Project Showcase & Calibration Help: Upgrading a Body Coach Water Rower

[kasary]

Hi everyone,

First of all, thank you for this amazing project! The installation of ESPRowingMonitor was really straightforward. I want to share my recent DIY upgrade of a water rower and also ask for some advice regarding calibration and fine-tuning.
Thanks to Gemini that brought up this Repo while i was considering selling my stupid rower.

My Setup & Hardware

Machine: Body Coach Water Rowing Machine (bought in 2021)

Microcontroller: ESP32 D1 Mini

Sensor: KY-003 Hall Effect Sensor

Magnets: 8x Neodymium magnets (10x2mm)

Mechanical Modifications
The original setup was quite basic: the old monitor and reed sensor relied on a single magnet that wasn't even attached to a flywheel, but directly to the pull strap winder.

Luckily, the shaft continued upwards with 13.8 mm of clearance. Here is what I did to upgrade it:

• Custom Magnet Disc: I 3D-printed a custom disc/flywheel holding the 8 magnets. If anyone is interested, I'm happy to share the STL/STEP files!
• Force Absorption: To handle the pulling forces, I added a ball bearing on top of my magnet disc.
• Sensor Placement: The magnet disc now spins with just a few millimeters of clearance past the KY-003 Hall Effect Sensor.
• Custom Housing: I modeled the upper wooden part of the rower in Fusion 360 to design a perfectly fitting custom housing for the ESP32 and the sensor.

Current State & Software

The software side was quick to set up.
The stroke rate is being detected correctly!

My Questions Regarding Calibration

For the Flywheel Inertia, I used 0.1 as a starting point, taking the middle of the "Typical Range" for Water Rowers mentioned in the documentation.

Since I want to make sure my data is as accurate as possible, I have a few questions:

Is 0.1 a plausible and safe starting point for my specific setup?

How can I best refine and verify this inertia value? What is the recommended workflow to dial this in?

Edit: To be honest, i don't quite understand what settings i need to set in the WebGUI :D Pls Help

I haven't had the time to look deeply into the end-to-end (e2e) simulation yet. Is using the e2e simulation the best next step, or should I start by recording specific logs (e.g., via SD card) and analyzing them in the Calibration Helper GUI?

Thanks in advance for your help and for providing such a great open-source tool!

[Abasz]

Thanks for sharing this. It look very cool! I am happy to help with the calibration and all that. We had one water rower: #22

ESP Rowing Monitor had one water rower so far but with https://github.com/JaapvanEkris/openrowingmonitor we have more experience on this.

If you have not done so I recommend consoluting with the interactive Wiki: https://deepwiki.com/Abasz/ESPRowingMonitor

You can ask questions and it will try to provide answers based on the documentation.

There is a FAQ which has a section on the inertia: https://github.com/Abasz/ESPRowingMonitor/blob/master/docs/faq.md#10-how-do-i-measure-flywheel-inertia

FAQ has also on the E2E. The basic idea is that from the WebGUI you can export the delta times CSV which is the raw impulses produced by your machine. If you do 20-50 strokes and send that CSV to me I can help you with the calibration.

The inertia value will as the docs say depend on the fill level. So here the rule of thumb recommendation is that you set a water level that provide a plausible feel resistance, then if possible you will need to compare to a machine which we know produce a good result (e.g. Concpet 2). You basically do the same/similar effort and you set the inertia until the metrics (power and speed mostly) matches what you have recorded on the C2.

To be honest, i don't quite understand what settings i need to set in the WebGUI

I am happy to help, actually there is a section on the DOCS for the settings but also a workflow in the FAQ. I would start with the later.

Also again, if you provide me with the settings you have I can run the simulation and make adjustments to the setting and send it back to you.

Thanks for your contribution it is very valuable!

[kasary]

Hey @Abasz, thanks for your really fast reply.

I looked into the Supported Rowers Documentation and found the "Joroto MR380" in this discussion.
The watertank and the pull system look exactly like my machine (see the manual here).

