A Serverless Weatherclock To Monitor My Favorite Kiteboarding Spot At The Lake

Jan 17, 2017

Hi, I'm Douwe Homans. I'm a trained medical doctor, software engineer and entrepreneur in the Netherlands. I recently decided to turn an old weatherclock into an IoT project using a Particle Photon and the Serverless Framework. In this post I'll share how I did it.


In the 80's, my grandparents received a homemade weatherclock as a gift from a friend. The device had a clock-like display that indicated the current wind direction near their home. The sensor part was an actual mechanical wind vane that used a magnet and reed switches to determine the wind direction.

They placed it several meters from their house on a high pole. It was connected with a multicore cable (a core for every wind direction) to the display inside. The display consisted of 8 LEDs mounted in a black acrylic plate. The plate was framed in a circular piece of wood.

After my grandparents moved to an apartment, we never found the space to put up the sensor. The clock hadn't been working since.


A lot has changed since then. Now we have the Internet! I thought it would be nice to hook up the clock to the Internet so we could get the current wind direction (and speed) from the web. This would allow me to get rid of the mechanical wind vane, and I could also add windspeed to the display in addition to the direction.

Getting Started

It came down to 2 steps:

  1. Connect the clock to the Internet
  2. Get the data to the clock

Connect the clock to the Internet

I decided to work with a Photon made by Particle.io. It's a microprocessor that automatically connects to the Particle Cloud once set up. The device can be programmed in C.

I connected each LED, using a current limiting resistor, to a separate pin of the Photon.

The cool thing about the Photon is that you can program it over the air from your browser. In your code you can define remote functions which you'll be able to call over the Internet once the code has been deployed to the device. (Particle Documentation).

I exposed two of those functions: setWindSpeed and setWindDir. The first one takes the windspeed in Beaufort (a commonly used scale in the Netherlands), the second one the windDirection.

The code on the Photon simply runs an infinite loop, similar to this:

} else {
    turn_on_led_for_corresponding_wind_direction(); // which has been set by externally calling setWindDir
    turn_on_number_of_leds_to_indicate_windspeed(); // which has been set by externally calling setWindSpeed

You can see actual code on GitHub.

The clock is pretty 'dumb'. It's not reaching out to the Internet to find, parse, and display data. It just displays the values for the windSpeed and windDir and those values get set by calling setWindSpeed and setWindDir. This helps to keep the code in the clock really simple and focused on one job.

Gathering weather data and getting it to the clock is not the concern of the clock itself.

So how do we get the data to the clock?

Once the clock is connected to the Particle Cloud you can (with the right credentials) connect to it through the Particle Cloud and call the functions you exposed in the code (setWindSpeed and setWindDir). I can login to the Particle Cloud, find out the ID of our wind-clock, and just use the particle command line tool.

$ particle call ID_FROM_OUR_CLOCK setWindDir "NNE"

And that makes my clock's N and NE LED's light up indicating the wind is blowing from North-North-East.

This shows me that the hardware is working, but of course I want the current wind direction for my favorite spot, pushed automically (and periodically) to the clock.

So basically I need to run the following steps periodically.

  • Get data from the web
  • Push data to particle-cloud


When I thought about how I wanted to run the code to update my clock I came to the following conclusion:

I really don't care that much where my code runs. As long as I know it does AND I can tune in to see it still does :).

I could spin up a server myself, I could use an EC2 instance, or Heroku, or a virtual server from a different vendor, or.... I don't want to spend time figuring out what is the best option.

This is where Serverless made sense to me. I just write my code and configure which events triggers my code to run.

The complete code is on GitHub. But it boils down to the following:

A handler.js which exposes an update function which:

  1. Gets Weather data (from Dutch Weather Institute)
  2. Converts M/S to Beaufort
  3. Connects and pushes data to the Particle Cloud

A serverless.yml:

service: windclock
  name: aws
  runtime: nodejs4.3
  region: eu-central-1

    handler: handler.update
      - schedule: rate(15 minutes)

That's it! I can run a $ serverless deploy to deploy my code and know that it runs every 15 minutes. If I want to tune in I simply run $ serverless logs -f update.

And just like that I can tell whether it's time to head out with my kiteboard.

Subscribe to our newsletter to get the latest product updates, tips, and best practices!

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.