update


(Test Runner) #1

(Johnny RCCSFC (D)) #2

I’ve noticed the kickr (previously) would get into some mode where the effort felt the same. I would usually stop the zwift software and start again and it would go back to being normal. This has happened to me 3-4 times… just not lately.


(D LS (OSI)) #3

No you’re not imagining things - I felt the same on my ride tonight. I too saw that a new update was installed and it sure felt much different than before the update. I use to be able to recover on the downhills and now it feels like I’m riding in molasses and have to put in a lot of effort just to maintain speed. The best way to describe it is it feels like the wrong value was entered in the program for the viscosity of air. It doesn’t feel similar to riding outdoors - the previous version felt much better to me (and frankly was more enjoyable even when suffering :slight_smile: ).


(Christian Wiedmann [X] 50) #4

Look at power rather than speed. Is your power staying the same as your cadence stays the same? If so, you’re experiencing the bug/issue where the Kickr resistance is no longer being adjusted by Zwift.

Are you using a USB extension cord to get your ANT+ dongle closer to the Kickr? If not, that might be worth a try. Also try turning off as many other wireless devices that might interfere (e.g. Bluetooth on your phone).


(D LS (OSI)) #5

@Christian - It’s different than the bug you describe. I’ve experienced that on a number of occasions. I have the ANT+ dongle on an extension cord. The reception in the TrainerRoad app shows “Excellent”.

This is different - something has definitely changed. The downhills require significant effort just to maintain speed, while before it felt more like it does riding outdoors.


(D LS (OSI)) #6

@Dickson - My ride felt the same today as well. Much much better, the downhills felt just like they use to. Very enjoyable training session for me.


(Steve Paltzer) #7

Short answer from Zwift is “No math was changed in the update”.

I was wondering the same thing, and posted in the Beta Feedback thread before I saw this thread. I am on rollers, not a trainer, so I can not comment on the resistance for the hills. However, my speed data seemed more accurate on this last ride. I am going to sprinting more on next ride to see how that affects my results.

Here a link to the thread I stated, with an official Zwift reply from Jon Mayfield:

https://support.zwift.com/hc/communities/public/questions/201339069-What-s-new-on-Zwift-Island-around-12-27


(Christian Wiedmann [X] 50) #8

@Dickson - Were you able to add the Kickr as the speed sensor? I was only given the option to add my Garmin speed sensor (Kickr was set to power) and it rode exactly like before - with the Kickr power determining speed. Without the rear wheel the Garmin sensor doesn’t detect any speed, of course.


(Christian Wiedmann [X] 50) #9

@Dickson, I’m actually quite familiar with the way ANT sensors work and how Zwift is calculating speed from power. It’s a further nuance I’m after.

As far as I understand it, Zwift is using the Kickr’s simulation mode to set the resistance. This means that in theory the Kickr is setting the resistance so that the speed it’s spinning at matches the power you’re putting in given air resistance and the grade you’re going up. I was hoping to tell Zwift to use this speed directly rather than doing the power calculation by setting the Kickr as the speed sensor rather than the power sensor. This is what I couldn’t figure out.

If one were able to do this, it should reduce the lag since the power measurement is averaged over time, but the speed is not (or at least much less so).


(Christian Wiedmann [X] 50) #10

@Dickson: I’m probably not explaining what I mean clearly enough. I’ll give it one more shot from the start. If I’ve used up your patience, feel free to ignore this post.

As you probably are aware, the Kickr has four resistance modes: erg, level, resistance, and simulation. Erg sets resistance so that power stays constant. Level has a resistance curve based on speed similar to a fluid trainer. Resistance mode has a linear resistance curve like a magnetic trainer. Simulation mode is the interesting one, and is the one I’m talking about.

Simulation mode on the Kickr lets you set rider weight, air resistance, rolling resistance, slope and wind speed and then adjusts resistance based on riding speed given those parameters (see http://www.dcrainmaker.com/2013/03/fitness-trainer-review.html for screen shots of the Wahoo app).

From other posts I’ve inferred (maybe erroneously) that Zwift is using the latter mode to set the resistance on the Kickr. They just update the slope as you progress around the course and the Kickr changes resistance in response to that.

Given that they’re using sim mode, and provided they’ve set the parameters correctly, they Kickr’s speed should match the speed a rider would have on the course at that point given the power being applied. Therefore, the Zwift could theoretically just read speed off the Kickr rather than power. That’s all I’m saying.

