DIY LED driver for reef lighting

[jgwinner]

John, the problem is that we want to be able to do analog dimming from a controller and also be able to turn off the lights by supplying 0 voltage to that pin.

Oh and RTFDS is like RTFM for data sheets

ROFL on the RTFDS.

Ok, got it.

You could use a comparator circuit, and a small voltage divider to establish 'near 0'. The output of the comparator would easily drive a small signal transistor. I think.

What I was going to do was either use another digital out (I'm going to use GHI's Panda II), OR just turn off the power supply. This is when I was thinking of using 20v power supplies for laptops - we get a number of them back I could cut up.

But I'm torn between that and a big 24 or 28v P/S for 'all' of the strings. I guess I see your point though, it makes sense to turn off an individual string.

Is disabling the LED driver necessary?

0 on the PWM (well, 1) gives how much wasted current through the string? I thought earlier you said they would dim all the way dark.

== John ==​

 

[terahz]

John, the problem is the LED string. You see, unless you turn off the driver/PSU or "open" the Iadj pin, the driver will send current the LEDs. I get 8mA when there is no voltage on Vadj on the board, which is enough to produce light. If you turn off the PSU, that's not an issue.

After some fiddling around, a transistor directly on the adjust pin ... just isn't going to work. Cheapest opamp I can find is $0.80 (TLV271SN1T1G), that's like 10% increase in cost . I'll order a few along with the needed resistors to set the voltages and will give it a try. Thanks for the idea (had no idea you can use opamps this way, things you learn on reef central ).

Meanwhile, there are 2 options:
Use one more pin from the controller to turn on and off the driver board.
Put the PSU on a relay and control that relay from the controller.

 

[kcress]

Maybe you can use a transistor to kill the power once the dimming signal drops below some threshold.

 

[jgwinner]

I get 8mA when there is no voltage on Vadj on the board, which is enough to produce light.

Ah - that's disappointing.

In the meanwhile then I may get a couple of BuckPuck's just to get the first bank running.

I was actually thinking of using relays to switch particular banks of LED's in, but it would work best if the bank could be dimmed to 0 first, otherwise there will be a slight visible gap I think when each bank gets switched off.

Op Amps are amazing - they can do all kinds of weird stuff. I'm more of a digital guy than an analog guy, and now really all software, but check out the Op Amp cookbook series.

== John ==​

 

[ghellin]

Well I am having a different problem altogether with my cat 4101 board. My 24v power supplies are putting out anywhere from 29 to 32 volts I used the power supplies from web-tronics has anyone else had this issue? I am working with kcress on trying to find a solution, but if that doesn't work out I might need to buy a new supply. So where did everyone get there power supplies?

Here is the link to the ones I ordered.

http://www.web-tronics.com/pospu1524vsi.html

 

[TheFishMan65]

That seems odd unless you are drawing very little current, but it sounds like you are in good hands.

I found one (actually 2) on ebay. But mpja.com was a popular supplier early on.

 

[der_wille_zur_macht]

All,

I am mainly browsing the forums from a smartphone these days so finding info in these big threads is awkward at best. I need to drive some XM-L at 2A. I am aware of terahz's 3409 design and am likely to use that, but fishman reminded me that the CAT4101 can be paralleled. Has anyone actually done that?

 

[kcress]

the CAT4101 can be paralleled. Has anyone actually done that?

.
.
.
.
.
.
.

 

[der_wille_zur_macht]

Not a fan of that concept are we?

 

[kcress]

Not a fan of that concept are we?

No. Not in the least. It's pretty kludgy IMO. CATs already have lots of T issues and this just sticks two of them on a board to roast in togetherness.


I'd be looking at an active solution along the lines of terahz's.

 

[terahz]

Ok, so transistor just doesn't work. Tried with a few, but there seem to be some tricky states with saturation and such which I'm not familiar with.

Next I tried a comparator with an LM741 opamp from radioshack (grabbed the cheapest one). I put together the circuit and it doesn't work! the lowest I was able to set the reference voltage was ~2V! I started looking online, simulating in LTspice... everything suggested that it should work with much lower voltages. Second trip to radioshack for an LM324 quad opamp. Plugged it in, set the reference to 10mV, ran the pwm ramp up/ramp down demo behind a low pass filter, and the thing worked!!! When the voltage goes under 10mV, it switches off, above that, it is on!

Now the only question is, if the opamp I'm considering (TLV271SN1T1G) will work with low voltages. I guess one way to find out...

So there is hope, this might actually work! The only drawback I see is that setting the low cutoff voltage would be tricky with fixed resistors and I don't see adding a $2 trimpot for that. So from here we have 2 ways of addressing this:

A. Require a single 5V input (should be easy, since the analog dimming is designed for 5V signal anyway) to power the opamp and set the cutoff voltage
or
B. Put together a list of possible resistor values based on different power supplies (24V, 48V etc).

I tend to lean towards A due to the availability of 5V anyway.

