices to make sure that isn't impacting my results. (i always do a manual zero-offset for each power meter at the start of every ride and i'll continue to do that.)

    during my next workout i can try doing the manual zero offset or different devices during recovery intervals to see how that affects the results.

    of course looking at the data from this last workout it seems like there was drift within the first 11 minutes of the workout (before my first backpedal). the powertap numbers started out below the vector numbers, but by 11 minutes in they were both reporting matching numbers.

    that said, do you think there's any downside to just doing a static weight test at the end of my next ride (regardless of the effects of the manual zero offsets)? i recall reading some youtube comments where someone recommended putting power meter components through a few stress cycles before doing a weight test, so presumably doing a static weight test immediately after a workout couldn't hurt?
  • If the pedals were substantially undertorqued on installation (say, below about 20 N-m or 15 lb-ft) then they may never settle in. They will just (like any pedal) slowly fret out the threads on your cranks, and give you inconsistent power data.

    There is no downside to doing the static weight test at the end of a ride. For Vector it is important that all the parts are at thermal equilibrium when you do the zero offset calibration, and if you've been riding for some time in a static setting then that should certainly be the case.

  • I verified that the Vectors were indeed torqued down more than 20 Nm.

    I did a static weight test after a ride and the scaling factors didn't change much (0.003%)
    Left: Pre-ride: 0.9781; Post-ride: 0.9809; Difference: -0.0028
    Right: Pre-ride: 0.9928; Post-ride 0.9905; Difference: 0.0023

    I disabled auto-zero and did a couple more indoor rides. I determined that doing a manual zero after a bit of warm up changes the values reported for both power meters. But past that I''m not able to draw any satisfying conclusions from the data and I still see my PowerTap consistently reporting notably higher values than my Vectors (well past the accuracy of both power meters added together), which just seems wrong.

    here's two rides that i've recently done and my notes about the rides.

    https://analyze.dcrainmaker.com/#/public/7bc54314-8146-480d-68b2-f1be8e3bcd65
    ---8<---
    TR PowerMatch with Vector3 was enabled. (this means that TrainerRoad would adjust the trainer power levels based on the power values are reported by the Vectors.)
    Auto-Zero was disabled for both power meters.
    I calibrated the Vectors and PowerTap before the start of the ride.

    At the start of the ride the Vectors were reading higher than the PowerTap, which makes sense. But by the first set of intervals, the PowerTap was reading higher than the Vectors, without any recalibrations.

    I recalibrated the Vectors after the 1st set (@ ~17:00:32), which resulted in an ~10 W drop in the PowerTap numbers. Since PowerMatch was enabled, this indicates that there was actually a ~10W jump in the numbers being reported by the Vectors.

    I recalibrated the PowerTap after the 2nd set (@ ~17:16:23), and that resulted in a ~27W jump in the PowerTap numbers.

    Take aways:
    - Power values at the start of the ride made sense (with Vectors reporting more than the PowerTap), but without any recalibration they both drifted to reporting equal values.
    - Recalibration increased the values reported by both power meters.
    - At the end of the ride, the PowerTap values are on average 12% higher than the Vector values, which doesn't make much sense.
    ---8<---

    https://analyze.dcrainmaker.com/#/public/d14a4585-0bfb-45a8-75c2-bb88bf77eea3
    ---8<---
    TR PowerMatch was disabled.
    Auto-Zero was disabled for both power meters.
    I calibrated the Vectors and PowerTap before the start of the ride.
    I pedaled for about 9 mins to warm up the Kickr and then at ~16:51:31 I did a Kickr spindown, I re-calibrated both power meters, and I started the TR workout.
    I re-calibrated the Vectors after the 1st set (@ ~17:19:32) and the 3rd set (@ ~17:54:59).
    I re-calibrated the PowerTap after the 2nd set (@ ~17:36:24) and the 4th set (@ ~18:09:27).

    During the warmup the Vectors were reporting higher power than the PowerTap (for example: Vectors @ 57W vs PT @ 35W), but after recalibrating both power meters and starting the actual workout (@ ~16:51:31), they power values reversed (for example: Vectors @ 92W vs PT @ 109W).

    The Vector recalibration at ~17:19:32 didn't seem to change anything. The Vector recalibration at ~17:54:59 resulted in a 3 W jump, but the PowerTap also showed a jump here, so i'm guessing that can be attributed to variance in the Kickr.

    The PowerTap recalibration at ~17:36:24 didn't seem to change anything. The PowerTap recalibration at ~18:09:27 resulted in a 5W drop, but once again the Vectors also registered a drop. So once again probably a change in the Kickr Power.

