Elite Turbo Muin 3.6 Sep 2020

[David N]

Hi folks

Was running v 3.5 with no huge issues with my Elite Turbo Muin

Installed 3.6 Sep 2020 build but speed (km/hr) is through the roof

Easy turning of the pedals and it's 250+ km/hr

I have wheel circumference set correctly

Wanted to plot my own virtual power curve which I guess I still technically can even if the speed is way off

Any ideas though on how to fix?

Thanks

[Eric Christoffersen]

In july I recoded the representations of all the trainers. Its possible I made a mistake with that one but I just checked and I believe it is numerically the same as before.

I was thinking perhaps your trainer was one that reported rpm instead of speed but it wasn't implemented that way before and isn't now either.

Searching this group for muin I see some people talk about 'special' wheel size that must be set.

https://groups.google.com/g/golden-cheetah-users/c/p0SgLyAHopA/m/yDuK19aMAgAJ

That thread suggests dividing wheel circumference by 13.2, but that wouldn't be enough to reduce 240kph to a casual speed.

Thinking though... since that is a wheel-off direct drive trainer it should probably have been coded to use rpm in the first place, that trainer doesn't know the wheel size so how could it report a speed?

You should be able to get a result you like today by playing with the wheel size you report during device definition. If you find a setting you like please open an issue and I will see about making it automatically correct for everyone.

[David N]

Thanks for the reply
Ya. I divide wheel circumference by 13.2 and it works perfectly on GC v3.5
Was messing around today on v3.6 with different circumferences but none made any difference.
I've tried all the different settings I can think of but can't figure this one out

[Eric Christoffersen]

Hi David,

I debugged with my bt40 kurt inride to see how this works. On the ant and bt40 controller types there is a setwheelrpm that is set from device data. The setwheelrpm method itself sets speed by calculating with wheelsize. I just tried this with my kurt inride and it seems to work like it should for me. I don't know the specifics of your device but I don't see why wheelsize isn't working for you to adjust speed.

Sorry I can't be more helpful. Is it possible you are using a different device connection type now? It is an ant device right? So should be simply reporting rpm which we then interpret as speed using wheel circumference.

Eric

[David N]

Thanks again Eric for your time and help

Yes, I'm using an ANT+ device

Tomorrow I'll reset the Misuro sensor in the turbo and see if that changes anything and I'll report back to this thread.

I have v 3.5 running on a desktop pc and 3.6 on a laptop.

I'll also uninstall 3.6 from the laptop and see what happens with v3.5

Thanks

David

[Ale Martinez (Please don't email or cc me)]

On v3.6 Sep 2020 there is a bug which duplicates speed readings from BTLE sensors: https://github.com/GoldenCheetah/GoldenCheetah/issues/3578, it is fixed in the last snapshot

But speed reading from ANT+ sensors has not changed, barring the introduction of simulated speed in slope mode, do you have it enabled in preferences? 

[David N]

Thanks Ale. I will check that out tomorrow also. 

[Eric Christoffersen]

I don't see a path where wheelsize won't be used on ant. Even with sim speed the wheelsize change should vary your speed. It seems like it is stuck so my guess is that there is something preventing your config from being loaded and wheelsize is falling back on the default of 2100mm.

Try moving your gc folder on that laptop so gc can't find it. Let gc build a new one, make a new athlete and see if wheelsize starts having an effect.

[David N]

Hi folks

It looks alright like this has something to do with athlete profile

If I navigate to Tool>Options in v3.6, the athlete profile folder location is a folder that doesn't exist on the laptop. I have no idea where the athlete library is getting stored.

I uninstalled 3.6 and reinstalled on the laptop. Rebooted. Created new profile again when prompted. But again, in Tools>Options it tells me athlete library is in a folder location C:/Program Files/Golden Cheetah Files. This folder does not exist on my laptop. 

So I then took a copy of the athlete library from v 3.5 on desktop PC and put it on the laptop. I pointed the athlete library in v3.6 to this folder and everything works OK.

But if I didn't have this backup from the PC I wouldn't be able to get 3.6 working correctly on the laptop

[Ale Martinez (Please don't email or cc me)]

El sábado, 24 de octubre de 2020, 5:57:46 (UTC-3), David N escribió:

Hi folks

It looks alright like this has something to do with athlete profile

If I navigate to Tool>Options in v3.6, the athlete profile folder location is a folder that doesn't exist on the laptop. I have no idea where the athlete library is getting stored.

