Shimano R9100P Accuracy

back to main Shimano R9100P Review

There are several aspects of accuracy we can look at. I’ll cover the absolute accuracy of power in the conditions/environments that I normally ride in and I’ll touch on power balance and cadence as well. Cadence is the easiest; you perhaps might not be too interested in cadence alone but it is an input into power calculations and sometimes you might find subtle variations of accuracy at different cadence levels…or not.

For those of us with power asymmetries then power balance has some importance but when I’m riding I’m mostly interested in the total power level being actionable.

Cadence

On this outdoor, bumpy Surrey outdoor group ride the cadence was +1rpm with the Shimano or -1rpm with the Assioma, depending on how you see it, which doesn’t sound like much but it’s over 1%. Sorry about the noise on this chart

I have found that same 1 rpm discrepancy on some, but not all, outdoor rides over both the last two firmware versions. I don’t why and I don’t know which device is the cause. However, I would speculate that it seems to be on stop-start or bumpy type rides where the discrepancy occurs – as a counter-example, here is a 90minute outdoor solo ride with very few cars or junctions (smoother roads) and this time the cadences were near-identical…or at least they were on the 90-minute constant effort portion, getting TO and FROM that portion had notable cadence variations.

On another day in the smoothness of my pain cave, the average cadences are identical with all 3 at precisely 96.5rpm

Here is another indoor ride and the average cadences were Wahoo Cadence (Kickr17) 92.6rpm, Assioma 92.9rpm, R9100P (92.4rpm or 92.5rpm if you remove the extra few seconds of recording at the end)

Test 1 – Gplama Test (1) OLD FIRMWARE

I followed gplama’s test protocol but added in my own warmup protocol before his planned, second calibration.

This first test was on the older v4.0.7 Shimano firmware which has been widely used for other tests in reviews.

• DIY Data Analyzer Link Here
• Source Data Files Here in the Shimano R9100P Folder

In total, the post-calibration power values (after 20:00) are R9100P 205w, Assioma 201.6w, Kickr 202w. If you assume a 2-4w drivetrain loss then that would make the R9100P and Kickr superficially agree.

Then looking at either of the 10-minute effort periods that same pattern (kickr+drivetrain loss equals R9100p) is repeated, and you can see that yourself if you dig into the data at the data analyzer (above)

Turning to the 4 under-over efforts towards the end then, again, what I see at the aggregate level could still be accounted for by drivetrain loss of 1-2%. Having said that, specific instants can be chosen where there is a great variation of up to 70w but instantaneous variations could be due to many factors such as simply setting an incorrect offset in the presentation of the graph. Here we are with 10s smoothing of those 4 effort periods and all looks sweet.

Then, unsurprisingly, the mean-max data looks alright too with the curves only starting to diverge slightly above 265w.

Drilling down further into the detail of the LR balance where I have looked at the variation from left to right for each of the effort periods in the test

• Period 1 the 10mins @200w – Shimano is 104.29(L)-97.27(R) =7.02w, and Assioma is 101.6(L)-96.6(R) =5w
• Period 2 the 10 minutes @250w – Shimano is 126.05(L)-125.76(R) =0.29w, and Assioma is 122.98(L)-125.16(R) =-2.18w
• Period 3 the 4x under/over efforts at the end @c300w – Shimano is 149.03(L)-149.93(R) =-0.9w, and Assioma is 145.43(L)-151.23(R) =-5.8w

So my asymmetry seems to go away at the higher effort levels according to Shimano and turns from left bias to right bias according to Favero. Both devices show what could be plausible data about my imbalances but, of course, they are not both right. The PMs could even change their power output behaviour at higher cadence/power levels too.

This chart is also interesting when the power balance on the 200w, 10-minute section is exploded out. OK, it shows my inability to maintain a constant 93rpm cadence but look how the LR balance of both the Assioma and R9100P diverge as my cadence initially drops.

Test 2 – Gplama Test (2) – LATEST FIRMWARE

OK here’s the SAME test repeated in broadly the same conditions. This time I had to repeat the second Kickr spindown/calibration at 20:00 as it failed a couple of times. I also used the latest firmware on the R9100P (v4.1.7) and I’ve seen comments reported from Shimano as saying that v4.1.7 increases accuracy despite the official firmware not saying that any accuracy changes were included.

• DIY Data Analyzer Link Here
• Source Data Files Here in the Shimano R9100P Folder

When looking at the total average wattage this time around, rather than the ASSIOMA and the Wahoo matching, it’s the R9100P and Kickr that match. Hmmm. Strange

Let’s look some more

Here we have the first two 10-minute periods, again don’t forget this is the latest firmware.

