Trainer Accuracy

[travispape]

The question has come up recently about how accurate the trainer feature is and I thought I would post some data & analysis that I've done in the last few days. I'm working on a feature for being able to measure trainer coefficients if you have a DFPM installed. You would read in the SRM/PT/Q/Ergomo/Polar data and iBike3 would fit the trainer coefficients. I tried a test protocol that worked well for my cheap trainer that I got last winter.

Here is the trainer ride result:

Trainer Ride.jpg (130.58 KiB) Viewed 15032 times

If you have been around a while, you know I have a bunch of power meters installed on my bike, including a PT and a CinQo. Here is how they compare:

PMC Overview.png (99.12 KiB) Viewed 15028 times

Here is a zoom-in on a sprint-ish effort:

PMC Sprint.png (96.39 KiB) Viewed 15030 times

Here's the comparison numbers:

Code: Select all

------------------------------------------------------------------------
               POWER METERS COMPARED TO EQUATIONS OF MOTION
------------------------------------------------------------------------
           Ride File   Min Pow   Avg Pow   Max Pow			
           iBike Pro	     0.00	   120.58	   923.62
Quarq CinQo (iB DFPM	     0.00	   121.44	   927.75
               PT SL	    -1.00	   119.55	   930.22

------------------------------------------------------------------------
                                  R^2
------------------------------------------------------------------------
                 |iBike Pro|Quarq Cin|    PT SL
        iBike Pro|    1.000|    0.938|    0.984
Quarq CinQo (iB D|    0.938|    1.000|    0.962
            PT SL|    0.983|    0.960|    1.000

R^2 calculation is row w.r.t. column.

Keep in mind that this iBike data is after fitting the ride data to one of the DFPM files, so the iBike data agrees by definition. The point is that once your coefficients are right, the iBike can produce very accurate power data on the trainer even when you accelerate for a sprint.

By the way, I did have a few speed sensor glitches, but they didn't mess up the overall analysis.

Here is the coefficient fitting result:

Trainer DFPM Fit.png (97.17 KiB) Viewed 15034 times

This protocol seemed to work very nicely. What I did was rode at target speeds holding each one for 2 minutes. I increased the target speeds by 3 mi/h steps. On the way back down, I inserted a couple of sprint efforts to extend the curve toward the higher powers. Before doing this protocol, I only warmed the trainer for about 3 or 4 minutes, so I was glad to see that there wasn't a big hysterisis effect on the way back down in power due to warming the fluid.

This was done on a Perfomace TravelTrac Century V fluid trainer that I got off of Craig's list. Yes, the coefficients I got were substancially different than the Century model shown in the trainer list. That might have been a different HW revision--the one I have has "V" in the name. That data in the trainer list came from Kurt Kinetic.

Travis

[Tom_Anhalt]

Travis,
That's cool stuff...but as a long time Polar user, I'd highly recommend removing the Polar from the list of DFPMs one can use to "calibrate" the trainer power function. Due to various reasons, Polar performance on a trainer is "hit or miss" as to the accuracy and can be wildly "off" at times...just the nature of the beast (as implemented). Just a suggestion

[nreimche]

What is this DFPM I have been reading about? What does it do? Do I need it? Thanks.

[travispape]

What is this DFPM I have been reading about? What does it do? Do I need it? Thanks.

Good question. DFPM = direct force power meter. A DFPM is a power meter that more-directly measures the torque of the rider in order to measure power; in other words, every other power meter out there on the market.

No you don't need one. A few of us in Velocomp ride with them so that you don't have to. We are constantly making improvements to the technology the iBike uses (measurment of the resistances of motion) so that you don't have to buy an expensive mechanical device that has to be inserted somewhere in your bike's drive train. With each new generation of iBike, there is less and less that separates the performance of the iBike to the various DFPM models that are available and nothing can compare to the iBike in terms of weight and the flexibility you have in your choice of wheels, cranks, etc.