2026-02-20 15-23-22 - session.csv
2026-02-20 15-23-22 - deltaTimes.csv

Is there a way to export the settings so sharing is easier?
Flywheel Inertia: 0,74 (Copied from the MR380)
Magic Constant: 2,8
Sprocket Radius: 7 (Copied from the MR380)
Impulsed per Revolution: 8

Sensor Debounce: 0 (Never changed)
Rowing Stop Threshold: 7 (Never changed)

Drag Factor Lower Threshold: 100 (Does this make sense)
Drag Factor Upper Threshold: 32000 (Copied from the MR380. Also saw that in the ORM documentation)

Drag Facotr Period: 7 (Does this make sense. Was provided from Gemini :D )
Drag Coefficients Array Size: 5
Goodness of Fit Threshold: 60% read that in the documentation

Stroke Detection: Both
Minimum Recovery Time: 500 (Copied from the MR380)
Minimum Drive Time: 400 (Copied from the MR380)

Hope you can help, thanks a lot

[Abasz]

Is there a way to export the settings so sharing is easier?

Its a good point! unfortunately now :D but it should be so something to the backlog.

Delta times look nice, so input data is correct. This is ok and can be calibrated.

Stroke Detection: Both

So essentially this should be set to Torque once you confirmed that you see stroke count increasing. This is because SLOPE based detection with a 0 slope comes always before the torque detects.

Flywheel Inertia: 0,74 (Copied from the MR380)

This will depend on the water level, but actually it influences the power that is getting output. So stroke detection is only impacted via the calculated torque amount.

I will do some calibration runs tomorrow for you and send back something that you can work with.

[Abasz]

Thanks for your patience. I have run the calibration. Your machine produces where clean data which is excellent (your magnet alignment is extremely good especially considering the number of magnets, well done!).

This helped to create a config that is very conservative and produces good data.

A few considerations:

• Inertia: I have set it to one as this produces approx 140watt per strok on average (which corresponds to 2:10 500m/min pace if you were on a concept 2). I don know what the effort here was so that probably needs adjusting for more accurate distance speed and power.
• Sprocket raidus: In the Faq you can find how to measure it, but for your machine this is the pulley that rolls on the handle chain. It does not need to be super accurate as it does not influence stroke detection or power or distance (it only matters for drive length in meter which is an upcoming tile on the web-gui)
• These settings represent this steady rowing and your stroke is well formed especially consistent in length which helps the algorithm. But for intervals or higher effort may need a bit more relaxed settings. So there fore I would kindly ask you that you do a full workout (with variable intervals if you possible but not a must) and share the delta times so I can create and finalise the profile and add it to the precompiled rowers (so it shows up with correct name as BLE device :)).
• Can you also share what ESP board are you using? By the look of it it seems a lollin/wemos D1 mini mwhich is known to have issues with multiple BLE connection. Are you using generic-dinamic firmware so far?

#pragma once

#include "../utils/macros.h"

// NOLINTBEGIN(cppcoreguidelines-macro-usage,cppcoreguidelines-macro-to-enum)

#define DEVICE_NAME Joroto MR380
// clang-format off
// NOLINTNEXTLINE(bugprone-macro-parentheses)
#define MODEL_NUMBER 07982-500
// clang-format on
#define ADD_BLE_SERVICE_TO_DEVICE_NAME false

// Hardware settings
#define IMPULSES_PER_REVOLUTION 8
#define FLYWHEEL_INERTIA 1
#define SPROCKET_RADIUS 7
#define CONCEPT_2_MAGIC_NUMBER 2.8

// Sensor signal filter settings
#define ENABLE_DEBOUNCE_FILTER false
#define ROTATION_DEBOUNCE_TIME_MIN 8
#define ROWING_STOPPED_THRESHOLD_PERIOD 7'000

// Drag factor filter settings
#define GOODNESS_OF_FIT_THRESHOLD 0.6
#define MAX_DRAG_FACTOR_RECOVERY_PERIOD 6'000 // Max recovery time in ms
#define LOWER_DRAG_FACTOR_THRESHOLD 100
#define UPPER_DRAG_FACTOR_THRESHOLD 25'000 // Magnetic rowers can have very high drag factors
#define DRAG_COEFFICIENTS_ARRAY_LENGTH 3   // Magnetic rowers DF jumps around a lot (as they are speed dependent), need longer average to smooth it out

