DIY LED driver for reef lighting

[SpacedCowboy]

Ok, so I got home and tried it at 24v. It lasted about 10 minutes before the temperature got too high and the chip started to periodically cut the power to the LEDs. IIRC, that means the chip got to 150 degrees centigrade.

Remember this board isn't particularly well designed thermally - it's effectively only got 1 side of 1" square copper to dissipate the heat. You'd probably get better results with a proper double-sided copper "heatsink". There's also no active cooling.

I also have some TO-220 heatsinks lying around and I'll try putting one on top of the chip with some thermal paste and see if that helps matters. This is purely for academic interest as far as I'm concerned though - I plan to run it at the lower voltage anyway.

Simon

 

[der_wille_zur_macht]

Ok, so I got home and tried it at 24v. It lasted about 10 minutes before the temperature got too high and the chip started to periodically cut the power to the LEDs. IIRC, that means the chip got to 150 degrees centigrade.

Interesting - the datasheet's example calculation based on 1oz double sided and ambient of 60C gives 3w max dissipation. You are probably dissipating around that, at a lower ambient, right? So I guess it confirms that your thermal management isn't as good as the datasheet's example but probably not THAT far off.

Though I agree that it's probably all academic since it seems 90% of people doing this stuff are using the highly adjustable MPJA supplies which would let you trim down the voltage.

 

[SpacedCowboy]

Interesting - the datasheet's example calculation based on 1oz double sided and ambient of 60C gives 3w max dissipation. You are probably dissipating around that, at a lower ambient, right? So I guess it confirms that your thermal management isn't as good as the datasheet's example but probably not THAT far off.

Yeah, I was actually a bit surprised at how long it lasted - the temperature rose pretty slowly, but steadily. None of the other components seemed to be affected, it just started cutting out after a while. Contrast that with the ZXLD1366 letting the magic smoke out of the resistor...

I think the board-copper is only 0.5oz/side on my prototype actually, and of course I'm really not seeing the benefit of the second side because there's no easy path through to that side for the heat to take. All told, I think it did pretty well, and I like the obvious failure-mode where the LEDs start blinking at you

Simon.

 

[kcress]

1/2 oz.. pffft!

Single side heat dissipation.. pffft!

Oh, undoubtably, but the vias in question are beneath the chip itself.

A TO-263? There's yer problem - don't bother with vias under the part!
S u r r o u n d the part with vias.

For a temporary fix go to your hardware store and pick up a sheet of brass shim stock. Cut a postage stamp piece and solder it to the tab sticking straight up.

 

[TheFishMan65]

Parts list

Parts list

Well I have my design finalyzed I think. I have an EE friend laying it out for me. These are the parts I picked out (futureelectronics) - some are listed as call so I used mouser. If you have time I was wondering what you think of them.

--------------------- Item ----------------------- Quantity - Price - $ ----- Digikey - Mouser
MSB02005 ------------ Block Master 5.08mm 2 term -------- 5 - 0.205 -1.025
CAT4101TC-T75 ------- 1 AMP Constant Current LED Driver - 4 - 1.90 --- 7.60 -- 3.46 --- 1.90
PV36W103C01B00 ------ 10k Pot --------------------------- 4 - 0.80 --- 3.20
CRCW0603-6040FRT1E3 - 605 Ohm 1% SMT0603 ---------------- 4 - 0.0376 - 0.1504
REA010M1H0511 ------- 1uf Cap --------------------------- 2 - 0.0136 - 0.0272
LM7805CT ------------ 5V Regulator ---------------------- 1 - 0.35 --- 0.35 -- 0.60 --- 0.35

Total ---------------------------------------------------------------- 3.3062

I am not sure of the board price, but I think between 13-15. so under $20 for 4 drivers.

[EDIT}
I used one 1uf cap rather than 4 .1uf caps on the 24v supply - anyone see that as a problem.

 

[der_wille_zur_macht]

iirc, the datasheet's example schematic shows a .1uF on the 5v pin and a 1uF from 24v to gnd, essentially across the LED string.

How will you be assembling? If hand soldering, that 0603 resistor is going to be tough. At least it would be for me, maybe you have better eyes and a steadier hand!

And you're probably going to want decoupling caps on the 7805 voltage regulator.

