I'm trying to wrap my head around the following puzzle.

I have Garmin Edge 800, cadence sensor, and I used to have a Powertap G3 in a 700x23 clincher. I never got around to put a magnet on the spoke, but distances always appeared accurate. As recently as mid-September, I had a long ride that concluded with reported mileage within 0.3% from the correct value.

Last month I took apart the rear wheel, pulled out the power meter, put it into a 700C tubular rim, and updated wheel size in Garmin (both direct measurement and the table in the owner's manual indicated that I had to enter 2130 mm circumference).

Since then I had two long rides, and in both cases Garmin seems to be consistently overreporting distance traveled by approximately 2%. For example, here are the distances reported to certain points by my GPS and by another person who rode the same route:

18.81, 19.17 (ratio 1.019)
31.19, 31.84 (ratio 1.020)
50.15, 51.18 (ratio 1.020)

My unit is the one reporting 51.18. Plotting the same route in Google Maps gives 49.9.

One logical interpretation of the events is that I somehow failed to inform the power meter of the changed rim size. However, if you think it through, that does not work. If the powermeter were reporting distances based on the assumption of the old wheel circumference of 2090 mm, distances would be low, not high!

There's a 50:50 chance that I can fix this by hanging a magnet on the spoke, but I'd like to understand what's going on first.

I think that you probably set the Garmin as the default source from which to measure distance when you input the wheel diameter, so that info is coming from GPS, and not the Powertap nor the GSC10 speed/cadence sensor (which I'm assuming you have since you mention the uninstalled magnet).

My understanding is that you can pull speed/distance in 3 ways: from Powertap (which calculates based on hub rotations, not wheel circumference); from S/C sensor (which uses magnets and will override GPS input when installed); from GPS, which the head unit pulls.

I think that, by programming circumference info into the Garmin, you told it to use GPS input and disregard the ANT signals from Powertap and S/C sensor, although without a wheel magnet the S/C sensor wouldn't have worked anyway.

What setup you were using before I don't know, but if it was more accurate, I'm guessing you had the Garmin receiving speed/distance via ANT+ from Powertap.

Hub rotation requires wheel circumference to get distance.

<count of rotations> x <wheel circumference> = <distance>.

I'm not sure, but the Powertap hub may report wheel rotations just like the Garmin speed/cadence sensor does.

The crank-based power meters can provide cadence (which is counting crank rotations).

You can get distance from GPS. GPS doesn't require wheel circumference at all.

The Garmins will/can automatically compute your wheel circumference fromm the GPS data.

===========================

If you have a wheel rotation sensor (which could be the power tap), the Garmins will display that rather than the GPS distance.

The Garmins keep track of both distances (the wheel rotation distance and te GPS distance). The fit file has fields for both.

Thanks for the clarification. What I was meaning to address was the OP's concern that the Powertap was not informed of the circumference size; the hub does not do calculations, rather *transmits* only hub rotations for calculation at the head unit into speed/distance based on circumference info there. That's correct, no?

Again, thanks for the clarification. I thought a Garmin would reconcile GPS distance with circumference value, but I guess it just gives precedence to a paired S/C sensor (which requires circumference value) when available.

Could the OP's wheel size setting have switched from Custom to Auto (after the wheel build) causing the discrepancies?

Yes, that's correct. There's no way of informing the Powertap about the circumference.

It's not clear what "reconconciling" would mean.

(If you don't have a circumference specified, the unit will use GPS to determine it.)

Yes, it does the simple thing. If the wheel circumference is accurate, the distance using that (and wheel rotations) is going to be more accurate.

Maybe.

In my experience, Garmins will use the wheel sensor over GPS if there's one present. They can be set to auto calculate the wheel circumference after first pairing a sensor. If you change something and want it to recalculate the circumference, you need to force the sensor to pair with the Edge again. You can also manually set the circumference. This is best done by doing an actual roll out with the bike weighted rather than relying on measuring the diameter or measuring around the tire with a tape.

I have all my activity logs going back a few months, so I wrote a script comparing distances traveled according to the GPS (directly using latitude & longitude) vs distance traveled according to whatever sensor was used by Garmin (since I had no magnet on the spoke, this had to be the power meter).

For rides longer than 10 miles prior to the rebuild, distances almost always agree with each other within 0.5% and usually agree within 0.3%. (For perspective, if you do a climb at 7% and compare distance actually traveled vs. distance traveled according to longitude and latitude, ignoring changes in altitude, the mismatch caused by ignoring changes in altitude is 0.25%.)

At the first ride after the rebuild, disagreement is 0.34% over 62 miles.

At the second, third and fourth ride after the rebuild (I'm guessing that these are the three rides after I manually changed wheel size in Garmin's settings), mismatch is 1.8%, 2.0% and 2.2% in the same direction.

Today I glued a magnet on the spoke and made sure it was detected by GSC10. I'll test tomorrow and report what happens.

I don't think your power meter has any effect on distance measurement but I'm interested in your findings.

It might be interesting to go look at the raw data and see if the distance being reported is the actual geographic distance between two coordinates (I know golden cheetah will display the raw data).

That's exactly what I did with the script.

https://pastebin.com/kircLiJa

Hanging the magnet on the spoke appears to have no effect. I still see a 2% disagreement between distance reported and the geographic distance. I guess I'll just have to set the wheel circumference back to 2090 again.

I believe that the spoke magnet only affects the display but I have no facts to back that up (or research). It will fubar your ride if your sensor isn't registering the spoke magnet though, so there's that, even though your GPS is clearly registering movement. (basically pauses data collection). It's not a very clever implementation.

One word of caution here, though... while Garmins wheel use the wheel sensor, the "auto calculate" figures will almost always be off, sometimes to the tune of 10 percent of total distance. I always enter my wheel circumference manually (2096 mm for a 700x23 wheel) and never have issues.

Data derived from the wheel sensor is written to the fit file.

The fit file (produced by the Garmin 800) contains the following fields.

1) Timestamp.
2) Time (relative).
3) Position Lat.
4) Position Long.
5) Altitude.
6) Distance (from the wheel sensor, presumably).
7) Distance GPS.
8) HR.
9) Cadence.
10) Speed (I believe this is sensor speed not GPS).
11) Power.
12) Temp.
13) Laps.

I just ran into this today. Set the wheel size to "auto" at the beginning of the ride, took about 2.3 miles for Garmin to calculate ... it came up with 2060 mm. :O Unsurprisingly, instead of being 2% high, distance traveled was 1% low.