Gradient accuracy on smart trainers

redvision

So, I have had the Wahoo kickr, kickr snap and tacx neo, all of which I ended up returning.
I currently have an Elite Direto which so far is brilliant, although noisier than I thought it would be.

Anyway, with each trainer I have used zwift, bkool sim, trainerroad, and also routes recorded on my gps - mainly focusing on hilly routes. And this has raised a question, just how accurate are the gradients?

I have been using the same training bike and power meter (have also used the device power meter) yet the gradients simulated on the turbo (all of them had similar results) seem to be different to outdoor. Basically i have smashed all my outdoor pb's despite using less power. I know there are other factors to take in to account, eg weather, road surface etc, but the time difference are too big for these to be the culprits.

Out of curiosity I asked Wahoo, tacx and elite the question - tacx didn't respond - Wahoo and elite said it is a common observation but the gradients are set from data collection, extensive research and testing and using clever software. Elite pointed me in the direction of the YouTube video of the drivo power calibration.

I am now starting to wonder if so much emphasis is on getting the inbuilt power accurate and the ride feel as realistic as possible, are the gradients/resistance being neglected?

N0bodyOfTheGoat

I don't know about other apps, but I've read that by default, Zwift hills are taken at 50% of their actual gradient. I've also read the max gradient at 100% simulation in Zwift is 17%, less than the peak gradient on one of my real hill rep climbs (Dell Road SO18).

redvision

Yeah I know, and I have always set it to the 100% gradient setting.

What I'm really trying to figure out though is why, when you ride a course that you've uploaded from your gps, is the gradient seemingly different on the turbo to outside? Is it because of the magnetic resistance? (I'm told this is largely still based on the same resistance 'steps' as a traditional stationary exercise bike)

mrb123

I don't have a definitive answer, but a few thoughts...

The turbo may not factor in rolling resistance, particularly significant if your climb in the real world is on a poor road surface.

You've presumably entered your weight into your app, but is the weight of your bike, water bottles, clothing etc accounted for somewhere.

Lack of wind resistance will also be a factor.

FWIW my times on my Direto have so far been pretty much in line with real world times for stuff like Sa Calobra. On steeper gradients (8 or 9% and above) it probably feels tougher on the turbo but hard to know how the speed compares. I may of course be using a different app to you.

redvision

MrB123 wrote:

I don't have a definitive answer, but a few thoughts...

The turbo may not factor in rolling resistance, particularly significant if your climb in the real world is on a poor road surface.

You've presumably entered your weight into your app, but is the weight of your bike, water bottles, clothing etc accounted for somewhere.

Lack of wind resistance will also be a factor.

FWIW my times on my Direto have so far been pretty much in line with real world times for stuff like Sa Calobra. On steeper gradients (8 or 9% and above) it probably feels tougher on the turbo but hard to know how the speed compares. I may of course be using a different app to you.

All good points, but would they add up to several minutes difference, at the same watt output, even over climbs as short as 10 mins long?

I figure it's either a program thing, or something to do with how the magnetic resistance is applied.

I guess I'm probably thinking too much in to it, but when you buy a product you expect it to deliver its spec/usp (such as the realistic gradient up to **%) and it does appear that the gradient perhaps isn't as accurate as the high end turbo manufacturer's claim.

mrb123

Are you having this issue with video routes on your apps or just routes that you've ridden in real life and then repeated on your trainer?

I'll say again that my Direto appears to be absolutely spot on with the video routes I've tried.

If yours is so far out perhaps there's a problem with your unit or something in the settings that needs adjusting.

redvision

MrB123 wrote:

Are you having this issue with video routes on your apps or just routes that you've ridden in real life and then repeated on your trainer?

I'll say again that my Direto appears to be absolutely spot on with the video routes I've tried.

If yours is so far out perhaps there's a problem with your unit or something in the settings that needs adjusting.

No, only routes repeated on trainer.
It was exactly the same on the kickr, neo, snap, and now direto. Basically for the same watt output I'm smashing my outdoor pb.

