DIY LED driver for reef lighting

[csa dad]

but still interested in setting up a diy led setu for my african cichlid tank.

 

[stugray]

DWZM,

"That means I can "fit" three on this size PCB in terms of thermal limitations."

Have you found any board houses that can do MCPCB ( Metal Core PCBs )?
If not ask around ;-) I'd be VERY interested.

Stu

 

[csa dad]

DWZM,

"That means I can "fit" three on this size PCB in terms of thermal limitations."

Have you found any board houses that can do MCPCB ( Metal Core PCBs )?
If not ask around ;-) I'd be VERY interested.

Stu

i like your avatar where did you find marvin

 

[TheFishMan65]

Our company has a metal core board they build. I could get more info, but they are very expensive I believe. I don't think you would get much cost savings. I will see what I can find out.

[EDIT]
And is that the Q36 space modulator

 

[der_wille_zur_macht]

DWZM,

"That means I can "fit" three on this size PCB in terms of thermal limitations."

Have you found any board houses that can do MCPCB ( Metal Core PCBs )?
If not ask around ;-) I'd be VERY interested.

Stu

Nope. And honestly for this particular application I'm not sure if it would really be worthwhile. I just laid out three of these on a 5cm square board, and there's really not a TON of room left, so even if I felt like the PCB could carry the thermal load of 4, I probably would just use three. I'm guessing this would scale to other board sizes - by the time you've done a "smallest" layout from a thermal perspective, it's probably a reasonably compact physical layout, too.

 

[TheFishMan65]

We order a complete board, but the estimage I got from an engineer is about $1 a square inch for metal core (in large quantity).

 

[stugray]

Who said it was JUST for this application? ;-)

Wouldnt it be cool to have one of your drivers on one side of a MCPCB, and 6 CREE Dies soldered to the other side?

Then we just make big bare thermal pads on the circuit side & screw down a little CPU heatsink behind each LED.

Instant "module" that you just hook DC power & PWM/enable to and go!
Then it would be cool to put a micro 8mm X 8mm fan on top, but they cost more than the whole driver....

Stu

 

[der_wille_zur_macht]

Stu, I agree, that would be super cool! But probably a little beyond what I want to do.

 

[TheFishMan65]

I agree, but you are going to have to make quite a few to make it worth while unless you are really good with surface mount parts.

 

[SpacedCowboy]

Simon, some more questions about your CAT4101 experiment:

1) Any chance you've tried to dim it? What are you feeding the EN/PWM pin?

I connected the leads to EN/PWM to the 5V supply to get 100% at the moment - remember this was the "why doesn't it work" PCB I made in case I had blown some of the SMT components [aside: you can actually see the SMT resistor (top) and capacitor (bottom) I was using on that first photo :-]

So no dimming yet, but I think the route in the datasheet / Fishman's approach is the one I'd take. I'm not sure if I'd use a 560R resistor (I *like* some headroom) but placing a fixed resistor in series with a variable one would be the way I'd go. If I wanted to get fancy I might even use an I2C-based digital pot

To be honest, I'm not expecting the PWM to be an issue - I think it'll be fine in terms of dimming for corals. I used to live in the Caribbean, and I still don't think the LED output is anything like the full power of the sun in those climes.

2) How are your lead lengths to/from the PCB? Thoughts about how extending them might cause problems? I know I've seen in several LED driver datasheets that additional capacitance is required across the LEDs and/or input if lead lengths are long. Thoughts?

One of the reasons I wanted to try building a non-SMT one was to see if that was an issue. In my experience, although the datasheets show best-practices, and following them is a *good* idea, they do often tend to be quite paranoid. I kept the traces relatively short (less than a centimeter) but I didn't try too hard to get them right next to the chip. It works just fine - well, the last time I checked it was late last night, and they were still working just fine - current draw was slightly down to 0.83A in fact.

[edit]
Ah, I think I misunderstood the question there: The leads from the PCB to the LEDs are about 3m long at the moment. This is my test setup so I figured it might as well be realistic. They're not any special wire either, just standard hookup wire coiling off the bench then over to the LEDs.