However, if you want to put a DFPM on your bike, the iAero will work together with it to measure your aerodynamic drag (CdA) live while you ride. This is a powerful tool for racers to use for optimizing their ride position in order to gain speed. The DFPM needs to adhere to the ANT+Sport wireless standard and the iAero unit will read it. It works with Quarq and the wireless SRM and will work with the new version of the wireless PowerTap that eventually will be released.

Getting back to the thread topic, measuring power on trainers is different for the iBike. On a trainer, you aren't moving and so measuring wind, slope, etc. doesn't measure power. Instead, the iBike uses the trainer's power vs. speed characteristic in order to calculate power. A DFPM can be used to measure that characteristic curve.

[travispape]

Travis,
That's cool stuff...but as a long time Polar user, I'd highly recommend removing the Polar from the list of DFPMs one can use to "calibrate" the trainer power function. Due to various reasons, Polar performance on a trainer is "hit or miss" as to the accuracy and can be wildly "off" at times...just the nature of the beast (as implemented). Just a suggestion

Good point, Tom. I know you are a Polar power expert. Does a Polar ever perform well on a trainer, or is it very unlikely? I'm wondering if I should disallow it or if I should just give a warning to the user that it might be futile. What I could do is make a judgement based on the goodness of the fit and throw up a message if there is a lot of scatter that it might not have worked.

Travis

[Russ]

I found in the power and or trainer forums awhile back (forget which) that most trainers have such light fly wheels that there are micro accelerations during the peddle stroke that do not exist on the road. This is because on the road, the higher weight/inertia smooths out the differences in torque as compared to the trainer were the speed actually fluctuates more.

I just posted a comment with this information on the fluid2 thread.

Since the Polar (if I remember correctly) uses a chain vibration sensor, I expect the variations explained in the first paragraph would drive the thing crazy in many cases. Perhaps the Kurt Kenetics Pro, I think it is, that has the large flywheel optionally attached would reduce that problem. I also should bring the perceived exertion on the trainer more inline with the road.
If someone has the big flywheel, would you care to comment on this?

Russ

[Tom_Anhalt]

Travis,
That's cool stuff...but as a long time Polar user, I'd highly recommend removing the Polar from the list of DFPMs one can use to "calibrate" the trainer power function. Due to various reasons, Polar performance on a trainer is "hit or miss" as to the accuracy and can be wildly "off" at times...just the nature of the beast (as implemented). Just a suggestion

Good point, Tom. I know you are a Polar power expert. Does a Polar ever perform well on a trainer, or is it very unlikely? I'm wondering if I should disallow it or if I should just give a warning to the user that it might be futile. What I could do is make a judgement based on the goodness of the fit and throw up a message if there is a lot of scatter that it might not have worked.

Travis

Well...it really "depends"...I've done constant power testing over a cogset on a friend's computrainer and it performed VERY well, but I've also seen incidences where I've been warming up on my Fluid2 trainer where the power reading was obviously "tweaked". My experience has been that when it goes "out" it really goes. One thing that doesn't help is if the chain tension is light since that tends to make the chain vibration signal "light" as well and makes it more likely that the inductive sensor circuitry locks onto other signals (such as the chain pins passing the sensor).

Yeah, some sort of "goodness of fit" determination probably would be a good idea. If you'd like, perhaps I could throw my bike on the trainer sometime and then toss you a couple of Polar files. Would that help any? I could follow the general protocol you outlined above. In fact, with my current setup, you could get accompanying CinQo and PT files for the same session too

[Tom_Anhalt]

I found in the power and or trainer forums awhile back (forget which) that most trainers have such light fly wheels that there are micro accelerations during the peddle stroke that do not exist on the road. This is because on the road, the higher weight/inertia smooths out the differences in torque as compared to the trainer were the speed actually fluctuates more.

I just posted a comment with this information on the fluid2 thread.

