ADC Raw Values

[MRobi1]

Hey, not sure if you still use this, I'm wondering if you remember how you viewed the raw values from the ADC sensor to determine the filter values?

    filters:
      - calibrate_linear:
        - 1.795 -> 0.0
        - 1.43 -> 100.0

I just finished watering to runoff and have the xiaomi plant sensor right beside the m5stack sensor. The xiaomi sensor is reading 82%, m5stack is reading 5%.

[rasclatt-dot-com]

These values were based on recommendations in other projects in m5 community.

My initial testing noticed that the m5 capacitive sensor (and similar capacitive sensors) take some time to register changes in soil moisture.

This is due to the nature of how capacitive sensors work, also how quickly the soil absorbs the moisture.

This is why I configured long delays between initial trigger of the pump before allowing the pump to run again to allow time for the readings to catch up, (and avoid flooding my plants).

[nborson]

The xiaomi sensor is reading 82%, m5stack is reading 5%.

@MRobi1 wow, that's a big difference! I didn't have another sensor to compare with, but using the calibration values in the sample code, my new sensor gave me ranges of 2% (in air) to 27% (in fairly dry soil before watering) to 73% (after watering thoroughly and waiting overnight), which passes the reality check. Edit: the reading in soil jumped from 73% to 96% after re-calibrating to - 1.53 -> 100.0. Could you be measuring the wrong pin, or using a different attenuation: value, or using 3.3V rather than the 5V output on the Atom's Grove connector?

how you viewed the raw values from the ADC sensor

Commenting out unit_of_measurement: "%" and - calibrate_linear: filter -- and increasing the accuracy_decimals -- gives the raw voltage at the pin.

sensor:
  - platform: adc
    pin: 32
    name: ${plant_1} Soil Moisture Value
    attenuation: auto
    id: plant_1_soil_moisture
    accuracy_decimals: 3
    # unit_of_measurement: "%"
    update_interval: 10s
    icon: mdi:flower
    filters:
    #   - calibrate_linear:
    #     - 1.795 -> 0.0
    #     - 1.43 -> 100.0
      - sliding_window_moving_average:
          window_size: 12
          send_every: 3
          send_first_at: 2
      - median:
          window_size: 12
          send_every: 3
          send_first_at: 2

I bought two new m5stack U101 sensors (the ones @rasclatt-dot-com describes), and I found these voltages from them on the Atom lite with attenuation: auto:

V (Min/Max/Avg)	Sensor 1	Sensor 2	Notes
Air	1.786/1.787/1.7865	1.789/1.792/1.7905	The difference between sensing dessicated air with ~10% RH vs room air with 47% RH was at most 0.001V. See screenshot below
Water	1.529/1.532/1.530	1.528/1.532/1.530	Immersed the probe in water to the top of the sensor strip
Vs. sample code (dry/wet)	-0.47%/+6.99%	-0.25%/+6.99%

Screenshot of History in Home Assistant shows raw Sensor 2 voltages while the sensor is in dessicated air (left half of graph) and 47% RH room air (right half of graph). The average is just 0.25% lower than the calibration voltage in the sample code.

Notes

• Attenuation: In the sample code, it's set to 11. I set it to that, then 12 (because of a compile message saying 11 was deprecated, contrary to the docs), then to auto, with no obvious difference in the readings. The calibration tests in the above table were done with attenuation: auto
• the m5 capacitive sensor (and similar capacitive sensors) take some time to register changes I found that when plunging Sensor 2 into water, the voltage plunged from 1.79 to 1.54 in < 2 minutes, even using the - sliding_window_moving_average: values in the sample code above. But it took another 90 minutes for the voltage to stabilize around the average of 1.530.
• some time to register changes in soil moisture. Yes, in my small potted plant, the measured moisture level went from 27% to 70% within about 30 minutes after watering, but continued rising overnight to 73% in the morning.

Conclusions

• For practical purposes, the calibration values in the sample code should work well enough, though I might suggest changing - 1.43 -> 100.0 to - 1.53 -> 100.0
• Variations in sensors and maybe boards may require a calibration like I did to get the best accuracy. Maybe repeatedly, as the sensors age? But is it worth the trouble?
• @MRobi1 has some other problem, like a misconfiguration or faulty sensor; miscalibration does not account for the 5% reading he is getting in nearly-saturated soil.