It’s documented in the wiki: https://github.com/GoldenCheetah/GoldenCheetah/wiki/FAQ#data

 

I uninstalled 3.6 and reinstalled on the laptop. Rebooted. Created new profile again when prompted. But again, in Tools>Options it tells me athlete library is in a folder location C:/Program Files/Golden Cheetah Files. This folder does not exist on my laptop. 

It’s weird it would have been interesting to check which folder GC was effectively using (informed in About...)

[David N]

I will reinstall Ale later today and check this out and report back

[Ale Martinez (Please don't email or cc me)]

Beware to remove GoldenCheetah.ini from settings folder if you want to reproduce a clean install, it is not removed by uninstall script.

[David N]

Actual Athlere Library folder located at C:/Users/David/AppData/Local/GoldenCheetah/

[Ale Martinez (Please don't email or cc me)]

El sábado, 24 de octubre de 2020, 11:07:21 (UTC-3), David N escribió:

Actual Athlere Library folder located at C:/Users/David/AppData/Local/GoldenCheetah/

That's the default on Windows, it should work that way

[David N]

I went out to try set it up again on laptop with v3.6

I pointed Athlete Library path in Tools/Options to the  C:/Users/David/AppData/Local/GoldenCheetah/  folder

I then tried to add devices in the 'Train' tab. HR monitor, Turbo Muin Power and Turbo Muin Speed/Cadence recognised instantly and picked up

Finished out adding devices on Train tab but I wasn't brought to the 'Train' screen

Exited out after 30s or so and tried again. 

Was brought this time to 'Train' page but once again Power registering as 80,000 watts and speed at 150km/hr+

I don't know what more to do at this stage 

I checked the config folder in the Athlete Library. 'train-layout.xml' was updated at the same time as I added devices 

There is a TrainDB file which was last updated some minutes before that

[Eric Christoffersen]

I don't know anything about the disk format of user profile or how versioning is supported. But looks like your user profile on your laptop is hosed and that has been the key issue all along. Maybe something with write permissions, maybe some files were corrupted by a crash at the wrong time. I would move all the gc state on the laptop's filesystem somewhere safe and try creating a new profile on laptop. Then you can feed your activity history back in.

Thinking:

1) It is not ok that writes to profile don't succeed, stored values are not actually saved. Could we harden gc against bad profile writes by checking the values after read? Give an error if the write didn't work?

2) Some sort of profile checker on load? Have gc give error if something doesnt pass muster.

3) GC command to 'nuke profile from orbit'?

Anyhoo, when you do get this profile stuff sorted I would love to hear what values work by default for your device, I'd like to see it just work out of the box. I'm very curious how well the defined speed power curves match up with reality.

Eric

[David N]

Hi Eric

I'm not overly concerned about the activity history. The reason I was upgrading to 3.6 was because the very odd time I get little freezes in GC mid training session. This might have absolutely nothing at all to do with GC and could be a PC issue. It's not that big a problem but just thought the upgrade might help.

Regards the speed power curves..... I have a friend supposed to call soon with a PM so we can calibrate Muin/GC power. I have a Muin bought in Jan 2020 and I find the virtual power curve from the 2013 Muin you have on GC v3.5 seems pretty accurate. Certainly much more accurate than the power curve on Elites own software.

I'll get back to you as soon as I have that calibrated.

Thanks

David

[Ale Martinez (Please don't email or cc me)]

El sábado, 24 de octubre de 2020, 11:44:01 (UTC-3), David N escribió:

I went out to try set it up again on laptop with v3.6

I pointed Athlete Library path in Tools/Options to the  C:/Users/David/AppData/Local/GoldenCheetah/  folder

I then tried to add devices in the 'Train' tab. HR monitor, Turbo Muin Power and Turbo Muin Speed/Cadence recognised instantly and picked up

Finished out adding devices on Train tab but I wasn't brought to the 'Train' screen

Exited out after 30s or so and tried again. 

Was brought this time to 'Train' page but once again Power registering as 80,000 watts and speed at 150km/hr+

I don't know what more to do at this stage 

These settings should go to configglobal-trainmode.ini in that folder

 

I checked the config folder in the Athlete Library. 'train-layout.xml' was updated at the same time as I added devices 

This is the config file for Video Window Overlays

 

There is a TrainDB file which was last updated some minutes before that

This is the DB of workouts, media and video sync 

[Eric Christoffersen]

Great. At the least we should fix the wheelsize thing. There is a scale value sitting there just waiting to be used.

WRT short hangs and freezes, I've no idea. I do get lots of issues with signal interference on my trainers, now I have an 8' usb cord with the ant dongle on the end, that helped my experience hugely but it still happens. The other big help was turning off nearby devices especially ipads and smart tvs.