Since the Polar (if I remember correctly) uses a chain vibration sensor, I expect the variations explained in the first paragraph would drive the thing crazy in many cases. Perhaps the Kurt Kenetics Pro, I think it is, that has the large flywheel optionally attached would reduce that problem. I also should bring the perceived exertion on the trainer more inline with the road.
If someone has the big flywheel, would you care to comment on this?

Russ

Actually, those types of fluctuations wouldn't have an effect on the power reading...and actually might help in some instances, since it would tend to "pluck" the chain span and help increase the vibration amplitude.

It's my (unproven) theory that the problem with the Polar at times on a trainer isn't so much resonances or odd frequencies, but actually that the environment is "too smooth". Out on the road, the chain span is continuously receiving basically random inputs from the road surface. This keeps the chains span vibrating at it's natural frequency and with a good amplitude. The Polar sensor is just a big inductive sensor like a bass guitar pickup and it merely detects the chain moving closer and further away. However, the circuitry will also detect any other "density" changes passing it, such as the chain pins (which is actually how the chain speed sensor on the derailleur operates) passing over the module. The problem when the bike is clamped to a trainer is that under certain conditions, the chain vibration amplitude will be small and the circuitry will mistakenly lock onto the chain pin signal and affect the power reading. Anyway...that's my take on it.

I've always thought that a neat test would be to strap a white noise generator to a trainer and see if it solves things

[travispape]

If you'd like, perhaps I could throw my bike on the trainer sometime and then toss you a couple of Polar files. Would that help any? I could follow the general protocol you outlined above. In fact, with my current setup, you could get accompanying CinQo and PT files for the same session too

Definitely, go for it.

If there is anyone else out there with a trainer and DFPM, send me your files too. Just do something like the protocol above. Send me your trainer model info and I'll update the trainer list with your coefficients too. You can send the files to: tpape |a|t| velocomp _ com.

Thanks,

Travis

[turbomentor]

I have an older Cyclops Fluid (not Fluid2!) trainer that I'd like to work out the trainer coefficient for.

I did an FTP test the other night on the trainer with my friend's bike equipped with a PT. I just purchased my iAero this wknd.
If I know my Avg pwr and avg speed from the ride the other night (with a DFPM) can I come up with a good trainer curve for my iAero? Sorry, I'm a brand new ibike guy and just getting used to the idea of indirect power measurement and using all the tools that come with the ibike software.

[riderofrohan]

I was just wondering how many of you do a separate FTP test, or fitness test (awesome gen III feature!) to establish training zones for training on your indoor trainer. I have a Fluid 2 and the numbers match up pretty well, but it seems an indoor test would be worth it for super accurate numbers during indoor winter training, especially since I don't have a DFPM to get it really dialed in. Worth it for the long winter months?

[aytchkay]

I finally tried out my gen 3 Ipro with my old Cycleops Fluid...I used the profile with the lowest watts at 20 mph....and, it seemed to be failrly accurate. At 20 MPH I was putting out about 185 watts...which is in the range indicated.
My question is, was this accurate? In other words, if I pick a profile that indicates 20 mph at 250 watts, will the Ipro be setup to have me put out 250 watts to get to 20mph??
I'm in Chicago, with a nasty winter so far and have not had a chance to do my coast-downs and cal ride. Are there any other adjustments I should do while using this trainer?

thanks...

aytchkay
chicago wind climber

[turbomentor]

Travis:

I can send you a PT file from my trainer. If I do, can you develop the coefficient data for me so I can use my IB on the trainer? My trainer (an older Cyclops Fluid) is not supported under the list of trainers.

If you can/will, just tell me the ride protocol and I can easily make it happy for you!

Thanks

James

[EDIT] OOPS! Saw that you offered to do so above. Will send my ride file along.

[dbrouse]

Travis (and anyone else)

I noticed that a while back Travis did testing with a Performance brand trainer...my trainer is the "Travel Trac Century V Fluid +"

Which trainer profile do you think I should use?

[Velocomp]

I will "ping" Travis to see if he has a recommendation.

You may have noticed that 4.0.2 just had some new trainers added to its list on March 15...