Multiplex question for multiple RGB LEDs

On a sunny day (Thu, 03 Apr 2008 07:46:49 +0100) it happened Mike G wrote in :

Your hardware would be much simpler with an FPGA.

You can drive the array with a 4-bit port. Continue with the timed interrupt at 255x100=25500 Hz, update the RGB outputs at the same time, and output a clock edge every 85 interrupts ( this means output 1 on

85th and then 0 on 86th) to produce 300Hz for a ring counter drive, like so: View in a fixed-width font such as Courier.

. . V+ . | . .-------+------. . | | | . | | | . __ | | | . R>---------------| \\ | | |< . | o-----|-------|----| . .---------|__/ | | |\\ . | | | | . | | | | . 100Hz | __ | | | . G>-----|---------| \\ | |< | . update | | o-----|-----| | . | .------|__/ | |\\ | . | | | | | . | | | | | . | | __ | | | . B>-----|--|------| \\ |< | | . | | | o---| | | . | | .---|__/ |\\ | | . | | | | | | . | | | | | | . | | | B-anode G-anode R-anode . | | | . | | | . ----------- . | Q0 Q1 Q2 | . 300Hz>------> | . | 4017 | . 25500 ----------- . = ----- ring counter . 85 . . .

Actually if you simply toggle the clock output for every interrupt at

25500Hz, that gives you 12750Hz, and the ring counter partitions the LED anode drives at 33% anyway. This makes things easier for the LEDs, because you have to drive them with 3x the current, or more, for same brightness as steady state, and the 12750Hz higher frequency means a pulse duration of only 78us at this elevated instantaneous power dissipation versus 3.3ms with the 300Hz. Transistor switching losses increase, but these can be made negligible at that frequency. View in a fixed-width font such as Courier.

. . V+ . | . .-------+------. . | | | . | | | . __ | | | . R>---------------| \\ | | |< . | o-----|-------|----| . .---------|__/ | | |\\ . | | | | . | | | | . 100Hz | __ | | | . G>-----|---------| \\ | |< | . update | | o-----|-----| | . | .------|__/ | |\\ | . | | | | | . | | | | | . | | __ | | | . B>-----|--|------| \\ |< | | . | | | o---| | | . | | .---|__/ |\\ | | . | | | | | | . | | | | | | . | | | B-anode G-anode R-anode . | | | . | | | . ----------- . | Q0 Q1 Q2 | . 12750Hz>---> | . | 4017 | . ----------- . ring counter . . . .

Try this... it's not PWM but you get the same effect with a much lower clock rate...

At each interrupt, ADD the brightness value to a counter. Set the led output to the CARRY. I.e (in pseudocode):

interrupt_handler: mov #0, r0 add red_bright, red_count setc 1, r0 add green_bright, green_count setc 2, r0 add blue_bright, blue_count setc 4, r0 mov r0, led_port iret

Note that this is the same algorithm as used to draw diagonal lines in pixels (Bresnehan's Algorithm) - it distributes the "on" bits evently throughout the cycle while maintaining the desired number of them.

So, your effective resolution is determined by the ratio of your interrupt frequency to 60 Hz - a 240 Hz interrupt is two bits, etc. However, your counters can still be 8 or 16 bits - whatever your chip is fastest at adding.

The extra bits in your counter cause the brightness to vary slightly from cycle to cycle, with the average brightness being as accurate as your counter.

I've got an R8C module that has 128 LEDs (half red, half green) that I do this with - each LED has an 8-bit brightness and I dither them all (8 rows, 16 columns) fast enough to have no flicker:

Thanks for your reply. I'll give it a try.

Regards, Mike

Thanks for your reply & also the URL. I'll give it a go.

Regards, Mike