I want to make this working now.

A 768 LEDs matrix!

I use 6x multiplex and serial buffers.
To save a little space, 4x 74hc164 are used for 32 LEDs each, and 1x 74hc595 to drive MOSFET sinks.

4x 16f1824 controllers are used because I want to implement 16 brightness levels for each LEDs.

1x PIC32 is used to generate the bitmaps. It is not even wired up yet.

I did not draw a schematic...

So it is a bit messy to complete the circuit. I figured out some connections are left out for the 3rd panel, for instance between the serial buffers. The MOSFETs are not wired as well.

Recently I fixed some broken LEDs for the first two panels.

All 4 controllers are flashed with a test firmware. I have not yet debugged the brightness control, only worked out the bitmap pattern.

Each data line has a 22uH damping coil at it's end- otherwise the data does not arrive (I use full speed).

Maybe I publish the firmware for you to examine and to help out. It does not look so nice.

Power supply is 3.3V from a 12V electronic transformer, so I don't need 128 LED resistors. Made a big cooler for the LM2576 chip but right now it is not really neccessary.

This was some months ago, rather complicate way of mounting. It is glued with epoxy on the bottom of the VGA cooler and mounted upside down.

 

nice one i will love to listen more.............

 

You are a master of patience and tenacity!

 

So if I've understood this correctly:

There are 4 separate panels each driven by a PIC16f1824, with 192 LEDs on each, wired as a 32x6 array. You use 4 74HC164 shift registers to drive the anodes, and MOSFETs gated by a 74HC595 to drive the cathodes. It runs on 3.3v and there are no current-limiting resistors.

It sounds workable but I'm dubious about the lack of resistors. What really does set the current through the LEDs? And it seems as though you must get transitions on the 74HC164 outptuts as data is shifted through. Why not use 74HC595 chips for all the shift registers? Then you could load up the data on all 5 chips and hit a strobe line to set all the outputs simultaneously.

 

So if I've understood this correctly:

There are 4 separate panels each driven by a PIC16f1824, with 192 LEDs on each, wired as a 32x6 array. You use 4 74HC164 shift registers to drive the anodes, and MOSFETs gated by a 74HC595 to drive the cathodes. It runs on 3.3v and there are no current-limiting resistors.

It sounds workable but I'm dubious about the lack of resistors. What really does set the current through the LEDs? And it seems as though you must get transitions on the 74HC164 outptuts as data is shifted through. Why not use 74HC595 chips for all the shift registers? Then you could load up the data on all 5 chips and hit a strobe line to set all the outputs simultaneously.

I use hc164 since they are a little smaller ICs. It is not a problem, since the sink drive is floated while updating.

The effective resistance of each 74xx I/O is about 150 Ohms. Have not observed any problems so far.

I am just busy with other things, but very close to make the 3rd panel working.

Yes you are correct about the way it is built.

I want 16 brightness levels, and the software does not look nice even in C, since the PICs memory is small, only 4bit are used.

It is not complete now, but 100% ON pattern works correctly.