DIY LED driver for reef lighting

I don't have a link to the paper, but I believe LED's work more efficiently when PWM dimmed than Analog dimmed (i.e. current control).

I would be really surprised if this was the case. Here is why:

At 350mA, the XPG is producing 350lm and has a Vfw of 3V. That is 1.05W or 333.(3) lm/W.
At 1500mA, the XPG is producing ~1155lm and has Vfw of ~3.45V. That is 5.175W or 223.2 lm/W.

So at higher currents the LEDs are less efficient.

Now, if we PWM that XPG at 1500mA to get 350lm, we'd have to do about 30% duty cycle. If a 5.175W led is on for 30% of the time and off 70% of the time it will use 1.5525W. And that is equal to 225.4 lm/W compared to the 333.(3)lm/W we got from a regular current dimming.

You can see that in this case PWM is more than 30% *less* efficient.

Now if someone managed to show that it is indeed more efficient, that would be an interesting read.
 
terahz,

Was your mosfet overheating that you opted for oversize heatsink?
I'm planing to run XMLs at 1.5A and wondering how hot does it get. My Vin =10V and Vout=6.4V (2 LEDs in series).
 
It wasn't overheating, just getting on the hot side with too much load. 2 LEDs at 1.5A shouldn't be a problem at all. I was heating it up with ~15 LEDs at 1A.
 
Not possible yet!
Anyway good thing someone is paying attention :) Of course all comes from the 350lm at 350ma :) Bad copy/paste. Sorry about that. Either way the % is correct.

Revised numbers:
@350mA - 130lm or 123.8lm/W
@1500mA - 130*330% = 429lm or 82.9lm/W
@1500mA and 31% PWM = 133lm or 1.60425W or again 82.9lm/W

So about 33% less efficient.

Will try to pay more attention when posting numbers :)
 
It all comes from the Relative Flux vs. Current plot in their datasheets. It isn't as bad in the middle of the current range. EG, 700mA, where it is only ~ 12% loss.
If you draw plot a line on top of the Relative Flux vs. Current plot at a 3.5:1 angle it will become more visible how much you're loosing at higher currents.
 
Interesting find terahz.

I think an important takeaway here is to not way overdesign your array with plans to to make up for it by dimming, at least if you are dimming with PWM...

I agree dwzm. I have some CAT4101 drivers setup to run close to 1000mA (560ohm Rset) and I was planning to use PWM dimming to lower the PAR to what I need. May need to rethink my design.
 
Interesting find terahz.



I agree dwzm. I have some CAT4101 drivers setup to run close to 1000mA (560ohm Rset) and I was planning to use PWM dimming to lower the PAR to what I need. May need to rethink my design.

You can always add a 10k pot inline with the Rset resistor, and adjust the current down to whatever output level you desire.
 
I don't understand how someone can know how to use EAGLE and still be sane!!!
sorry, just had to express my frustration... As it looks like everything else in regards to design was piece of cake comparing to Eagle.. :( :(
 
so how do we find that "money" spot? Of course i decided to use chinese LED's with no data sheet. (was in china at the time, and price was right .50).. manufacture says. 700mA, 3.5-4 volts per 10k LED, and 3.3 for the blues.
 
ok, so apparently im going to need to buy more resistors, i bought the 768's, how do i decide what resistors to buy for my LED's?
 
Buy a handful - they're cheap and easy to swap. Or, put a higher-current resistor on, dim until you find what you think will be the max you typically want, then swap for an equivalent resistor.

It's going to be hard to guess ahead of time given that you have "mystery" LEDs.
 
if you are going to use PWM, which looks like everyone is

Not I! Switching to your I2C boards, or something very similar to it.

I'm thinking about putting a spot on the board for resistors/fuses for parallel strings but the catch there is you'd be designing for a specific number of strings and it's not really a big deal to have those components external to the board. Does anyone else think this would be useful?
 
Back
Top