Help me design a cheap XM-L driver!

Hi guys,
New here and looking for some advice. I keep finding myself on this forum because, well, I’m cheap and like rigging things together. Looks like most of you do too.

Currently building an off-road light bar for my enduro bike and coming up with about a million questions. My test version is using an LM338T regulator with 3S XM-L’s at 2.5A. With the engine on and producing about 14V, this setup is about 70% efficient. All that excess goes away as heat.

I’ve decided with the latest low Vf XM-Ls it’s now possible to run 4 in series, as long as I’ve got a very low dropout driver. To stay in regulation we’re talking less than .2 or .3 volts. I don’t care about output with the engine off. Also I don’t need any fancy modes, just on/off and reliability.

The only one I know of and have played with is the AMC7135, a very popular choice among hobby guys. The issue is it’s very sensitive to excess voltage. I cooked one tonight by letting it get too hot. It started to go into thermal protection mode, and then poof, suddenly it’s giving WAY more current than it should.

Anyway, I got this idea. I’m a complete noob at designing circuits. Any input?

There are a bunch of multiple emitter driver designed to work in that voltage range. Try places like LCK-LED on eBay, (example) and intl-outdoor (another)

You'll definitely want to filter the supply chain though.

Hm well you have a point about buying a pre-fabbed unit. That second one looks pretty well made and is what I would consider cheap. Any idea how I’d remove the modes on it though?

I also just enjoy experimenting with stuff and learning how it works. I think that’s why we’re all here. When you say filter the supply chain, are you referring to a cap, or a transient voltage suppressor? I’ve seen both used to filter input signals of dangerous voltages.

A 12V zener across a 14V supply will just start a fire.

Lol, will it really? I was confused about that.

From what I learned on Wikipedia, 14.1V, the “minimum breakdown voltage” is where the diode starts to conduct excess voltage to ground. And then 16.5V is the absolute maximum, where anything above 16.5V will be grounded. I guessed that voltages between 14.1 and 16.5 are only partially dissipated.

Can you explain why this is horribly wrong? As you can see, the intended goal is to cut off any voltage that might damage the driver board. Maybe I’m just doing it wrong.

A zener regulator isn't the right choice for something that needs high current. Since the zener dumps any excess to ground there has to be a resistor inline to limit the current, which is exactly what you don't want in the supply to your LEDs.

What about a cheap switching DC-DC board to get your (sorta) variable supply voltage down to a constant safe voltage for the lights? Actually, if you pick the right one, that can be your driver, since you only need one mode. (not necessarily that one from that seller, just that particular model)

Output current & voltage independently adjustable. Wire the LEDs 2 in series, 2 series strings in parallel. Set the module's output current limiter to a little above where you want it, then use the voltage adjuster to set the actual current. It'll be running at 6.6-6.8v output, and double whatever current you want to each LED, since there's two strings in parallel. Easily within the capabilities of that converter board.

That DC-DC board looks like one of the nicer cheap jobs I’ve seen. The issue is it requires 1V greater supply voltage to stay in regulation.

In my diagram, the zener is intended to be a TVS. I guess they are the same thing? Ideally, it would be invisible during normal operation, and only sink excess voltages to ground. Factoring in some real world wiring losses, it should stay out of its “breakdown voltage” for 99% of the time.

The ultimate goal is to be able to run 4 XM-Ls in series at about 2.5 amps. It’s turning out to be rather difficult, especially not knowing as much as I’d like to.

Hi Veena!

Looking at the XM-L specs for 2.5A, VF is around 3.3V. For 4 XM-Ls you need ~13.2V DC power. If my calculation is right. :slight_smile: :beer:

That's why I recommended a 2S2P setup for the LEDs instead of 4S, lower output voltage but higher current. This would also give you the headroom to use more efficient XML2s instead of older XMLs and still have full regulation. For 2.5A per LED, you'd need an output from the converter of 6.8v & 5A, instead of 13.6v & 2.5A. Either setup would require the same input power and give the same light output.

Good point. I could do that too, since it’s a switcher the voltage difference isn’t quite as critical.

Yep! Which means in order to regulate 4 of them in series from 14.1V you need a very low dropout!

So, something interesting I discovered.

It is often assumed that when an engine is on and the alternator is powering the system, the voltage will be 14.1V or so.

