Some weeks ago I got some Adafruit NeoPixel, a compact RGB LED module. Each LED module has a WS2812 chip that only needs three wires: power, ground and data. The data protocol is self-clocking, at a rate of 800KHz. The LED modules can be chained to build longer LED stripes. (For more details see the Adafruit NeoPixel Überguide.)

Adafruit provides an Arduino library on github to control the LEDs, but there is no direct support for Microsoft .NET Microframework (NETMF) controllers like the Netduino Plus 2 that I was going to use.

Actually, there is an advanced guide to use NETMF controllers with the WS2812 LEDs, but it requires a custom firmware, because it seems that the GPIO pins of most NETMF controllers are not fast enough to switch the signal at the needed rate (there is a timing needed in the range of 0.4μs to 0.8μs and some websites found out that the GPIO pins have a latency of 17μs). As I nearly bricked a Netduino before, I didn’t want to dive into custom firmwares at this point.

So I found a different way to send the control codes to the LEDs at the needed speed as the following photo proves:

Netduino with WS2812 LEDs

How does it work?

I thought that there must be a faster way to get signals off a Netduino Plus 2 because it has an Ethernet port and SD card slot, and I suppose they need to be adressed faster than 17μs (I haven’t evaluated this, so maybe I’m wrong with that, but it led me in the right direction). Looking at the ports that the Netduino provides my attention was drawn towards the SPI interface.

SPI (Serial Peripheral Interface) is a four-wire master-slave interface for full-duplex communication between chipsets. It has a clock wire (SCLK), output (MOSI) and input (MISO). Normally it can be used at a frequency up to some MHz, so the ports must be really fast.

My assumtion was, that I could only use the output (MOSI) pin, configure the frequency accordingly and send the needed control bytes on that wire to the LEDs. I tried some settings and data packets, but first without any success, but luckily I wasn’t the first one to try this, so I found some code snippets for different microcontrollers, that pointed me into the right direction, e.g. that a timing of 6.666MHz is recommended.

I’ve written two classes and uploaded them to github, so if you have a Netduino Plus 2 (or similar microcontroller) and don’t want to use a custom firmware to control WS2812 RGB LEDs, then maybe you want to give my code a try.

Code on github:

Usage example:

NeoPixelSPI neoPixel = new NeoPixelSPI(Pins.GPIO_PIN_D10, SPI.SPI_module.SPI1);
Pixel pixelRed = new Pixel(255, 0, 0);
Pixel pixelGreen = new Pixel(0, 255, 0);
Pixel pixelBlue = new Pixel(0, 0, 255);
Pixel pixelWhite = new Pixel(255, 255, 255);
Pixel[] block1 = new Pixel[] { pixelRed, pixelWhite };
Pixel[] block2 = new Pixel[] { pixelBlue, pixelGreen };
int waittime = 1000;
while (true)

The library code is not optimized (yet), and I didn’t write a helper class for easier handling of long LED stripes (because I don’t own one), so feel free to extend the code. I’m not quite sure if this approach is really a good one, but it worked for me.

Also feel free to leave a comment below if you have any remarks, ideas, etc.

As always: use this information and the code on your own risk. ;)

Tagged with:

7 Responses to Netduino Plus 2 (NETMF) and WS2812 RGB LEDs

  1. Darrek Kay says:

    Thanks for this alternative solution…

    Your bit-bang interface works pretty good for my strip of 600 WS2812s. I get a few glitches located different distances down the strip depending upon the SPI clock speed I select and the 1/0 bit patterns.

    The custom firmware solution runs an order of magnitude faster since the bit-bang data doesn’t need to be stored in memory and marshaled to the SPI native code. For my 600 LEDs, the 14,400 bytes of bit-bang data that must be calculated takes 395ms + another 25ms to write at 9.6MHz. Whereas the custom firmware can drive the same 600 LED strip flawlessly using only 55ms of time.

    I think your code will allow a lot more people to experiment with the WS2812s since a custom firmware build is not required.

    • jCoder says:

      Thank you for both giving the code a try and providing some valuable timing feedback.
      I didn’t want to test a custom firmware myself due to previous issues with nearly-bricked hardware, so I really appreciate your tests.

      Also, I guess that the code is not really optimized for longer chains of LEDs (I only have a few of them at hand), and I can think of a different API implementation that caches the complete data to send when the color value of an individual LED is set (which maybe reduces the overall calculation costs) – the current implementation does not make any assumption at first about how many LEDs will be set, and the transfer data is calculated right at the time when it is needed on the wire.

      The code is “only” intended as a starting point for other coder’s experiments, so I am thankful if it provides any value for other users out there :)

  2. Marco Cordeiro says:

    Thanks for this solution, but I’m having some trouble compiling it, maybe it is just a dumb error.

    I can’t reference this: Microsoft.SPOT.Presentation.Media

    • jCoder says:

      Have you tried adding a reference to Microsoft.SPOT.Graphics.dll ?

      In addition, I will check my test project later to see if there are any other useful hints for this topic.

      • Marco Cordeiro says:

        That did it, I said it was a dumb problem.
        Sorry about that, no too familiar with all this libraries.

        Thanks for the help