3D printed anemometer for breezedude

[khbenjamin]

Hi,

I designed a full magnetic sensing 3D-printed anemometer based on a printed board inserted through the shaft.

There are the CAD files: https://github.com/windnerd-labs/Anemometer-3D-files

So far more than 50 have been built around the world and I'm getting very good feedback.

Now a serial sensor module is available and open sourced: https://github.com/windnerd-labs/Windnerd-Core

It should be a few lines of code away from being compatible with the breezedude, so it might interest you as a cheap and easily maintainable wind sensor!

[mikechristiansen]

Awesome!
Wonder if we could modify your pcb and chuck a rak3172 module on it, and have an integrated low power weather head transmitting FANET or other LORA protocol. Endless possibilities!

[khbenjamin]

I tried different things with LoRa dev boards and custom PCB, last iteration was a STM32 board with sx1262 module and 14500 cell. It can fit in the same plastic body with an elongated bottom part.

It is OK but I didn't find an elegant way to fit a solar cell on the anemometer body itself. Actually a small box attached on the mast, like you do for the breezedude, allows to use a more standard 18650, fits any antenna, doesn't perturb the wind and makes everything more convenient. So for the LoRa use case I will just publish an example project for a WindNerd Core connected to a Heltec CubeCell, this dev board is cheap and has one of the best radio according to my experiments.

[thezenox]

Very nice project!
No problem, I can include reading the sensor.

A few thoughts on that:

• If cable length allows (<1m) use a high baud rate (115200), that saves CPU time = power
• The very best is to send binary data with a CRC in the end.
• Add a wake byte or few stuffing bytes in the beginning of the message. When in deepsleep, I miss a few bytes during wakeup when running on 115200 baud. I need to check the exact timing, I think it was like 2-3 bytes.

We just need to specify a protocol.

You can try to get better in terms of low power. 1.5mA is quite good for always-on, but still a lot. I did not check the exact hardware capability of the STM32G031 yet, but it should be possible to:

• use deepsleep, enable RTC for timing and wakeup
• do pulse counting in hardware with Timer/Counter
• put everything in sleep that does not need to be on when not used (STM32 peripherals and external magnetometer)

Regarding hardware:

• I am from the alps. Here we don't want to mount anything except a lightning rod on the top of a pole.
• The biggest problem of the cups is braking in hail or in winter (even the Davis ones)
• How do your big bearing behave in therms of minimal speed?

---

I actually started something similar in the beginning of this year, but did not find time to finish it, yet.
This is the first concept just copying the proportions of "a well known model". My plan was to make it completely passive (so no internal MCU) using a ultra low power hall effect encoder with analog/pwm output and a magnetic latch.

This uses 3x7x3mm bearings, 3mm steel shaft and brass shaft couplers that are cut in half.

printed in ASA

I have seen things like this, could be interesting cause they look way more robust, but now sure how good they perform in low wind conditions.

[khbenjamin]

The factory firmware has a lot of trade off toward user friendliness including human readable protocol, having LED indicating vane status on board and trying to use only Arduino library. If we want to optimize power consumption, I reckon 1mA would be easily reachable, 0.5mA should be possible if we throw Arduino off to work with Low Layer drivers from ST.
Like you suggest, it will be necessary to make some kind of standard binary protocol, I will work on that and it would be nice if you can review.

Regarding hardware it behaves quite well in low speed, side by side the starting speed is similar, perhaps slightly better than the Peet Bros Pro (big cups). These plastic/glass bearings have lowest friction and inertia than their steel couterparts, yet from the endurance test I made, they should have a fairly long life.

Your model is nice and should work well! The focus for mine was to remove the arm, avoid internal wiring and have something reproducible mainly with 3D printing.

[khbenjamin]

Well, I looked into power optimization and discovered that the magnetometer wasn’t entering sleep mode properly, which caused an extra 0.6 mA of consumption.
The average power consumption is now 0.9 mA, and I was able to reduce it further to 0.6 mA by lowering the vane angle sensing period to 500 ms and turning off the diagnostic LEDs. I think this is satisfying for something running on a solar cell.

I drafted a sketch for a simple binary output of instant wind every 3 seconds with a checksum: https://github.com/windnerd-labs/Windnerd-Core/blob/a80c87502a8beef9634c6766940924fa4993c331/examples/breezedude/breezedude.ino#L19

@thezenox would you be interested if I shipped you a free WindNerd Core kit to test it with a Breezedude?

[thezenox]

@khbenjamin that sounds great. I will add a parser for the data. I send you an email ;)