Hello everyone,

I am building a 300w-400w led system for a salt water aquarium, which will be using 54 10w leds and 3 sets of 18 LM317 each, for multiple current regulation.

In page 9 of the datasheet (http://www.ti.com/lit/ds/symlink/lm317.pdf) a single control for multiple LM317 is shown.

Does this apply to current regulation also?

In the case of the full power configuration, I can have up to 12.5 amp (31-32V) current handled by the 18 LM317, so the one point for regulation should handle at least 16 amp.

Another possibility is for example 3 points of regulation, 5 amps each...

Does this make sense?

Thanks.

 

hi dm,
The circuit on page #9 of the d/s is for setting the voltage output of the multiple LM317's, it will not give any current control.

E

 

For brightness control, you would be better off using PWM. You could use a MOSFET to control each string of LEDs and drive them all with the same PWM signal.

Bob

 

I agree. There's too much wattage here to be wasting more by using linear control methods.

 

Maybe using 3 old atx power supply are more easy to reach what you want, in series 3 psu and adjust the voltage one of them to meet 31~32V.

 

As already stated multiple LM317 chips in CC mode is a horrible choice for that much power..
I would highly recommend meanwell LDD drivers or similar with suitable cv power supply.. They can be dimmed with a 5V PWM signal which is fairly common in reef controllers/led controllers now.

Your led specs are what..
1000mA @ 9-12Vf ??

You could get a few 10A+ 48V supplies and put 4 x 10W leds on each meanwell LDD driver..

How many gallons/tank dimensions? What type of coral? SPS or just LPS/softies?
What color temp LED's are you getting..

I build/design/recommend LED systems for reef tanks all the time and have seen it all..

Be aware that the color temp on Chinese 10W LEDs varies GREATLY.. You could order 2 x 10K from the same vendor and get 2 totally different colors... Most of us now are using a mix of cool white...royal blue plus some neutral or warm whites... a few reds/greens.. and some UV to get a more "full spectrum".. Back in the day it was only cool white and royal blue. But a more full spectrum build brings out colors that you didn't get with only cool whites/royal blue combos

 

You could set up a free running 555 timer with adjustable pulse width to drive power FETs (one for each LED). You could also use something like the SG3525 set up with a trimpot to adjust pulse width. Use a power resistor in series with each LED so the FET turns fully on each pulse and the resistor limits the current.

 

For brightness control, you would be better off using PWM. You could use a MOSFET to control each string of LEDs and drive them all with the same PWM signal.

Bob

Thanks, I know PWM with a MOSFET for each string would be better, I'm not there yet, so I have to go with what I know I can do and have been doing.

Maybe using 3 old atx power supply are more easy to reach what you want, in series 3 psu and adjust the voltage one of them to meet 31~32V.

I have a voltage booster with adjustable voltage so I can get the exact voltage into the leds.

As already stated multiple LM317 chips in CC mode is a horrible choice for that much power..
I would highly recommend meanwell LDD drivers or similar with suitable cv power supply.. They can be dimmed with a 5V PWM signal which is fairly common in reef controllers/led controllers now.

Your led specs are what..
1000mA @ 9-12Vf ??

You could get a few 10A+ 48V supplies and put 4 x 10W leds on each meanwell LDD driver..

How many gallons/tank dimensions? What type of coral? SPS or just LPS/softies?
What color temp LED's are you getting..

I build/design/recommend LED systems for reef tanks all the time and have seen it all..

Be aware that the color temp on Chinese 10W LEDs varies GREATLY.. You could order 2 x 10K from the same vendor and get 2 totally different colors... Most of us now are using a mix of cool white...royal blue plus some neutral or warm whites... a few reds/greens.. and some UV to get a more "full spectrum".. Back in the day it was only cool white and royal blue. But a more full spectrum build brings out colors that you didn't get with only cool whites/royal blue combos

The driver tip is great. I will check the prices for those. It drops 3V, right? Does this means that it "wastes" 3w at 1000 mA? If so, not much different from the LM317? The LDD can manage up to 4 10W leds, the LM317 can manage up to 3 10w leds, but at a far lower cost. What did I miss?

It's a 200 gallon tank with intention of growing SPS, I think. My led + glass lens system can output quite a bit of light, here's a 6500K 3 led 26-27w test:

I have been getting fairly uniform 10w leds, no complaint on that. The planned setup is 24 10000k + 20 blue + 10 red. I'll post photos when I have it ready.

I know an Arduino would be far better, just haven't gotten there yet... I'm still in analog stage...

You could set up a free running 555 timer with adjustable pulse width to drive power FETs (one for each LED). You could also use something like the SG3525 set up with a trimpot to adjust pulse width. Use a power resistor in series with each LED so the FET turns fully on each pulse and the resistor limits the current.

