Order of Actions in Automation: Light-On / Picture / Light-Off

Configuration
1

Dear community

I am reading the analog thermometer of our hot-water boiler using an ESP-Cam and ppf.angauge. The picture-to-temperature conversion works well, but I am struggling with more basic things: The thermometer is in a dark place. Therefore, I use the built-in LED of the ESP-Cam to illuminate the scene.

To take proper pictures, I need the illumination to turn on, first. Then the camera is supposed to take a picture. And afterwards the illumination turns off.

However, I need to use a ridiculously long delay (5 seconds) between light-on and shot to reliably avoid getting dark frames. To use ppf.angauge, I write a custom_component in python. However, the same problem happens with a normal automation (see below).

I assumed that HA executes the actions in the order I give them. I understand that HA works asynchronously: It may do something else while waiting for the ESP-Cam to turn on the LED. But nevertheless I would expect it to execute the next action in my sequence only after completing the first one.

Am I wrong?

Or does the ESP-Cam not actually trigger an exposure when I tell it to, but merely sends the last frame it had already taken earlier (maybe even before I turned on the light)?

Any help or tipps on how to diagnose the problem further are appreciated.

actions:
  - type: turn_on
    device_id: ...
    entity_id: ...
    domain: light
    brightness_pct: 100
  - delay:
      hours: 0
      minutes: 0
      seconds: 5
      milliseconds: 0
  - action: camera.snapshot
    metadata: {}
    data:
      filename: >-
        /media/boiler/{{ utcnow() | as_timestamp |
        timestamp_local('0000-12-31T00:00:00') }}.jpg
    target:
      device_id: ...
      entity_id: camera.boiler_kamera
  - delay:
      hours: 0
      minutes: 0
      seconds: 0
      milliseconds: 100
  - type: turn_off
    device_id: ...
    entity_id: ...
    domain: light
2

Can’t answer your question, but I have 1 of my own.

Are you sure the dark frames are due to the initial delay & not the subsequent one?
Asking because 100ms delay might start turning off the light before the camera is done taking the picture.

3

For reliability and performance you can also consider creating a script directly in ESPHome — to run directly on the ESP, but triggered from HA.

4

Hi ShadowFist

I’m sure, yes. I started from extremely large delays (on both ends) and reduced them until I got dark frames. The values I posted are the minimum values I can get away with without getting dark frames.

5

You are referring to just the part related to LED-on / capture / LED-off, I guess? That sounds very promising (although I would still like to know what is going on with my automation). Is that possible on ESPHome-yaml grounds or would I have to use C++?

6

Yes

The actions you have invoked remotely from HA is natively available in ESPHome too – so you can do this in YAML.

7

:+1: Going to try.

8

I still think this is a promising approach, but I’m struggling a bit. I have no problem toggling the LED and doing delays in a script. Also, I can link this to a button that I can press from Home-Assistant. But how am I supposed to capture an image and send it to HA?

esp32_camera:
  name: Kamera
  ... <shortened>

light:
  - platform: binary
    output: boilerLED
    id: illumination

script:
  - id: capture
    then:
      - light.turn_on: illumination
      - delay: 1s
      - light.turn_off: illumination
    mode: queued

button:
  - platform: template
    name: "capture image"
    on_press:
      then:
        - script.execute: capture
9

Hmm, you got me there. After some research, it doesn’t look like my idea is as simple as I first thought. ESPHome doesn’t have the equivalent of HA’s file camera component to expose a stored image as a new camera, but I found something else. I cannot test this, unfortunately.

The ESP camera has an on_image automation, so you could do something like this which should then run automatically when you take a snapshot:

esp32_camera:
  ...
  on_image:
    then:
      - light.turn_on: illumination
      - delay: 1s
      - light.turn_off: illumination
10

That wont work: on_image is called after the image has been taken: The docs say that the image is available in a variable:

1 Like
11

Bugger, you’re right… Back to HA then, I guess.

12

Going back to your original issue, there is something else you could try, at least for debugging. It’s ugly for me, so I wouldn’t stick with this, but hopefully it can give you some info.

You can still have a script within ESPHome, but make a call back to HA to perform the actual snapshot. Seems there’s no way to get a stored image otherwise, except to push it via HTTP to a custom web server.

Something like this:

script:
  - id: snapshot_with_light
    then:
      - light.turn_on: illumination
      - delay: 1000ms
      - homeassistant.service:
          service: camera.snapshot
          data:
            entity_id: camera.my_cam
            filename: >-
              /media/boiler/{{ utcnow() | as_timestamp |
        timestamp_local('0000-12-31T00:00:00') }}.jpg
      - light.turn_off: illumination

The other thing we don’t know here is how long it takes for that camera’s exposure to stabilise after a lighting change.