Is hard to believe that a manufacturer would screw up a power curve on a device they built, at least in pictures it looks like well engineered unit. Something I know from my own fluid trainer is that they might take 10-15 minutes of steady riding to warm up and settle. Would be interesting for me to see the difference in speed->power for a cold and warm trainer.

Good luck!

[David N]

The  configglobal-trainmode.ini  is not changing when I add the devices. At least the update time of that file isn't changing, so I assume the file content isn't changing. 

[David N]

Their power curve is ridiculous beyond belief!! Using Elite's own power curve my FTP if I remember correctly was 380watts :-) Reality is somewhat different, unfortunately!!! But I found the 2013 Muin virtual power curve inbuilt in GC to be about what I'd expect. I'm really interested myself to see next week how it compares to data from a PM

I have read on different forums that the trainer needs a 10-15 minute warm up. Elite themselves dismiss this assertion and I have to say in my experience I don't notice any difference in speed/power for a cold or warm trainer. I don't doubt that this is a real experience for some people and this trainer but it's not an issue I've ever encountered

[Eric Christoffersen]

380... woo. That is awesome, congrats. You are so humble, why are you selling yourself short? :)

Hey you should figure out your user profile damage and get the trainer working, report back and close the issue you opened if speed can be made reasonable.

[David N]

In the Athlete Library folder on the laptop with v 3.6 there are two  configglobal-trainmode.ini files

One is in C:/Users/David/AppData/Local/GoldenCheetah/  and the other inside  C:/Users/David/AppData/Local/GoldenCheetah/David

On the PC with v3.5 there is only one configglobal-trainmode.ini file. It's in the equivalent C:/Users/David/AppData/Local/GoldenCheetah/ on the PC

Are the two configglobal-trainmode.ini present in v 3.6, albeit in different folders, causing a conflict or error of some sort?

[David N]

Just checked the configglobal-trainmode.ini  in v 3.6  (the one located in  C:/Users/David/AppData/Local/GoldenCheetah/  )

The wheel size is not updating. It's set as 2100. So when I change it to 158mm during device setup in the Train tab it's not updating. Issue found! 

[Ale Martinez (Please don't email or cc me)]

It looks like a bug introduced when global config was separated from athlete config in v3.6, I will take a look.

[David N]

Apologies. The default DeviceWheel1 sets to 0, not 2100

I deleted the  configglobal-trainmode.ini file on v3.6 and created a new one. DeviceWheel1 set itself to 0

When I manually edited the file and set the DeviceWheel1 = 158 (for Turbo Muin) my speed came back to normal.

Thanks for the help folks

Very much appreciated

[Ale Martinez (Please don't email or cc me)]

El sábado, 24 de octubre de 2020, 16:54:10 (UTC-3), David N escribió:

Apologies. The default DeviceWheel1 sets to 0, not 2100

This is a bug in v3.6: when you set a custom wheel perimeter, DeviceWheel1 is set to 0 instead of the value you enter,

it is reproducible and it doesn't happen when you select one of the predefined rim+tire combinations.

It is not present in v3.5 and, I think, it was likely introduced in the last major change on add device wizard: https://github.com/GoldenCheetah/GoldenCheetah/pull/3410

[Eric Christoffersen]

Uh! It *is* my fault. Sorry! I've made a fix here:

https://github.com/GoldenCheetah/GoldenCheetah/pull/3643

[Eric Christoffersen]

I was thinking about this trainer's setup, is helping me reason about all the virtual trainers we have defined.

I don't think that it makes sense for anyone to need to ever set the wheelsize when using this trainer.

The wheelsize option really only makes sense for wheel-on trainers where the user can change the physical wheelsize, their tire size, etc. The bt trainers all report rpm and gc converts that to speed using the selected wheel circumference. Its an option that is necessary to support different user setups.

But in the case of this Muin trainer it is direct drive, there is no wheel so the trainer can't possibly care about speed, it has only rpm. I believe this trainer should be redefined to be an rpm-based trainer. If you find that the wheelsize number "158" works, that means we should change the definition to:

            "Power - ELITE TURBO MUIN 2013",

            // Power curve fit from data collected by Ray Maker at dcrainmaker.com

            PolyFitGenerator::GetPolyFit({ 0, 2.30615942, -0.28395558, 0.02099661 }, (158./1000.)*(60./1000.)),

            true

So... when you get a chance to compare power readings we can update this trainer's definition, could be the current polynomial is as good as it gets but we should still redefine the trainer to be rpm based.

I've opened an issue for this:

https://github.com/GoldenCheetah/GoldenCheetah/issues/3644

[David N]

Eric

I have 2 calibrations done this evening. Fairly significant differences between the PM and GC.