I use another program called veloreality that does just this. It gives a much more connected feel to the ride than Zwift has when reading power instead, probably because the power reading is laggy because of the averaging involved.


(Christopher Pallotta) #11

“Zwift is using the Kickr’s simulation mode to set the resistance. This means that in theory the Kickr is setting the resistance so that the speed it’s spinning at matches the power you’re putting in given air resistance and the grade you’re going up. I was hoping to tell Zwift to use this speed directly rather than doing the power calculation by setting the Kickr as the speed sensor rather than the power sensor. This is what I couldn’t figure out.”

This can be confusing, but that’s not how it works. It’s actually more like the opposite. The Kickr’s speed or RPM is irrelevant. The Kickr does not tell Zwift anything, except report back the current wattage. Zwift tells the Kickr what the current grade, air resistance and your weight is and the Kickr uses this to simulate the correct resistance based on those parameters. 

When using a Kickr in sim mode, speed is best to be calculated by the application controlling the simulated environment. That’s Zwift in this case. Zwift knows all the factors needed to compute a realistic speed. How many RPMs the Kickr is turning is irrelevant to speed.


(Christian Wiedmann [X] 50) #12

I understand this isn’t the way it works in Zwift. However other simulation software does exactly this. Both Zwift and the Kickr are running simulations. In theory it might be possible to get the parameters synced up enough that the two simulations match close enough (or at least as close as the variance in different power measuring methods). If Zwift were feeding the exact parameters to the Kickr (which it will be able to do when the settings screen is introduced) and the two models were the same, in theory the Kickr speed should match Zwift’s speed. Hence, it would be possible to use speed as the input from the Kickr instead of power.

The big advantage of doing it this way is that you don’t get the cadence vs. speed disconnect that the current simulation has. There are a few places on the island where my cadence and speed are changing in opposite directions (i.e., cadence increasing when speed is decreasing and vice versa). On almost all transitions there is a significant lag during which my speed changes but cadence stays constant.

I understand that in practice there are many hurdles to getting the two models synced up. There would also be issues like how to handle drafting - maybe as a tailwind. I’m not holding my breath for this feature, but based on experience riding both models it really does feel significantly better the other way.

This is no longer my highest priority issue in any case. My new pet peeve is the flywheel compensation issue with the Kickr. There are places where the grade changes where the Quarq and Kickr wattages can momentarily differ by as much as 100 watts, it seems (350 Quarq vs. 250 Kickr when cresting the first 10% hill after the green jersey sprint). Unfortunately this is probably in Wahoo’s arena to fix.


(Christopher Pallotta) #13

I disagree with the notion that the Kickr works differently with other simulation apps or that the RPM of the Kickr should always match speed. It’s about power not RPM. There are instances outdoors where one could be spinning faster, but going slower. Again, it depends on the power, grade and gearing, not  RPM. 

Where I do agree, though, is that the Kickr is slow to update its power and I notice the same issues as you. After cresting a hill, my SRM power will quickly respond and update power correctly. Watching the Kickr power alongside this in Zwift, power always drops considerably and consistently after cresting a hill and things flatten out. Even if I maintain the same power (or more) the Kickr’s reported power always drops in those cases. 

Kickr power in general is out of wack and this issue just adds to the problem. I’d like a mode where Zwift controls the Kickr’s resistance for a realistic road feel, but the speed and displayed/recorded power is done with an external power meter. Trainer Road has this capability. A further improvement on that will be to allow the Kickr to be controlled by an external power meter and having its internal one overridden. 

It’s all in the works. 


(Christian Wiedmann [X] 50) #14

Let me try a different way of explaining: if the simulation were completely accurate, the Kickr’s speed should always match simulated speed, because the Kickr would be applying the resistance necessary to make you generate the power that matches the speed. If that weren’t the case, it’s effectively making  a virtual gear change.

There is a free trial on veloreality.com that lets you try their software (if you can run Windows software). They have a slope scaling feature that can be turned on or off. When the slope is being scaled, they use power from the Kickr. If it isn’t being scaled, they use the speed. More tuning would be necessary to use the same technique for Zwift since there is a speed change when it shifts from one mode to the other (the scaling can be set to turn on or off depending on grade), but that might be achievable by tweaking the simulation parameters.

As to the external power meter solution: unfortunately I have a Powertap, so I won’t have the ability to use power meter linking. I happen to have a “good” Kickr, though. Its power is close enough after warming up that the numbers are usable (verified yesterday using a borrowed Quarq). It reads considerably low when cold, though, so until it warms up I’m working harder than indicated. I’m hoping they’re not just going to give up on solving the drift issue (which seems temperature related) or the flywheel issue.