Like I say, Shane Miller has made reference to the difference but doesn't see it as a problem. I guess it's not, its just a bit frustrating if you want to train inside repeating a 10% hill and assume the resistance is accurate, whereas in reality it isn't.

sungod

given system mass (you+bike etc.), and the distance, time, and average power, over a known vertical ascent, you can test the trainer (and if you like, calculate back to check the 'gradient' that the trainer simulated)

mass * ascent * 9.81 = energy = power * time

for instance, raising 80kg through 100m requires 78,480kj

the power * time reported by the trainer should roughly match that (a few % will go in drivetrain losses), if you see a big disparity then you know something is wrong

dabber

My own experiences for what it's worth using a Bkool Pro trainer + plus BigRingVR rides and application....

I've used several of my own (French) rides on BRVR - using mp4 video filmed on GoPro+Garmin recorded tcx file.
On the virtual rides I've created a few segments on Strava and if I compare my real world times to my VR rides the times are pretty realistic and consistent to my RL experience, in addition, when riding them they feel "right". On some segments I'm slightly faster in VR, other I'm slightly faster in RL.

When riding BRVR's own rides I can only make a couple of comparison to RL as the BRVR rides are generally fairly classic big name rides. However, if I compare my RL ride on Luz Ardiden, my ride on BRVR is about 15 minutes faster but, in RL although i wasn't trying to take it too easy I was enjoying the fantastic experience and making sure I made it to the top without incident. On the BRVR, sitting in the comfort of my house I was trying.

A similair situation on the Tourmalet although I can't make so direct a comparison as in RL I took the Voie Laurent Fignon diversion rather than the current main road route... but again, I was trying hard on BRVR but in RL I was definitely making sure I got to the top

With both the Luz Ardiden and Tourmalet Vr the gradient/resistance experience was totally consistent with how it felt in RL. I can't make power output comparisons as I don't have a power meter.

Machoman121

I haven't factored in water bottles with water, backpack, spare tubes, pump, sunglasses, helmet, car keys etc.....all these will be at least 2 KG extra......maybe I need to add these into the bike weight too. Which could explain why some people find indoor easier than their outdoor rides?.......

redvision

sungod wrote:

given system mass (you+bike etc.), and the distance, time, and average power, over a known vertical ascent, you can test the trainer (and if you like, calculate back to check the 'gradient' that the trainer simulated)

mass * ascent * 9.81 = energy = power * time

for instance, raising 80kg through 100m requires 78,480kj

the power * time reported by the trainer should roughly match that (a few % will go in drivetrain losses), if you see a big disparity then you know something is wrong

Thanks for this info. Will use it over the next few training sessions to check what's going on.

Machoman121 wrote:

I haven't factored in water bottles with water, backpack, spare tubes, pump, sunglasses, helmet, car keys etc.....all these will be at least 2 KG extra......maybe I need to add these into the bike weight too. Which could explain why some people find indoor easier than their outdoor rides?.......

Good points as well. I had thought about the weight of the bike but not the additional weight of the helmet and water bottles. I don't think they would equate to more than an extra kg but would still make a bit of a difference.

sungod

redvision wrote:

Thanks for this info. Will use it over the next few training sessions to check what's going on.

i should throw in the caveat that it needs to be 'steep' enough that speed is below, say, 20 kph, otherwise the trainer will (should) ramp the resistance to simulate aerodynamic drag

Machoman121

While it's always harder on the kickr....there's no coasting....i always thought I seem to be cycling faster with higher speeds compared to real life. Perhaps there's no air resistance and tyre resistance on the Kickr. So if i add the extra weights of the stuff i normally have on and carry, maybe it'll bring the performance back to as close as possible to real life.

ilja.booij

Smart trainers will factor in air resistance. Part of the ANT+ FE-C protocol allows for an application to communicate wind resistance coefficients, wind speed and drafting factor. The trainer will calculate the applied resistance from these factors, combined with rider speed of course and the resistance needed for the specific grade of the road.

hojo

It is true that the ANT+ FE-C (simulation mode) has the ability to transfer several important parameters to the trainer to emulate slopes in a realistic fashion.