However I noticed that when you load the system it can actually drop quite a bit. For example I noticed that I was only able to get about 12.5 to 13.5V with an 8A extra load. This may be approaching the limit of the alternator. 4 LEDs in series with even a direct drive would still fall out of regulation.

Yes, exactly why you'd be better off with 2S2P instead of 4S. (do you understand what I mean by that? - I keep saying it, and you keep going back to the problems with not enough voltage overhead... ask and I'll happily draw up a diagram) :)

Oh yes :slight_smile: I totally understand. See it’s not necessarily that I need to use 4 LEDs - I just want the most efficiency possible, whether its 1, 2, 3, 4, etc. LEDs in series. I think since 4 is basically out of the question, 3 is the next best option right? That would minimize the difference between supply and load voltage which is generally a good thing.

I really want to make my own driver still, except a real buck driver seems like the next best option now rather than the funny linear driver on the first post. Sure I can buy one but then you can never trust the quality, reliability, operating limits, or availability. It’s also fun to learn stuff! I want to learn how to make Gerber files with EagleCAD and design a PCB. What do you think?

I think there is no problem at all as long as you abandon the idea of a LED config that's too close to your supply voltage to allow proper regulation.

output: 6.8v @ 5A = 34W

Then work backwards to see what kind of input your desired output would require:

34W / 12.5v = 2.72A, plus a fudge factor for the converter's efficiency, let's say worst case 85% (they are much better than that in reality): 3.12A. Easy, simple, cheap, well within what your bike's electrical system can reliably supply.

Thanks for the schematic. Hmm, I sorta like this. I guess I never ran the numbers running 2s2p. A buck driver is really quite good compared to linear drivers. Maybe I’ll pick one of those drivers up to play with.

One thing though … they also make a constant voltage only version. I’m a little lost at why I should set the current first, and then set the voltage on the one you linked. Can you explain why this is? I would have assumed you set the current to what you want, and then the voltage you can leave maxed out.

Well, you would think that... but cranking up the voltage even though the current limit is set where you want makes the LEDs unhappy. The voltage adjuster will still increase the voltage even though the current remains limited to whatever you set. I think the reason is that it's either CV or CC, but not both at the same time. Before connecting the LEDs, just set the current to around 5.2A, then connect LEDs and adjust voltage to whatever required to get 5A output. Or, start with the current turned full up and voltage turned full down, increase the voltage until you get 5A, then turn down the current pot to find the spot where it starts reducing the current, and increase it by 1/4 turn.

There are cheaper, CV-only LM2596 boards out there too, but they're limited to 2A as-is or 3A with additional heatsinking, so your 4 LEDs would require two boards. The more expensive 10A version would be a more compact & easier solution.

On that 10A version the only hot parts are the two FETs on the backside so it's easy to mount to an aluminum plate or inside a sealed metal box, using the box itself as the heatsink. At only ~35W it won't make a great deal of heat anyway, but since it needs to be waterproof you might as well do it in a way that also keeps it nice and cool at the same time.

Maybe you can answer this for me. How does one go about designing one of these driver boards from scratch? Seems like there are hundreds, or thousands of variations. Is it a matter of finding a “main” component, and following its manufacturer datasheet for required components, or is every design simply a variation of one basic model?

I wish I knew... I think I'm pretty good at un-designing things, and sticking parts together at random until something interesting happens. But when it comes to laying things out from scratch (or writing code, especially that) I just go blank and my eyes glaze over. :tired:

Haha, I have a tendency to do that too. What happens when I encounter something I don’t understand, is I research it a LOT and constantly get that lost/sinking feeling when I don’t understand what I’m reading. I think it just takes time.

I think I’m going to start with a very simple circuit that I’ve been wanting to build, before trying something more complicated. Once I can do the process of designing a basic circuit I will expand it to something with 10-20 components. Optimization comes with experience I guess.

Thanks for helping me with this!

By the way, the current method I’m using to drive 3 XM-Ls from a 12-14V source is with an LM338T regulator. I’ve done testing to show that this is at most 75% efficient, and at worst 65% efficient. It’s a really stable method of current regulation. The current only fluctuated 0.4% at ~2500mA over an 80 deg F range (60 to 140F) and is unaffected by increased voltages over the minimum regulation voltage. It just makes a lot of heat.