(Christopher Pallotta) #15

Christian, I think I understand where you’re coming from. I believe software developers such as Zwift have some latitude as far as incorporating the Kickr’s slope features into their own environments. So, I guess my response would be that I think it may not necessarily be correct to assume that the Kickr’s interpretation of speed at a certain weight/power/slope is “completely accurate” as you write compared to what Zwift determines is accurate. 

Given that we’re all riding in Zwift’s environment, I think Zwift needs to make these determinations provided that they’ve done the best to equalize performance across the many trainers being used. 

Zwift works on the principle of power determines speed. It’s the power output of the Kickr that is important to determine speed and it’s how speed is determined for every other rider on Zwift. Expecting Zwift to disregard the Kickr’s power and instead use its speed is asking to have Kickrs play by different rules. 

I imagine someone’s response to what I just wrote could be, “But I don’t have a power meter and Zwift is using my speed sensor.” True, but remember, even in this case, Zwift is not using the raw speed number from the speed sensor. It’s taking that raw speed number and performing a calculation based on the known-power curve of your trainer. It is this calculated power that determines speed, not the speed read at the speed sensor.

Another response could be, “But the Kickr’s power is determined by the slope calculation and if Zwift is using a different calculation then the power isn’t going to be right.” While it’s true that a different slope calculation would create a different resistance for the same grade, this does not in any way affect the power numbers from a Kickr in sim mode. You can achieve most any power by shifting your gears. That’s the way it’s intended to be used. 200 watts is still 200 watts whether or not the Kickr says the speed should be 20 mph and Zwift says it should be 18. 

“But it’s not just power that determines speed in Zwift,” you say. “It’s also determined by weight, CdA, and other factors. Compared to this other software, Zwift is doing this wrong.” Again, yes, weight, CdA are all factors, but Zwift needs to calculate this so that all riders are subject to the same calculations and not those of other apps. All that matters is that Zwift gets the power from your trainer and it applies the other factors fairly and equitably to all riders.

Of course there are all sorts of problems with getting accurate power from all the different trainers and power meters out there. That’s a real issue. Addionally, even if you can get all the differerent trainers to closely match as far as average power, there are inherent problems getting them to match linearly when sprinting, coming off a grade, etc.

Yes, there are issues. But any discussion that forgets that speed is a factor of power in Zwift and not a factor of how fast your wheel is turning, or what some other simulation says is correct is always going to be flawed at a fundamental level in my view. 


(Christian Wiedmann [X] 50) #16

Actually my use of “completely accurate” was just to illustrate the point that there is a definite relation between a smart trainer’s wheel speed and the speed in Zwift.

My main question is if the simulation can be accurate enough if the Kickr’s model is used. As you say, there are many variables already in the equation. In particular if you throw in the calibration issues with zPower (and smart trainers too), there is going to be a significant variance to the power input to Zwift’s model. The GIGO principle will apply here.

Therefore, given the significant benefits (in feel of the simulation) to using speed from the Kickr as an input instead of power, isn’t it worth the cost of having slightly less fidelity in the power to speed conversion? I believe that by tweaking CdA and/or weight inputs you should be able to tune the models to be fair even if not identical. I think this would be at least as fair as comparing zPower to Kickr to SRM to Stages.


(Christopher Pallotta) #17

Again, there is no direct correlation of a Kickr’s wheel speed to power. It’s about power and how the simulation applies that power. Higher wheel speed does not correlate to higher power either in the real world or in a simulated world. Higher power creates higher power.

There are definitely problems with the way a Kickr reports power and in cases, it’s true that that creates speed problems in Zwift. For example I’ve see power go up (as reported by an external meter) while and Kickr’s does not. But again, this is the fault of the power measurement of the Kickr. It has nothing to do with and shouldn’t have anything to do with the Kickr’s speed. 

There is no benefit to using a Kickr’s (estimated) speed to effect performance on Zwift when everyone else is being judged on power not wheel speed. That’s almost like suggesting that Zwift should use the gravity characteristics of the moon to calculate one person’s climbing speed while others will use the parameters of the earth. You can’t use the parameters for one environment in another environment.

Now, if you want to say there’s a problem with Kickr POWER and even a problem with the way Zwift is interpreting that power, you won’t get a disagreement from me. I believe improvement is necessary in that regard. But to say Zwift should disregard power and use some other speed caluculation rather than their own, that’s another story.