Stellarcore

I recently saw a cool project that a company is making: http://www.becausewecan.org/LED_Coffee_Table_The_Wave

I'm moving out next week and will be needing a coffee table, so I thought it might be cool to make something like the table in the link.

I've been planning a bit and think I can do a decent job. Unlike the table in the link, I dont think its neccessary to make a huge PCB the size of the entire table (must cost them a fortune!). Instead, I'm opting for several smaller pcb's (that I can make in the free version of eagle) with flexible conduit that runs the LEDs and phototransistors under the table.

My project would be controlled digitally by a PIC. I've got a bunch of I2C 16 bit I/O expanders from Microchip that I'd like to use. My design would split the entire project into 8 PCBs, one master board (with the PIC and power circuitry etc..) and 7 slave boards. Each slave board would be identical to the master except certain parts will not be populated. Each board will have five 16 bit I/O expanders. Four I/O expanders will control four rows of LEDs and two rows of phototransistors would be polled by the fifth I/O expander.

Because I can only put eight I/O expanders on a single I2C bus, I will run need nine I2C busses between the boards, as well as power, ground, and a PWM signal that will be used to control the brightness of all the LEDs. All 8 of the phototransistor expanders will be on one bus, and the the LED expanders will share a bus between two boards. Each slave board would have some jumpers on it so that they can be addressed differently.

Because my project would have up to 512 white LEDs rated at 25mA each, I've been trying to think of a way to supply enough current that I can turn all the LEDs on at a time if I wish (the advantage of it being digitally controlled will be that I could use the whole thing as a giant sign board if I wanted or play neat animations, etc...). The total current needed (plus any current dissipated by the buttloads of switching FETs that I'll need) seems to be around 512*25ma = 12.8 Amps @ 5 Volts.

Does anyobdy have any idea what I could use to provide that much current that would be relatively quiet? Also, what gauge of wire do I need to run that much current between the boards? How big should my power traces be?

things

i saw some miniturised versions of those at a science museum i went to. looks like a cool project! as for your power supply, maybe you could use a computer power supply, they are capable of something like 8 or so amps?

__________________
The 3 laws of science....
1. If it smells bad, It's chemistry.
2. If it's mushy, It's biology.
3. If it doesn't work, It's physics.
www.laserpointerforums.com A forum for disscusion of lasers and laser shows!
www.laserchat.org IRC chat for laser and electronic related stuff.

Andy1845c

You could lower the total current a little by not running your leds at the 25ma rated current. How do they look at 20ma or even 15ma? I have found sometimes the human eye can't detect much differance.

Very cool project though! Be sure and post some pictures here when you finish it

I'd love to have somthing like that in our living room, but I doubt the wife would approve!

things

just make it neat, then the wife will approve

__________________
The 3 laws of science....
1. If it smells bad, It's chemistry.
2. If it's mushy, It's biology.
3. If it doesn't work, It's physics.
www.laserpointerforums.com A forum for disscusion of lasers and laser shows!
www.laserchat.org IRC chat for laser and electronic related stuff.

pc88

I don't think you'll need that much power. By multiplexing you can easily cut the power requirement by a factor of 10, perhaps even more. Your PIC will be working 10 times harder, but I'm sure it has the spare cycles required to do the job.

Stellarcore

I think that I'll probably end up using a small form factor computer power supply.

Multiplexing is an option, but its just added complexity. Especially if the only reason to do so is to essentially control the brightness. I can achieve that with the PWM common ground.

Also, my large current demand will only be there when ALL 512 LEDs are on at the same time; an event that is most likely never going to happen except maybe in a test pattern.

After discussing the table with my girlfriend, I think this could work out really nicely. We are going to look into getting a nice piece of frosted glass to neatly cover everything up.

Mike, K8LH

Multiplexing often reduces complexity. It doesn't necessarily reduce the power requirement though as it pulses each column or row, depending on how you're scanning the matrix, at a much higher peak current to make up for the reduced duty cycle.

Attached Images

Matrix Display (10 pin design) temp.PNG (37.6 KB, 90 views)

IRL3103

My guess is that they just capacitor couple the photo transistors to the LEDs to light on motion. You should be able to have that and digital.

JeffreyMcGrew

Actually it's 12" x 14" tiles of PCB's. One table is a 2x4, the other a 2x3 grid. They are just butted together very tightly, and use zero-Ohm resistors as jumpers between them. It looks like one huge board, but it's not.

However, getting the custom boards printed wasn't that cheap IIRC. If you e-mail the Evil Mad Scientist folks directly they might be able to send you circut diagrams. And at some point I think they are planning on making a do-it-yourself kit for the tinkering set out there.

It was made just as you describe, I think they even used an GPL circut designing program. It doesn't use photo-transistors actually, for then it doesn't work in the dark. It uses an IR sender and receiver instead to detect when something's over a node. Just might want to think about that.

This one's all analog. But doing it digitally instead would allow for some other fun things. The analog circuit though allows for perfect fades with the lights, there no twitching or flickering at all, which makes for a nice behavior.

These tables used two small off the shelf power supplies. Windell's redesigning them to use a single supply I believe, so if you get in touch with him he can maybe give you some pointers: http://www.evilmadscientist.com/

Hope this helps,

Jeffrey McGrew
Because We Can, LLC

Stellarcore

Hey Jeffrey!

Thanks for the tips, I'm glad you mentioned the details about the IR diodes and detectors.

I'm not sure how mine will look with digital on/off LED behaviour as you describe, but I think it will still be cool. I could always try and PWM the LEDs from 0 duty cycle up to full duty cycle to mimic a smooth fade on. I've tried that before on an LCD backlight which makes for a really nice effect (much like the pulsing sleep light on a macbook pro).

I hope to start actually tinkering on this in the next week or so; things are kinda hectic with the move coming up and all.

Thanks for the feedback! I hope you follow my project as its being done.

Robot builder 101

You may be able to buy a xbox 360 power adapter. They have a thick cord with a max output of 16.5 amps. You would just need to figure out the connections, but should be able to work for you. Try ebay or somewhere where they sell adapters for game consoles.

Sceadwian

__________________

Robot builder 101

Wow. I am speechless. Lol try finding the schematic to that!

Stellarcore

All right, I've got a very preliminary schematic drawn up. The schematic is probably missing quite a few things, but I'd like opinions and/or suggestions.

http://www.stellarcore.com/version1.png

The part that I am mostly unsure about is the big mess of LEDs and nFets in the top right. Once I've got this circuit settled, I'll replicate it for the other three MCP23017s (the fifth MCP23017 is for the IR sensors).

I think the nFet I chose should be good for each individual LED, but I'm not sure if it will be adequate for the BRIGHTNESS_PWM signal. It is rated for 0.68A when pulsed (and if I put 25mA through each of the 16 LEDs that works out to .4A so it should be ok, right?).

I havent chosen values for the resistors yet, but that will come later. Also do you think the resistors between the gates of the FETs and the MCP23017 neccessary? If I could eliminate those resistor networks I think it would simplify my PCB a bit.

Final question: Is there a suitable chip or something that I can get that has a whole bunch of fets in one package? Again, it would really simplify the PCB.

Thanks for your help, and I'm really looking forward to reading what you all have to say.

pc88

Do you really need a FET for each LED? That's going to be a lot of FETs!

I would first shop around for your LEDS and experiment with how you want to power them. Choosing LEDs is kinda like buying paint - LEDs from different suppliers are going to look different. Test the LEDs at various currents and with different types of frosted glass to see what will work best. That will give you an idea of what your real power requirements are. If, for example, you can get the drive to each LED down to an average of 8 mA, you won't need the FETs.