What I assume is the problem is what calculations the microcontroller(processor) in the trainer is capable/allowed to do and the general tolerance and calibration possibilities of the parts that provide the actual resistive force. Could be that the math is simplified (e.g. by piece-wise linearisation of non-linear phenomena) to make it possible to execute on the limited processor/memory.

The realisticity is not easy to measure. Have done some attempts on my old genius smart. And the result was mixed. Some usecases are dead on and others are +26% off. Have spent considerable time on getting power meter pedals to match the power reported from the trainer by calibration and then aligning it to the power meter with the roller-tension knob. But when comparing these values to values from the basic suggested equations (ANT+ FE-C Profile Specification (4.2) simulation mode) I get these mixed results.

Anyone else out there who has tried to do some measurements on realisticity???

And yes I'm waiting for answers from support...

FatTed

On BRVR & RL I have ridden the Telegraph, Galibier, Alpe D'Huez and Stelvio, the Gradients are realistic my times in VR are faster.

58585

What do you mean by "gradient" if you are riding indoors on a trainer?
Not being negative, but riding a bike clamped in a fixed trainer indoors is never going to be the same as outdoors riding when you are up and out the saddle, moving the bike around. You also have the effect of road surface, weather, traffic, braking into corners, effect of your position on the bike etc to consider if you are trying to "compare" times.

frisbee

58585 wrote:

What do you mean by "gradient" if you are riding indoors on a trainer?
Not being negative, but riding a bike clamped in a fixed trainer indoors is never going to be the same as outdoors riding when you are up and out the saddle, moving the bike around. You also have the effect of road surface, weather, traffic, braking into corners, effect of your position on the bike etc to consider if you are trying to "compare" times.

The trainer simulates the resistance provided by the gradient. You can dream up all the factors in the world but in my experience indoor and outdoor factors generally balance out and my trainer provides a pretty good simulation of an outdoor route. The tough 20-25% sections are out of the saddle struggling to keep the lowest gear turning.

PinarelloPrince

To the OP. I’m looking to buy a Smart trainer. If you have the time and inclination could you tell me why you returned the trainers?

kingstongraham

NitrousOxide wrote:

I don't know about other apps, but I've read that by default, Zwift hills are taken at 50% of their actual gradient. I've also read the max gradient at 100% simulation in Zwift is 17%, less than the peak gradient on one of my real hill rep climbs (Dell Road SO18).

This is only partially true in Zwift. The default setting is that the additional resistance of the trainer is set to 50% of the true gradient, but the speed at which you go up the hill is based on your power, and the true gradient.

You should go up at the speed that your power suggests. It seems pretty close to me - a bit faster because of no slowing in corners, no variable wind resistance etc.

fixedgeartester

One thing that most people don’t take into consideration fully is weight. Smart trainers running with Zwift ask for your body weight to be entered. Let’s say for example, you are 75kg (I wish!) and your bike with drink bottles and saddle pack weighs 9kg. If you don’t include the weight of your bike this represents a 12% inaccuracy. This is enough to dramatically ‘improve’ your performance on the trainer. If you add to this rolling resistance, wind resistance etc, then this will only add to the apparent improvement in performance. Out on the road you have to haul you, your bike, your clothing and shoes etc up any incline. None of this is factored in to a VR programme like Zwift.
I think my point is, don’t worry about it too much, just enjoy Zwift for what it is, and it is bloody good fun.
Graham

kingstongraham

Zwift's calculations do include the fact you are riding a bike. That weight is included in their estimations. They also make assumptions about how aero you are, and the resistance changes with the different bike and wheelset you select. Also, the different road surfaces have different resistance assumptions.

hojo

There are a handful of control message communcation pages that the VR program sends to the trainer over ANT+ or any of the BLE flavors out there. Each control message contains a subset of the relevant parameters needed to calculate the "right" resistance that the trainer should provide. If data pages are not sent or values in them is not set correctly default value(s) will be used.

