(Help) choose driver for my DIY led board.

[Jellypowered]

Hey all!
Hopefully this is the right place.

I need help choosing a driver, I don't want to use a power supply and buck converter, I'd like streamlined goodness.

An 84 LED panel, custom made by me.

Using these LEDS:

Looking at some information online,
A single string of series wired LEDs, the total voltage is the sum of all LEDS in series.
So 7 leds at 6.3v should come to 44.1v
With Parallel you add all the current together.
If i'm driving at 150ma, with 12 strings in parallel, that should be 1800ma.

So, If i'm learning anything, I need a power supply that will output
44.1v @ 1800ma. per board? Technically, if I add more than one board, the MA requirement goes up by 1800ma each time?

I was thinking about using the meanwell LPF-90-42 or LPF-90-48, (they also make PWM versions)

If I use the LPF-90-42 will the output of 42v hurt my performance in any way? I need at least 44.1 based on calculations.

With that being said, would the LPF-90-48 hurt my LEDS being that it's putting out 48v?

Thanks for any info, I'm pretty new to LEDS so be kind but blunt.
Edit: found Sosen SS-100H-48 100 Watt LED Driver 36-48V Output 2100mA Constant Current as well. I don't know which one to use lol.
Jelly

 

[nfhiggs]

Hey all!
Hopefully this is the right place.

I need help choosing a driver, I don't want to use a power supply and buck converter, I'd like streamlined goodness.

An 84 LED panel, custom made by me.
View attachment 3957568
Using these LEDS:
View attachment 3957569
Looking at some information online,
A single string of series wired LEDs, the total voltage is the sum of all LEDS in series.
So 7 leds at 6.3v should come to 44.1v
With Parallel you add all the current together.
If i'm driving at 150ma, with 12 strings in parallel, that should be 1800ma.

So, If i'm learning anything, I need a power supply that will output
44.1v @ 1800ma. per board? Technically, if I add more than one board, the MA requirement goes up by 1800ma each time?

I was thinking about using the meanwell LPF-90-42 or LPF-90-48, (they also make PWM versions)

If I use the LPF-90-42 will the output of 42v hurt my performance in any way? I need at least 44.1 based on calculations.

With that being said, would the LPF-90-48 hurt my LEDS being that it's putting out 48v?

Thanks for any info, I'm pretty new to LEDS so be kind but blunt.
Edit: found Sosen SS-100H-48 100 Watt LED Driver 36-48V Output 2100mA Constant Current as well. I don't know which one to use lol.
Jelly

I would not use a constant current supply with multiple parallel strings. Use a constant voltage supply.

Why not use Samsung 561C diodes? much higher efficiency with them....

 

[Jellypowered]

Why not use Samsung 561C diodes? much higher efficiency with them....

Too invested in this setup right now. In the future I may in fact go with 561C.

I would not use a constant current supply with multiple parallel strings. Use a constant voltage supply.

Since i'm doing everything myself, I had to pay someone to design the led form factor for use in Fritzing so I could make my own layout, having 10 bare PCBS made (Gerbers submitted, no going back) the cost of equipment and materials, only thing I haven't purchased yet is the actual Leds.

So constant voltage. can I use 48v or will that fry all my LEDs. Thanks for the reply!

 

[nfhiggs]

Too invested in this setup right now. In the future I may in fact go with 561C.


Since i'm doing everything myself, I had to pay someone to design the led form factor for use in Fritzing so I could make my own layout, having 10 bare PCBS made (Gerbers submitted, no going back) the cost of equipment and materials, only thing I haven't purchased yet is the actual PCBS.

So constant voltage. can I use 48v or will that fry all my LEDs. Thanks for the reply!

A 48V supply will generally be adjustable at least +/-10%

 

[Jellypowered]

Pwm-90-48 looks like that's my Best bet then.

 

[NoFucks2Give]

Not a good idea to run strings in parallel. Each string will draw different current and the light will be unbalanced. It could get bad enough where some of the strings shut down. Depends how warm they run. I've attached two PDFs that explain the problem and solutions. The forward voltage is both temperature and current sensitive. By the nature of LEDs, the strings will have different forward voltages so some strings will have more current, some less. This changes the temperature and currents which in turn changes the forward voltages. Whether a string will become a current hog is random chance.