Time to go to bed now, I'll order that tiny opamp tomorrow morning, hopefully it will come before the weekend so I can test and revise the board.

 

[liquidarts]

All,

I am mainly browsing the forums from a smartphone these days so finding info in these big threads is awkward at best. I need to drive some XM-L at 2A. I am aware of terahz's 3409 design and am likely to use that, but fishman reminded me that the CAT4101 can be paralleled. Has anyone actually done that?

DWZM,

I came across a design that had two channels paralleled and 4 other channels separate..I'll see if I can find it for you if your looking to try it out.

 

[O2Surplus]

Not a fan of that concept are we?

DWZM-

I've run the Cat4101's in parallel to drive some big BridgeLux BXRA-C4500's at 2 amps. I have one that's been running for nearly a year now, without any problems whatsoever. :thumbsup: I've got a DIY fixture retofit with 6 of those same leds driven by 12 cat4101's. I haven't completed it yet, but I did a 24 hour "burn-in" and it seems to work just fine. Here's a photo of the fixture. Each led array is rated for 5k lumens at 2.1 amps.

and the Cat4101 driver,( each driver board controls 2 led arrays)

 

[der_wille_zur_macht]

Thanks for the info guys. For now I am going to stick with terahz's design. Partially because I have been really interested in the 3409 for a while and just want an excuse to use it.

 

[terahz]

v0.4 of the LM3409 board is ready.

Changes:

• Added MCP6001T opamp with a couple of resistors to set the low voltage to about 20mA which is about 4 PWM steps. To get a lower low voltage put a smaller R9 (2K for 10mV, 1K for 5mV etc).
• Analog dimming requires 5V reference voltage (to power OpAmp, set low voltage limit etc).
• Changed screw terminals with molex KK connectors (from 3.5mm to 2.54mm spacing) to save a bit of space.
• BOM now includes the molex connector part numbers.
• Board is now ~2mm taller (wider?).

http://joro.geodar.com/code/LM3409 v0.4/
http://joro.geodar.com/code/LM3409 v0.4.zip

As before, for analog dimming, all components are required. For PWM dimming R4, R6, R7, R8, R9, C6 and U2 should be omitted.

To get a fairly good analog voltage out of an Arduino PWM signal, make a RC filter with 40K resistor and 20uF capacitor. That's what I've been using to test and the signal seems to be smooth enough (albeit slow).

 

[der_wille_zur_macht]

Every time I look at the 3409 and terahz's design I get more excited. This thing has tons of potential, we are just scratching the surface. It strikes me as incredibly flexible. Too bad I have a big stack of 4101 boars for my 360, because I think I could run the whole thing off two or three of these with parallel strings and a high drive current.

 

[Spaced Cowboy]

Well, I'm going to put my stash of 4101's to use soon(ish, probably Thanksgiving) when I get around to the lighting part of my new build.

One of the things I want to do is simulate sunrise/sunset across the tank, so I'll be arranging the LEDs in strips running front-to-back of the tank, and allowing each strip to be dimmed individually. Probably 12 LEDs per strip, 20 strips across the tank (it's an 8' tank, though only 24" deep).

I have some interesting ideas for the controller (both tank and LED) as well... But that would be telling

Simon

 

[der_wille_zur_macht]

Simon, looking forward to seeing your implementation. I am on the verge of selling my 4101 boards on the classifieds, as I could run my whole tank on two or three of the 3409 drivers if I spec'd them carefully. Beats having 8 or 10 4101's.

 

[Spaced Cowboy]

Simon, looking forward to seeing your implementation. I am on the verge of selling my 4101 boards on the classifieds, as I could run my whole tank on two or three of the 3409 drivers if I spec'd them carefully. Beats having 8 or 10 4101's.

Right, but you lose the fine-control if you bundle the LEDs together, right ? I haven't looked at the 3409 circuit, so tell me if I'm wrong

I'm going to have 5 or 6 "pods" over the tank, each pod will house 4 strips of 12 LEDs (ergo 8 CAT4101's) or possibly 4 strips of 6 LEDs (therefore 4 CAT4101's), depending on how much light I think I'll need. Irrespective of the number, there'll be a control module per pod which can daisy-chain down the tank. The user will then identify which pod is where in the array and be able to select control effects (storm lightning, sunrise, sunset, clouds, anything-else-I-can-think-of...). The system will figure out which LEDs should get which brightness at which instant based on its now-known location.

Simon

 

[der_wille_zur_macht]

You are correct with respect to bundling of control. I am accepting a low control resolution on this tank because (and this is clearly personal preference) when I experimented with things like cloud cover and sunrise with a high resolution (ie illuminating slamm sections of the tank individually) I thought the result was not worthwhile. So I am basically defaulting to two channels for the main lighting (blue andwhite) plus a third channel of accent lighting, which will be a small tight group of strategically placed LEDs at a big angle to produce a "shaft of light" effect during certain events. But now *I* am giving away too much.