Paddle erg and high stroke rates

[wyattwww]

Hi everyone,

I have a paddle erg (KayakPro) and am trying to retrofit the openrowingmonitor as a digital console. I've wired up a reed sensor to pick on a magnet attached to the flywheel axle and that is working well. But I am having trouble figuring out the rower specific settings - I can't get an accurate stroke rate. It's possible and common to crank out high stroke rates on paddle ergs (90+SPM) but the openrowingmonitor doesn't output any higher than 38-40 SPM regardless of the settings.

I have attached an example raw CSV, can anyone please help me with the settings?

PS: I am the creator of Swelldone, an alternative to EXR and other virtual paddling/rowing apps. Happy to support this community, please let me know if I can help!

2023-08-03_19-54-45_raw.csv

Thanks!
Wyatt

[Abasz]

Hi, That is awesome. Actually I've been working on a kayak profile for the Kayak First ergo for month now. The biggest challange is the number of datapoints per stroke that is significantly less than for rowing. This means that the data is even more dependent on data accuracy (i.e. on magnet placement). Actually I was not able to get anywhere near good data if I used one magnet. I needed to install atleast 3 to have sufficient data for good stroke detection. Also the minimum stroke and recovery time needs to be set to a very low number. I use 3ms.

I had a look at the raw data, you can see that you generally have 4 points only for the recovery. Nevertheless the data seems good (i.e. its clear when the change in acceleration/deceleration happens) so in theory this sholud work.

Can you please post your settings?

[Abasz]

Just a side note, my reference to good stroke detection is the advanced stroke detection not the simple acceleration/deceleration one. That should work well with few data points. I wanted force curves and accurate torque based stroke detection and to get that I needed more datapoints.

[wyattwww]

Thanks for the suggestion. I just tried to add more magnets (up to 4) and it looks like that actually produces more erratic measurements - likely because I have the magnets attached to the axle/spindle and it's causing the reed switch to stay on with the magnets passing quickly when the flywheel is spinning fast.

I'll try to better analyze the data with one magnet and post back some working settings, stay tuned.

Thanks!

[Abasz]

If you post your settings I can test the few strokes you provided in the raw data file as ORM has the ability to simlate previously recorded activities.

On the erratic data, if you increase the number of magnets you need to be aware that magnet placement becomes very important (actually for the Kayak First ergo I had to design and 3d print holders and staff to achive acceptable results).

Also reed switches have a rating for how long a cycle (open and close takes). For instance the once I encountered were reated for 400-500hz that means they are capable of handling singlas with 2ms. But they also debounce and have more noise compared to hal effect sensors. Any way on the discussions here there are plenty of resources on this topic.

Just for reference. I am pretty sure this should be doable. Below is the chart of a 2k paddle on the Kajak First. the columns that go to near zero on the chart are the detected strokes. This is an early test with tweaking I was able to get near perfect assuming stroke style is consistent.

Of course if you look closer there are missed strokes as well as ghost strokes but overall consistent.

The thing I really recommend is to use mainly the acceleration/deceleration (i.e. slope based) stroke detection method as with one magnet that is the only reliable one. Also do not increase the flank length above 3-4 as you only have 4 datapoints per recovery that really not much. I would expect ORM to struggle with drag factor calculation with so few datapoints, but I know that there are water rowers that produce very little amount of data points and its know to work.

[wyattwww]

Good call on the reed switch rating, I will check and get appropriate ones. Here are my current settings (mostly guesses) for KayakPro. I took the dragFactor from the KayakPro's console downspin calibration.

KayakPro: {
numOfImpulsesPerRevolution: 1,
dragFactor: 47
minimumTimeBetweenImpulses: 0.001,
maximumTimeBetweenImpulses: 0.15,
smoothing: 1,
maximumDownwardChange: 0.9,
maximumUpwardChange: 1.75,
flankLength: 3,
numberOfErrorsAllowed: 0,
naturalDeceleration: -2.45,
minimumDriveTime: 0.002,
minimumRecoveryTime: 0.002,
flywheelInertia: 0.155,
magicConstant: 2.8
},

CSV file attached. 2023-08-04_14-11-02_raw.csv

Image of the physical setup:

Thank you!