This would waste energy on the resistor side, right? I think CC with the LM317 is a better choice. About the pulse, no doubt it will be the next thing to do.

Thanks for all the valuable opinions.

 

I just cut my 3 sets of LM317s to one set, I'll be using this idea from eeweb.com:

http://www.eeweb.com/blog/circuit_projects/programmable-current-source-by-lm317

 

Your original post mentioned 54 LEDs and 54 LM317s each (in 3 groups of 18 each). If you are now down to one set, does that mean you still have 18 LM317s? If so, some thoughts...

You're looking at 54 BCD thumbwheels and 216 resistors. Seems like a lot.

The 1.55 ohm resistor will be dissipating 1W, so you will need at least a 2W body for that one, and 1W bodies for a few others.

The switch for the 1.55 ohm resistor must be able to switch (not just carry) 0.8A. Again, could be a problem, especially at DC.

I see two problems here. First is that while the 317 can be tricked into being a CC regulator, neither end of the CC adjust component is tied to GND, where you can manipulate a bunch of them together. Second is that 100% of the output current flows through the CC adjust component.

In app notes for LED driver chips from LT and Maxim, one common theme is to have the regulated output drive multiple strings of LEDs in parallel, and each string has a single current sense resistor from the last LED cathode to GND. The feedback to the control chip is taken from only one current shunt, the theory being that all of the LEDs and all of the shunt resistors are similar enough that if you regulate the current in one string, the others will be close enough.

That might be your best hope here, one big power supply driving all strings in parallel, and feedback from one string turning everything into a CC loop. Vary the feedback, vary the brightness.

ak

ps. Still not clear on the original setup. 400W / 54 LEDs - 7.4 W/LED. Even at 4V Vf per LED, that's 1.9A per LM317, exceeding its spec. Am I missing something here, and can you post a basic schematic of your proposed hookup?

 

Your original post mentioned 54 LEDs and 54 LM317s each (in 3 groups of 18 each). If you are now down to one set, does that mean you still have 18 LM317s? If so, some thoughts...

You're looking at 54 BCD thumbwheels and 216 resistors. Seems like a lot.

Ok I get it, I know it looks like a mess, the layout is in my head but I don't know how to draw the schematic.

What I do is group the 54 leds in 18 groups of 3. Each group will require around 30V from my power supply, solved, but I run them with LM317s in current regulation mode.

So, for the 3 light levels I use, I would require 3 sets of 18 LM317s, activated by only 3 SPDT relays. It's 72 resistor groups (one parallel group per LM317) in current regulation mode.

This system works fine in my own aquarium, it's just 12 leds for a 4 feet 90 gallon planted tank.

The 1.55 ohm resistor will be dissipating 1W, so you will need at least a 2W body for that one, and 1W bodies for a few others.

The switch for the 1.55 ohm resistor must be able to switch (not just carry) 0.8A. Again, could be a problem, especially at DC.

I see two problems here. First is that while the 317 can be tricked into being a CC regulator, neither end of the CC adjust component is tied to GND, where you can manipulate a bunch of them together. Second is that 100% of the output current flows through the CC adjust component.

I use 1w resistors in parallel, exceeding the 2W rating.

This is my problem, not being able to manipulate them all,correct. I am looking at using transistors to select from 3 resistor values tied to the adjust pin as in my last post.

2N2222 may be the answer (I may need a higher rating transistor for the 700-800mA current).

There would be 18 LM317s, 54 2N2222 and 54 resistors or resistor groups.

In app notes for LED driver chips from LT and Maxim, one common theme is to have the regulated output drive multiple strings of LEDs in parallel, and each string has a single current sense resistor from the last LED cathode to GND. The feedback to the control chip is taken from only one current shunt, the theory being that all of the LEDs and all of the shunt resistors are similar enough that if you regulate the current in one string, the others will be close enough.

That might be your best hope here, one big power supply driving all strings in parallel, and feedback from one string turning everything into a CC loop. Vary the feedback, vary the brightness.

This could be risky and I haven't tried that.

ps. Still not clear on the original setup. 400W / 54 LEDs - 7.4 W/LED. Even at 4V Vf per LED, that's 1.9A per LM317, exceeding its spec. Am I missing something here, and can you post a basic schematic of your proposed hookup?

Each 3 leds will run at max 700 ou 800mA, at about 30V, no more than that, I am not exceeding LM317's rating.

I know this is getting complicated, and there is a far shorter path, I have to go digital!

 

I just read my own answer again...

This is getting ridiculous. I need to pulse an IRF540 with an arduino and deal with all the leds at the same time, about 13-14A.

It will be better to invest a few hours programming and testing the board, than soldering dozens and dozens of resistors and transistors.

After that is accomplished, I can divide the color channels and program whatever I want as I go along.

Thanks for all the help.