DIY LED driver for reef lighting

[terahz]

the 1uF ceramic on the outputs are sometimes hard to find/expensive in through hole.

Moser has a bunch of these at a good price if anyone is looking: http://mouser.com/ProductDetail/TDK/FK24Y5V1H105Z/?qs=sGAEpiMZZMuAYrNc52CMZAWzPNiQZc//V1wOjcVueZE=

Or for few more cents these:
http://mouser.com/ProductDetail/TDK/FK22X7R2A105K/?qs=sGAEpiMZZMuAYrNc52CMZAWzPNiQZc//6UgC%2bEoqfeA=

EDIT: the LED labels are a bit strange, aren't they? L2+ L2+ L11 and L1- ?

 

[TheFishMan65]

Also shouldn't there be a few more vias under the CAT4101?

 

[der_wille_zur_macht]

Oops. :lol:

Also swapped the cap packages, since the .1uF is more likely to be the smaller size and the 1uF is more likely to be the larger (as seen in the two examples terahz posted).

 

[cyrusthevirus]

I finished my LED driver.
Here is a pic of the circuit:

Here is a pic of the board:

It drives 16 LEDs at 750mA. It has a connection for a PWM signal and accepts a 58V 1A power supply. The circuit is very efficient and only requires a small heatsink on the MOSFET. I just ordered the PCBs and as DWZM says wou will have a few extra. I ordered mine from Gold Phoenix for only ~$109. I should get them within the next week. I only need 8 but will end up with more than 35. Hey they make great coasters.

I also attached the files gerber files.

 

[Obi-dad]

That's the color "DSO Nano" you're referring to, correct? Seems like they never have it in stock.

FWIW they also sell Jyetech's DSO kit. I like the concept of jye's scope better than the DSO Nano because it's a little more hackable. If you don't mind some obscene SMT soldering, it's only $33 in kit form. I just got one, put it together last night. It's pretty wild. Now I'm wandering around the house looking for things I can watch on it. This DSO is also available pre-assembled or partially assembled from a bunch of other vendors (Sparkfun, nkc, jyetech themselves, etc.)

That looks cool and can't beat the price. What box did you put it in?

Was the soldering any harder than the SMD components for the cat4101 drivers?

 

[der_wille_zur_macht]

It comes wit "panels" so it ends up kind of enclosed, no box required.

The soldering is definitely harder than on the CAT4101 drivers. There are about 30 SMT cap and resistors, mostly 0805, some 0603. There's also two relatively fine pitch ICs, though those aren't really THAT bad. It's definitely not impossible but you'll want a good magnifying glass, good lighting, a steady hand, and a way to hold extremely small parts (sharp tweezers and/or the blue-tac method).

cyrus, nice work! Next time you only need 8 of something, you should check out seeedstudio's prototype service - they sell in batches of 10 and the pricing is very reasonable.

 

[cyrusthevirus]

cyrus, nice work! Next time you only need 8 of something, you should check out seeedstudio's prototype service - they sell in batches of 10 and the pricing is very reasonable.

Wow, never heard of them. Any idea of their lead time? I can get them from Phoenix in about a week.

I also reviewed the circuit and it will work for any power supply as long and the remaining voltage after the forward voltage is between 1-2 volts. Anything greater would require a larger heatsink. The PWM is also logic level so it will work fine with as little as .8v

 

[der_wille_zur_macht]

I've used them for 6 or 7 projects in the last few months. Good quality, but turnaround is about a month. They don't manufacture themselves, they're just a middleman. They sell two "sizes" of boards, under 5cm square or under 10cm square. They are $20 or $40 for 10 units each respectively, which is DIRT CHEAP when you work out the per-area cost. Hence all the designs I've posted here are more or less designed to fit in the 5cm square size.

 

[nowlan]

dwzm: can you rotate the cat4101 up down or left/right?
Or do the chips not get that hot?

Also is that single sided board?
If not add some wire links?

 

[der_wille_zur_macht]

Dunno exactly what you mean by trying to rotate the part. As long as all the connections are there (including the ground tab), it should be fine.

