Economical way to drive a 16x16 LED matrix

[BoFH]

I'm thinking about building a large LED screen out of 16x16 LED panels. I'm trying to find out if I can do 240x96 LEDs (15x6 panels) reasonably economically. I have sourced the LEDs but am trying to find the driving circuitry without breaking the bank. I'm thinking of using shift registers. I'm thinking of running the LEDs at 20mA on 1/16 duty cycle. I'm therefore looking for a column shift register that can source 20mA per output and 320mA in total. Also something to sink the row at 320mA a pin and 320mA in total. Does a shift register exist that can sink that much or do I need to use a driver IC? Can anyone recommend which chips to use? Would I be better off sinking the rows with power transistors?

I have taken a look at an instructable that used a HEF4794 and a mic2981 but these become somewhat expensive when you need 180 of them, plus something to scan the rows.

Can anyone suggest a good and economical way to do this?

Regards,
BOfH.

 

[betwixt]

There is an idea I had for something like this, and I have not taken it further so no plans or software exists, but it might be worth following up.

Rather than use high current drivers or sinks to power the LEDs, use cheap SCRs and power the whole thing from a full-wave rectified, but un-filtered supply line.

The advantages are that the flicker rate would be 100Hz or 120Hz, depending on the supply frequency so it wouldn't be noticed, to turn a row or column on you would only need one brief pulse to trigger the SCR so the potential is there to operate a lot more LEDs from a single controller and small SCRs cost no more than other driver circuitry. It might be possible to dispense with multiplexing and drive each LED individually so graphics are possible as well as text.

The drawbacks I see as being strobing and syncing the SCR trigger pulse to the rise of the supply so full brightness is achieved.

Just an idea...

Brian.

 

[banjo]

Several years ago, I designed LED signs for a company. We started out using driver ICs from Allegro. When these became long lead time items, we switched to TI parts, the TPIC595. These are a serial shift register followed by open drain outputs. I believe ST Micro also makes a pin compatible version. The MOSFETs in the ST parts cannot sink as much current though.

The SCR idea has some challenges. First, you have to sink to to power line frequency or the brightness will vary. Second the AC input must have a known and controlled peak voltage or the brightness will vary. Third, for very bright displays, we eliminated the current limiting series resistors for the LEDs and just drove them directly. We would typically drive 5V across some LEDs. To do this we had to limit the duty cycle or the LED would burn out. (Within reasonable limits, the important parameter is the average LED current, not the peak current.) The SCR method seems to imply the duty cycle would be either 0% or 50%.

 

[betwixt]

You are quite right, that's why it's only reached the idea stage!

Some time ago when living in the US I purchased a domestic clock radio which worked perfectly well from 115V/60Hz. When I moved to the UK I tried running it through a 240/115 transformer to see if it used an internal oscillator or was mains locked which would be seen as it running slow. I was quite surprised to see that not only did it run slow but the hours and minutes flashed alternately, something I had not seen before.

Being curious, I opened it up and traced the clock schematic. The way it multiplexed the LED display was to use a center tapped mains transformer with a rectifier in each end, one going to the hours digits, the other to the minutes digits. It was a neat trick to save on multiplex drivers which I thought might be adaptable to larger displays. the drawback as I discovered was that with lower AC frequencies, the flicker became obvious.

Brian.

 

[funky2x]

do you guys suggest replacement for the HEF4794 and mic(UDN)2981 for the original project 8x8 led ?