High power led matrix connection

[bmandl]

Hy guys.
I need your help. I am building a high power led matrix and I need advice on how to connect individual leds. I have a 260x160mm heat sink, where I will mount star PCB leds. PCBs are 20mm wide and I need to space them 3mm apart from each other for lens to fit on. So I can put them 11x6. It doesn't really matter for conection, but anyways... I have a 200W 6A 18-34V led driver for my matrix. Leds are 400-415nm so forward voltages are somewhere between 3.5-4.0V, and current is 700mA. So how should I connect LEDs to a power supply in the meaning of serial-parallel connection, so that I won't damage any led and that all would be as bright as possible?
Thank you.

 

[KlausST]

Re: high power led matrix connection

hi,

with a quick view I see no other way as to connect them 6S x 11P.

then each LED gets 545 mA only.

To come close to the 700mA limit you should connect 9 LEDs in parallel.
But connecting 7 in series makes 63 LEDs only. 3 LEDs unused.

***

If you manage to get 6 x 12 LEDs on your heatsink, then connect them 8s x 9P

Klaus

 

[bmandl]

I think that 545mA would be enough, as luminosity doesn't change that much in this range, correct? But what happens with voltage on individual leds, when connected this way? Wouldn't voltage be too high?

 

[KlausST]

Hi,

6 in series makes 3.5V x 6 = 21V.
If your supply is a true current limiting LED supply then everything is OK.
You worte 18..34V. So 21V is within it´s range.

Klaus

 

[FvM]

You should check in the LED datasheets if parallel connection achieves sufficient uniform currents without dedicated balancing means (e.g. series resistors) or binning. Direct parallel connection is generally usual in LED lighting, but in some cases need binned LEDs.

 

[bmandl]

Unfortunately, I don't have any datasheet for these leds, because I bought them from a Chinese seller. So I guess I'll have to try.

 

[bmandl]

@FvM, after you mentioned uniform currents, I started to investigate in this direction. The problem is, if one led in any string fails, all leds will be over driven without protection, and eventually every led will fail. So I started digging around some current balancing solutions and came across current mirror based circuit, but I can't find any details on designing such a circuit: https://www.ledsmagazine.com/articles/print/volume-6/issue-2/features/led-design-forum-avoiding-thermal-runaway-when-driving-multiple-led-strings-magazine.html. I like it, because it has current protection besides current balancing properties.
Then I also fond this chip from Texas Instruments: https://www.ti.com/product/LM3466/description.
Would this work with my case?

 

[Audioguru]

It seems that you will have a massive amount of heat. Is your heatsink cooled with liquid hydrogen?
Will all those high powered LEDs blind anybody who sees them?

 

[bmandl]

No, heat sink will be additionally cooled with 4 fans, 2 from every side (two for inlet air, two for outlet air). Why would I need liquid hydrogen? Leds won't blind anybody, because this will be a closed box.

 

[Audioguru]

What plants will be growing in the closed box?

 

[bmandl]

I am building LCD based 3d DLP printer.
So, what do you say about mentioned ICs from texas instruments? Would it be better to build my own circuit with transistors?

 

[KlausST]

Hi,

I don't think the effort is worth it.

How I'd do it*
* Make series LED strings.
* add a series resistor to each string
* connect all strings together.
Then start with placing zero ohm resistors to the strings.
If you have, then start with low current. If you use 100mA total current fir example, then the imbalance is more visible and you don't risk damage.
If the brightness is about equal with 100mA, then I see no problem with high current.
Even the temperature drift shouldn't harm the LEDs

Klaus

 

[bmandl]

I just have 6A constant current supply, so I can't test with lower current. I have constant voltage supply with current regulation thou. But it is useless here. But what if some day one led in one string dies? Then whole matrix will die eventually, wont it?

 

[KlausST]

Hi,

Not sure what configuration you decided to use...
But if you use 7s, 9p with 6A of current.

If one string fails, then I assume it will fail open circuit.
This means one string is dark, the other strings become brighter.
Instead of 9 string there are only 8 strings carrying the 6A.
This means 750mA per string. This is just 7% over the rated current...not likely that this kills the LEDs.

If you want a more safe circuit...then use a resistor in each string that drops about 1V in normal condition.
Use a BJT with base resistor to detect current flow in each string.

Use a NAND logic to generate an alarm if NOT ALL strings carry current. Simple, reliable.

Klaus

 

[c_mitra]

Direct parallel connection is generally usual in LED lighting, but in some cases need binned LEDs.

