Drive an RGB led with software PWM

This project illustrates how to use an ATTiny microcontroller and an RGB LED to create a coloured light that cyles continuously through random colours, but changes colour dramatically at the press of a switch.

Motivation

Why? My kids had this $2 novelty pen that did this employing a teensy circuit along with 3x3mm LEDs and some alkaline coin-cells. They broke it almost instantly, losing some critical small parts.

"Daddy, my light-up pen broke! WAAAAAAAAAAAAAAHHH!!!"

Having deprived themselves of the captivating coloured lights, they then drew on the walls with the instrument-formerly-known-as-the-light-up-pen. Then they experimented with strangling each other with the pen's narrow ribbon lanyard. I decided to replicate the pretty light without the other problems.

Circuit Overview

This circuit produces coloured light by using an 8-pin ATTiny13 microcontroller and Pulse-width-modulation (PWM) to vary the brightness of three LEDs (Red, Green and Blue) that are closely located, or share a common package. I used a common-anode 5mm RGB led, which looks just like a normal 5mm LED but has 4 leads instead of two.

Since the ATtiny13 only has one timer with 2 PWM channels, it cannot control 3 PWM channels in hardware. Instead, I do the PWM logic in software, timing the pulse modulation via a regular tick (interrupt) from the ATTiny's hardware timer. (Nowadays you can get the ATTiny45 and friends which have 3 PWM channels, but I could not obtain them when I first built the prototype for my middle child a couple of years ago). My youngest child is now 2, so I dug this project out of the mothball bunker for her to play with.

WARNING: modern multi-colour LEDs can be PAINFULLY bright. If children will play with this, choose one of the following: ensure your LED is not a superbright, or use a diffuse led, or diffuse a clear one (eg cover it with half a table tennis ball, or enclose it in resin or wax. You could also use higher-value resistors to limit current, or lower PWM_OC_MAX to limit the PWM duty-cycle (LED brightness) via software.

Source code and schematic are available as usual from github.com.

Pulse-Width Modulation provides colour choice

The microcontroller should be configured to obtain the system clock from the 8MHz internal RC oscillator. The ATTiny13 factory fuse settings include a divide-by-8 of the RC oscillator giving 1MHz, so remember to change that fuse to get the original 8MHz signal..

The overflow interrupt from the hardware 8-bit timer triggers an interrupt service routine (roughly 32000 times per second) which increments each of 3 separate 8 bit counters. These counters, and corresponding "Output Compare" (OC) values are used to generate 3 Pulse-Width-Modulated waveforms, with each counter being associated with a particular output pin. Every time a counter reaches zero, the corresponding pin is turned on. When the counter reaches its PWM_OC value it is turned off. When the counter reaches its PWM_TOP value (255) it wraps around to zero and the cycle repeats.

The effect of all this is that each of the 3 output pins is driven with a square-wave having a wave-legnth of 256 timer ticks (about 1/128 of a second), which is equivalent to a frequency of 128Hz (cycles-per-second). The OC value controls the "duty-cycle" (ratio of on-period to off-period) of the wave, with 0 giving always off, 255 always on, and any intermediate value giving a wave that is on for x/255 of a cycle and off for the remainder of each cycle. If an electric light is powered by such a square wave, with a frequency of more than about 50Hz, the human eye does not percieve any flickering---instead changes in the duty cycle are perceived as changes in brightness (this is how most household dimmers work). With 3 leds in Red, Green and Blue, and the ability to vary the perceived brighness of each, you can produce many different colours, exactly as is done in modern TVs and projectors.

So using 3 PWM outputs, we have a coloured light that we can set to any brightness and colour (denoted by a particular [R=x, G=y, B=z] tuple) that we desire.

BTW, it's not strictly necessary to use 3 separate counters. A single counter could be used with 3 different OC values to control each output pin independently. I chose to implement 3 distinct channels (rather than 1 channel with 3 Output-compare values) out of some voodoo idea that running the channels out-of-phase will lower the peak current draw and give slightly longer battery life. For example, if you set PWM_TOP to 240 and PWM_OC_MAX to 80, then if the PWM counters are initialized at 0,80 and 160, the 3 output pins will never be on simultaneously, and peak output current will never be more than 20mA. This also means you can get away with ONE resistor instead of 3.

Colour change by drifting PWM parameters

A coloured light isn't very interesting. But I have it on good (2 year old) authority that a pulsing and constantly changing light is captivating.

For each of the 3 channels, at the end of every PWM cycle (when the counter reaches 255), the duty cycle of the particular channel changes by a small amount. This causes the colour to change. The rate that each of the PWM channels changes duty-cycle is chosen semi-randomly such that an ever changing progression of colour is produced.

User interface

A pulsing coloured light is nice, but when you're 2, a BUTTON TO PUSH is best of all.

If the CHANGE button is pressed, the LEDs are reset to all off (OCx=0), and the cycle-rate of one channel is changed to a new random value, giving an abrupt change of colour, followed by a new and different progression of colour changes. Small children love this and will sit and frob the button for ages, before they break it or lose it. Then you can make them something else.

A second button acts as an "on/off switch" (actually triggering a low-power sleep mode, and waking from same).

Both buttons trigger interrupts via the pin-change interrupt facility. The buttons are "debounced" in software, by masking further interrupts for a certain interval after any button interrupt.

Hardware

I prototyped this on a solderless breadboard, and programmed the ATTiny with an AVRUSB500v2 programmer, which has the useful advantage of using a 1x5 connector instead of the more common 2x3 or 2x5, allowing simple in-circuit programming of breadboarded AVRs.

Once I had the software working, I built a soldered prototype on a scrap of pad-per-hole proto-board. With only 5 data pins, and 2 power, you can just plonk the components down next to the chip socket, and either use point-to-point wiring and/or chains of bridged pads. If you're bridging pads, use the cheap phenolic protoboard, the high-end green fiberglass boards with solder masking are commendably resistant to solder bridging.

You can use any simple voltage regulator (or none). The ATTiny chips come in a 5v version that runs to 20MHz, or a low voltage version that will work off as low as 1.8v (but is limited to 4 or 10MHz) . I used an LP2950 low-dropout regulator, which will easily give you 5v from 4 alkaline cells, or 3.3v from 3 cells. If you're using a 9v battery, the cheap and common 78L05 (which needs an input of at least 7 volts) works fine too. You could even leave the regulator off entirely, if you use a 4.5v alkaline pack or a 3.7v lithium. The low-votage ATTiny13v is remarkably forgiving (but remember to set the brown-out fuses to an appropriate low-voltage cutoff).

I haven't bothered to design a PCB yet, as it wouldn't be significantly more compact than the protoboard version. I suppose an SMD version would be interesting---it could be made not much larger than the coin cells that power it, giving some interesting swallowing-hazard fun to be had.