Driving NeoPixels with Z80

I

've long been thinking about a version two  RC2014 LED matrix module. I've had a matrix with a MAX 7219 on a module. It's a nice enhancement.

But there's only so much you can do with a single-colour LED array right? Wouldn't it be cool to have RGB LEDs?  At Liverpool MakeFest I saw a wall-sized ping-pong ball NeoPixel display and picked up some NeoPixels with the intention of making one. Possibly driven by my RC2014. 

I enjoy learning about protocols and have had some SPI devices working with the RC2014 - bit-banging SPI works really well because it doesn't care about timing.

NeoPixels really do care about timing though. From Adafruit's web page about their 8x8  NeoPixel matrix:

If there's one thing I want to get across in this blog post, it's don't just accept what you're told. Question everything. Learn about what's going on and find out why you're being told something isn't possible. Get creative with workarounds.

I've learned a lot about Z80 timings during this project. I had assumed that because the clock speed I'm using (7.3MHz) was close to the speed required, that I'd be able to drive a line high and low at close to that speed, but no. An 'out' instruction, for example, takes a whopping 11 or 12 clock cycles. 

So a little bit of extra hardware is required. An interface if you will. And some careful programming. This article was a huge help. It reinforces my message about questioning what you're told. Some parts of the timing are very tight. Some are really flexible.

So for this solution (actually the one that worked after a few that didn't) there's not much required. Besides the '138 for decoding the port address (I'm using port 3 - which seems not to clash with anything else on the RC2014, Spectrum or Minstrel 4th)  there are just a couple of logic chips and passives here.

The first time I had a program switching on a pixel was a great day!

Once I'd got the hardware working reliably, and some of the fundamental code working, then it's very easy and fun to write a few demos. (See video further down the page.)

The breadboard wasn't terribly robust (I tend to buy cheap ones)  

.. and so a pcb was the next logical step.  The tall one (for the Adafruit 8x8 matrix) went through a couple of versions before I added the standard-sized one. 

In both cases there are headers for daisy-chaining more pixels (and here we're back to my ping-pong ball display, which is now closer to being a reality) and you can include or bypass whatever pixels are on the board. The very clever protocol means that you simply daisy-chain your pixels and address them by index number in your code. Once a pixel has received 24 bits of data, it simply passes any further data to the next one along.

All of these pixels follow the WS2812B protocol. I've now bought a bunch of pixels in various formats from sources ranging from reputable to less reputable and they all work. The only minor thing is that most are GRB, some are RGB, which requires a little tweak to the code depending on which you're using. (I have some WRGB but haven't tried working with those yet.)

Here's a video in which I talk about the project and show off a whole bunch of demos that I've written - some in BASIC (using a blob of machine code for the low-level stuff) some in assembly but most in C.

So the next challenge was to see whether this could be made to work with a Spectrum. This has the same Z80 processor but it runs at around half the speed of the RC2014. Because of the way my adaptor board works, and again exploiting that flexibility in some parts of the timing, I have been able to make it work. It took a tweak to component values on the adaptor, and a tweak to the code too. But in principle it's using the same adaptor board, and working in the same way. Thanks to Z88dk which can target RC2014 and ZX  Spectrum, I've easily been able to build my existing C and Assembly demos for Speccy.