This is perhaps not needed if you make (say for example) first 12 LEDs in series (there will be 6 such series) and they will have a total of 3.5V x 12= 42V. If the original LEDs had a spread of (say again for example) 4V-3.5V (with a variance of 0.5) then the variance of the string of 12 LEDs will be (0.5/12)- rather small. The probability that all the LEDs in the string will be 3.5V or 4V (extreme value) will be very very less.

It is therefore always wise to make the series and then connect these strings in parallel.

- - - Updated - - -

I think that 545mA would be enough, as luminosity doesn't change that much in this range...

I am not so sure.

 

[bmandl]

Hi,

Not sure what configuration you decided to use...
But if you use 7s, 9p with 6A of current.

If one string fails, then I assume it will fail open circuit.
This means one string is dark, the other strings become brighter.
Instead of 9 string there are only 8 strings carrying the 6A.
This means 750mA per string. This is just 7% over the rated current...not likely that this kills the LEDs.

If you want a more safe circuit...then use a resistor in each string that drops about 1V in normal condition.
Use a BJT with base resistor to detect current flow in each string.

Use a NAND logic to generate an alarm if NOT ALL strings carry current. Simple, reliable.

Klaus

Klaus, i am afraid, that these Chinese leds won't stand any temperature drifts and over-currents. So, I would like to make some kind "bullet-proof" design, analog circuit preferred. Could 4 transistor current mirror be used in someway here, to balance currents and also protect leds from over current. If for example only one string is connected, to bypass excess current and only 700mA through leds alowed, the rest of 5300mA flows through transistors.
I know that all this is not worth the effort, but I would like to try such a circuit.

 

[KlausST]

Hi,

i am afraid, that these Chinese leds won't stand any temperature drifts and over-currents.

It sounds like you think the LEDs work from 0...100% of rated current without change of temperature and without fail.
...and they will be immediately be killed above 100.1% of rated current.
--> This is not the case

I don't understand your current mirror solution.
If one LED fails, it most probably fails OPEN. How can the current still flow? How the bypass can work?

In my eyes you put too much effort in the current control circuit:
My approch:
* use a constant voltage output power supply
* connect n LEDs in strings with a series resistor per string.
* connect those strings in parallel directely to the power supply.
The only thing is to calculate the resistor value.

With correct resistor value there will be about no drift with temperature.
You avoid overload of the LEDs
If a LED fails...the complete string is OFF, but this doesn't change the current of the remaining LEDs
It is very simple, cheap and rugged.
Where do you see the drawback?

Klaus

 

[bmandl]

I have a constant current supply ready for this led matrix.

If one LED fails, it most probably fails OPEN. How can the current still flow? How the bypass can work?

If one led fails, the current doesn't flow through this string. So current has to flow through remaining strings, because I have constant current power supply. But I want to limit this current, so excess of current would flow through some parallel transistor if you can imagine what I am trying to say. But I don't know, how to implement such a circuit. I am imagining this like some PNP current mirror, with additional transistor parallel with reference transistor to bypass excess of current. Can this be implemented in some way?

 

[KlausST]

Hi,

I don´t know a simple solution for this.

The problem is that then there are two "constant current" circuits in series. One in the power supply and one at each LED string.
This will always cause problems, because there are two regulation circuits working on one problem.

Imagine:
you have a balance weigthing scale. And two persons - one at each side - but the can´t communicate with each other.
Either person can fill in water or pull out water from it´s side. Either person tries to get the weighing scale in balance.
It will become difficult.

It´s much easier to fix the ammount of water in one side (in your case: work one part with fix voltage) while only one person brings the weighing scale into balance (in your case current regulation).

Klaus

 

[c_mitra]

I understand that you want a bullet proof design; but that does not mean that things will be 100% fail-proof.

A reasonably simple solution will be to reduce the current in each LED string and that will reduce the failure rate *drastically*- but you will lose some intensity.

The LED brightness is approx linearly related to the current and you can compensate by increasing the number of LEDs in a string.

Yes, that means you will need more number of LEDs for same brightness but that is the price for greater life expectancy.

You may still need to install one series power resistor in each string to limit the current inequality in current sharing. You may also monitor the voltage across this resistor to know the actual current in the given string.

You also need to consider complexity vs reliability. More complex systems may actually reduce reliability (they are introducing another part that can fail!) in some cases. That is what was pointed out in post #19. It will be desirable to have independent power supplies for each string with independent controls but you must make them talk to each other. Hence it is imperative to improve on the reliability of the existing parts.