Looking for a specific DC CC driver for LEDs

[theatrus]

Actually, I have plenty of friends and colleagues who match that description. I also like Shenzhen- great seafood and only a short ferry ride away from HK.

A tad industrial though

I doubt that these particular chips deviate much in terms of thermal properties from normally good chips. They're rejects from the same lines - unusable by industry but should be adequate for my uses.

I'll start by finding published C/W for these kinds of chips - maybe a distribution can help. I also have a Chinese contact in one of the factories, but he may not have this kind of data.

If I get no leads, we could cross-section one to get a measure of the thickness and materials in the construction. Maybe cross section a few. Well, not the expensive ones I plan to use in my final setup- those are $160 a chip.

I always walked across the border and took the train. Not a boat person Debugging production line problems is also a whole other ball game through interpreters.

I wish you luck in your dubiously cheap Chinese electronic overdriving endeavors

 

[karimwassef]

Are you going to characterize your dubiously cheap LED?

I'm planning to test your board today - I hope.

 

[karimwassef]

So 0.5-2C/W isn't unrealistic

http://www.luminus.com/resource/application_10_3117559341.pdf

But that's a pretty wild range.

 

[karimwassef]

To digitally dim the boost (that comes with an analog pot), I'm thinking of using an Analog Devices digital potentiometer AD5171 from my Arduino...

https://www.arduino.cc/en/Tutorial/DigitalPotentiometer

http://www.ebay.com/itm/DC-DC-boost...ower-supply-250W-10A-LED-Driver-/181940997254

 

[theatrus]

For those curious, test sample #1 of a "100W" "royal Blue" $10 multichip:


Vf is 33.95V at 3A drive
Vf is 34.09V at 3.1A
34.27 / 3.2
34.43 / 3.3
34.59 / 3.4
34.77 / 3.5


CC is driven by a single HP6644A linear power supply. Vf is measured at the connection terminal on the LED.

I cap out at 3.5A as my second supply is busy in a test.

And for good measure:
28.8V at 1A
30.17V at 1.5A
31.4V at 2A
32.51V at 2.5A

 

[karimwassef]

That is excellent! Thanks.

<a href="http://s1062.photobucket.com/user/karimwassef/media/karimwassef002/0_zpslab8evck.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/karimwassef002/0_zpslab8evck.png" border="0" alt=" photo 0_zpslab8evck.png"/></a>

I plotted it with my data. It looks like the same characteristics but with a different intercept and resistance...

Do you have a PAR meter? I'm curious to see if the additional power in my chip is just resistive power loss or if it's also radiating more.

 

[karimwassef]

I'm going to try the other chips I have.. maybe a $2 chip is different than a $5 or $10 or $20 or $40... There really are $2 chips...

I've gotten samples at different prices - let's see what the data shows.

I couldn't get them all at the same wavelength though. I was going to use them for a different kind of experiment.

 

[karimwassef]

For those curious, test sample #1 of a "100W" "royal Blue" $10 multichip:


Vf is 33.95V at 3A drive
Vf is 34.09V at 3.1A
34.27 / 3.2
34.43 / 3.3
34.59 / 3.4
34.77 / 3.5


CC is driven by a single HP6644A linear power supply. Vf is measured at the connection terminal on the LED.

I cap out at 3.5A as my second supply is busy in a test.

And for good measure:
28.8V at 1A
30.17V at 1.5A
31.4V at 2A
32.51V at 2.5A

Oh! forgot to ask! how are you cooling it again? 100W heatsink and fan?

 

[theatrus]

Do you have a PAR meter? I'm curious to see if the additional power in my chip is just resistive power loss or if it's also radiating more.

Have the fancy new SQ-500 now (along with an old SQ-120) - getting an efficiency curve would also be a great idea. :thumbsup:

 

[theatrus]

Oh! forgot to ask! how are you cooling it again? 100W heatsink and fan?

Yup. I let the system reach equilibrium (measured via a thermocouple at the heatsink package boundary in some thermal paste).

I didn't capture notes on the actual temperature though - sloppy lab work

 

[theatrus]

Also, you likely have a decent voltage drop over the supply wire at these currents - where are you measuring the voltage in your system?

 

[karimwassef]

I'm measuring about 4" from the LED. I'll make a resistance measurement.

 

[karimwassef]

I got a reply from my contact in one of the factories. Apparently, my $5 chip is basically just burning power. The current throughput correlates to the light output. They measure it in Lumens, but the concept is the same.

So if I can run 3A at 50V and you can run 3A at 30V, the radiation should be close, but the power dissipated is also much higher on my 50V LED. The difference of 20V x 3A ~ 60W is just heat. That further complicates things...

His design guidelines for thermal resistance were on the order of 10C/W! That's 10x what I was expecting... So with the increased Q (power throughout), and worse thermal conductivity, it won't be feasible to get the PAR/W even with chilling.