The boards (my original 3x design and this dual design) are both double sided, but you could easily make them single sided by replacing all the bottom layer traces with jumpers.

 

[cyrusthevirus]

I finished my LED driver.
Here is a pic of the circuit:

I got a couple of PMs so here are the parts and costs:

Here is a list of the parts from Mouser.

Q1-844-IRL520PBF $1.34
Q2-512-2N3904_J18Z $.066
Q3-512-2N3906J61Z $.07
D1-512-1N5250B $.05
Heatsink for Q1 - 532-6237B $.36
R1-660-MFS1/4DCT52R1003 $.06
R2-282-0.82-RC $.19
R3-660-MF1/4DCT52R4701F $.05
R4-660-MFS1/4DCT52R1004 $.06
2P-Phoenix Blocks-651-1729128 $.72
3P-Phoenix Blocks-651-1729131 $1.17

Total: $4.136

Plus you will need a power supply. I use a 58V 1A one for $19.95.

Not bad. $5 for the board, $4.136 for the driver and ~$20 for the power supply. So ~$29 to drive 16 LEDs at 750mA with PWM support.

 

[cyrusthevirus]

I've used them for 6 or 7 projects in the last few months. Good quality, but turnaround is about a month. They don't manufacture themselves, they're just a middleman. They sell two "sizes" of boards, under 5cm square or under 10cm square. They are $20 or $40 for 10 units each respectively, which is DIRT CHEAP when you work out the per-area cost. Hence all the designs I've posted here are more or less designed to fit in the 5cm square size.

Have you ever used BatchPCB? I think their pricing and lead time are about the same.

 

[terahz]

Have you ever used BatchPCB? I think their pricing and lead time are about the same.

I used BatchPCB for the PH circuit that's on the Hydra and their price is higher because you get only 1-2 boards of your design vs 10 from seeedstudio. I used them because the size of the board I wanted didn't fit in 10cm x 10cm. Turnaround for me was about 6 weeks (I used them twice).

 

[der_wille_zur_macht]

Yes, I've used batchpcb. When you work out the price per square inch, batch PCB is at least four times more than Seeed. For the 5cm square size, Seed's price works out to $2 for 3.8 square inches, or 52 cents per square inch. For the 10cm square size, the price is like 25 cents per square inch. Batchpcb is $2.50 per square inch regardless of size, and plus there's the $10 handling charge per order.

Just about the only time I use batch PCB is when I really only need ONE of a small board. Otherwise, even if I only want two or three of something, seeed's price is better. Or, if I wanted like 20 or 30 of something, I'd probably use golden phoenix. But Seeed's prices are just so insanely good for the majority of what hobbyists want (small boards, small counts) IMHO they're a great choice for this stuff.

 

[kcress]

I often use pcb4u.com. It's not real clear but down on the bottom right they have a permanent special. For smaller boards it's $10 a piece up to about 20 sq-in. You have to order a minimum of 10. So for a $100 you get ten in about a week.

 

[Obi-dad]

Cool! Just got my PC boards from Hong Kong, less than 2 weeks since I ordered. Now the soldering fun begins.

 

[stephenrwright]

Cool! Just got my PC boards from Hong Kong, less than 2 weeks since I ordered. Now the soldering fun begins.

Thats good to hear, I just ordered some yesterday and was wondering just how log it will take to get them, I hate to wait, but I am cheap too.

 

[trag]

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.

I still have 18 pages of this thread to read, so I'm posting with the risk of being completely irrelevant....

Have you looked at the documentation for the Rebel LEDs? They have a recommended PCB layout with a pattern of plated through holes to a cooling back plane. It's interesting. Drill size is also covered.

See, e.g., page 11 of DS56.pdf or DS65.pdf (they both exist), which are datasheets for Luxeon Rebels. For more detail see their application brief AB-32.

While that design is aimed at cooling the Rebels, I imagine that the choice of drill hole size and hole spacing is probably close to the most efficient choices regardless of the specific application.

Also, note that they recommend .032" thick board and 2 oz. copper as being much more efficient thermally, than the more common .063/1oz.

I agree about keeping the back of the board smooth. In my mind, the back of the board should be something that one could bolt to the LED heat sink along with the LEDs. to keep it extra cool, but perhaps you have discussed that in the next couple of months and 18 pages. I'm not sure how well that idea will work in practice since vias create a risk of shorting if the solder mask doesn't have perfect coverage. Of course, one may always use an electrically insulating thermal pad.

I'm reading this thread because I have an idea for a heavy duty/versatile (as opposed to most affordable) DIY driver. But I want to see if my idea is already redundant. No point in reinventing the wheel.

Back to reading. See ya' soon...

 

[trag]

Yes I finally found it in several place. I reread while I had a chat session with them. The first answer was figure 1 shows only 25 volts. I referred to the power dissipation which is a function of voltage and current on the LED pin. I got him to admit it is confusing. So he is forwarding the question to a product engineer. So I will post when I get more info.

Okay, this time at the risk of being a month and a half behind...

The 4101 is not a true buck driver. It appears to be quite similar to the STCS1.

For the STCS1 the relevant bit in the datasheet is this equation:
===========================
The power dissipation in the device can be calculated as follow:
PD = (VDRAIN - VFB) x ILED + (VCC x ICC)
===========================

some discussion I posted in Nano-Reef
=================================
Vfb is always .1V (.2V for the 4101?) so we'll ignore it for our purposes here. Also Vcc X Icc is never much because Icc is small. So the meat of the equation is:

PD = Vdrain X ILED

In the case of running three LEDs off of a 24V supply, one would have:
Vdrain = 24V - 3(~3.5) = 24 - 10.5 = 13.5V

And at ILED = .7A, 1A and 1.5A this would result in a power dissipation :
Pd = 13.5V X .7A = 9.45W
Pd = 13.5V X 1A = 13.5W
Pd = 13.5V X 1.5A = 20.25W
========================================

Now, I know one of you posted the 4101's power dissipation equation from it's datasheet, but as I recall, they omitted the Vdrain X ILED term in the 4101's datasheet.

Editing: Ah, checked the 4101 datasheet and see that what they are calling VLED is what ST calls Vdrain. And you don't have a Vfb because the Rset pin on the 4101 operates rather differently than the FB pin on the ST part does.

This is the power dissipation you're fighting when your power supply voltage is very much above your summed forward voltages.

I (no longer) believe that the datasheet is in error. It is merely unclear.

 

[gringabot]

Okay, this time at the risk of being a month and a half behind...

The 4101 is not a true buck driver. It appears to be quite similar to the STCS1.

For the STCS1 the relevant bit in the datasheet is this equation:
===========================
The power dissipation in the device can be calculated as follow:
PD = (VDRAIN - VFB) x ILED + (VCC x ICC)
===========================

some discussion I posted in Nano-Reef
=================================
Vfb is always .1V (.2V for the 4101?) so we'll ignore it for our purposes here. Also Vcc X Icc is never much because Icc is small. So the meat of the equation is:

PD = Vdrain X ILED

In the case of running three LEDs off of a 24V supply, one would have:
Vdrain = 24V - 3(~3.5) = 24 - 10.5 = 13.5V

And at ILED = .7A, 1A and 1.5A this would result in a power dissipation :
Pd = 13.5V X .7A = 9.45W
Pd = 13.5V X 1A = 13.5W
Pd = 13.5V X 1.5A = 20.25W
========================================

Now, I know one of you posted the 4101's power dissipation equation from it's datasheet, but as I recall, they omitted the Vdrain X ILED term in the 4101's datasheet.

This is the power dissipation you're fighting when your power supply voltage is very much above your summed forward voltages.

I believe that the datasheet is in error.

how likely is it that you would run 3 leds off a 24v supply , 6 would be more likely giving a vdrain of 24 - 6(~3.5) = 24 - 21 = 3V