3) Thoughts on heat dissipation? If I follow my standard practice of cramming as many as I can "fit" on to a 5cm square PCB, that means I'm working with a little over 3 square inches. The datasheet uses 1 square inch (double sided) for heatsinking when calculating power dissipation. That means I can "fit" three on this size PCB in terms of thermal limitations. But if it's not really getting THAT warm, maybe 4 wouldn't be a stretch?

I think you'd be ok. I have ~1" square of copper on the top-side of the PCB, but there's very little thermal path through to the other (1" square) side of the PCB (no plated-through vias). I did place the via holes themselves, but that's not going to work as well as a real copper path. The chip was warm to the touch but not hot, and from what I recall, I'm running them at a slightly higher power rating than you intend to.

My personal use of them will be on a 1" x 1.5" board, because of how I intend to place them on the fitting. I think though, that by the time you've added the caps, the resistor and any connectors, you'll see that one square inch of board is roughly what you need anyway

Oh, and recall that the power-supply is matched perfectly in my set-up. It's a bench PSU, so I can turn a knob and see exactly where the current stops increasing as I increase the voltage; I run the circuit at that voltage. If you're running them at 24v (for example), you'll probably find the heat characteristics to be different...

which leads on to ...

4) Any chance you've cranked up the voltage to see what happens when the IC is dropping more than the bare minimum? I know that's not best practice to run it like that, but I'm wondering what it'll do.

Not for any period of time. I probably will, just so I have a good idea of the failure characteristics, but I'm still in the "can it run reliably for a week or so at constant output" stress-mode.

My plan is to find the voltage that works best for the LEDs, then trim the MPJA PSU down to that voltage - for the Rebels that's 20.7v, it'll probably be slightly higher for the Crees.

Simon.

 

[TheFishMan65]

Simon

You and I think a like. I figured I would really use about 600 for the resistor, but did not want to confuse anyone. I also thought of IIC pot, but did not think anyone else would be interested enough to mention it.

I too am interested in a 24v test to see how hot it gets, so when you get a moment ...

[EDIT]
I just looked at your picture. Can you tell me where (and part number) you got the terminals. Thanks

 

[SpacedCowboy]

Simon

I just looked at your picture. Can you tell me where (and part number) you got the terminals. Thanks

Those I bought from Farnell about 6 years ago when I was still living in the UK...

I did happen to be pricing out new ones though this weekend (I'm slowly running out of the old ones ) and I found these for $0.28 each, which look good.

Simon

 

[SpacedCowboy]

Thinking of using the ZXLD 1366 as i have already procured 16 nos of them. apart from the costing perspective, is there anything wrong with the ZXLD?

Nothing wrong with them as long as you fine-tune the input voltage so there's no waste - it's not particularly graceful about coping with over-voltage on the input. Any extra power just gets dumped into the input-side resistor which is easy to overheat and cause failure. The CAT4101 seems easier to work with, for me.

Simon

 

[terahz]

TheFishMan65 and DWZM,
Interesting thoughts on the PWM. I'm currently using regular buckpucks and according to their datasheet the rise and fall times of the output will be at 25 microseconds when you use them in strobe configuration which I'm guessing is the same as PWM from a controller. However I couldn't find the on/off times of the XR-Es. If they are faster than my controller speed (standard Arduino PWM is about 500Hz if I'm not wrong) then there probably is 0%/100% cycle out of the driver. However if they are slower than that (2ms - which I highly doubt) then there shouldn't be a problem.

Then again, as I said, my corals seem to be doing just fine and they've been going at 1A on 60% PWM for almost 4 months now so I'm happy.

 

[TheFishMan65]

Simon I think you and I are thinking a like. Is this the design you have in mind?

Then replicate the everything but the 7805 and C1. How many times are people interested in? I figure and even number and have two PWM (white and royal blue). Seems like a waste to have to add the 5V circuit every time, so I lean torward a larger number, but most people are getting the 24V at 6.5 amps supply from mjpa. Which might support six if not at a true 1 amp (upto about 866 ma). Any other good power supplies out there?

