DIY LED driver for reef lighting

[der_wille_zur_macht]

Any device capable of generating a PWM signal lower than 1khz in frequency and an "on" voltage between 2.4v and 25v will work fine. That leaves you with a VAST range of choices - all sorts of timers, microcontrollers, AD converters, purpose-build PWM generators, fan controllers, even other LED drivers. I mentioned the two specific methods above because that's what I have experience with. If anyone else reading this has specific ideas, please share!

 

[SpacedCowboy]

I know I keep promising this, but once I'm done testing the prototype of the driver, I will start a thread about arduino use for LED dimming. I've soldered the prototype up, and while it works, I'm going to make a few small tweaks (some of the hole sizes aren't QUITE right, for instance, and I'm going to turn up the isolation for the ground planes, since it's too easy for sloppy solderers like me to bridge to it). If anyone else out there is already making one of these, or thinking about making one, you might want to wait a day or two till I get the revised eagle files up - or, proceed at your own risk and double-check hole sizes and package outlines against the parts you'll be using.

Glad to hear it works - I was beginning to worry that re-routing it had screwed up somehow

If a few of the hole-sizes are only very slightly too small, that could be my fault. I think there was one drill required that I didn't have - 28mil IIRC. The closest drills I had were 31mil or 24 mil, and I went for the lower because I didn't want to drill away the entire pad... Needless to say, any board-house will have a larger selection of drills I don't think there were too many holes of that size, though.

Always good to increase separations between ground and traces (or traces and traces for that matter )

Simon

 

[der_wille_zur_macht]

The holes for the schottky diode were quite a bit too small - enough that I had to drill out the whole diameter of the pad to get it to fit. Other than that, there were a few components that were "squeaky tight" in their holes, so I'm going through the whole design top-to-bottom double checking hole sizes against the datasheets for the particular parts I'm using.

I want to "burn in" this thing for a few days, under a few different configurations (350mA, 500mA, 700mA, different inductors, etc.) before calling it "good" and ordering the PCBs from the board house, but yeah, so far it's working. The camera on my cell phone is freaking out right now but as soon as I dig out a real camera I'll post some photos.

 

[der_wille_zur_macht]

Thanks to silly parallel-resistor math mistakes, I can now state that both XR-E and Rebel LEDs will take 1.4A for a few minutes (the time it took me to ponder over why the LEDs were vastly brighter than they should be, measure the current on the LED string, re-do the math, and quickly turn the power off), and live to tell about it. :lol:

 

[stugray]

"I can now state that both XR-E and Rebel LEDs will take 1.4A for a few minutes"

Not in my experience. It was only 10 or so seconds with a Meanwell without turning it down first ;-)

That's why during my experimentation I have learned to use those Tyco connectors I have shown so many times.

All my LED strings have the same connectors so I have a little "harness" with two Tycos and two banana jacks.

When I plug it into a LED string & a DMM I read the current directly.

Stu

 

[der_wille_zur_macht]

Well, maybe it was quicker than that - but the math and the multimeter both showed 1.4A, and the LEDs are still viable.

At any rate, while resoldering the resistors, I must have made a solder bridge somewhere, because when I just plugged it back in, both ICs and one of the caps went poof. The ICs are burnt adjacent to their GND pins which is what makes me suspect a short somewhere.

I think it roasted my Arduino too, since it's acting funny (no surprise if I managed to put 24v on it's gnd pin I guess).

 

[kimimaro_kiddo]

Any device capable of generating a PWM signal lower than 1khz in frequency and an "on" voltage between 2.4v and 25v will work fine. That leaves you with a VAST range of choices - all sorts of timers, microcontrollers, AD converters, purpose-build PWM generators, fan controllers, even other LED drivers. I mentioned the two specific methods above because that's what I have experience with. If anyone else reading this has specific ideas, please share!

thanks a lot for the reply.. i have a doubt in mind to clear. please lead me.

you use NCP3066 in boost topology right? I have seen the spreadsheet of NCP3066 from onsemi.com. the spreadsheet given is in boost topology. when i first started, i was using NCP3065 before i realised it does not have built in on/off function. i followed the spreadsheet of NCP3065 which is in buck topology from onsemi.com for the external devices. then i read your post saying that NCP3066 and NCP3065 are interchangable.. my question is, can i use the spreadsheet of NCP3065 which is in buck topology to decided the external component values for NCP3066?

thanks

 

[der_wille_zur_macht]

The spreadsheets should have both topologies - look at the tabs in the bottom left corner. But yes, the chips are interchangeable other than the on/off pin - I've made a driver for one, then swapped the other in and it worked fine.

 

[kimimaro_kiddo]

I see.. thanks alot for your help..