But a little caution needed here yet!!

I have a Muin purchased Jan 2020 and was using the virtual power curve from the 2013 Muin. 

I'm going to get the PM over again next week and have GC set up with the power curve from the 2015 Muin.

I tried changing the DeviceWheel diameter and while we could get the power equal between Muin and PM at the lower watts, when we hit 600+ the differences began to get pretty big.

When I get a chance tomorrow I will create a little excel file with to compare the data or alternatively I can send you on the 4 files from the 2 tests this evening. I'm not sure what use they'll be though and have a feeling I'll get much more accurate results from 2015 virtual Muin power curve when I test next week

Regards

David

[Eric Christoffersen]

Hi David,

From reviewing the equations for built in devices I was pretty skeptical of some. I think Instead of using a pulldown choice for your muin you should program your own curve using the custom virtual power feature. Literally enter a set of speed/power datapoints from your powermeter runs and configure (order, variance) to make a reasonable curve. Then run trainer and pm side by side to validate that custom curve. Once you've got it 100% perfect, or maybe close, share your datapoints with me and I'll translate the math and update the gc defintion for your device.

One thing I'm confused about, I see something called a 2020 Muin B+ that seems to be a smart trainer that reports power. If so that trainer should connect with ant+ and fec and not need virtual power to work. Can you tell me exactly what trainer you have?

Thanks!

Eric

[David N]

Ale

Where exactly is the GoldenCheetah.ini file. I can't find if in C:\Program Files\Golden Cheetah or C:\Users\David\AppData\Local\GoldenCheetah

Thanks

David

[David N]

Edit - Found a GoldenCheetah.ini file in C:\Users\David\AppData\Roaming\goldencheetah.org

[Ale Martinez (Please don't email or cc me)]

El viernes, 30 de octubre de 2020, 19:46:33 (UTC-3), David N escribió:

Ale

Where exactly is the GoldenCheetah.ini file. I can't find if in C:\Program Files\Golden Cheetah or C:\Users\David\AppData\Local\GoldenCheetah

https://github.com/GoldenCheetah/GoldenCheetah/wiki/FAQ-DATA#where-are-settings-located

> Read the Frequently Asked Questions before posting 

[thomas luckner]

I've the same problem with Trainer Elite Direto XR using Bluetooth connection. 

Speed > 250 km/h - I can enter in the  wheel circumference  what i want. It doesn't matter

Thanks 

David N schrieb am Freitag, 23. Oktober 2020 um 13:06:27 UTC+2:

Hi folks

Was running v 3.5 with no huge issues with my Elite Turbo Muin

Installed 3.6 Sep 2020 build but speed (km/hr) is through the roof

Easy turning of the pedals and it's 250+ km/hr

I have wheel circumference set correctly

Wanted to plot my own virtual power curve which I guess I still technically can even if the speed is way off

Any ideas though on how to fix?

Thanks

[Eric Christoffersen]

Hi Thomas,

This was due to an error in my ui code where wheelsize that was not specified by dropdown was ignored. It was fixed in the sources a week ago so you should find that it is fixed if you pick up a newer build.

I have been working on an improvement for folks that use the virtual power feature, it adds the ability to specify the trainers inertial mass so trainer can report power more accurately.

I'd really like feedback. A windows install with this change can be found here:

https://ci.appveyor.com/project/Joern-R/goldencheetah-knhd8/builds/36105184/artifacts

Let me know if you need help specifying a number for the inertial moment.

Thanks!

[stijn goossens]

Hello,

I have a Elite turbo Muin since today and wanted to give golden cheetah a try:

- on ubuntu I didn't get bluetooth to work and my ant+ stick didn't seem to recognize anything, will try a bit more later.

- switched to windows  using the GC 3.6 - sep 2020, At least I've got some numbers appearing in training mode, but they were extremely high (using a bluetooth dongle)

- discovered this thread, deleted everything from previous installs on windows and downloaded the release mentioned above ( https://ci.appveyor.com/project/Joern-R/goldencheetah-knhd8/builds/36105184/artifacts), unfortunately I still have wicked high values (see screenshot below).

(On my watch however, I receive perfectly normal power and cadence value.)

[stijn goossens]

(apparently you can't edit messages)
- I've got my bluetooth dongle to run on ubuntu as well using goldencheetah v3.5, got the same behaviour (extremely high powers).
Let me know if this issue is still linked with the discussion above or if I should create a ticket on github.

[Eric Christoffersen]

I'd love to help you get this working.

