Hello all,

I'm trying to make a PCB for a LED-based lighting panel and I'm trying to wrap my head around the best way to layout the circuit. After a decent amount of scanning through the web, it seems like using voltage regulator(s) is a better way to go over the old LED array with resistors. I'm hoping that someone on here can let me know if I'm going down the right path or I things need to be changed/adjusted.

Basically, I have 3 different 12vdc inputs (one for each color) feeding into an LM317 regulator since it's going to be powered by a car; I wanted to try and make sure that the output is close to 12vdc, not the 12-14.4vdc that it can usually fluctuate. Then it goes through a resistor (to try and get a constant current output) and into two rows of 4 LEDs (each LED is 350mA) and finally into the resistor to drop the voltage down some and into ground.

I'm a mechanical kid, so the electronic's side of the fence isn't the easiest for me to follow/catch on to. I basically know a little bit to get myself into trouble, but definitely not enough to get myself out of it. If anyone could review this/point me in a decent direction, it would be much appreciated. I've included a screenshot of the circuit that I have thus far.

Thanks!

 

Welcome to AAC!

LEDs are current driven devices. You need to regulate the current, not the voltage.
Hence you need a 350mA constant current supply for each row of LEDs.

Look up LM317 constant current supply. For example:

https://www.bristolwatch.com/ccs/LM317.htm

 

Welcome to AAC!

If anyone could review this/point me in a decent direction, it would be much appreciated.

You don't have enough headroom for 4 LEDs in series. The forward voltage is going to be over 3V each at 700mA, the ballast resistor will drop 3.5V, and the current set resistor will drop 1.25V. The LM317 won't have a high enough input voltage to function.

You're going to have to remove 2 LEDs in each string and make the ballast resistors smaller.

EDIT: Red isn't as bad, but still won't work...


From Nat Semi:

At 700mA, the dropout voltage is going to be close to 2V.

I'd use a current source that didn't use ~1/4 of your supply voltage.

 

LEDs are current driven devices. You need to regulate the current, not the voltage.

Which is what his schematic shows. But it will put out 700 mA instead of 350.

Bob

 

If you want a really simple solution, see
https://www.onsemi.com/pdf/datasheet/nsi50350as-d.pdf
One of these on each string would replace all the resistors and LM317s

 

each LED is 350mA

For how long? It is not usually a good idea to run components at their maximum rated current if you want them to have a long and happy life. I doubt you'd notice any brightness reduction if you ran the LEDs at, say, 300mA.
What heatsinks will you use?

 

These are some high power LEDs.

Specs for White:

 

If you want a really simple solution, see
https://www.onsemi.com/pdf/datasheet/nsi50350as-d.pdf
One of these on each string would replace all the resistors and LM317s

Interesting device, if you can find someone with stock and not taking the p*** on price! - a few on eBay selling 1 for the price of 10!

 

Yes - they are a bit scarce!
Findchips.com says RS have some at 65p each, but they are fibbing!
I've just used some of the 25mA version, which are available.

 

Yes - they are a bit scarce!
Findchips.com says RS have some at 65p each, but they are fibbing!
I've just used some of the 25mA version, which are available.

RS claim back order fulfilment on Sep 21st, but everyone else quoting end 2021 through mid-22!

However, this LT3092 200mA programmable device is available, and 2 in parallel configured for 175mA each would give 350mA and do the job with only <1.6v dropout, so at 12v in you can get >10.4v compliance voltage, possibly enough for 3 LED if carefully chosen.

 

If you want a really simple solution, see
https://www.onsemi.com/pdf/datasheet/nsi50350as-d.pdf
One of these on each string would replace all the resistors and LM317s

I wouldn't mind using one of those at all; unfortunately Digikey, Mouser, and Newark are all out of stock and not expecting more for the next 6 months. I did find a TS19601 that might work...

For how long? It is not usually a good idea to run components at their maximum rated current if you want them to have a long and happy life. I doubt you'd notice any brightness reduction if you ran the LEDs at, say, 300mA.
What heatsinks will you use?

Didn't exactly think of the life span, but I will have to. The idea was to mimic emergency lights for background lighting when we are filming; so even though the lights will be flashing, they'll still put out some heat. I'd love to just buy the real McCoy, but our tiny budget won't allow it. Besides, I'm hoping that a more 'modular' design will allow us to use these "garage builds" for other projects as well.

As of now (and everything is up for change with this), the heat sink would be roughly the size of the board; so 8" L x 2.25" W x 0.5" D with approximately 64 fins per square inch.

 

A 24W LED emergency light? Most of them are 3W.

 

A 24W LED emergency light? Most of them are 3W.

From my understanding, they're using 3w chips in the design, for example, a Whelen Legacy light bar averages close to 11A (2.7A per corner). But either way, not all of the lights would be used at once. In the case of this PCB, the power is split for each color and it would be using an external flasher. Less banging my head against the desk (don't even want to think about building a circuit for that), plus I could pick and choose which color string to use for different scenarios.

 

I see now - not that sort of emergency light, an emergency vehicle light!

 

The TS19601 looks like a good find. On a quick look it has the lowest drop-out of all options so far...

 

Welcome to AAC!
You don't have enough headroom for 4 LEDs in series. The forward voltage is going to be over 3V each at 700mA, the ballast resistor will drop 3.5V, and the current set resistor will drop 1.25V. The LM317 won't have a high enough input voltage to function.

You're going to have to remove 2 LEDs in each string and make the ballast resistors smaller.

EDIT: Red isn't as bad, but still won't work...
View attachment 247534

From Nat Semi:
View attachment 247533
At 700mA, the dropout voltage is going to be close to 2V.

I'd use a current source that didn't use ~1/4 of your supply voltage.

That is interesting and surprising to me (the forward voltage drop for different colors). Being that blue is the highest frequency and photon_energy is proportional to frequency. I expected blue to have the highest (Joules/Col) not green or yellow.

 

The LM317 in the schematic is a tiny little LM317L in a small TO-92 package that has a maximum allowed output of only 100mA if it is cooled well.

An LED has a range of forward voltage, the graph shows only a typical one. You must allow for a maximum forward voltage printed in its datasheet plus its change with temperature.

Using a normal LM317 in a TO-220 power package, maybe with a heatsink, then its maximum dropout voltage and voltage across its current-setting resistor must be observed.

I did not lookup your LEDs to see if they need heatsinks for cooling since your current is so high.

 

That is interesting and surprising to me (the forward voltage drop for different colors). Being that blue is the highest frequency and photon_energy is proportional to frequency. I expected blue to have the highest (Joules/Col) not green or yellow.

Joules / what?
That graph is wrong - "normal" green and yellow have forward volt drops around 2V and 2.2V respectively.
"normal" green being the type of green we always used to have. There is a new 525nm GaN green as well these days, which has a higher volt-drop, presumably because it is GaN.

 

Yes, the modern very bright green LEDs use the same chemistry (InGaN or GaN) as blue and white LEDs so they have the higher forward voltage.

 

That is interesting and surprising to me (the forward voltage drop for different colors). Being that blue is the highest frequency and photon_energy is proportional to frequency. I expected blue to have the highest (Joules/Col) not green or yellow.

How did we end up with an 'incorrect' graph?