// Stroke phase detection filter settings
#define MINIMUM_POWERED_TORQUE 0
#define MINIMUM_DRAG_TORQUE 0
#define STROKE_DETECTION_TYPE STROKE_DETECTION_TORQUE
#define MINIMUM_RECOVERY_SLOPE 0 // Only relevant if STROKE_DETECTION_TYPE is either BOTH or SLOPE
#define MINIMUM_RECOVERY_TIME 500
#define MINIMUM_DRIVE_TIME 400
#define IMPULSE_DATA_ARRAY_LENGTH 10
#define DRIVE_HANDLE_FORCES_MAX_CAPACITY 45

// NOLINTEND(cppcoreguidelines-macro-usage,cppcoreguidelines-macro-to-enum)

EDIT:

One more thing, I see you are using the KY-003. This normally uses an allegro 3144 Digital Hall which is 4.5v as per the data sheet. I see that you have soldered this to 3.3v. Would you mind if you have measure the 3.3v pint to which you have soldered supply to with a multi meter how much voltage you have on it? I am puzzled how this works with such a low supply voltage :D

[kasary]

Hello, thank you so much for your effort :)

• Inertia: I have set it to one as this produces approx 140watt per strok on average (which corresponds to 2:10 500m/min pace if you were on a concept 2). I don know what the effort here was so that probably needs adjusting for more accurate distance speed and power.

Hard to tell, i was a steady pace, not hard going.

I'll test it and get back to you.
Is there anything specific you'd like regarding the interval?

I'm using the dynamic genirc firmware.

Thats also something i was kind of confused with.
Seeing 40 assets in the releases was kind of overwheling. That brings up the import/export again in my mind.
But lets not mix up the dicsussions, i'm not that deep into the repo to understand it completly.

I also read that the KY-003 needs 4,5V, but its kind of a operation range 3,3V to 5V
The supplier's instruction manual specifies 3.3 to 5 V.

I use this ESP32 D1 Mini
Datasheet.pdf

[Abasz]

Is there anything specific you'd like regarding the interval?

No, technically not, but the more variety is in it the more it could tell me in terms of the setting. But its up to you and your fitness, my aim is not that you hurt your selft but rather that there is warmup then some intervals where there are slowdown (or even stopping) between. 2-3-4 is fine.

So for reference (but I mean I am not suggesting you do this in any way just so you see what variety I mean) with the Kayak erg I am working on and which is being tested with professional athletes we do:

1. 4 min warmup,
2. 3 min normal economy
3. 2x2 min 2000m pace
4. 3x1 min 1000m pace
5. 4x30sec 500m pace
6. 20 sec max out but from a flying start from 1000m pace

1-1.5 min rest between.

Seeing 40 assets in the releases was kind of overwheling. That brings up the import/export again in my mind.

Technically you dont need to see that :). Because when you use the flasher tool it auto loads the profiles and you can select it from a drop down after connecting your board to the usb serial. And from there on in the GUI the auto updater takes care of matching your board based on its model number. That is why it is a matrix of boards and profiles. So you can use the D1 Mini (which the docs says do not work properly with multi connection BLE by the way) and others can use the S3 (which is much faster and modern chip) and so on.

You see, there is value in the pre-compilation. There is performance reduction with the dynamic settings because the compiler can do less assumptions as it does not see what values are there. Based on bench marking dynamic settings adds 5% overhead on average and even more on the max execution time (this is the time it takes to process one impulse and does not include tasks related to the rest of the application). This does not seem a lot but on an air rower where we can have an incoming impulse on every 5 millisecond this becomes important. Also settings update after profile tweaks are automatic but that is just convenience.

I also read that the KY-003 needs 4,5V, but its kind of a operation range 3,3V to 5V

Yeah I mean the specific hall's data sheet says otherwise, but this is wired. It is possible that they use something else on it. Any way thanks for letting me know that it works.