Look at your displayed speed. 146kph? Almost certainly the issue is that the 'wheelsize' is not set correctly. Above responses to this thread say you should be setting circumference to "actual wheel circumference / 13.2". I think david above said it worked well with circumference set at 158mm. That will reduce your speed and thus the derived power.

If you find this 158mm value works well, tell me and I can rejigger the device definition so it is set automatically.

With the artifact you pointed to above there is now also an inertia setting. I'd love to know what value you settle upon.

Good luck!

Eric

[stijn goossens]

Apology, seems like I misunderstood what was included in the specific build. I'll tinker some more over the weekend.

What I don't yet understand is why this power curve is needed at all, my watch can directly read out the power from the trainer without programming any curves. If useful, would capturing some specific bluetooth data help with reading power directly in golden cheetah instead of going through a curve?

thanks!

[Eric Christoffersen]

Hi Stijn,

How exactly does your watch do it? I've got no idea.

Either the trainer broadcasts power or the trainer broadcasts its 'speed' and the app determines power. We have the code for the kurt kinetic inride pod (another 'dumb' trainer that provides power) and it is the code we run that converts the received rpm to power. The user doesn't see it as 'virtual power' in the vendor's app because the app is hiding how it works.

I don't know exactly what trainer you have. Elite makes a lot of different models that all sound similar to me. They make a lovely FEC smart trainer, in which case you shouldn't use virtual power, you should be connecting as ble device. I know they do make a 'dumb' trainer that looks the same called 'muin', no idea what data it provides.

The virtual power feature is for trainers that can only provide speed.

Assuming your trainer doesn't broadcast power then I found a review of something that might be your trainer. The review claimed a 6kg flywheel with a diameter of 197mm. Thats big!

If the flywheel is a solid disc your I should be:

I = 1/2 m r^2

I = 1/2 6 0.1^2

I = 3/100

That is a reasonable value.

BUT: for you to determine a working value for I, you need to know how fast the flywheel is actually turning. Almost certainly the flywheel is turning lots faster than the rear cassette, looks like there's a big wheel attached to the cassette that drives the flywheel via a belt. Supposing the flywheel goes 'R' revolutions for a single turn of the rear cassette then the I you should use in the inertia field should be:

I = (3 R^2) / 100

So if your flywheel turns 10x for every turn of the cassette the right value might be:

I = (3 10 10) / 100

I = 300 / 100

I = 3

Maybe email Elite and ask them?

Eric

[Eric Christoffersen]

I just misspoke, got confused. Correction to my previous post:

The kurt inride pod actually broadcasts computed power over a private channel. This lets the device do a better job because it can super-sample its rpm data. That super-sampling is also how it can 'see' cadence.

[stijn goossens]

Thanks for all the information and support,

In case you are interested:

the sensor used inside the elite trainer that I have (and in many more elite trainers) is this one: https://www.elite-it.com/en/products/home-trainers/sensors/misuro-b#
You can configure it by pushing on the one button, to match it with - what should be the power-speed curve / inertia - inside the trainer.
So I believe this device has embedded what we try to replicate here in golden cheetah.

some more useful information: http://www.olympum.com/sports/the-curious-case-of-mr-elite-turbo-muin/

I'll see if there is an easy way to record some bluetooth packages.

[stijn goossens]

In case it's relevant for someone with more knowledge about bluetooth to obtain power messages directly from the trainer, I've recorded a few things using bluetoothctl on an ubuntu computer:

- bluetooth_services.txt: An overview of all the bluetooth services / characteristics from the device

- bluetooth_power.txt: an output example of the specific "Cycling Power Measurement" while pedalling on the bike (and standing still at the end)

I'll try the inertia and ratio value out later this weekend.

thanks!

[Eric Christoffersen]

Everything I read says to me that your trainer with that misuro sensor pod should be broadcasting power, and your attached logs seem to confirm that.

Golden cheetah already can accept power from the cyclingpower service:

        if (service->serviceUuid() == QBluetoothUuid(QBluetoothUuid::CyclingPower)) {

            has_power = true;

            service->discoverDetails();

        } else if (service->serviceUuid() == QBluetoothUuid(QBluetoothUuid::CyclingSpeedAndCadence)) {

Have you tried creating a bluetooth device for the muin? Once you create it, it should just connect and read power automagically.

The whole "virtual power" thing is for devices that can only report rpm or speed, its a way to derive power for devices that can't do it themselves. Your sensor should be able to do a better job than the virtual power feature.

[stijn goossens]

"

The whole "virtual power" thing is for devices that can only report rpm or speed, its a way to derive power for devices that can't do it themselves. Your sensor should be able to do a better job than the virtual power feature."

ok lol, selecting "none" made life indeed easier ... :).