Zwift have apparently halted their elite Zwift World Series due to some edge-case trainer accuracy issues.
4 Likes
I would love to get @GPLamaâs take on this, maybe he needs to do laps of the radio tower at a low cadence to test all these trainers again 
2 Likes
Smart trainers have always had what I call a âGoldilocks zoneâ of operation when it comes to accuracy and capability.
The Lama Lab Test protocol has a number of tests that usually push a trainer outside of these bounds. Lower zone ERG and flywheel speed tests in ERG being two that a lot of failures occur in. I can confirm after testing 100s of trainers, none are perfect. Some are a lot closer to perfect than others. 
Accuracy drift at low RPM (and lower flywheel speed) at steep SIM gradients combined with thermal issues with elite racers pushing big power numbers is definitely an interesting problem trainer companies will have to address.
Requiring on-bike meters for accuracy is one quick way to resolve this from a product standpoint. Although this is not an ideal solution for the sport due to the commercial arrangements/sponsorships of eRacing.
Iâm on the road this week and within literal earshot of some key players in this space. All of this and more will be discussed. The goal is how to resolve it, or to be part of the resolution. Not how to blow this up into a massive controversy for social media / YouTube views (which Iâm sure will happen).
16 Likes
There are some fundamental things that can be done at the platform level to alleviate the impact of some of this inaccuracy, rather than everything being put on the hardware providers. It doesnât make a lot of sense for manufacturers to invest a lot to resolve these edge cases, if it doesnât ultimately lead them to sell more trainers (which is to the masses, who want close-enough accuracy for training and everyday platform use and donât care about any of this).
Examples - matchmaking (so accidental or deliberate fake watts simply mean you face tougher opposition), in-game DQs (sticky watts, microbursting, misconfigured trainers), âstrictâ races versus free for all races (where you know what you are signing up for).
That ticks off the 99.9% of the userbase, then the elite racing problem can be considered in isolation - where there may also need to be an amount of investment and commitment from the riders as well.
4 Likes
Failing to fix it will lead to selling fewer trainers once this blows up on all the usual channels
1 Like
I agree there are some things that can be implemented on the platform, but thatâs still like trying to put out a fire by spraying water on the flames.
I donât think it will.
Itâs hard to see itâs a niche, when youâre in the niche.
I am sure wahoo will sort it out from goodwill and maybe a fear of potential negative coverage (which probably isnât real).
I can see some benefit to a certification system IF it actually makes a difference for the wider more casual audience.
This isnât a Shimano PM. The data isnât junk. Itâs just unfair for a tiny number of people racing at an elite level (<100).
This would never even be spotted at community level where the issues are FAR broader and more impactful (ski erg ZRL anyone?)
2 Likes
I use an unaffected device and Iâm far from elite. But if I were in the market Iâd not buy something with serious accuracy issues.
1 Like
I might be looking at this from a different perspective as I see this as being EXACTLY like the Shimano PM issue where the data coming from the unit isnât correct. The title of this thread says âTrainer Accuracy Issuesâ, which is what this appears to be. The issue will never be resolved unless itâs done at the source.
Looping back to the circus act Shimano are performing with their power meters that adjusts the right side scaling by ~12% depending what chainring youâre in (by pairing your power meter to the rear derailleur (master unit))⌠thatâs the kind of complex knot that happens when the root cause of a problem isnât addressed.
Anyhow - Iâm off topic somewhat. And not trying to be argumentative. These are good discussions to have as theyâll ultimately lead to a solution (or at least a better understanding of the problem).
Iâll watch with a keen interest how this plays out. I hope at the end of the day the result is better performance from the products people are buying off the shelf.
7 Likes
My differentiation is that Shimano PMs are useless for pretty much everyone.
These issues are confined to a specific set of circumstances and a very limited set of people. Iâm no Wahoo apologist, but they were also exposed by virtual shifting which isnât something they could have exactly designed and tested for.
Shining a light on it is good, the best way to get stuff improved, of course. However I still struggle to understand the business case for a trainer manfacturer to invest a lot of time and money in to such marginal adjustments when the weight of the market doesnât care.
This thread is a good indicator. Very little engagement (also the case where itâs reported elsewhere). Check out todayâs thread on the racing score changes to see the difference in interest. Both about fair racingâŚ
3 Likes
Iâve been following the trainer issues since March this year when I had an open ticket for ghost watts while coasting. I was advised it was an unknown and unique issue and to do a factory spin down which cleared it for a while, then a few months later it showed up again and a replacement trainer was provided under warranty.
It started shortly after the implementation of virtual shifting and several threads started to pop up in different forums siting virtual shifting issues.