You could have regulated each string using a CCR like the On-Semi NSV45090JDT4G at a cost of $6 per board. Cheaper than using the TI LM3466. Then if you tweak the voltage correctly the boost in efficiency would pay you back.

Definitely the LPF-90-48 over the LPF-90-42. The 42 would not work if your LEDs Vf run over typical. Not likely, but possible. Also the voltage may end up outside the constant current range. The 48 will self adjust to the required voltage, it will work fine.

The better bet is the HLG-80-48A. Cost less ($36 at TRC) and 7 year warranty vs 3 year. Plus you have the voltage and current adjust. 44.1 x 1.8 A = 79.38 Watts. You'd be running at nearly 100% capacity which would be maximum efficiency. The HLG always will exceed their rated current if your forward voltages run high.

BTW if you are going to layout more PCBs, learn KiCad.

 

[Jellypowered]

@NoFucks2Give Thank you for the material and lessons!

Why KiCad? It's interface is stupid complicated to me, I can't wrap my head around it. I guess on the next board I will give it a shot though. It may take me a while but I suppose with tutorials and patience I could learn to use just about anything.

Also, Thanks for the driver suggestion I hadn't seen that one in my searches yet.

 

[NoFucks2Give]

Why KiCad?

Because it's free and very capable. It does have a steep learning curve especially the first day.
I do not like reading manuals and figured it out on my own.
Now I find it very simple to use.
I do have a lot of experience with other commercial packages over many years.
Getting the part libraries setup at first was troublesome.

 

[nfhiggs]

@NoFucks2Give Thank you for the material and lessons!

Why KiCad? It's interface is stupid complicated to me, I can't wrap my head around it. I guess on the next board I will give it a shot though. It may take me a while but I suppose with tutorials and patience I could learn to use just about anything.

Also, Thanks for the driver suggestion I hadn't seen that one in my searches yet.

One way to mitigate the current imbalance issue, is to hand sort the diodes so you know exactly how much forward voltage on each one will produce your 150 mA target current. Once you know that data its simply a matter of building the strings so each one has the same total forward voltage.

 

[Jellypowered]

One way to mitigate the current imbalance issue, is to hand sort the diodes so you know exactly how much forward voltage on each one will produce your 150 mA target current. Once you know that data its simply a matter of building the strings so each one has the same total forward voltage.

Can you describe a test procedure for verification of forward voltage?

 

[nfhiggs]

Can you describe a test procedure for verification of forward voltage?

A small variable (0-10V) power source, and a voltage/current display. You can get the voltage/current displays on ebay for 2 bucks or so. Start at 0V and turn the voltage up until you get 150 mA through the diode. Write down the voltage at that point. Sort them by the voltage so you can make strings that have the same total voltage.

 

[nfhiggs]

A small variable (0-10V) power source, and a voltage/current display. You can get the voltage/current displays on ebay for 2 bucks or so. Start at 0V and turn the voltage up until you get 150 mA through the diode. Write down the voltage at that point. Sort them by the voltage so you can make strings that have the same total voltage.

cheap benchtop power supply:
http://www.ebay.com/itm/132161977387

Edit: I'm gonna have to pick up one of these, I need a little benchtop power supply...

 

[NoFucks2Give]

ts simply a matter of building the strings so each one has the same total forward voltage.

I'm going to go out on a limb here. I'm guessing you never did many parallel strings of LEDs.

If you read the two papers on this topic I posted earlier, neither of them recommended doing that. Why? Because it does not help. It's a myth. Many don't believe there is such a thing as thermal runaway. Especially on this site.

Let's say you do bin the LED by Vf. at 150 mA and 6.3V you're looking at one watt per LED. What going to happen to the forward voltage once they are fired up? It's going to get hot an Vf is going to change.

Then at least one string is going to get more than 150mA. That one is going to get really hot.

The problem is exacerbated by heat and more than a few LEDs in the string. At 6.3V there is more than one LED.

This board has no thermal management. No heat spreader, no thermal vias, it's going to get very hot.

