Entropy Reduction

Personal log of yet another digital native.

---

Project maintained by andre-abadi Hosted on GitHub Pages — Theme by mattgraham

Purpose

• Detect bed occupancy for home automations
• Possibly measure weight as well
• Possibly measure number of occupants

References

• Original Smart Bed occupancy sensor on Thingiverse by Lukáš Porubcan
• Self made remix:
  • Thingiverse: Bed Occupancy Sensor Shim
  • Github: Bed Occupancy Sensor Shim
• Original Wemos D1 mini case on Thingiverse by holda
• Self made remix:
  • Thingiverse: Wemos D1 Mini Case for HX711 Bed Occupancy Sensor Assembly
  • Github: Wemos D1 Mini Case for HX711 Bed Occupancy Sensor Assembly
• HX711 + 4x 50kg Load Cells

Actions

• Printed all of the parts for the Smart Bed occupancy sensor
• Remixed the Wemos D1 mini case by holda to make room for a four pin header coming from the HX711
• Printed my remix
• Wired up the load cells in a Wheatstone Bridge, per the picture in the original thing
• Got weird values of -8388608
• Researched and found out this was due to incorrect wiring of the Wheatstone Bridge
• Corrected wiring
• Mounted load cells in holder assembly using super glue (cyanoacrylate)
• Calibrated using a heavy book
• Drilled into centre of bed frame
• Realised the plates wouldn’t be high enough to take any load from bed slats
• Remixed the plates from the Smart Bed occupancy sensor and created the Bed Occupancy Sensor Shim
• Measured distances between original plates and slats, then printed in various heights by scaling in the slicer
• Glued the shims onto the plates to achieve contact/load
• Taped loose wires to prevent snagging and to neaten up appearance
• Calibrated
• Found one side shims were too thick so reported loads to high
• Resorted so safe maximum for a binary sensor
• Automated within ESPhome to abstract high volume of sensor readings
• Interpolated raw readings to prevent incorrect readings from triggering binary sensor state change
• Final binary sensor was responsive to change without giving erroneous readings

Pictures

Code

sensor:
# actual input HX711
  - platform: hx711
    internal: true
    id: weight
    name: "${friendly_name} Weight"
    dout_pin: D3 # DT
    clk_pin: D4  # SCK
    gain: 128
    update_interval: 1s
    unit_of_measurement: kg
    # raw value of -29528 means sensors are effectively disconnected from hx711
    filters:
      #- calibrate_linear:
      #    - 230405 -> 0     # unladen with only 3d printed toppers
      #    - 132925 -> 4.726 # greys anatomy weight
      #
      - calibrate_linear:
          - 201800 -> 0    # mattress, sheets and summer blanket
          - 89100 -> 76.7 # point in time human male
          # 95000 = 95kg
          # 77500 = 90kg
          # 89100 = 85kg
          # 91300
      - sliding_window_moving_average:
          window_size: 10
          send_every: 10
          #send_first_at: 5
    # https://esphome.io/components/sensor/index.html#on-value
    on_value:
      - if:
          condition:
            sensor.in_range:
              id: weight
              above: 10
          then:
            # https://esphome.io/components/sensor/index.html#lambda-calls
            lambda: 'id(occupied).publish_state(true);'
      - if:
          condition:
            sensor.in_range:
              id: weight
              below: 10
          then:
            lambda: 'id(occupied).publish_state(false);'


# binary sensor to be modified by hx711
binary_sensor:
  - platform: template
    id: occupied
    name: "Master Bed Occupied"
    #device_class: presence

Observations

• HX711 readings vary and do have frequent spikes requiring software smoothing, 10 sample moving average worked well
• Using template automations to toggle the binary sensor was good as it abstracted raw values away from home assistant