    My takeaway here is that:
    - The values from the warmup make sense, with the Vectors reporting more power than the PowerTap
    - Any subsequent re-calibration after about 10 minutes of warming up don't seem to make any difference.
    - At the end of the ride, the PowerTap values are on average 6% higher than the Vector values, which doesn't make much sense.
    ---8<---
  • I did a static weight test after a ride and the scaling factors didn't change much (0.003%)
    Left: Pre-ride: 0.9781; Post-ride: 0.9809; Difference: -0.0028
    Right: Pre-ride: 0.9928; Post-ride 0.9905; Difference: 0.0023

    I disabled auto-zero and did a couple more indoor rides. I determined that doing a manual zero after a bit of warm up changes the values reported for both power meters. But past that I''m not able to draw any satisfying conclusions from the data and I still see my PowerTap consistently reporting notably higher values than my Vectors (well past the accuracy of both power meters added together), which just seems wrong.


    The relevant take-away here is that you are getting the expected result (within tolerance) from your static weight test on the pedals, and it's not changing through the ride. You can do a similar test on the Powertap hub, but I am not sure what specific tool (i.e. how to configure a head unit) is currently required to do that. I don't know how you validate the Kickr.

    I recalibrated the Vectors after the 1st set (@ ~17:00:32), which resulted in an ~10 W drop in the PowerTap numbers. Since PowerMatch was enabled, this indicates that there was actually a ~10W jump in the numbers being reported by the Vectors.

    Can you post the .fit files for the pedals from these tests? Or email them to [email][email protected][/email]. It may be possible to see exactly what is changing when you did these calibrations.
    I would not expect a large jump after unless something was affecting the pedal installation even after you had put in several rides. Possibilities would include severe undertorquing but also debris between the pedal spindle flange and the crank mounting face or the threads.
  • My Edge 1030 reports torque values during calibration for the PowerTap,
    so i did a static weight test for that unit as well. I used the same
    procedure as with the Vectors. (But you have to divide the result by
    the gearing ratio.) The measured results were 1.5% below the
    calculated. Unfortunately there is no way for me to set the slope on
    the PowerTap unit. When i did the test the unit was cold though, so
    I'll try doing another test after a workout to see if the accuracy
    changes.

    My kickr snap is a pre-2017 model, so the accuracy is supposed to be +/-
    5%, way lower than either of the power meters. so generally I don't
    trust the numbers from that unit. But once it's warmed up, calibrated,
    set to a specific power, and i'm pedaling smoothly at that power, i
    generally expect it not to drift from that power. Looking around online
    it seems that most folks say that the reported kickr power values run
    high.

    The .fit files are downloadable via the dc rainmaker analyzer links.
    Just look for "Download Set Files", this has has a .zip with all the
    .fit files, and then look for a file with a -fenix5x.fit suffix. (I
    always just my Fenix 5x to record the Vector data.) If you'd prefer
    that I email them just lemme know.

    When installing pedals I normally remove any old grease from threads and
    the cranks and then add fresh grease. I normally don't use a torque
    wrench to install them, but I do use the Park Tool PW-4 pedal wrench
    (the largest one they have), which allows me to crank them down pretty
    good. (And I did verify with a torque wrench they are installed with >
    20 Nm force.) For the Vectors I also always reset the install angle
    (via a head unit) after installing them.
  • Not wishing to hijack the issues reported but I read with interest as I've experienced something similar. I serviced my pedals at the weekend following the Garmin video instructions. I did a static load test on them after servicing and found the left pedal was within 0.1% and right pedal within 0.4% so did not bother altering the Scale Factor. Yesterday I did a recovery ride on Zwift at a steady 156 watts for an hour also using my Tacx Neo which I have always been consistent with. The result was exactly the same from both the Neo and the pedals. Then last night I did an hour long race on Zwift pretty much at Threshold. The Neo reported 277 watts average for the duration and the Vectors reported 260 watts a difference of 6.5% something which I have never previously experienced. Admittedly I did a cold calibration at the start of the race but I have never done any different and I don't possess the means to torque the pedals up as I generally just nip them up as I understood torque was less relevant to Vector 3 pedals. I will clean the the threads, re-grease and tighten up and look at doing a sprint warm up prior to calibration and will then do a Step Test to see how well it tracks.
  • Admittedly I did a cold calibration at the start of the race but I have never done any different and I don't possess the means to torque the pedals up as I generally just nip them up as I understood torque was less relevant to Vector 3 pedals.

    The Vector 3 spindles are less sensitive to the installation torque but they do still need to be installed "tight enough". All that you probably need to do is to give them a firm pull with your pedal wrench and you should be fine. You will need to do a zero calibration after that.
    You should do the zero offset when the bike is at thermal equilibrium and making sure there is no load on the pedals. Besides that it matters little whether it's done at room temperatuer or outdoors.

    If you have the .fit file from your 260 watt session available I can see if there is an issue other than a possible zero offset discrepancy.
  • Thank you for the swift response attached is the full .fit file within the zip folder from Garmin Connect. The race event is Lap 2.

  • For comparison here is the Zwift .fit file for the same race generated by my Tacx Neo.