 

[TheFishMan65]

I was going to put a 1uf cap on 5v and 1 on 24. I need to go reread the schematic from what you say I need four on the 24V.

[EDIT]
I picked on of the few resistor that you could buy by the ones. I may go to radio shack and see what they have. I didn't think I needed 5000.

[EDIT 2]
THe schematic that I did above is wrong it shows 2 .uf caps. When I referenced the data sheet I then transposed them. So how about 1 on 5 volts and 4 on the 24 volts.

 

[der_wille_zur_macht]

FWIW my prototypes put three circuits on one PCB, no voltage regulator onboard since I'll have 5v from elsewhere. I'm building the prototypes with the following components:

3 x 698-CAT4101TV-T75 CAT4101 $1.90 ea
3 x 660-RK73H2BTTD7680F 768R 1206 SMT 1/4w 1% resistor $0.05 ea
3 x 77-VJ1206Y104KXAMC .1uF 50v X7R 10% 1206 SMT ceramic capacitor $0.05 ea
3 x 80-C1206C105M5R 1uF 50v X7R 20% 1206 SMT ceramic capacitor $0.26 ea

I have headers for the prototypes; for "production" runs I'll probably use 3.5mm screw terminal headers, so maybe another $1.50.

The boards are costing $2.50 each (bought 5 at a time, more like $1.75/ea in high quantities.)

So using the "production quantity" cost for the PCBs, the cost for the driver is $5.51 for three circuits on one PCB, which will drive 18 LEDs. So around 30 cents per LED. If you include power supply cost (around 25 cents per LED), this design is HALF THE COST of the next cheapest, which is my original NCP3066 dual driver. Thanks Simon for digging this chip up!

 

[der_wille_zur_macht]

[EDIT 2]
THe schematic that I did above is wrong it shows 2 .uf caps. When I referenced the data sheet I then transposed them. So how about 1 on 5 volts and 4 on the 24 volts.

I'm kinda curious to see what other people have to say about this. I've just literally duplicated the circuit three times, so each IC has a .1uF cap on the 5v pin, and a 1uF cap across the LEDs. When you're cloning a circuit multiple times on a single PCB I'm not really sure what the "best practice" is.

 

[kcress]

I checked on metal clad boards. I should've thought about the pricing a little more first.

The stock costs about the same as regular board. The problem is that you need special drills and the entire etching process is V E R Y different since you aren't etching off copper you're etching off aluminum. doh!

Very few shops actually do that type as it means whole new production lines and cutters.

Willie; the only rules at shops is if they think you're doing an end run on them they will get torqued off at you. If they think you can just shear the boards and end up with three they often refuse to do the job.

 

[TheFishMan65]

That was one reason I used one large cap on the 5 volts. The same thing could be done on 24. In digital circuit each chip usually has a bypass capitor to keep the voltage to that chip solid. I don't think what we are doing is quit a critical as long as you keep the board small.

Where are you getting your boards for around $2 and in what quantity? Also where did you find parts? Mouser? I went with the businesses my company uses, but they may not be the cheapest.

 

[TheFishMan65]

Double post - sorry

 

[der_wille_zur_macht]

Where are you getting your boards for around $2 and in what quantity? Also where did you find parts? Mouser? I went with the businesses my company uses, but they may not be the cheapest.

seeedstudio.com for the boards - it's an electronics shop out of China that caters towards the Arduino crowd. They have a really obscure pricing structure that can work out really cheap or really expensive depending on quantity and board size. My design was slanted towards optimizing their pricing structure. I bought 5 for my prototype run, they cost $2.50 each more or less. For a "production" run of 30 - 40, it would be under $2 each, which is like 12 cents per square cm, about the best I've seen for counts this low.

Mouser for the components - part numbers linked in my list above. All parts are available in 1's. Almost all my other projects have been from digikey but mouser was SIGNIFICANTLY cheaper for this design (like 40%).

kcress - no complaints from the board house about it being three circuits on one PCB. They're sorta "tied" together since there's only one set of terminals for gnd, 5v, 24v, and pwm - so you couldn't just cut them into smaller boards.