I was testing a board of 16 Luxeon Rebel White with a string voltage of 42v, or 2.625v per LED without being attached to the heatsink. I would periodically increase the current starting at 80mA. At 375 mA I had to stop the test because the case temp was at 129° C meaning Tj was near 150° C. The wattage at 375 mA was very close to these LEDs @ 150mA.
The measurements from my testing.


I wish the best for this board but it has potential to be a disaster. Binning the LEDs is not going to help enough.

I ran this board in parallel with another just like it where the Vf difference was 0.2v. Like 42.6 and 42.8. Thermal runaway at 150 mA. It would take a day. When I'd get up in the morning one board was very bright and the other was barely on and just flickering a little bit looked dead. Turned it off and on and everything appeared to be fine. The next morning, same thing. I could get the boards to alternate as to which failed and which burned bright by powering them up one at a time instead of simultaneously and alternating which was powered first.

I also began doing some testing on parallel strings. I was going to do this over with a better test setup. One of the values i measured is wrong because I forgot to write it down and guessed at what it was. The results of this test was worse than expected.

 

[Jellypowered]

In the interest of friendly discussion, @NoFucks2Give if parallel series strings are so bad, why do quantum boards work so well? There's no visible current limiting resistors that I can see, and it doesn't appear to be an aluminum based pcb, yes there are large heat sinks but I'll be using heat sinks and thermal epoxy as well.

 

[nfhiggs]

I'm going to go out on a limb here. I'm guessing you never did many parallel strings of LEDs.

If you read the two papers on this topic I posted earlier, neither of them recommended doing that. Why? Because it does not help. It's a myth. Many don't believe there is such a thing as thermal runaway. Especially on this site.

Let's say you do bin the LED by Vf. at 150 mA and 6.3V you're looking at one watt per LED. What going to happen to the forward voltage once they are fired up? It's going to get hot an Vf is going to change.

Then at least one string is going to get more than 150mA. That one is going to get really hot.

The problem is exacerbated by heat and more than a few LEDs in the string. At 6.3V there is more than one LED.

This board has no thermal management. No heat spreader, no thermal vias, it's going to get very hot.

I was testing a board of 16 Luxeon Rebel White with a string voltage of 42v, or 2.625v per LED without being attached to the heatsink. I would periodically increase the current starting at 80mA. At 375 mA I had to stop the test because the case temp was at 129° C meaning Tj was near 150° C. The wattage at 375 mA was very close to these LEDs @ 150mA.
The measurements from my testing.
View attachment 3958214

I wish the best for this board but it has potential to be a disaster. Binning the LEDs is not going to help enough.

I ran this board in parallel with another just like it where the Vf difference was 0.2v. Like 42.6 and 42.8. Thermal runaway at 150 mA. It would take a day. When I'd get up in the morning one board was very bright and the other was barely on and just flickering a little bit looked dead. Turned it off and on and everything appeared to be fine. The next morning, same thing. I could get the boards to alternate as to which failed and which burned bright by powering them up one at a time instead of simultaneously and alternating which was powered first.

I also began doing some testing on parallel strings. I was going to do this over with a better test setup. One of the values i measured is wrong because I forgot to write it down and guessed at what it was. The results of this test was worse than expected.

View attachment 3958218

A few counterpoints:

I read both of your links and neither one mentioned anything about hand selection of diodes to create matched Vf strings not being an effective mitigation strategy.

Heat sinking to stabilize temps is just a given, IMO. Not doing so is a certain path to failure.

COBs are constructed of multiple series LED strings in parallel, no current mirrors or other current management. They work, no current imbalance.

Quantum Boards work. No current mirrors or other current management. They consist of multiple series strings in parallel.

I have a bunch of Samsung Acuity 56 diode strip boards. They consist of 8 parallel strings of 7 diodes each. No current mirror or other current management. They work.

I keep hearing people say "parallel strings don't work without doing x, y, and z" and referencing links such as the ones you did...and yet there are myriad examples of it working.

 

[NoFucks2Give]

They work, no current imbalance.

I'll agree with part one, they work. I doubt part two is true.
CoBs are a different story. The dies are so close to one another they are all at the nearly same temperature. They are also designed to work imbalanced as there is no way to actually see any imbalance.

For example I have these two CoBs, the same part number. One a bunch of it's LEDs are melted and it still preforms identically to the undamaged CoB.