Tacx Flux S - Gearing / accuracy

Good evening,

First of all, my native language isn’t English. So if i made a typo or grammar error, please excuse me.

Once every while I check my data between my Tacx Flux S and my Precision 4iiii left power meter. Just to see if everything is on par.
In the latest workout I noticed a large difference between the data when shifting gears.
After I saw this, I recorded a workout with both the 4iiii and the Flux and tested if the power difference is there across all gears. This isn’t the case. There are just a few gears where the wattage (ERG mode) is close enough.

I’ve tested with the workout Carillon +1. After the warmup I’ve calibrated both devices, and begin testing. I’ve started with the 34 in front and the 22 in the back.

Every next step (every 2 min’s), I’ve shifted one gear. In the first interval, 34/25 the tacx couldn’t deliver the power so I’ve ran the 22 for the first 4 minutes.

"Interval 1: ERG

Interval 2: ERG
Interval 3: ERG
Interval 4: ERG

In case you want to run the data in DCR Analyzer:

Only the 34/16 en de 50 / 16 provide the same power numbers.
Is this just a flawed Tacx Flux S? Or is there something else I’m missing.


The power accuracy of many trainers vary with spindle speed (the equivalent of the wheel speed). So the same power at a different gearing will result in a different accuracy. In general, the faster the spindle is turning (i.e. the longer or “harder” gear you’re using at a given cadence), the more error there will be.


As this is becoming more widely shown, I wonder if trainer makers can review the issue and maybe work to dial in that variation at higher speeds? I don’t know if it comes down to them establishing power data at fixed wheel speeds outside of what we see in these examples or another reason.

But it sure seems like there is a potential for these variations to be closed to some degree and make them more accurate across the speed range.


I know Elite has sent users custom firmware mods to bring the two lines closer; that’s quite a lot of fiddling honestly (the concept of sending data points to someone in Italy who modifies a firmware load and sends it back is such a low-tech/high-tech combo, all that’s missing is using a fax in the process). An “Advanced” calibration mode that would allow a power meter user to enter a few set points and set a calibration curve would do miracles. It’s not rocket science.


Can this be a difference of like 40 watt ?

34/22 -|Avg Power| Tacx 248.44 w| 4iiii 238.49 w|

50/12 - |Avg Power| Tacx 254.84 w| 4iiii 280.99 w|

If so, does every trainer has this issue?

1 Like

Not “every”, but certainly many.

One of the reasons why it’s generally recommended to do your ramp test and your main intervals in a similar gear selection. That eliminates the variation - you set your training level and measure your workouts in the same “precision” range.


I can’t recall, but @GPLama & @dcrainmaker , do you guys ever take this into account when reviewing a trainer unit ?

I check for it, but I haven’t seen anything like you’re showing in a long while. That’s odd. Usually the issue I see with ERG mode and using the big ring is set-point accuracy responsiveness (how long it takes to stabilize), not power accuracy per se.

In your case, having a left-sided only unit makes it tricky to know for certain.

Note that while some trainers do still struggle with high flywheel accuracy issues, but for the most part I see that more in SIM mode than ERG mode. Most of my ERG-related high flywheel issues are around set point responsiveness/accuracy.

1 Like

I thought of a L/R imbalance, but my RPE and HR went way up during the 50 - 12 and 13 sets. So yes, its tricky to know for certain, but i’m pretty sure it’s an accuracy “issue”.

Although i’m a bit bummed by this, I will make sure for future workouts that i’m in exactly the same gear. So that my training and testing is identical measured every time. 34/16 in my case.

You’ve chosen to take the red pill… welcome to the unpleasant truth.

In Short:
It’s related to flywheel speed on the trainer. Since you’re using a left only meter (4iiii) to compare to total power (Flux) it’s impossible to conclude anything from the smaller variances. Having said that - things go well off the rails in the 50/14-13-12 gearing… so I’m sticking to my story on flywheel speed.

