24x12 matrix bicolor LEDs

[Ttelmah]

It'll be interesting to hear what the next version is. As it currently stands he seems to just be talking about turning all the LED's on together. Easiest way, forget the PIC entirely, have the LED's connected as blocks of something like 16 in series, with a simple constant current driver, and just switch this on/off. Something like four transistors for every 64 LED's, and running off perhaps 40v, the current will be nice and low (<100mA) for each block.

[tEaM]

Thank you all for the help.
The project is well defined from the beginning! You associated to graphics displays, but no. The project is an indicator LED panel.

See the old project:



(22 column LEDs x 12 row LEDs ) x 2 painels

The old project was not made by me. He had 528 red LEDs and had a TPIC6B595 of 8 LEDs.

Now I'm going to do a project with 1152 bicolor LEDs (red/green).
(24 column LEDs x 12 row LEDs ) x 4 painels
Each LED will change color individually.

[VernonAMiller]

At this point, it doesn't really matter how many LED rows/columns/panels are being driven or what they are being used for. There's a lot of good information and ideas in this thread giving several approaches to think about, as well as cautions relating to current draw and chip package limitations and so on. It's the same stuff regardless of whether you are driving 8 LEDs or 8000 - only the scale changes.

That's my opinion anyway.

VAM

[Jerry I]

From your picture, it appears to be an indicator panel.

Most of the designs that have been discussed are based on some type of message/graphic board. I don't think it has to be so sophisticated. On the indicator panel, what are below the LED's, are they switches?.

What data and from what, is fed to the panel to indicate status??.

What event will change the data on the LED from RED to GREEN or vice versa.

The easy way to do this may be just use the WS2812 like Ttelmah suggested, connected all in a string then load data for required colour. Its a lot of data but who cares how long it takes. It would still be < 1sec.
This way no extra drivers are required. Just power up the WS2812 LED, send data the output of the first LED will feed the next LED and so on.

[Mike Walne]

[tEaM]

This panel has 1152 bicolor LEDs panel and 1152 switch panel.

The LEDs have two functions:
- Flash to indicate that the switch to turn on / off;
- Green / red to indicate the status of switch.

This is controlled by PIC that will communicate with the PC via TCP / IP.

The problem of using WS2812 is that this panel has 3mm holes for standard LEDs panel.

[temtronic]

Since you supplied more info and I hab a kold...I've been researching this project a bit...

...found an interesting solution using the LT8500. It has 48 PWM outputs. So it will directly control a 'row' of 24 bicolor LEDs.Simply pair two outputs to each bicolor LED
You'll need 24 pieces to do the entire display panel.
Simple SPI interface so only one PIC needed for the project. However you'll need to have an SPI buffer to get the data all the way to the last LT8500.
If you choose a PIC with 2 hardware SPI peripherals, it'll be a bit easier as one does say the 'left' side, the other the 'right' side of the panel.

The 'fun' or challenge' is to design the PCB for each LT8500 and 24 LED connections. Also 'do the math' and get a BIG power supply capable of turning all LEDs on.

just another way to do it....

Jay

[ezflyr]

OK, so we are now three pages into this thread, and there still isn't a shred of hardware or code???

John

[tEaM]

LT8500 with 48 PWM outputs is quite interesting to control the brightness of the LEDs. But seems to be too "complex" to just control on / off.

I was thinking something simple. I like TPIC6B595. But instead of directly connect one output to each LED, created a matrix 8x8.

@ ezflyr
Before any scheme or code, it is interesting hear the opinions of others.

[temtronic]

Actually the LT8500 is super easy to setup and control.set a '00' and the LED is off, 'FF' and it's ON, NO different than sending a '0' or '1' to a 595.
The key to using the chip is that it is designed for this purpose and will handle the 24 LEDs with ease. The 595 approach requires a lot of math and additional components. Also you need 2 595s(and more parts) to replace just 1 8500.
The wiring is super easy with the 8500. two wires to an LED.Board layout is equally easy 1 part and LED pads to layout.Given today's PCB programs, I's suspect the modules could be created in an hour or so. Actual wiring maybe another hour per group of 24 LEDs.
I don't know if anyone else is still interested in this project but it could be 'up and running' in as little as 3 days once PCBs are made.

If there's a 'better' solution I'd like to see it offered.

hth
jay

[VernonAMiller]

The problem with learning about a cool chip like the LT8500, is that now I have YET ANOTHER project idea in my head! And I already have more than I can do!

[tEaM]

LT8500 is very nice. But for 1152 bicolor led, I need many chips.
I liked the WS2812 solution. I've never worked with this RGB LED, but I'm interested in testing.
And to enjoy the hole 3mm, can I use a light pipe:

[temtronic]

Before you commit to 1152 ws2812 get a few and see the 'fun' at getting them to work !! It's unlikely that you'll get a string of 1152 of them to actually work due to the timing requirements. You haven't said how big the display board is but you MUST calculate whether the WS2812 signaling can travel the distance. The display looks like it's 6 feet wide and 4 feet tall, so from top left to bottom right it's 72 linear feet in distance ! My gut feeling says it's NOT possible to do, others might chime in if they've used them, I do know timing is very, very critical !
My last post was wrong, you'll need SIX 595's to control the 24 RG LEDs not just 2. The LT8500 is about $5. Actually a great deal since it converts easy SPI code into PWM LED. Overall probably the cheapest, easiest method to control 1152 RG LEDs on the display board.
Frankly if a client gave this project to me, I'd use the LT8500 (or similar), spend 1 day designing PCB (or find a 'breakout board for it), day two - cut test code, day 3 - assemble test PCBs and test, day 4 - assemble 11 other boards into 'display unit', day 5 - revise 'test' code and do a 24hr burn-in.
Since 99% of the heard work is DONE by the LT8500, the PIC code is simple....select a 'pixel pattern', send via SPI to LT8500 units..done! For test purposes almost ANY PIC with a hardware SPI peripheral could be used, though I do like the 18F46K22 these days. Once the LT8500 'driver' is created, it's simple to design the 'front end' code to decide which LEDs should be on, off, flash.

hth
jay

[Ttelmah]

I must admit I wouldn't cascade 1152 ws8212's.

The way these work, is they are SPI, but are sequential and 'cascade forwards'. Each needs 24bits of data, so a 'line' of (say) 128, would require 384bytes (3072bits) to drive!. They do contain SPI amplifiers, so the line can go on 'forever', but the update rate starts to get a bit silly!.... The largest module they offer using these is a 32*32 matrix, which requires over 24000 bits (3072bytes) for an update.

Looking at the design, consider a 'line'. A line across half the panel is 24 LED's or one across the double matrix is 48. Both relatively easy. Remember that SPI can be generated on any pair of PIC pins (supporting output), so 12 double line modules, could be directly driven by a single PIC. Also consider looking at chips like the HT1632C. This multiplex drives 24*16 LED's. Requires 10 external transistors, for high power LED's, but for the currents being talked about here, can direct drive (row resistors only needed). Has an SPI interface (with a CS, so having several driven from one PIC can easily be done). With your 'double' LED's, you could use a single one to drive 24*8 lines. Just three would do the entire module!.... Annoyingly the internal 'flash' capability is for the entire module, so flashing a single LED would have to be done in software.

[Jerry I]

Are the switches below the LED also going to be sampled by the same LED controller?.