Oartec Slider profile that seems to work

[tigrrrlily]

!!! caution, while this works flawlessly in the UX sense, now looking at individual data points once magnified out there's definitely some screwiness with points clustering in groups of three that is masked somewhat by the chart's line drawing algorithm.

Been using these or very similar settings for a while now, with a Seeed eTerminal CM4 with a built-in touchscreen that is currently running Bookworm. The raw data readout looks rock solid, I couldn't have drawn nicer spikes myself. This machine was clearly well constructed in terms of magnet placement. The original head unit supplies 5V to the sensor on one wire, the other two are ground and signal. The internal pull-up from the Pi results in beautiful square pulses to ground (checked with an oscilloscope, with me yanking on the handle the pulse width went down to about one millisecond). The moment of inertia of the flywheel was measured directly with the wheel removed from the machine. I could have made a miscalculation, but the data passes a basic sanity check.

Regarding physical integration, I am powering the unit with PoE going to a splitter hidden in the head unit holding arm. Due to the unique nature of the machine, I had to use a drag chain to run the Ethernet cable in to connect the floating part of the machine to the base. Sized right, it just stays out of harm's way when the machine is stowed upright on its end. 5V and ground are split between the sensor and two pins each on the GPIO header. Sensor signal goes to GPIO 17.

Anyway, here's the profile I'm using in the hopes that it will help somebody. Caveat: we're two aging women here in this house who weren't particularly strong to begin with so there might be situations where these settings don't perform as well and I wouldn't know.

  Oartec_Slider: {
    numOfImpulsesPerRevolution: 3,
    sprocketRadius: 1.4148,
    minimumTimeBetweenImpulses: 0.01,
    maximumTimeBetweenImpulses: 0.06,
    smoothing: 1,
    flankLength: 13,
    minimumForceBeforeStroke: 10,
    minimumRecoverySlope: 0,
    minimumStrokeQuality: 0.33,
    autoAdjustRecoverySlope: true,
    autoAdjustRecoverySlopeMargin: 0.05,
    minimumDriveTime: 0.300, // minimum time of the drive phase
    minimumRecoveryTime: 0.900, // minimum time of the recovery phase
    dragFactor: 150,
    autoAdjustDragFactor: true,
    dragFactorSmoothing: 3,
    minimumDragQuality: 0.9,
    flywheelInertia: 0.1057,
    maximumStrokeTimeBeforePause: 6.0,
    magicConstant: 2.8
  }

2026-03-12_18-29-26_raw.csv.gz
2026-03-12_18-29-26_rowingData.csv

The above raw file, massaged a bit to make it pretty:

[tigrrrlily]

OK so setting the triggered flank to "Both" and taking the raw data from that, and taking the midpoint of the up and down detection times cleans up the line quite a bit. So it's not so much that the magnet placement is that far off, it's probably more down to mixed north/south orientation or just plain field strength. I don't know what kind of filtering is performed in the engine, but it seems to take care of this so I'm going to stop obsessing here.

[JaapvanEkris]

Many thanks for sharing, I'll include it as an officially supported machine for release 0.9.7 which will be released soon.

When I test with your data, I can bring down the flanklength to 6 datapoints (I do use the new systematic error correction filter, which helps remove structural noise). This makes the force curve a bit more responsive instead of that nice bell curve. Any reason you set it to 13 I'm not seeing?

[tigrrrlily]

2023 me would love to tell you, but is unavailable at this time. I'll try it with six, thank you!

[tigrrrlily]

new systematic error correction filter

Is this something that's already in the main branch? Do I do anything special to enable it?

[JaapvanEkris]

new systematic error correction filter

Is this something that's already in the main branch? Do I do anything special to enable it?

It is in the 0.9.7 branch. You can install it via the update script. You will to adapt your config to:

Oartec_Slider: {
    numOfImpulsesPerRevolution: 3,
    sprocketRadius: 1.4148,
    minimumTimeBetweenImpulses: 0.01,
    maximumTimeBetweenImpulses: 0.06,
    flankLength: 6,
    systematicErrorAgressiveness: 0.975,
    systematicErrorNumberOfDatapoints: 220,
    minimumForceBeforeStroke: 10,
    minimumRecoverySlope: 0,
    minimumStrokeQuality: 0.33,
    autoAdjustRecoverySlope: true,
    autoAdjustRecoverySlopeMargin: 0.05,
    minimumDriveTime: 0.300,
    minimumRecoveryTime: 0.900,
    dragFactor: 150,
    autoAdjustDragFactor: true,
    dragFactorSmoothing: 3,
    minimumDragQuality: 0.9,
    flywheelInertia: 0.1057,
    maximumStrokeTimeBeforePause: 6.0,
    magicConstant: 2.8
  },