 

[va7zt]

OK, here is a silly question in regards to power supplies.
When pwm dims the lights to off, is there still any draw on the power supply? Reason I am asking is I am debating on putting the power supplies on a relay to kill the power to them if the LED's are all off. If the draw is tiny then I won't bother but if they are still sucking a lot of juice I want to shut them off.

Built my first driver yesterday. Seeing spots still!!! I can't get over how bright 8 LED's can be! Thanks Willi for the assistance and ideas! Now on to building the other 14 and then I will tackle the arduino a little more in depth!

 

[der_wille_zur_macht]

The driver will still draw a very small quiescent current when fully off. I think it's in the datasheet. FWIW, I am planning on putting mine on a relay anyways.

Did you build a driver using this design?

 

[jma1978]

Won't the relay use more current during the time the lights are on and it's energized, than you'll be saving by killing power to the drivers while the lights are off?

 

[va7zt]

yup, I built them based on this design. I don't have the pwm hooked up yet. I was going to get them all built and tested first before I hook up the pwm and start testing the dimming.

Going with your boards will save a tonne of time. I'm doing it the hard way and it's a rather slow process and certainly doesn't look as neat as a proper board will. I was debating on using your boards but since I want 7 drivers per board I decided to do it the hard way since my breadboards fit the cases I bought.

Checked with the electrician at work today and he confirmed that when the lights are off there will be very little draw on the power supplies (just a little draw from the transformer) so I don't think I will bother with the relays on the power supplies at this time.

Built the second of 15 drivers tonight and managed to blind my neighbour who showed up just as I tested it!! I'm starting to see multiple benifits of these lights!! LOL

 

[AcroSteve]

Won't the relay use more current during the time the lights are on and it's energized, than you'll be saving by killing power to the drivers while the lights are off?

Use the normally open set of contacts so that when the relay is off, no power can flow.

Energize the relay to activate the lights.

 

[noobtothereef]

are you soldering this or using solder paste?

 

[noobtothereef]

is simon in this thread? i was wondering if he could prototype a pcb for me

 

[SpacedCowboy]

is simon in this thread? i was wondering if he could prototype a pcb for me

I am, but I'm not planning on becoming a board-house [grin]

If there's a one-off that's different to DWZM's design, and if it's small, and if it's easy to do, and (crucially) if I have the time, I might consider it. Ain't no promises, though . PM me, and let me have a look at the eagle files (.sch and .brd, because I'll probably have to re-route some stuff to make it milling-friendly. I usually do)

Be aware that I don't have plated-through-hole equipment (some day, [sigh]), so you have to manually do any vias on the board (I usually use an arbor-press to deform wire through them for prototypes, but you could solder wire through for more-permanent connections)

Simon.

 

[SpacedCowboy]

Some more info for the designers...

Some more info for the designers...

See http://www.national.com/vcm/NSC_Content/Files/en_US/Power/national_power_designer126.pdf

Sure, it's marketing for nat-semi, but it's technical marketing, and I think the graph on page one is interesting. From that, it seems that driving the LEDs at 500mA would put out little-to-no heat (compared with the higher currents - it's not a linear relationship). These look to be different LEDs (with lumens hitting 1000) but they're probably just the collections-of-LEDs-on-a-die, ie, just parallel versions of the ones we're using, so the current relationship is probably still valid.

Just sayin', 'cos it's easier when you don't have to actively cool things...

Simon

 

[der_wille_zur_macht]

See that, but don't think it's showing the whole picture. The blue area could be described with an x^n function, and I'm SURE there's a first-order n*x component (that they're not showing with the crossmarked blue) as well - right? At the least, I can definitely confirm that these LEDs will still be putting out enough heat at 500mA (and even 350mA) that I want goot heatsinking.

At any rate, the shape of the curve matches the curves in the datasheets for the LEDs we're using, which backs up that efficiency drops as current increases. But based on the values shown in the datasheets, we're probably talking about a 10 - 20% difference in efficiency from 350mA to 1000mA.

Simon I'm sure you've seen these but for the benefit of others:

Flux v drive current. Look at the slight downward curve. This matches the shape in the graph Simon is referencing:

Flux v junction temp. In a build where all you're changing is drive current, this will contribute to make the downward curve in the first graph more pronounced (junction temp will increase as current increases):

It's interesting that blues do better than whites w.r.t. junction temp vs. drive current. It's my understanding that a white LED is basically a blue LED with phosphors added to the lens, so that must contribute?

 

[der_wille_zur_macht]

are you soldering this or using solder paste?

One of the goals of this design, back on the first page, was to keep it "newcomer friendly" - hence it's through-hole components designed to be hand-soldered with a soldering iron. That's how I'm making mine, can't speak for others. If I had an oven and was going to reflow solder it, I'd just go all SMT.