This data set does NOT support the assertion that the R9100P notably increases or decreases the right side balance. Why? Well, there is a skew to the right but the ASSIOMA also shows a similar skew which, if anything, is MORE pronounced. My understanding of GPLAMA’s tests on the Shimano R9100P is that the R9100 does NOT affect Assioma. Clearly here the Assioma resutls are skewed as well as the R9100P, which to me suggests it’s my pedalling and not the crank.

Why is that? Well, I’m not comparing like-with-like on numerous levels. Obviously the firmware has changed and there are two more Kickr spindowns on the second attempt. On the second attempt, I also missed the start of the first 10 minutes by a couple of seconds and I also stopped the test immediately after the 4x under/over sprints so omitting some low wattage data. I initially thought it was this last factor that made the difference as the only time I noticed a difference in what head units displayed was when I was cycling at low levels (150w ish), whereas in the 200s of watts it was usually hard to see any material difference on the head units. But if I zoomed in on the period from the start of the 10 minutes to the end of the last under over effort, the average power between tests was similar and maybe it was a 2rpm difference that was more of a potential culprit?

Test 3 – Gplama Test validations – LATEST FIRMWARE

I wasn’t happy with those results. So I repeated the two 10-minute periods in Gplama’s test but this time with the higher effort first (preceded by  a further calibration)

• DIY Data Analyzer Link Here
• Source Data Files Here in the Shimano R9100P Folder

Here is the overview of the test and it ‘looks’ a bit noisy.

Looking at the mean/max curve you can see why because this time around the curves clearly diverge. Nothing has changed here with the hardware or general protocol. Neither the pedals nor the crankset has been adjusted/removed in any way.

Maybe the temperature of the room was different? It probably was but, nevertheless, the devices had 30 minutes just standing there before I started. In any case, I could see the variation in the first 20-minute period and so performed several re-calibrations of the Assioma and R9100P before the second 20-minute period.

Unsurprisingly the LR balance is a mess.

That last test was probably a complete waste of time…I just don’t know why. I’d like to understand what was so different from previous occasions?

Looking at all 3 tests, I’m not sure I can conclude anything from their data. Generally (apart from that last test!) Assioma’s overall claim of +/-1% accuracy and Shimano’s overall claim of +/- 2% accuracy seem reasonable.

Test 4 – Gplama Test validations – LATEST FIRMWARE

I wasn’t happy with the previous results either. Maybe there was some temperature-related factor? or some other spurious factor? This time I left the bike overnight in the pain cave so that temperature couldn’t be a factor. I also thought I’d up the cadence a tad and use the big ring (still with a straight chain run)

• DIY Data Analyzer Link Here
• Source Data Files Here in the Shimano R9100P Folder

The results looked more reasonable with the Assioma and R9100P showing similar wattages but the Kickr showing MORE watts than both those two. #Sigh.

I’ll add some chart in, maybe. But you can look at the charts now in the link above (or here)

Other Rides

I don’t want to dwell on other rides. However, here is a summary indication of how the Assioma and R9100P differed on other rides.

* At the time of writing, there’s an additional 400km of ride data not shown here

** Right Power Balance (% R9100P) less the same %age figure from Assioma

• The R9100P usually but not always shows a slightly higher power
• The R9100P always peaks at a higher level
• For me, the R9100P usually but not always UNDERstates right power more than the Assioma
• R9100P and Assioma agree less on cadence than I expected (I will investigate further outdoors with an extra cadence sensor)

Specifically commenting on LR balance issues, I would concur with this from dcrainmaker who said, “In any case, I can’t specifically say that I’m seeing broad balance issues across the board, despite looking at all my files. Partially because it’s also really tough to, with 100% certainty, say that either Garmin [ASSIOMA] or Shimano is perfectly correct.”

Thoughts

• Mean/Max Power curves seem highly similar below 300w, perhaps less so below 200w.
• Higher cadences seem to increase agreement between the PMs

Further Investigation

I’ve spent more time on this than I planned. I’m talking some more with Favero about the data from the Assioma as there are some issues thrown up that either weren’t there before or I hadn’t noticed them.

I’m going to continue to get more R9100P data as it will stay on my Cervelo R5 for at least another month or so.

I will be looking at the Stages G3 Ultegra alongside this on my S3, so perhaps some patterns will emerge there too showing performance on Shimano cranks and against the Wahoo Kickr 17.

Summary

Over these tests and other usage, I’ve no reason to dispute stated manufacturer accuracy tolerances.

Other than that, I came to several tentative conclusions about certain aspects of the data in my mind only to relatively easily find several reasons why such conclusions could be wrong.

I’ve provided ‘standard’ test results that can be downloaded and which might be useful for someone else as part of a larger data set.

back to main Shimano R9100P Review