Zwift have now stated âThis season in the Zwift World Series, we have seen, and we continue to see, a much higher volume of issues with hardware accuracy than we ever have in the pastâ
Iâve considered and tested many variables (interference, connection method, software platform, mobile device, belt tension, cold starts, chain ring combinations, rider pedalling imbalances, flywheel balance), my isolation of these variables has led me to think that virtual shifting is placing increased strain on the trainer internals especially when climbing at low cadence with large physical gearing. i.e the trainer is having to apply a lot of magnetic resistance to slow the flywheel and during the next sprint and coast there is a un-resolvable mis-calibration.
Iâve been as diligent as I could be with racing, and always check for coasting watts, In my experimentation I found that it would most likely show up the next time I loaded Zwift after sprinting at the end of a prior race, after a long climb out of the saddle and then coasting for rest, or after a change in user (on a shared bike/trainer).
I would be very interested in eventually finding out the root cause, I have burned rather a few mental calories and had several late nights testing and looking for the cause so that I could eliminate it. The best solution I came up with was avoiding virtual shifting and going back to mechanical gearing.
I was very hopeful with recent firmware updates and they seem to have worked for most part.
9 Likes
Is it wrong for me to wish I was capable of producing enough power for this to be a problem that I had to worry about?
Not sure how this would affect trainer sales. One the one hand, I bought my Direto XRT in part because of its 1.5% accuracy was the standard for elite racing (though Iâve never been elite) until like a month after I bought it when the standard became 1% (and I got pissed off that my trainer would no longer let me join races Iâm not qualified to join anyway). On the other hand, everything Iâve seen says that, if I had to buy a new trainer today (and I could get one), I should buy a JetBlack Victory and that has only 2% accuracy.
So 2 different things pulling there - wanting something that would be good enough to race on should I magically gain 2 W/kg, but on the other hand, wanting something thatâs got the most features for the lowest price even if itâs not what the pros would use. Maybe it means that, if I had $2k to blow on a trainer, Iâd be looking at the Neo instead of the Wahoo even though the Neo seems like much older tech, maybe it means Iâd ignore them both for a Victory.
3 Likes
Exactly⌠are you are a very engaged and considered zwifter. 95%+ just want something that does the basics very well.
(We know that even the number of people racing is a tiny proportion of the zwift userbase, let alone those that can spend time on these things).
3 Likes
Probably not, but think of it this way, the issue wonât help you on a real bike in the real world.
1 Like
How many people race on Zwift for the purpose of being better at racing outside?
1 Like
5 years ago it was almost all. Nowadays it is the opposite.
That is for racing on Zwift âseriouslyâ.
At a high level indoors and outdoors are not compatible anymore. You have to commit to one and take the hit on the other.
For people commiting to online it is unlucky that there are no standards and not much accountability. In the early days I was hopeful for a well executed sport but it has gotten worse and worse.
It is not easy to implement reality in computers but even really basic stuff is not done. Maybe Zwift should spend less time vendor locking the virtual shifting and do more in implementing basic checks and balances.
1 Like
I think you might be on to something in relation to the virtual shifting. Just thinking about it makes kinda sense. The elite racers ride at relative high power outputs for prolonged time. If they use virtual shifting, that would mean the trainers would constantly be braking the flywheel at a high power, meaning the flywheel actually getting a lot hotter than previously when they were using actual gears. Then they only had harder braking when going uphill, but their gearing would mean the flywheel speed would slow down. With virtual gearing the flywheel speed stays at a high speed, while the braking power fluctuates the whole time. I think the trainer designers never even considered this scenario.
2 Likes
Iâve not fully been able to wrap my head around virtual shifting.
In order to produce torque/force at the pedal, you need a opposing force. Your opposing forces are Mechanical Load from the drivetrain (Rm), and Electromagnetic Load (Re). The Electromagnetic Load can be thought of as rolling terrain load (Re) and then a function of virtual gear and trainer difficulty x(Re)
Assume you are in a mechanical gear ratio of 1 (small chain ring and big cog). Your rear cassette and freehub is mechanically coupled to the cranks and rotates at the same RPM as your pedals.
You are pedalling at 90 rpm, in a mechanical gear ratio of 1, and incline of 10% at 100% trainer difficulty. Firstly, you have a mechanical advantage by using a low gear ratio. The mechanical resistance Rm is low, so the manufacturer has to add electromagnetic resistance to model the terrain. Rm + x(Re)
So in this case Re needs to be large to provide the feel of 10% at 100% difficulty. But at the same time, you request virtual gear 1. This has a ratio of 0.75 according to Zwift tables . So it has to feel easier to pedal. Therefore x will needs to be below 1 to reduce Re.
