>
> > > I am considering building a small LED matrix sign, probably
> > > using 12 stackable 8x8 LED matrix blocks (approx 2"x2").
>
> > >  This
> > > could be done with 96 high side drivers sourcing 20 ma each,
> > > 96 current limiting resisters, and 8 low side drivers sinking
> > > about 2 amps each (fairly hefty).  Or reverse this, and have 96
> > > low side drivers (at this current, could be PICs directly) and 8 x
> > > 2 amp high side drivers.
> >
> > Wait I'm missing something. Aren't these displays internally
> > multiplexed?
>
> No, I believe that the 64 LED's in each of the twelve blocks are
> arranged in an 8x8 matrix.  16 pins.  8 are cathodes for columns,
> 8 are anodes for rows (also available vice versa).  To me
> the obvious way to scan them is to power one anode (of 8) at a
> time, while driving the appropriate subset of cathodes (depending
> one which LEDs in this row should be lit currently).

What you just described is internally multiplexed. A non multiplexed display
would have 64 anodes and 64 cathodes - a packaging nightmare.

I usually do the opposite - drive one cathode and control the anodes.

>
> For 96 wide by 8 high (27" x 2", twelve x8x blocks), this means 96
> individually controlled column drivers, and 8 massive row drivers.

You're still thinking in terms of driving an entire 96 element row. It's
not necessary I believe but of course the biggest display I've built only
had 40 columns. I drove one column at a time (max 7 LED on at a time using
5x7 displays). Looked great.

>
> There have been some great leads regarding low side drivers (for the
> 96 sinks).  What about the 8 high side (in this regard) high current
> drivers, many amps each?  I now read that the peak current for some
> of these LED blocks (per LED, 1/10 or less duty cycle) can be 80-100
> ma!  (by 8, would be 640-800 ma per block, via the row anode, times
> twelve is up to 10 amps!).  I don't have to have absolute maximum
> brightness, so I could compromise this to much less current.  But it
> seems that I still need 8 high current transistors for the
> rows/anodes.

Once again think smaller and faster. Like the post yesterday on the TI
TPIC6B595. This part has a shift register and latch with 8 low side drivers
that can sink 150ma continous, 500ma peak/pulsed. All in a 16 pin DIP.
Cascadeable too.  You'd need 12 of them and pulse one at a time (simply a
shift and latch).  Then you can drive all of the anodes with 8 transistors
2N2222 and 100 ohm current limiting resistors. The display will work fine
at 1/96 duty cycle. The 1/8 cycle isn't important and designing with
multiamp circuitry to get it is tough work. If it turns out you need
a faster duty cycle, then just find a duty cycle that works for you then
duplicate the anode circuitry.

Find a solution that works for one column (the sort side) then extend. Once
one column works, then you can have as many columns as you like.

Also consider how long it going to take the PIC to load up 96 bits of data
for each row. For one column all you need is a single 8 bit latch and 4-5
instructions to load it. 96 bits will requre either 96 bits shifted in or
require 12 8 bit latches. By working with one column at a time, you reduce
your data load capacity and time.

>
> I have read that logic gate power MOSFETs when used high side
> (between power and the LED anodes) take several volts (eg: 5)
> above the power supply to switch!  That makes it hard to simply
> interface them (as high side drivers) to PIC outputs.

Exactly. And you only need to do that for multiamp solutions. A one column
at a time solution will be less that 400 ma tops and that's pulsed.

BTW what are the peak/pulsed currents on the LEDs? And the duty cycle? That
should help determine a solution...

>
> How would I use a PIC output to switch approx 5 volts at 2-10 A
> (most to least compromise) into each of the 8 anode/rows?  Only
> one (or none) at a time is needed.  At LED multiplexing speeds;
> no relays please :-)
>
> I guess this also brings up the question of appropriate multiplexing
> speeds.  Would a 16 msec cycle, with 2 msec per row, be a good
> frequency?  Or does it need to be faster to avoid visible flicker?

I find that flicker is free at about 500 Hz. So a 96 column display would
need to be driven at about 48 khZ to be effective.

>
> Thanks for the help!

Thnk about 1 column at a time. You can control 96 columns using only 3 bits
of the PIC and a single shift an load for each column.

BTW Marshall Electronics has samples of the TI TPIC6B595. I plan to test
my next display with them.

BAJ