He also sent me some real datasheets. He doesn't sell these cheap LEDs. His are in the $200-$400 range and run 250W-500W for real industrial applications for OEMs. But his explanation for the sources of the cheap scrap product lines up with our measurements so far.

I don't think I've seen radiation efficiency or PAR/W discussed in other multichip DIY threads? I think it would be useful.

 

[oreo57]

"Watt efficiency" always drops w/ power increases.. Pretty SOP for LED's.
Though few have measured this at the extreme ends of the power curve.
Goes to the old story more chips driven at less power or less chips driven at more power.

Boils down to the cost of the chips and ones expectation of lifespan..

Live fast, die young applies to LEDs as well..

And, yes eventually one will exceed the possible thermal transfer inherent in the aluminum base heat removal properties.. allowing the chips to cook themselves..

 

[karimwassef]

I didn't see the explanation above anywhere else

When I published my IV characteristic, I looked for references to the resistance (slope) and no one had anything. I don't see binning / scrapping by resistive loss anywhere either. This is a biproduct of manufacturing variability...

What this does, though, is allow DIY makers to run the cost/benefit analysis of lower efficiency and compare it to the cost of premium pre-made fixtures. Those guys all publish their radiation efficiency since it's SOP, right?

 

[karimwassef]

"Watt efficiency" always drops w/ power increases.. Pretty SOP for LED's.
Though few have measured this at the extreme ends of the power curve.
Goes to the old story more chips driven at less power or less chips driven at more power.

Boils down to the cost of the chips and ones expectation of lifespan..

Live fast, die young applies to LEDs as well..

And, yes eventually one will exceed the possible thermal transfer inherent in the aluminum base heat removal properties.. allowing the chips to cook themselves..

It's not lifespan, it's actual power dissipated on an ongoing basis. There are plenty of threads discussing the comparison of LED to MH to T5, and the consensus is that LEDs are much more efficient in generating PAR vs power. Plenty of DIYers use a conversion ratio - 400W MH ... That means 120W LED should do the trick... But this shows that that's not the case. In some cases, they measure PAR and determine that their lack of growth is because they didn't have enough LEDs... I'm saying that it's not the number or power, but the grade of the LEDs. So- it's not "longevity" that's compromised: it's upfront "effectiveness".

 

[karimwassef]

Back to the original purpose of this thread...

Now that we know that a drop of more than 34V at 3A (or 34.8V at 3.5A) indicates a high resistance, poor efficiency LED (still asking and collecting PAR data ), the original sure buck or theatrus bus should work.

I'm still getting the boosts to experiment with higher voltage higher current chips but don't need dimming right now.

 

[karimwassef]

Another learning is that higher power LED chips (5A) will run at equivalently higher voltages. My contact shared two UV chips at 365nm with the same output (in Lumens) but one operating at 62V (310W) and the other at 82V (410W). The 310W being significantly superior, but (or maybe "and") much higher cost.

So 310W ~ $320 and 410W ~ $250... If you equate power with light output, you'd go for the cheaper "better" deal. But since they both generate the same light at 5A, the excess 100W is pure operating cost (setting reliability aside for a sec).

So running 18hr/days x 100W x $0.11/kWh x 365 days ~ $72 extra operating cost in heat alone. So the cheaper chip is the more expensive chip...

That's without adding back the cost of removing the heat from the LED, then the room, then carbon footprint, etc... And then the higher cost of replacement due to the higher operating junction temperature...

This isn't to say that the most efficient coolest LED is always best- it does say that the only way the lower cost parts make sense is if they're much lower. So, if the lower cost chip was $100 ($220 less), the answer reverses. It's now possible to replace twice as often and accept the cost of excess power.

Hard to assess cost of carbon footprint and wasting energy though.... That's a political battle though.

 

[karimwassef]

photoshop is acting funny. Here are the missing pictures from above



<a href="http://s1062.photobucket.com/user/karimwassef/media/0_zpsttugqseg.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/0_zpsttugqseg.png" border="0" alt=" photo 0_zpsttugqseg.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/0_zpslab8evck.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/0_zpslab8evck.png" border="0" alt=" photo 0_zpslab8evck.png"/></a>

 

[karimwassef]

Ok. Here's the data so far.

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/11_zps7sy4m9jj.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/11_zps7sy4m9jj.png" border="0" alt=" photo 11_zps7sy4m9jj.png"/></a>

I put it in a table and added some assumptions and calculations

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/00_zpsyvaixsxc.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/00_zpsyvaixsxc.png" border="0" alt=" photo 00_zpsyvaixsxc.png"/></a>

I still haven't switched over to theatrus's buck or measured PAR. I still have one more of the original LEDs that I haven't tested because I'm a heatsink short.