Now imagine the same scenario; but with a mechanical gear ratio of 4. The rear freehub is now spinning at 360rpm. That is 16x the energy in the flywheel as before, because energy in a flywheel is proportional to the square of its rotational speed.
The manufacturers canât physically stop the rear freehub spinning at 360rpm since its mechanically coupled to your legs. So now the manufacturer has to deal with 16x the energy when thinking about any electromagnetic control.
In this case Mechanical Load, Rm is 4x higher than before, Total load still needs to model a 10% gradient at 100% difficult at a 0.75 gear ratio. So here less electromagnetic resistance would be applied than before.
In both scenarios as the cyclist you are pedalling at 90 rpm and âfeelingâ resistance that is appropriate for 0.75 gear ratio on a 10% gradient.
However, the big difference is that in the first scenario, the rear hub is spinning at 90 rpm and more electromagnetic load is applied to balance the total load.
In Scenario 2, the rear hub is spinning at 360 rpm with 16x the energy in the flywheel and increased mechanical load.
Letâs take this to extremes, you put your mountain bike into a 32t front and 52t rear mechanical setup. Your gear ratio is 0.615. So at a cadence of 90rpm the rear freehub is spinning at 55rpm. You then ask for virtual gear ratio of 24 (a ratio of 5.49) to brute force a standing climb at 10% trainer difficulty. The manufacturer has to deal with a scenario where the freehub has very little energy, its hardly spinning! but you have asked for a very big gear so you can apply force at the pedal. Not having any mechanical load to work with, all the manufacturer can do is apply a very large electromagentic load which may well be out of tolerances for what they anticipated and the rider may feel the pulse like oscillations from the magnets.
And visa versa, you put a roadbike on with 60t chainring and 11t cog. You have a mechanical gear ratio of 5.5. You ask for a virtual gear ratio of 1 to spin up a climb at 100% trainer difficulty. This time at 90rpm the freehub is spinning at 490rpm and it sounds like your turbo trainer might take off.
You have significant mechanical load as you try and turn the 60t chainring and it likely already feels too hard even without any electromagnetic load applied.
This is all a very complex set of real-time calculations to solve with terrain changing all the time. I would love to know how they have accounted for these scenarios.
5 Likes
I was curious about the ratio of the flywheel to the pulley wheel and belt. It looks like a 10:1 ratio.
So based on the two scenarios above you could have the flywheel spinning at 500 rpm or 5000 rpm while requesting the same virtual gear.
1 Like
I like your analysis. But youâre confusing two things. One is the kinetic energy of he flywheel, which causes inertia, and the other is the resistance.
When you ride outside, and are moving quickly (on the flats, for example), and let up on the pedals, your speed will decrease slowly, due to inertia (mostly the momentum of moving you forward, far less for spinning the wheels). Similarly when you increase force on the pedals, the pedals donât start spinning faster instantly: they slowly increase rate of rotation, due to inertia.
When youâre climbing (or riding into a headwind), speed is lower, so inertia is less, and so the pedal rotation rate will respond more quickly to changes in force. This âlow inertiaâ condition is why a lot of riders can produce more power climbing than on the flats. Itâs easier to apply force thru more of the pedalstroke.
Inside, you are not moving foward, so all of the inertia comes from spinning the wheels (low inertia) and the flywheel (more inertia). The heavier and larger the flywheel, the more the inertia. This affects the feel of the pedals. This is why buying a more expensive trainer with a heavier flywheel has a more realistic feel.
So if you just stay in the same mechanical gear with the same cadence and power, using âvirtual shiftingâ to adjust, then the inertia is always the same. The flywheel is spinning at the same rate. This is different than outdoors, where even if you shift to maintain the same cadence and power, inertia changes, and climbing and the flats feel different.
If you use mechanical shifting, then itâs more similar to outdoors. On the flats you spin the flywheel faster. On the climbs you spin the flywheel slower (less inertia).
But this doesnât have anything to do with resistance. Resistance as you point out is applied to the flywheel. So this has to be done whether youâre using virtual shifting or not. The âresistanceâ from rotating the flywheel at whatever speed is low: you can see that on coastdown tests. It takes a substantial time for the flywheel to come to a stop, even adding in the mechanical resistance of the cassette ratcheting.
To apply a given power load = torque x rotation rate of the flywheel, less torque is needed if the flywheel is spinning faster. Iâm not sure how significant this is. What matters is if the system is heating up. And heat = power. So the more power youâre putting into the trainer, the faster it will heat.
I do wonder about the claim âonly elite athletesâ need to worry about this. The examples cited are women, who produce similar power to a lot of amateur men (who likely weigh more, but that doesnât matter). And if the affect is at all linear, even 50% of 8% = 4% is a big deal. Thatâs around 2 minutes up Alpe.
2 Likes