DIY Drivers


I would appreciate some feedback on making a driver for a headlamp I am planning on (I explore abandoned mines, I want as much light as I can get).

While my current 2x XM-L lamp is good, I want something brighter, and to keep this as a reliable backup. I currently use a TaskLED HipFlex to drive the 2x SM-L T5’s from a 3S2P/3S3P 18650 packs (Balanced on an iMax B6). It is very good, but overkill for what I need. I do like the 5 power levels, and the ability to switch into high and back to the previously set power level. Other than the voltage status LED, that’s all I want really. I plan on driving 4-6 XM-L2’s, or possibly run a pair of bare MP-G2’s with 2XM-L2’s on tight optics (MP-G2’s for walking around, close up photography, XM-L2’s for long-range throw).

I do have some electronics experience, and looking around, most things seem to be based on multiple AMC 7135’s, which strike me as rather inefficient compared to an inductor. That said, I know nothing about inductors, other than a hand-wavey basic overview. If I go down the Linear driver route, is there any reason I would not want to use say 2x 7150 instead of 8x 7135’s? I assume these are happy to have the Vcc pin PWM’d?

Any thoughts/ideas/links/totally different approaches appreciated.

I thinking “MP-G2” is a typo. Do you mean cree MT-G2?

Each 7135 gives 350 or 380mA of current, you will need a bunch for 4-6 XM-L2.
7135 drivers are used because they are cheap, easy & common. The attiny that does modes via pwm can also be reprogramed. Other microcontrollers can be substituted for the attiny if needed.
MT-G2 is higher voltage (6v) then most LEDs people use, a 7135 driver wont work for it.

You can modify the 7135 ATiny drivers to work with the 6V MT-G2. You swap the polarity diode for a 200 ohm resistor and stack a 4.3volt zener diode over MCU input capacitor. Perhaps not the optimal setup, but it can work well especially for lights that have limited driver space (17mm).

But what do you power the modded 7135 driver with? 2s is 8.4v and the 7135 supply voltage range is 2.7v-6v. Even within their range if 7135s don't have enough heatsinking the current is suppose to drop. Buck down to 6v and add much better heatsinking maybe?


I see you did a MT-G2 amc7135 build. Have you been able to check the current or driver temp?

7135 only sees the difference between Vf and supply voltage, which is 2-2.5 volts. They have no connection to the positive side of the circuit. The Vdd signal comes from the MCU, and if the zener/resistor trick drops the MCU voltage to a safe level, the Vdd will be at a safe level as well.

Upcoming wall-o-text warning…

I did indeed Typo the name, MT-G2 (6v) is what I meant. Powering the 7135’s (Should I go that way), is easy enough, many ways to regulate the voltage (DC-DC converter’s are the likely suspect However…

One of my goals is efficiency. I only have as much light as I can carry (Limited by power source, weight, bulk etc). Now, while I’m a big enough bloke, and I usually run a 3S2P pack (2.2AH Sanyo 18650’s) on the back of my helmet, and carry spares (3S2P or 3S3P), there is a distinct possibility that, should someone get into trouble underground, I could be there for a _ lot_ longer than I intended. It’s rare, but it does happen, and therefore I want to get as much from my cells as possible. Linear IC’s strike me as very inefficient compared to Inductors, as they convert any “power” sloughed off to heat. If I am on extended trips (think 2 or 3 days), or a “hard” trip (think nasty tight and crawly, long distances), then extending runtime’s and reducing weight become very large priorities. However, my current knowledge of Inductors is rather limited.

What I would like to do, (ignoring MT-G2’s for now), is to have 2 banks of 2S or 3S XM-L’s (Spot and Flood), and I want to be able to 1) Switch each bank in or out, and 2) Vary the apparent brightness of whichever banks are powered on. The driver will likely be mounted in a side box on my helmet, as my TaskLED one is currently, so adding a bank switch in addition to the normal mode switch is not a problem (Whereas it would be in a flashlight). Heatsinking is not an issue, as the box itself is aluminium, and I can add fins if needed. I would rather avoid this, as more heat=less light, and fins have a habit of getting clogged if there’s mud about.

While I do have some electronics knowledge, a lot of it is from school, which is 2 decades past, and I never used much of it out in the “real world”. I do remember some stuff, but looking at my notes for my A level project, I realise how much I have lost, as there isn’t a hope in hell I could even design a brief outline of my project, let alone design it (Glass density spectral analyser). I can do the basic’s, but Inductors are not something I’ve encountered since I started using/designing basic electronics again a couple of years ago.

I think in a “oh crap, this could kill me if it breaks” situation, the Taskled drivers are pretty much all I’d consider. They’re efficient, reliable and have all the features that you need. Compared to the overall cost of the light (LEDs, batteries etc), $40 or so on a driver isn’t much for all that :slight_smile:

If you’re not getting the output/ beam/ runtime that you want, it might be worth thinking how those aspects can be improved with the light you have or a similar sort of set up. Runtime could be increased simply by switching to newer cells, like the 3400mAh NCR18650B cells from Panasonic. Output and efficiency could be improved with the latest LEDs on copper, paralleling the LEDs (more LEDs at a lower current draw are more efficient than a couple of LEDs at a high current), stuff like that. Beam is something completely subjective and I’ve no idea what works underground, but there are an awful lot of options. Having multiple LEDs makes your job easier as you can then mix and match optics to get the beam you want.

It’s not quite as much fun as designing and making your own driver, but being stuck down a hole with no light doesn’t sound much fun either :slight_smile:

I agree with mat, I’d first look at updating emitters and cells to get the best you can from your current set up, or consider updated cells and emitters along with one of the *flex drivers for a new build, I’d also look at copper mounted emitters just to help get all the heat away from the emitters the best way possible, it sounds like really you just want to optimise what ever set up you settle on for efficiency, I thought the *flex drivers were supposed to be pretty efficient if you stick to the design parameters, it might be pretty hard to improve on that efficiency.

Thanks for the reply.

For reliability, at least initially, I’ll have my normal light with me. You are 100% right in that the TaskLED stuff is fantastic, and I, quite literally, trust my life with it.

I do want to make my own though, and as such I’m enquiring here to get a few pointers. I am looking intently at the Noctigon triples, should help with the thermal transfer as well. I have already switched the existing XM-L’s to parallel. they are each being driven (I presume, no measurements taken) at 1400mA (ish), as the current is limited to 2.8A by the hipflex. It gives a nicer light, but it’s not anywhere near bright enough, hence wanting a new light.

If it had the LEDs in series before, each LED in the string was getting 2.8A. Switching to parallel it's still putting out 2.8A but it's then divided equally between the parallel LEDs.

nickelflipper, dave_, pilotPTK and texaspyro have all designed their own drivers, so they’d be the ones I pester first. By all accounts, boost and buck drivers are quite a bit trickier to design than simple linear drivers, especially when they’re operated at low drive currents. I still haven’t even reprogrammed an ATtiny13 (yet!), but the guys I listed above should be able to help.

One place to start would be to look at a bunch of driver control chips and see what their reference designs are like. That might give you some clues aswell as narrowing down the field a bit. There’s a good chance that they’ll suggest particular controller chips too (often PICs), so you could start genning up on how to program those, if you don’t already know.