LED matrix with a remote PSU

I am building a 16x11 LED matrix which has 176 cells.

Requirements:

  • The power supply unit must be placed 20~25 meters away from the LED matrix.
  • The size of each cell is 30cm x 30cm. Each cell contains 2~4 WS2812 units.(datasheet)
  • A micro-controller which is placed near the matrix to send signals to the WS2812 LEDs

The challenge is to come up with a proper power distribution design and how to provide a stable ground reference for the LED matrix and the micro-controller.

Here is what I am planning to do, a diagram is attached:

  • Assign one PCB to each cell. The PCB contains a bulk converter(step down to 5V), 2 x WS2812 LED, and two RJ45 connectors to connect each PCB together (see attachment for a schematic of the PCB). The reason to use a buck converter is to eliminate the voltage drop caused by the long power cable as well as to make sure each LED has enough power for full brightness.
  • Each led at full brightness consumes about 0.5W, one row has 16 cells, each row consumes about 0.5w x 2leds/cell x 16 cells = 16W. Adding other factors such as wires and bulk converter. I can assume each row consume about 24V x 1A = 24W
  • At the location of the PSU, which I choose a 24V, 350W unit, split the power into 12 channels: 11 channels to 11 rows and 1 channel to MCU. Each channel is protected by a 2A PTC resettable fuse. I try to avoid high current power wires. This is probably a bad design. It would mess up the ground reference since each row doesn’t consume the same power when the matrix is doing its thing. How should I distribute the power?

PCB.PNG

I think your power estimates are too high, probably closer to double the actual consumption. If you are using 4 ws2812 to each cell, it would be about right I think, and 4 per cell seems like a good idea in such large cells.

Where does requirement #1 come from? It would be more efficient to have a 5V PSU located close to the matrix. Is there water involved, or something?

I would think about maybe one 5V converter per row, not per cell. They are probably less efficient at low currents.


PCB.PNG

PaulRB:
I think your power estimates are too high, probably closer to double the actual consumption. If you are using 4 ws2812 to each cell, it would be about right I think, and 4 per cell seems like a good idea in such large cells.

Where does requirement #1 come from? It would be more efficient to have a 5V PSU located close to the matrix. Is there water involved, or something?

I would think about maybe one 5V converter per row, not per cell. They are probably less efficient at low currents.

It is a requirement to place all of the PSU in one control room. 4 LEDs per cell is a good idea.
Assuming the power efficiency is not the priority, how can I make sure the each row shares the correct ground reference.

Why not simply connect all the ground wires together? And the 24V wires, for that matter.

Also, you could use ws2811 chips, one per cell, with 6 X 6-pin "dumb" RGB leds, and 3 suitable series resistors per cell. This could run directly off 24V, so no 5V converters needed (except one for the Arduino).

Or you could use off-the-shelf 24V addressable strips. These can be cut after every 6th led, allowing you to place one section in each cell.

PaulRB:
Why not simply connect all the ground wires together? And the 24V wires, for that matter.

Can I simply connected 12 channels' 24v and ground together near the matrix and then split the power/ground to the rows? Will it cause any ground loop problem? Looks like there are two star ground points: one at the PSU, and one near the matrix.

PaulRB:
Also, you could use ws2811 chips, one per cell, with 6 X 6-pin "dumb" RGB leds, and 3 suitable series resistors per cell. This could run directly off 24V, so no 5V converters needed (except one for the Arduino).

Or you could use off-the-shelf 24V addressable strips. These can be cut after every 6th led, allowing you to place one section in each cell.

The suggestions do surely make sense. Unfortunately, the requirements cannot be changed.

Well, there are many things going on here, but perhaps the most important thing is that everywhere a data line goes from one WS2812 LED to the next, the ground must also run together with it from one to the next.

This is important at every step, but particularly so where you show the data jumping from the end of one row back into the adjacent one. Because this may mean some significant current can flow in that ground connection because of differing voltage drops along the chains, the ground wire will need to be as heavy as any of the power supply lines.

Paul__B:
Well, there are many things going on here, but perhaps the most important thing is that everywhere a data line goes from one WS2812 LED to the next, the ground must also run together with it from one to the next.

This is important at every step, but particularly so where you show the data jumping from the end of one row back into the adjacent one. Because this may mean some significant current can flow in that ground connection because of differing voltage drops along the chains, the ground wire will need to be as heavy as any of the power supply lines.

After doing more research and adapting Paul__B's suggestions, I drew a new diagram:

New design details:

  • Run 2.5mm^2 wires for 24V and GND from the PSU to the location of the matrix
  • Split the power and ground using ground terminal blocks, the splitter power acts as a power injectors (this is often done when running a really long led strips to prevent voltage drop. To me, it seems to help me with correcting ground reference)
  • These power injections are applied to first column of each row
  • Cells are connected to cat6 cables to transmit 24V, GND and LED signals

Would this design raise any concern?

You have not indicated the strapping of the grounds from the end of one row to the next where the data crosses over where it says "11 rows".

And just as critical, between your "MCU" and the start of the first strip.

Paul__B:
You have not indicated the strapping of the grounds from the end of one row to the next where the data crosses over where it says "11 rows".

The CAT6 cables now run across the rows and transmit 24V, GND, and data. 24V and GND both use 23AWG x 2 inside the cables.
The power injection connectors should connect 24V and GND of each row together.
Does this sound right...? :roll_eyes:

Paul__B:
And just as critical, between your "MCU" and the start of the first strip.

Right, I will modify the connection from the MCU to the first cell so that it uses the CAT6 cable and the same 24V, Gnd, and data connection.

elginian:
The power injection connectors should connect 24V and GND of each row together.
Does this sound right...? :roll_eyes:

Yes, just as long as the grounds are connected together where the data goes from the right hand ("far") end of one row to another.