Most VR programs do not expose the user to all of the possible settings. Instead things are automatically set (or "hidden") not to "invite" users to get lost in the settings.

And yes one can achieve astounding performance results by tweaking these parameters...

But I would prefer to get reality-like behavior when I put the correct values to these parameters and not being forced to investigate which in-correct parameter value one should use to get reality-like behavior.

imafatman

Confused as to why this matters on an indoor turbo trainer. Turn your legs to achieve the wattage required of your training plan. The end. If you are riding an indoor trainer trying to simulate your real world outdoor performance then I think you are slightly confused and doing it wrong. The app can't simulate you in real life with real life conditions. Trying to achieve that would appear to be pointless.

hojo

So you mean that the simulation modes available in e.g. ANT+ FE-C, BLE FTMS, and the others are of no use? I disagree.

imafatman

hojo wrote:

So you mean that the simulation modes available in e.g. ANT+ FE-C, BLE FTMS, and the others are of no use? I disagree.

I didn't say simulation has no use. Those are just communication protocols and are very useful for transferring information between devices.

If you follow the thread, the op is wondering why he's rolling faster on a trainer compared to outdoor.

Speed on a trainer is just a virtual representation. It's estimating how fast you should be going with some fixed drag/weight/size figures. It's not accurate because a) it doesn't really need to be and b) it's very difficult to accurately model everything.

If you train with power then speed is irrelevant.

rdt

There are likely to be a number of factors at work, most touched on already, which can all add up....

Regarding weight, I imagine most people will materially under-state this because of (a) optimism and (b) not fully taking account of clothing, shoes, seat bag, pump, food and liquids, which can all add up to rather a lot on top of (optimistic) bike + body weights.

I'd guess that drag also gets under-stated by not fully accounting for blustery and varying conditions. And it's an advantage on the trainer not having to think about your immediate environment (road positioning, other traffic, etc) - you can just attack the workout without reservation or distraction.

But I reckon a large factor must be the inability to fully simulate or take account of all the "micro-variations" in gradient and road surface that you experience on a bike, which entail the rider repeatedly making tiny micro bursts of additional effort in order to push through them, accumulating additional fatigue compared to the smoother load presented by a trainer attempting to simulating the same climb.

Climbing outside, we may try to steer a line seeking out the smoothest possible bit of surface and least variations in gradient because we know intuitively that not doing so requires additional effort that all adds up. In my experience, the steeper the climb the more noticeable these gradient and surface micro-variations become, and therefore the less truly representative the trainer simulation becomes, even though it is nominally reproducing the hill gradient experienced.

hojo

imafatman wrote:

hojo wrote:

So you mean that the simulation modes available in e.g. ANT+ FE-C, BLE FTMS, and the others are of no use? I disagree.

I didn't say simulation has no use. Those are just communication protocols and are very useful for transferring information between devices.

If you follow the thread, the op is wondering why he's rolling faster on a trainer compared to outdoor.

Speed on a trainer is just a virtual representation. It's estimating how fast you should be going with some fixed drag/weight/size figures. It's not accurate because a) it doesn't really need to be and b) it's very difficult to accurately model everything.

If you train with power then speed is irrelevant.

I am aware of the original question/statement.

Simulation mode is not just a communication protocol. It also suggests a set of equations. It is the most basic ones for the bicycle power and speed relation.

It appears to me that we are discussing two kinds of simulation errors at once.
The one that I think we should be addressing _today_ is the FAILURE of current trainers and apps to get even the most basic physics resonably right and utilize the possibilities in the simulation modes.
Then in the future I hope that one can go on to discuss and refine these protocols/models/equations.

So yes it is impossible to get it perfectly right but that should not stop us from getting it less wrong!!

Speed is accurate on my trainers! But yes I am aware that some trainers/apps (occasionaly) recalculate power into virtual speeds when appropriate.

And finally: NO speed is not irrellevant when you train with power. Speed is related to inertia and inertia matters a lot when looking at pedalling dynamics (but inertia modelling is not always correct either in trainers). Yes it is possible to get the same power number with different speeds, but you don't train the same thing when using different speeds.