I reread this, and L/R balance could be at play. It is really impossible to know without a total power PM as a control. Here's the thing... while the power may be the same, the way your body produces the power is very different. At a low flywheel speed, the trainer is using a much higher resistance, so you fee resistance much more across a broader range of the pedal cycle. At a high flywheel speed, the resistance is light, and your power delivery is concentrated over a smaller range of the pedal cycle. Gplama measured this and illustrated the differences.


This is exactly the type of scenario that causes pedal L/R balance to vary. So, at a light resistance, maybe your L/R balance is at one point, but then as the your power delivery phase changes, your L/R balance changes too.


The nut is that you are trying to compare accuracy between a single sided PM that estimates 1/2 the total power to a PM that measures total power. That is pretty much impossible. I would vote in favor of the trainer's power reading in this case.

But that's kind of my point. It's easy to point at L/R balance, one sided power meters etc etc. and say what you have just said. I know, I did it for long enough and I read enough forum posts to convince me that this was the case.

What I then did was flip the question on its head and asked myself - what if my power meter is right and the Hammer is wrong? What would cause that? (OMG my FTP is worse than I think!!!)

When you take that approach and start to think about the basics of calibrating and that pretty much anything you calibrate has a calibration range, and if you are outside of that range then there is a chance that accuracy is lost. Then things start to look different.

Before ERG mode, a trainer would only spin as fast as your gearing and engine (you) could make it. My guess is that that influenced design - why would a company who wants to make money from a mass market trainer be all that interested in speeds that only a small % of cyclists could reach? Look at the resistance curves that Cycleops put out - they all stop at 30mph.

If you have your head in that particular design space and have been knocking out trainers for years, a design with a window of power accuracy between say 10mph and 25mph might be considered reasonable. Then ERG mode comes along and it tells you that it is holding you at 100w whilst spinning the gears at the equivalent of 35mph.

Put this into context. Bike calculator tells me that 100w should get you around 15mph and 500w should give you 27mph (it won't give me anything higher), yes it's an algorithm, but you get my point - how can you expect a trainer to be accurate in that circumstance?

But we assume that because we are in ERG mode and we spent a fortune on a trainer that is "PowerTuned using PowerTap technology for accurate power readings" then surely it can't be wrong...

I might be completely wrong (after all, I am not PowerTuned ) and as I say, much more testing can and will be done.

I lean toward Occam's razor when speculating on complex problems with unknowns. On one hand, it is possible that smart trainers were not designed to maintain power accuracy at high flywheel speeds. On the other hand, you tested against a single-sided PM that very likely introduced errors into the comparison.

You did not say what the initial difference was beyond "a few watts." So, I will assume it was 5W, 2% error at 205W.

Let's say that you have a natural 55%/45% power balance at low flywheel resistance (higher speed), but you pivot to 50%/50% at high flywheel resistance (lower speed / climbing). So, that is a 10% swing in power accuracy, or 20W at 205W.

You observed a 25W difference at high speed. Using simple and reasonable metrics, we just accounted for your entire 25W difference. Both devices have a 2% accuracy level, and it is possible that your power balance swing could be more than that simple illustration.

This post above provides a similar scenario to the OP's measured differences. https://forum.slowtwitch.com/...ost=6879521#p6879521 The thing is, the comparison is practically meaningless until you can do it with a total power PM.

Adding some thoughts to a good discussion. I class myself as a decent cyclist, more-so long distance TT, but do most of my 'training' in a general road ride group where I'm near the front of the front bunch. Context last IM bike split was 5:15ish. So not elite but competent.

My quarq had been telling me I was normally 48:52% left / right on long rides, drifting a bit late on which was no surprised as I was carrying / managing some running injuries. Anyway, in lockdown then my tri-club started some zoom based trainer sessions that I went to for social benefit mostly. I was always pretty happy with my coaches structured sessions and zwift for indoor stuff.

Anyway, in the first one we did some one legged drills. I've done these before (several years) and was always a little stronger on my left than right, but not a lot in it. Anyway a month back I got 30seconds into a 1min set on the left leg and failed totally. Right leg was fine throughout. In addition, the power reading (from quarq which needs to estimate the left / right contributions) was also really illuminating. Basically showed that I was way better at pedalling in circles on my right (ie it showed a 15-20% contribution from left leg when it was unclipped) as opposed to the left leg pistoning.

So not the same as the situation you are in exactly, but does, for me, show that a single side power meter would be way out even if I had been able to calibrate it to myself when I first got it, and also changes through a ride.

But your assumption is that is is a "reasonable" to assume a shift from 0% imbalance to 10% imbalance at a relatively modest workload, and that this balance swing is entirely (?) the result of the greater inertia of the flywheel. I would question that and to be honest I think Mr Ocam might well accuse you of "multiplying entities without necessity"...

So keeping with that theme, which is the simpler explanation?


1. One device is operating (way) outside its calibration
2. One device compounds any input error.


No.2 is known, but whilst input error is possible (even probable) neither its existence or exact manifestation can be assumed.
No.1 is assumed and remains to be proven.

I think both have interest, I think both have the likelihood to influence, and at the end of the day both may well play their part in the end result.

It's an interesting discussion and I apologise to the OP that I (we) have hijacked this thread. What we do seem to wholeheartedly agree on is that this is frankly a fools errand . But whilst lock-down is still in place and I am spending so many hours staring at these two different power values I guess, for my own entertainment at least, I'll carry on playing the fool for a bit longer...