 

[SpacedCowboy]

FWIW my prototypes put three circuits on one PCB, no voltage regulator onboard since I'll have 5v from elsewhere. I'm building the prototypes with the following components:

3 x 698-CAT4101TV-T75 CAT4101 $1.90 ea
3 x 660-RK73H2BTTD7680F 768R 1206 SMT 1/4w 1% resistor $0.05 ea
3 x 77-VJ1206Y104KXAMC .1uF 50v X7R 10% 1206 SMT ceramic capacitor $0.05 ea
3 x 80-C1206C105M5R 1uF 50v X7R 20% 1206 SMT ceramic capacitor $0.26 ea

I have headers for the prototypes; for "production" runs I'll probably use 3.5mm screw terminal headers, so maybe another $1.50.

The boards are costing $2.50 each (bought 5 at a time, more like $1.75/ea in high quantities.)

So using the "production quantity" cost for the PCBs, the cost for the driver is $5.51 for three circuits on one PCB, which will drive 18 LEDs. So around 30 cents per LED. If you include power supply cost (around 25 cents per LED), this design is HALF THE COST of the next cheapest, which is my original NCP3066 dual driver. Thanks Simon for digging this chip up!

Ok, I'm probably being dense here, but I don't understand your maths...

Component costs for 1 driver are ($1.90 + $0.05 + $0.05 + $0.26 + $1.50 + $2.50/3) = $4.58 (or $4.34 if you use $1.75/3 instead of $2.50/3).

If you put 3 of the CAT4101's together it comes to 3*$1.90 = $5.70, which is more than $5.51 on its own. :hmm3:

When I was working it out, for PCB, shipping, and parts (inc the terminals ) it came to $5.10/driver in quantity-10, $4.28 in quantity-100 and $2.41 in quantity-1000 (!) - I wasn't getting the crackerjack deal on the PCB that you're getting though...

Simon

 

[der_wille_zur_macht]

Crap, you're right, I can't add. :lol:

Still, I like this design the best. So the price per board of three is more like $12.50, which means cost per LED is around 70 cents. Plus 25 cents for the PS, and total cost to drive each LED is just under a buck, which makes it ~15 cents per LED cheaper than the NCP.

 

[TheFishMan65]

Thanks dwzm, I checked my board house and there size is under 10 inches. So am going with a least 6 and maybe 8 drivers per board. I may have trouble finding a PS if I use all 8 at 1 amp, but I can always not populate some of the drivers. Anyone know if these supplies a can be run in parallel? Or maybe have 2 24 volt supplies - board house may question it, but probably the better option? I think I have a minimum order of 5 boards. So if anyone is interested in one send me a PM.

 

[DmitriyY]

FishMan - you got PM

 

[marspeed]

FishMan

what are the details of your board also i think four drivers would be more usful and there would be no power supply problem 4x6 =24 6x6 =36 6x8 =48 LEDs eight drivers would only appeal to hobbyist with large tanks leaving many others out four drivers would make it more scalable (did i spell that right)

Marc

 

[der_wille_zur_macht]

FWIW I will also likely have extras once I order my production run for the big tank, if people want in. Especially those people who asked about NCP boards when I was planning on using those. No hard promises until I get the prototypes I've already ordered up and running for a while though.

 

[TheFishMan65]

marspeed,

The board is either going to have 6 or 8 drivers whatever I can fit on the maximum board size (might be 4, but I think more will fit). You can only put in as many drivers as you need. So just because it has N does not mean you have to have that many.

Not sure what you meant by power supply problems. It will always run at 24V (ok, read the thread more like 21-22). The current supply is the problem. If you drive all 6 at full power you would need 6 amps with a little for to 5V (.1 amp is plenty). So that is 6.1 amps, but from what I read we should drive the PS at 70-80% or it gets too hot. Which means a PS in the range of 7.6 -8.7 amps. I have heard MJPA makes an 8.3 (which would probably work), but I have not found it (link please if anyone has it).

I am planning on running mine much lower current so the 6.3 amp PS will work for me. I haven't done the math, but by running them cooler I can save on the heat sink and buy more LEDs. The XR-E at 350 ma puts out about 100lm at 1000ma about 200lm. So three times the power for twice the output, also as it gets hotter (a good heat sink will prevent this) they loose efficiency so they would not really put out 200lm. Also more LED will give me better coverage. Got all that?:spin2: