Help making a high powered RGB Lighting system for my highschool

YAY! Budget increase.

Now I have enough to manufacture my own PCB!

I won't be driving the LEDs in parallel anymore, I'm going to use 12x 500mA drivers at 350mA and then split the PWM signals from the arduino. I'm currently in the process of drawing up the circuit, I'll post it here when I am done.

EDIT: Is it okay to use a 350mA driver to drive 350mA? I know it's always a good idea to underrate electronics but I can't find a chip that does exactly what I need it to for with more current. I need an IC that can handle 24V input, 350mA output with a PWM frequency range that supports both the 500Hz and 1kHz speeds on the arduino pins.

It seems that the driver I listed in my original post is my best bet. Their datasheet seems to point to the fact that it was designed for use at 350mA.

Besides, this unit isn't going to see more than a thousand hours of use anyways. Not that I'd want to make it un-reliable, but if I reduce the IC'd MTBF to 40,000 hours from 80,000 it isn't a huge deal.

Here is a tentative circuit representing one of the six color channels.

Will this work? I'm pretty sure it will but I want to make sure before I get to work designing a PCB.

Thanks.

@neurostar: "In theory, driving power LEDs in parallel is a bad idea." The design being worked from the first page of the thread had current limit resistors in each string, thus they are not actually in parallel.

@charliehorse, This design is total overkil for a limited use stage lighting setup. Running left & right lights with brightness matching controlled by PWM adjust from ardiuno would have been plenty. Logic level mosfets with 0.005 Rds and 350mA going thru them would dissipate <2mW and would have no cooling needs. Now you have a ton more parts to deal with, you've got a pile of NPN transistors that may need cooling as well, dissipating 250mW each, a controller that you are using at 100% capacity, no derating, and with 0.95mm lead spacing which will be a bear to solder hand. http://www.onsemi.com/pub_link/Collateral/CAT4201-D.PDF

If you still want to pursue something besides MOSFET switching, consider one of the parts in this flyer I just received via e-mail notice. http://www.maxim-ic.com/design_guides/en/LED_LIGHTING_1.pdf MAX16822 or MAX16832 for example. Higher current flow, fewer parts, and with a student/school e-mail you are eligible for free samples. www.maxim-ic.com

Perhaps I am not fully understanding the mofset idea...

Looking at the circuit you posted earlier it looks as though when the PWM signal is ON, the LEDs are simply in run directly off 24V with only a resistor to reduce the voltage. If this is true, that resistor could be dissipating up to 10W when the circuit is on.

10W?
P=IV = .35 * (24V-14V) = 3.5W. So use 5W resistors. 10V/.35A = 28 ohm for Red,
(14 v worse case for Red)
4V/.35A = 68 ohm for blue/green. P=4*.35 = 1.4W, use 3W.

Depending on your PWM setting from Arduino, power dissipated will be 1/255 to 255/255 of that.
Or use a high wattage rheostat for even more fine tuning of the light levels
Say a 50 ohm resistor in series with 25 ohm 7.5W rheostat if you can find a place to get them - digikey has the part numbers to look for.
I would go with fixed resistors and PWM trimming, see how it works & add rheostats if its not quite what you want.
http://www.ohmite.com/catalog/pdf/rheostats_wirewound.pdf

Check that part about 50 ohm & 25 ohm rheostat - make the fixed resistor the full value so that if the rheostat is turned to 0 ohm and PWM is at 255 (full current) you don't smoke the LEDs. Just add more resisitance on top to dim the overly bright side. Digikey has 12.5W in 25, 50, 75 ohm. If you use 25 for Red and 75 for blue/green, install on the brighter side LEDs, that will let you tone down the current by half compared to the other side, perception of brightness is another matter.

$26 each, so I would try PWM adjustments before adding these. Just leave yourself room to adjust the circuit later.

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=RES50RE-ND

I have a 10W RGB LED that you are trying to use. These are exact voltage drops of the LED. Red : 6.39V Green : 9.38V Blue : 9.22V All at 350mA

If you want to make your own LED driver, why don't you use my design which is exactly fit with what you want to do. I published all Eagle cad files. http://code.google.com/p/neuroelec/source/browse/#svn%2Ftrunk%2FHP_RGB_LED_Shield If you design switching power first time, there are a lot of things to think about it carefully. The LED driver part of my design will cost less than $30 for 3 channels. You will have very efficient drivers and easy of PWM control that are proven to work. I will send you exact manufacture part numbers and BOM, if you need them. You will only need three drivers to drive 9x 10W RGB LED with 36V power supply such as this. http://www.powersupplydepot.com/productview.asp?product=16034+PS

If you don't want to or don't need to make your own driver, I would shamelessly recommend my RGB LED shield. You can use rest of your budget for led heat sink and housing.

3 x RGB shields : $37 x 3 = $111 + $3 shipping 9 x 10W RGB LED : $11.49 x 9 = $103.41 free shipping 36V/4.2A Power = $40.75 + $11 shipping Total: $268

4 x RGB shields : $37 x 4 = $148 + $3 shipping 8 x 10W RGB LED : $11.49 x 8 = $91.92 free shipping 24V/5A Power = $19.33 free shipping Total: $264

@neurostar: "In theory, driving power LEDs in parallel is a bad idea." The design being worked from the first page of the thread had current limit resistors in each string, thus they are not actually in parallel.

I was talking about his initial idea. Even though it is not ideal, in practice you can drive leds in parallel with same leds which is the case for his setup. I tried, it worked with +-5% current difference though.

Thanks for all the input.

As this is my third term project for my grade 12 I need a little more depth than simply buying a premade solution in order to get a good mark I will be designing my own custom PCB. Since I will be designing the PCB anyways it's better to use an actual driver IC than mofset switching.

That Maxium IC you linked to looks like an excellent choice.

I have no experience soldering chips onto PCBs, I have mainly only used my soldering iron to fix broken cables and connect wiring between the parts in my computer. Is there a service where you can get the PCB prototyped and assembled? I would think there is as it would be difficult if you wanted a complicated board with a lot of small ICs.

OK. that was it. I think you see all these drivers are basically very similar. Since designing is a part of your education, I would recommend chips from national semi or linear tech, simply because you have a tool to simulate your circuit with the chip. You can learn a lot by checking and poking signal here and there. Many of those signal are hard to measure even with good oscilloscope. Once you understand one chip, you will understand rest of them easily. Designing proper value of parts need some calculations. Write down all the equations you see from datasheets. Later you will also find all those equations are same regardless of manufactures. Start from a reference design such as evaluation kit of a chip. PCB design of switching power is also very important. Careful on the SW - diode - GND (input capacitor) which need to be short length. You want to have big GND pour bottom of those switching circuit to minimize EMI. Thermal vias around the chip is also critical. If you don't pay for your PCB, 4 layer PCB is also good idea. For current sensing resistor, check Kelvin trace. That's what I can think of right now. I hope that helps little.

@charliehorse55, Yes there are PCB prototype and assembly houses. I don't know if that would bust your budget. You are in the US? You could make a call to www.moderndevice.com in RI, see if they'd do it for you. I don't have any means to do surface mount either, I use thruhole for what I do and struggle hand-soldering the occasional surface mount part on a DIP adapter (actually have my wife do it, she has better eyes than I for the really small stuff). 0.05" spacing was not too hard to do. 0.025" (4 pins in normal DIP space of 0.1"!) was very difficult.

SMD soldering is not difficult at all, when you have a proper PCB footprint. All you need is good magnifier. You will be amazed with how good your hands are under magnifier or low-power microscope. Instead of making one big PCB for four drivers, make four pieces. Order 5-6 PCBs, there are almost no price differences. You have something to practice. If you are still not confident with SMD soldering, send them to me as long as you are in US. I'll reflow them for free.

me too? I just got a couple of MAX6921AUI with 0.025" spacing, thought I was getting MAX6921AWI with 0.05" spacing. Need to get an adapter too. Waiting for this to come from Thailand, http://cgi.ebay.com/SMD-CONVERTER-ADAPTER-PCB-SOT-TO-MSOP-SIP-DIP-14_W0QQitemZ120628243230QQcategoryZ36327QQcmdZViewItemQQ_trksidZp3286.m7QQ_trkparmsZalgo%3DLVI%26itu%3DUCI%26otn%3D2%26po%3DLVI%26ps%3D63%26clkid%3D7581984690700539118 not seeing a 28 pin adapter on it.

Why not.. Once you get them, send me email. sooner is better.

Cool, thanks neurostar. Trying to find some adapter boards that aren't too expensive yet at the same time not wait a month for delivery from Thailand.

neurostar: OK. that was it. I think you see all these drivers are basically very similar. Since designing is a part of your education, I would recommend chips from national semi or linear tech, simply because you have a tool to simulate your circuit with the chip. You can learn a lot by checking and poking signal here and there. Many of those signal are hard to measure even with good oscilloscope. Once you understand one chip, you will understand rest of them easily. Designing proper value of parts need some calculations. Write down all the equations you see from datasheets. Later you will also find all those equations are same regardless of manufactures. Start from a reference design such as evaluation kit of a chip. PCB design of switching power is also very important. Careful on the SW - diode - GND (input capacitor) which need to be short length. You want to have big GND pour bottom of those switching circuit to minimize EMI. Thermal vias around the chip is also critical. If you don't pay for your PCB, 4 layer PCB is also good idea. For current sensing resistor, check Kelvin trace. That's what I can think of right now. I hope that helps little.

That's what I've been doing. Haven't been using the simulators though, just been using the reference PCB design and modifying it to my needs. I like you idea of 6 seperate PCBs, one for each PWM channel. It seems like a much nicer and neater idea.

I live in Canada but I think I will just use a PCB assembly service. While I want to keep costs down, the school recognizes that the professional equivalent of this lighting system would cost a lot of money so as long as I justify all of the costs there won't be a problem. A moderate budgetary increase to go with a PCB assembly service is easily worth saving the potential failure of the project if I have trouble with the soldering. Additionally, even if I did manage to solder the chips I am leaving to U of T Computer engineering next year and thus I won't be able to fix the board if a mistake in my soldering causes a short. Finally, if the school wants to expand the LED array it would be very easy for them to simply order another copy of my design from the supplier.

Any suggestions for a good Canadian PCB assembler?

Try posting the question as a new thread, maybe a fellow Canadian can offer a suggestion; its bound to me missed buried 3 pages down in this thread. Also try asking the folks at dipmicro.com, or solarbotics.com, see if they can do it or offer suggestions.

Switching plans again. I found this beauty of an LED driver:

http://www.newark.com/linear-technology/lt3496ife-pbf/ic-led-drvr-tssop28/dp/07P8434

Three individual 750mA PWM outputs.

I will construct one PCB per light, and then use a terminal block to split the PWM signal from the Arduino. That way, If the lighting array needs to be expanded you can simply add another board and wire it up.

I'll also be using 30W RGB LEDs, they take 28-33V in, so I will use a +36V PSU.

Thoughts?

The LEDs use 33V max, so you plan to use 36V? Seems like planned overstress to me. Something has to dissipate that extra voltage.

Pay attention on the recommendations for the unused channels also (for your future expansion).

Watch this in your PCB layout: "The exposed pad on the bottom of the package must be soldered to a ground plane. This ground should then be connected to an internal copper ground plane with thermal vias placed directly under the package to spread out the heat dissipated by the LT3496." "The exposed pad of the package is the only GND terminal of the IC and is important for thermal management of the IC. Therefore, it is crucial to achieve a good electrical and thermal contact between the exposed pad and the ground plane of the board."

charliehorse55: Switching plans again. I found this beauty of an LED driver:

http://www.newark.com/linear-technology/lt3496ife-pbf/ic-led-drvr-tssop28/dp/07P8434

Three individual 750mA PWM outputs.

I will construct one PCB per light, and then use a terminal block to split the PWM signal from the Arduino. That way, If the lighting array needs to be expanded you can simply add another board and wire it up.

I'll also be using 30W RGB LEDs, they take 28-33V in, so I will use a +36V PSU.

Thoughts?

Obviously, you didn't read what I wrote carefully. LT3496 is the chip that I used in the high power RGB LED shield. Again, I already published the design of circuit. You can just use driver part of the shield. http://arduino.cc/forum/index.php/topic,51887.0.html. You can not hand-solder LT3496, it has to be reflow due to GND. You have to find someone can do it for you. If you make only three of the driver, total cost of whole thing (PCB, parts, and assembly) will be likely higher than cost of three shields.