In Long:
I’ve found a lot (most) smart trainers have a known/good/tested working flywheel speed zone for ERG… and also power accuracy. I don’t recall any manufacturer publishing this information, or discussing it… but if you’ve seen “Use the small chainring for ERG” messages, this is what’s behind it. This is mostly about a trainer being able to ‘hold the right watts’… but the evil side is power accuracy can also go wonky too.

The Neo will be well outside +/- claimed accuracy if the flywheel is >40km/h or so. It’s easy to test this in ERG, not as easy in SIM (unless loading a util and throwing a -5% gradient at it and doing a steady-state interval). Recent firmware updates on the 2T have addressed flywheel speed/accuracy somewhat after this was discovered. I can’t recall if Tacx/Garmin have done any similar fixes on any other trainer.

Back to your data: When overlaying your Flux flywheel speed with the power reported, it seems to backup what I’ve seen elsewhere.

When the flywheel hits >32km/h or so it looks like the accuracy goes to hell too.

As always, more testing needed if you want to have a solid understanding of what’s happening. As for a resolution… I don’t think there is one.


Thank you all!! Very helpful to understand the problem.

I’m just glad it’s not only me! I’ve got the Flux S and the power readings are next to useless. I certainly cannot be bothered to test a whole bunch of gear combinations in erg mode to work out which is most accurate. Well done to you for doing it, but we shouldn’t have to.

It’s slightly depressing that in 2020, so many years after power meters came to the market, and a fair number of years after smart trainers came to the market, it seems that we’re still in a position where, a lot of the time, data produced by smart trainers is unreliable if not downright useless.

I run a left-only power meter (I couldn’t afford a dual power meter at the time) and regularly wonder just how far off my readings are. :pensive: I know that’s just the nature of left-only power meters, and prices have come down a lot recently, but smart trainers do seem to be variations on rubbish for power reporting.

1 Like

Hey, as mentioned on the other thread I’m getting very similar results on my Neo.

Very interesting.

I had very similar issues with a neo2 and a 4iiii left only. Pretty similar in small ring, Neo much lower in big ring (and HR suggests I am putting out more power in big ring for same erg watts).

It seems to be less of an issue at higher watts though. I wonder if the accuracy issue is not just high flywheel speed but high flywheel speed combined with low power. I can’t quite put my ginger on why, but I think this kind of makes intuitive sense to me for some reason. Maybe because if you have low watts and fast flywheel speed then resistance of the trainer must be very low, and that makes large proportional errors in resistance more likely?

I was thinking about whether this is likely to apply in slope mode as well.

If it’s just a high flywheel speed issue then presumably it would apply in slope mode when flywheel speed is high - say above 40km/h. If it’s high flywheel speed and low power, maybe it only happens downhill or possibly drafting on the flat drafting. But I guess it could be that the ERG mode makes getting the resistance correct more difficult because it’s trying to chase a target in which case it might just be an erg issue. I’m not sure why this might be from a physics/engineers perspective but it seems plausible.

1 Like

This is your problem. You have two sources of power. If you had just one source of power, then it would always be right :rofl::rofl:

On a serious note, I think tracking down the answer to your issue will be a frustrating effort. There’s a lot of potential variables at play. Here’s an example of one - does your power source use average or instantaneous angular velocity in calculating power.

This is why people generally recommend to use a preferred source of power, and stick to that.

Separate note: this issue is another reason why no one should take Zwift racing results too seriously!


Ha! Or maybe I need N plus one plower meters :slight_smile:

That’s really interesting.

So I guess it could be that big or small ring (ie high or low flywheel speed) affects how angular velocity is distributed around my pedal stoke and makes the 4iiii read high.

In the example they give the difference is 4.5% and the average angular velocity PM reads low. From what I can see, this is because the real angular velocity is greater on the down stroke when strain is higher so this error will typically manifest itself as the PM reading low. I guess in theory it could work the other way, causing the average angular velocity PM to read higher but that would mean very odd pedalling going slower on the down stroke.

Also, the difference between PMs goes from not much to 18% and that seems a large difference to put down to this issue.

Who knows though! I just find the stuff quite interesting

Agree. I was highlighting that issue as an example of one of a number (many?) things that makes it challenging to accurately diagnose what’s going.

1 Like