max7219 cascade

Other Hardware Displays
1

:o Been wanting to cascade some MAX7219 matrices for a while and as a result have bought quite a few kits etc off the internet.

I have not had problems with coding or connecting but with some shortcomings of the various kits.

Firstly I need it so I can easily have say 52 or 83 etc which means the schematic should fit in a slightly smaller than 30mm * 30mm square.

The 30mm * 30mm thus makes everything a tight fit!

Most if not all kits like this use the smaller SIOC SMD MAX7219 which I had problems soldering. I would also like it so I can replace dud chips easily so are using the DIP24 version. The matrix is using a socket so I can swap/replace the LED matrix.

The clock I'm making will have 20+ so I want to be able to easily replace or fix individual units if they fail.

I have an initial schematic and board and due to chip sizes etc some components will be placed on the reverse side. The DIP24 with the socket will be on the bottom as will the connectors for the IN and OUT wires. (VCC GND CS SCK MOSI)

At the moment I have also opted for 23 connectors (with 1 nc) instead of the 51 as I found from practice trying to join OUT->IN when too close is horrible when boards are not perfect. Much easier to have them apart a bit which puts less strain on the cable.

At the moment I have 2 SMD parts but may make through hole to make it more hobby friendly (which is important for me).

I envision making the outline of the board corner rounded and slightly smaller than the 30mm * 30mm as the final placement should be determined by the matrix and NOT the board. I will also change the current design to have a fastening hole more in the middle of each matrix instead of 4 corners.

Have attached the current board layout that has pending changes but shows it's actually possible to mount everything in such a frugal space.

7219.PNG
7219.PNG638×643 57.7 KB

2

Sounds like a whole lot of work just in case of 'dud' chips. There already are footprint sized boards that use the SOIC so I'm not sure what soldering you have to do with chips. Power/SPI connections are all that need soldering and if you use headers, you can just use jumpers and whole modules will be easy to swap out should either matrix or board fail.

3

INTP:
Sounds like a whole lot of work just in case of 'dud' chips. There already are footprint sized boards that use the SOIC so I'm not sure what soldering you have to do with chips. Power/SPI connections are all that need soldering and if you use headers, you can just use jumpers and whole modules will be easy to swap out should either matrix or board fail.

I've also had problems linking them cleanly as the headers are out just a bit and causes the whole base to be under stress.

I will 1st test the single matrix and if it's ok make similar 2x1 3x1 and 4x1 pre-joined modules which will keep external wiring down and keep longer runs straight.

Also, I have noticed there is no provision for a pullup on the CS lines which is usually desirable at startup. Might also allow for an optional pullup on the CS line.

A lot of the kits provide a means of joining with jumpers which seems fine in theory but in reality board imperfections means a small error in each keeps adding up.

4

I know what you mean. But put on a respirator and take the board to a grinding wheel and it'll work out just fine.

5

OK - got a 4*1 design I will get fabricated and test.

Got a lot of 88 LED matrices to test and they seem quite well made and very little variation in the 32mm32mm claim. Unlike all the other modules on EBay etc the dimension of the actual board will be a bit smaller than 32mm*32mm to ensure the PCB never stops the LED matrices joining.

When spacing the 4 matrices on the PCB there is a "small tolerance" between them (too large and will be a noticeable gap, too small and ... won't be able to make them fit together).

If all goes well will make a 42 and a 44 array of MAX7219's which should make connecting multi-panel segments with more than 1 row a lot easier and cheaper.

These arrays will cut down on joiner cables and pin headers required when people want to join many. Also it means no clamps, glue, bolts, jumpers etc to try and get a heap of smaller ones joined. (a 4*4 will mean 16 are joined on a single pcb).

So, potentially a 124 array of matrices giving 9632 pixels will mean joining 4 4*4 panels instead of 48 individual modules.

The design has it so the LED module can be clicked in and the MAX7219 is the DIP version which can also either be soldered in place or inserted into a socket. Among other things, this ensures faulty ones can be replaced without soldering or the parts re-used as required. (Means a 44 panel no longer used can have all the MAX7219 chips reclaimed along with the LED 88 matrices).

I was also looking at 4 layer boards but will go back to that later if the 2 layer works well. The modules have a potential to need a nice power supply so was looking at 4 layer to ensure health power/GND and signal layers over longer runs.

The other thing I have looked into is the linking between adjacent pcb's. Use a standard 5 pin male header then use a 10cm female/female dupont cable to connect. If the headers to be joined are too close or too far pressure and strain is put on the headers. For a 10cm cable a "nice" distance is just a bit more than a single module (32mm).

All on a single PCB will have a common GND, VCC and SCK which will mean if chaining to the same module only the CS/DOUT need to be linked to CS/DIN.

Each initial row will have an optional jumper and 10K resistor that can be used as a CS line pullup If wanted.