Trainer Calibration in Version 3.5

[Mark McCain]

I could not get my Magnus to connect to GC 3.4 no matter what I tried.  I down loaded the latest developer's build and it connected to everything perfectly.  GC is showing the target, the speed from the trainer, and the power and cadence from the Quarq.  I was very surprised and happy.

I warmed everything up and then ran the calibration.  That failed several times before finally calibrating successfully.  The numbers reported are similar to the numbers Rouvy reports for my wife.  Happy again.

I then ran a canned program.  The GC target was 70 watts.  The trainer was controlling at 125 watts consistently - Is shifted up and down several times and each time the trainer adjusted back to about 125 watts.  The program jumped to 117 watts and the trainer still held the resistance at 125 watts.  The program jumped to 175 watts and the trainer held 210 watts. I shifted several times and the trainer adjusted back to 210 watts.  It then jumped to 235 watts and the trainer held 272 watts.

Since I'm running a developer's build I'm not real concerned, but is GC controlling the trainer off the Quarq or off what the power the trainer is reporting?  What would be the best way to force GC to control the trainer resistance to match the target?  Or should I just manually adjust the program difficulty and wait for the final version of 3.5?

[stsc...@gmail.com]

GC can not currently use your power meter measurement to adjust the load on the trainer to match the target. It tells the trainer to set the load to the programmed value and that is it - the error you are seeing is entirely due to the (in)accuracy of the trainer in hitting that power target.

Unfortunately, as you noticed from your tests, the error is usually 1) non linear and 2) has a significant temperature/time-dependent or entirely unobservable (think tire pressure) components. So making the trainer resistance match the power target based on power meter measurements is actually a complex control theory problem that so far no one has written the code for GC to solve. If you're looking for a project, a good starting term is "PID controller" ;)

[Mark McCain]

I find this explanation perplexing.  I'm not disputing it because it is obviously written by someone that knows smart trainers and control software much better than me, but consider this:

My wife uses the same tire at the same pressure on the same smart trainer.  She has a different bike, but the same hardware otherwise.  She is running the manufacturer's software to control the trainer (Which I seriously dislike).  Because of the limitations of the software it can not see the C1 on her bike, only the trainer and her heart rate monitor are connected.

After a 10 minute easy pedal that program pauses and does a spin down calibration.  Just like GC 3.5 is doing, only it does it automatically.  When it is complete the power being displayed on the computer screen is within a few watts of what is showing on a 920TX coming from a C1 power meter.  There are no variations and control is a little sloppy but actually pretty close.  The trainer is tracking to the C1.  Her power range is much smaller than mine, but between 70 watts and 200 watts there is no drift. 

If I put my bike on the trainer and run her program (Rouvy, BTW) it will track to my Quarq Riken perfectly all the way up to 550 watts.  At the very top of my power limits the trainer is 10 watts low to the Riken.  Again, the actual control is a little sloppy compared to GC 3.5, but it is pretty close.

Again, I'm not disputing what has been explained, but the engineer in me is having problems resolving the conclusions.

[stsc...@gmail.com]

I was only going from "is GC controlling the trainer off the Quarq or off what the power the trainer is reporting". That is definitely not happening - there is no feedback loop, GC only tells the trainer to set the resistance based on the workout selected. And from your description, Rouvy should work the same way.

I can believe that you can get somewhat close when doing a calibration after 10 minutes warmup. The calibration support in GC is recent, so it might be something wrong with that. Can you verify that you are really reading the values from the Riken? Lots of trainers simultaneously broadcast power.

Beyond that, to diagnose a problem with GC, it would be helpful if you could provide the antlog.raw from your athlete directory. This has all the ANT+ communication. It can be found in C:\Users\stefan\AppData\Local\GoldenCheetah\Test (when your Windows username is "stefan" and your athlete is named Test).

[Mark McCain]

I will find and upload the file tonight when I get home.  I am sure that Rouvy is not looking at the on bike power meter.  It clearly shows what is connected and it only shows the Magnus and the HRM.  As for GC it is displaying the power reported from the Riken because it is matching the power displayed on my Garmin, and the Garmin has never been given the chance to pair to the trainer.  In the devices display it is showing the power and cadence coming from the ANT+ID that is shown on my Garmin, and for speed and FE-C is showing a different ANT+ID, so I'm assuming that is the Magnus, since I don't have a wheel speed sensor on my training wheel.

As for the spin down calibration, the first time I did it I had warmed everything up for about 10 minutes (it took a while to figure out what the new button on the control panel did).  That took several tries to complete - Like a dozen or more.  It will go into calibrate mode and show the speed, but when it tells me to let the speed decrease to zero it freezes at some speed value and locks up, even though the wheel stops.  After several tries it finally worked and gave a trainer power calibration reading in the 64,000 range.  Last night I soft pedaled for about a minute and then ran the calibration, so not really warmed up.  The trainer calibrated on the 3rd try with a similar 64,000 reading.

[Mark McCain]

Attached.  The file is dated the 5th but my last ride was last night the 7th.  The clock on my computer is correct so I'm not sure what is up.  There are 3 files, this one and 2 from November of last year.

Thanks For Looking

On Thursday, February 8, 2018 at 2:28:37 PM UTC-5, Mark McCain wrote: