I am building an 8x8x8 LED cube and have already made the cube. I have 8 cathode layers and 64 anodes for the columns.

The trouble is I cannot find any shift registers that have constant current that can source. All the ones I find are current sink. This obviously will not work for my cube.

The only one I can find that will source current is a Toshiba TB62708N, but this has 1.77mm pin spacing, I really need a DIP package.

I can make do with 8 bit or 16 bit shift registers.

Anyone know of any current source ones available?

 

OK, I've kind of sorted it now. I'm going to get 8 x MIC581YN chips. These are 8 bit serial latch registers capable of sourcing 500mA. I will need to use resistors on the outputs of these, but as I would have had to use 64 small link wires on the LED board, it is just as easy to put resistors there. And after all, resistors are pretty cheap.

 

Please post the datasheet for the chip.

ak

 

OK, I've kind of sorted it now. I'm going to get 8 x MIC581YN chips. These are 8 bit serial latch registers capable of sourcing 500mA. I will need to use resistors on the outputs of these, but as I would have had to use 64 small link wires on the LED board, it is just as easy to put resistors there. And after all, resistors are pretty cheap.

I'm kind of dubious on your claim. 8 x 500 mA is a great deal of power for a DIP. You need to read the datasheet VERY carefully.
I also get no hits on MIC581YN. You might want to exercise more care in your postings.
Did you mean MIC5891? If you did, then:
Look carefully at the table of allowable duty cycles. It specifies 200 mA and for 8 outputs at 50°C ambient you get 53% duty cycle.
Also look at the power derating chart for the package. I think you will find that you need to temper your expectations.

 

Yeah, sorry, I did mean the MIC5891YN, datasheet:

http://www.micrel.com/_PDF/mic5891.pdf


The setup I have will be fine with this chip. Each layer has 64 LEDs. All the LEDs have the cathodes connected together which will be switched with MOSFETs. So I will have 64 anode columns which will have 27 ohm resistor in series with them. 5v supply - 3.0V (LED forward voltage) gives me 74mA. Those shift registers should be fine.

Wish I had known that nearly all shift register LED driver chips sink current before I built the cube

Anyway, here is the cube so far. Getting the 64 LED legs through 64 1mm holes in the black acrylic was a pain, but I'm very happy with the result so far.

 

I'm surprised someone hasn't come up with a modular approach for the LED matrix. Replacing LEDs that die (e.g. from infant mortality) will be a pain...

 

It was bad enough replacing a couple of LEDs after soldering a layer together (guess the heat killed them), can't imagine trying to change one now that the cube is complete!

I have tested them all and they all work.....at the moment. Just got to make sure the multiplexing code I write is bomb proof as the LEDs are being overdriven because of the multiplexing.