 

[SpacedCowboy]

Simon I think you and I are thinking a like. Is this the design you have in mind?

Pretty much, yes, modulo sharing the 7805 as you indicate below. The chip appears to take very little power to operate, so it ought to be safe to share many of them across one 7805.

Then replicate the everything but the 7805 and C1. How many times are people interested in? I figure and even number and have two PWM (white and royal blue). Seems like a waste to have to add the 5V circuit every time, so I lean torward a larger number, but most people are getting the 24V at 6.5 amps supply from mjpa. Which might support six if not at a true 1 amp (upto about 866 ma). Any other good power supplies out there?

I have two of the 8.3A MPJA supplies, and I happen to need 18 strings of 6 LEDs, so (9 x 0.85) = 7.65A, which works well for me. I might prefer slightly more headroom on the PSU, but I think that's enough.

Simon

 

[der_wille_zur_macht]

Just ordered a bunch of parts to try this IC myself. Will report back when I get it working.

Oh FWIW I switched from digikey to mouser for this order - the IC itself is like half the price, and some of the other components were cheaper, too.

My approach is that I've got abundant, well regulated +5v thanks to the Arduino I'll be using to control this whole mess, so no onboard voltage regulator for me.

Simon, I'm going to have slightly less than twice the number of LEDs you'll have, so I'll probably be right in the "nice" zone for four of those power supplies. I don't have any of the 8.3A units, but the 6.5A units really don't seem to like being loaded beyond 70% (they start getting VERY warm!)

 

[SpacedCowboy]

Simon, I'm going to have slightly less than twice the number of LEDs you'll have, so I'll probably be right in the "nice" zone for four of those power supplies. I don't have any of the 8.3A units, but the 6.5A units really don't seem to like being loaded beyond 70% (they start getting VERY warm!)

Hmm - well I guess I'll see when I start hooking them up to the real load. I can always get another one I suppose.

Oh FWIW I switched from digikey to mouser for this order - the IC itself is like half the price, and some of the other components were cheaper, too

Yeah, I mentioned that earlier - worth saying again though. For the 10-off price at mouser, you have to go to 10,000-off at digikey (!) Given the cost of the driver, it's actually cheaper (and not significantly more difficult to make) than the resistor-based one in this thread and you get the PWM option Really happy to have found this chip

Simon

 

[kcress]

perikaruppan; You don't want a linear. They give off lots of heat because they are very inefficient. Heck, you live in the land of power supplies. Probably 90% of the switching supplies on earth come out of India. There is NO problem with ripple going to LEDs. None what-so-ever. BTW Linear supplies are also non-trivial to design correctly, due to the major thermal dumping and the crest-factor.


TheFishMan65; I take your point on epileptics/response. However that's much different because they are responding to the 120Hz flicker and PWMing LEDs is done at many kHz.


Metal clad boards: Stu your driver/HP LEDs on a board is where I wanted to go but a few calls to board houses stopped that in its tracks. For a few boards the price was always way over $1,500 which is too rich for my tastes. I really don't understand why the MASSIVE premium exists. Isn't it just a different kind of feed stock? Why the big hit? I am sending a board out today. I will ask them why.

SpacedCowboy; You can poke any size copper wires thru those drilled vias and solder both sides and you will get better back-to-front cooling than a standard plated-thru via. You could do one board and you will see the difference front/back makes.

 

[SpacedCowboy]

SpacedCowboy; You can poke any size copper wires thru those drilled vias and solder both sides and you will get better back-to-front cooling than a standard plated-thru via. You could do one board and you will see the difference front/back makes.

Oh, undoubtably, but the vias in question are beneath the chip itself - not so easy to solder copper through them and still get a level surface for the chip to rest on [grin].

What I was trying to say was that I don't have a great cooling setup on the board I have, and the chip is still running pretty cool. All I have is a 3x6 grid of 0.018" holes, and I'd have thought you'd get better cooling if those holes were plated-copper rather than just bare holes.

Simon