High current shift registers for led matrix

[midgetfidget]

http://www.sparkfun.com/commerce/product_info.php?products_id=734

The above link was posted somewhere here a while ago.

I'm planning a 16x16 led matrix and i've read about the current limit of the standard DOUT shift registers.

This chip can supply 150ma constant per channel. So if can play around and get leds that draw a little under 10ma it should be fine to run 16 x 16 I guess?

I'm looking for opinions on whether this would be the best way to do this?

Would this chip be a direct replacement for the standard chip?

[pay_c]

I´m doing stuff with LED matrices, too. And I do have exactly the same probs as you.

First of all: Yes, that *would* work. But as you have 16 multiplexing steps you should load one LED with appr. 160 mA (or let´s say 100 mA) so it is not loosing to much light due to multiplexing 16 times. LED´s can be loaded with up to appr. 250 mA of *pulsed* (and only *pulsed*) current. So you need at least appr. 1,6 A of current (this IS much).

First solution: low current LEDs running at 3 mA (nearly a must for 256 LEDs). There you only need 48 mA for one LED at one multiplex step. That´s 800 mA in the whole if all LEDs are lit.

This also is ok for your IC. Unfortunately it´s not pin-compatible with the 595. You can easily see that in the datasheet. Also possible: Using a 74HC595 plus either transistor driver array or an driver IC as the ULN280x and UDN2981. Both solutions include much layouting work (I tried both). But it´s possible. I would try re-layouting the DOUTx4 with your IC, seems to be the most easy solution.

Anyhow: Say goodbye to an "easy" solution with a pin-compatible IC. I searched my a** off and didn´t find anything.

Greetz!

[midgetfidget]

Thanks for your input, i'm still really novice at this electronics thing.

The transistor array thing you're talking about would that be along the lines of suckow/monolake rgb circuit?  (http://www.suckow.de/ralf/ledmatrix4x20/index.html)

I'm thinking maybe use more DOUTS and break up the matrix into smaller sections like 4 times 8x8. The leds have to be reasonably bright to go through the silicon pads.

[pay_c]

Monolake did use the ULN solution for his stuff. Check out the schematic.

Here is some example for the transistor solution: http://www.mikrocontroller.net/wikifiles/6/64/LED_Matrix_8x5.png (only using npn types e.g. BC337 or BC547)

So: IC solutions are normally a little more area-efficient to say it in a not to clear way. *Sooooometimes* you are off better with the transistor solution, but in 80% the ULN / UDN solution is far more easy.

The breaking up is also a solution, right. You would have 8 shift registers instead of 4 but whatever, it´s the same chip count I guess. :)

Good luck and greetz! :)

[TK.]

Have a look into Wilba's schematics for his MB-6582 (-> http://www.midibox.org/dokuwiki/doku.php?id=wilba_mb_6582&s=6582), , he uses transistors as current sink. This only requires a small change in the firmware (outputs connected to transistors need to be inverted)

Best Regards, Thorsten.