[wyattwww]

Currently using these reed switches: https://www.amazon.ca/dp/B07VRBN1M9?ref=ppx_yo2ov_dt_b_product_details&th=1

Maximum power 10W
Maximum switching voltage 100VDC
Working temperature: -20 degrees to 80 degrees
High voltage 250VDC
Maximum switching current 0.5A
Maximum load current 1.0A
Maximum contact resistance 100M
Minimum insulation resistance 1010
Action time 1.0ms
Bounce time 0.6ms
Release Time 0.4ms
Response frequency 4000Hz
Maximum operating frequency 500Hz
Sensitive Value 10-15AT
Minimum reset Min Value 4FT
Maximum contact capacitance 0.7PF
Electrical life (load) 108 (5VDC, 10mA) 105 (24VDC, 400mA)
Test line circle number / Resistance 870/5000Ω

[Abasz]

KayakPro: {
numOfImpulsesPerRevolution: 1,
dragFactor: 47
minimumTimeBetweenImpulses: 0.001,
maximumTimeBetweenImpulses: 0.15,
smoothing: 1,
maximumDownwardChange: 0.9,
maximumUpwardChange: 1.75,
flankLength: 3,
numberOfErrorsAllowed: 0,
naturalDeceleration: -2.45,
minimumDriveTime: 0.002,
minimumRecoveryTime: 0.002,
flywheelInertia: 0.155,
magicConstant: 2.8
},

Please note that you are missing a comma at the end of the dragFactor line! So this config is invalid.

Also, I dont think naturalDeceleration or numberOfErrorsAllowed (as well as maximumDownwardChange, or maximumUpwardChange) is used anymore. Are you using the v1beta or better yet v1beta_update? I recommend using @JaapvanEkris's upgraded version as that is much better at handling errors and noise and low number of impulses.

Have a look at this how to install: https://github.com/JaapvanEkris/openrowingmonitor/blob/v1beta_updates/docs/installation.md
Then how to upgrade: JaapvanEkris#17

With the updated version I am able to produce 28 strokes even with your config that, is defaulted in many cases for the stroke detection algorithm, with your test data. If you provide me with a longer session I can test that too.

Actually, the updated version of ORM is more forgiving with more magnets. So if you have been using the original version, I recommend testing 3 magnets with the new algorithm.

[wyattwww]

Thanks @Abasz, I will try the alternate versions and let you know. Will record a longer file too.

[wyattwww]

Great success, thanks for your pointers @Abasz!

I'm now using the JaapvanEkris fork with v1beta_updates. I also upgraded the hardware to use a hall sensor with two magnets, much better readings and no bouncing.

Here are the settings for KayakPro that produced very close results to the physical console monitor:

KayakPro: {
numOfImpulsesPerRevolution: 2,
dragFactor: 1025,
minimumTimeBetweenImpulses: 0.00005,
maximumTimeBetweenImpulses: 0.2,
smoothing: 1,
autoAdjustRecoverySlope: true,
autoAdjustRecoverySlopeMargin: 0.035,
flankLength: 3,
minimumDriveTime: 0.2,
minimumRecoveryTime: 0.3,
sprocketRadius: 0.3,
flywheelInertia: 0.355,
magicConstant: 2.5,
autoAdjustDragFactor: false,
}

Attached a raw CSV file too if you're interested: kayakpro.csv.

Next, I'm going to try your ESPRowingMonitor... I'll need to implement BLE FTMS though. :)

[Abasz]

Great to hear! A side not on the ESP rowing monitor: I did not implement FTMS as my garmin watch does not support it so I had no real use case for it :) But I would be happy if you would do it. It would be a good test of the modularity and the extentionablility of the code :) If you have any question please let me know, on the Q&A of that repo.

One more question: Did you try autoAdjustDragFactor? I mean, 1025 as drag factor for an air flywheel seems extreme, while the flywheel inertia seems rather on the low end, considering the size of the KayakPro. Have you actually measured this inertia number of just guessed?

[wyattwww]

The numbers are guesses by trial and error and I converged on this set. I will probably refine further as I do more testing. I will try autoAdjustDragFactor on future iterations. I agree the dragFactor seems high, but then I saw that Sportstech_WRX700 had a dragFactor of 32000... I'll update again if I confidently find better settings.

I will definitely check in with you over at the ESP repo when I get to it. Thanks!

[Abasz]

Sportstech_WRX700

That is a water rower so you can imagine the difference of drag between water and air :)

For reference: Concept 2 air rower tops out around 250 max. KayakFirst kayak machine is around 160 at the top.
Inertia for Concept 2 is, I think 0.1016 or something around that
Inertia for KayakFirst machine is 0.02527962 (I extracted this figure from their android application)

But trial and error is mostly fine. There is a post how to measure inertia, but you would need to take the flywheel out i.e. take the machine partial to parts.