How hard can it be - fully regulated 3A buck driver

Flashlight Modding and DIY Parts
1

There are some very talented EE types on the forum.

I am only dangerous in the field but I know how to stuff things into small places.

I also know enough to understand the limitations and the challenges.

Here is my most important factor for a high power LED light; CONSISTENCY!
I “want” my 6V, 3A MT-G2 to get constant input until the last few minutes of battery life.
I have accomplished this with the standard NANJG drivers running a max of 1.05 amps.
I get full output into an XP-G or an XM-L until the very end.

How can we do this same thing with the MT-G2?

I am going to assume the following:

  • The 6V MT-G2 requires 6.2V to sustain 3 amps.
  • A 0.6V margin for the regulation circuit is insufficient for the V-drop across the regulators (nP2S).
  • Regulation can be sustained until each source cell reaches 3.4V where a good cell will drop off quickly (80%).

Conclusion with this information would suggest that a 2S arrangement for a 6V MT-G running full rated power without sagging would be challenging.

What about the 9V MT-G2 which requires nearly 9.3V to sustain the full 2 amp draw? Will 3S 11.1V work? Being only 2 amps draw, the cell will sustain 3.45V to the 80% mark, providing 10.35V or a full 1V for drop across the regulator. Are we now in the ballpark for this challenge?

Personally, I don’t mind a 3-cell fat body flashlight which also provides a nice cavity for the driver. Maybe even a 2P extension if runtime becomes an issue… and of course, the FATBOY if 3X 26650 is desired.

A smart design would also allow different configurations where it could drive multiple XM-L2’s with a simple component change. After all, we are talking 18 watts total capability which will drive 2X XM-L at full rated current. In series, 2 XM-L2’s is pretty much equivalent to a MT-G2.

So, what do you say… does anyone have some conclusions to these assumptions? Is pursuit of a 9V 2A regulated driver feasible for driving the MT-G2 within the Cree specified usable range?

A project like this is only reasonable if the technology can support it. And the one thing I cannot stand is a flashlight that doesn’t regulate output to the end (80% of cell capacity).

If someone is up to the challenge to design the circuit, I can follow up on packaging it.

Note: I am specifically leaving out the hot-rods here which the direct drive ambitions are already addressing. I am looking for a true current regulation circuit. Preferably a buck driver that is easy to configure.

2

Going to a 9v MTG2 and 3 cells puts you in exactly the same no-overhead boat as the 6v/2S. 6v with 3S is really the configuration to start with.

3

http://www.taskled.com/b3flex.shtml

http://www.taskled.com/h6flex.shtml

It's super easy when you don't mind dropping $40 for a single driver.

4

A lot’s been done this year in pursuit of a BLF buck driver but it takes time to start from the ground up as they are a quantum leap in complexity and difficulty over linear drivers. Some different threads are going on using 105C controllers piggybacked onto existing buck drivers with the stock control disabled and 2 members are trying to build up buck drivers with new boards. One is from the ground up and the other is more of a component swap to a new board design but in both cases there seem to be some areas where the board layout itself has a dramatic affect on performance and some other areas where a bit of well placed advice could be helpful. Since much effort goes into making one work reliably it would not surprise me at all if someone managed to make one and wanted to sell it instead of open sourcing it. The advantage of open source work is you can get others to help in design and testing and that can be costly in boards and components.

5

I am following the other builds with interest. It is almost like the 6V and 9V are a mismatch with the Li-Ion outputs.

In the linked video, the linear current regulators show about a 0.5V loss across 3 AMC7135’s at 1 amp.
I cannot envision the curve as Vf and C are varied. All I know is that you want to minimize the V-drop across the regulators.
At 6.2Vf (3A) across the emitter with 11.1V source just seems extremely inefficient, not to mention 1.8W across each device.

I do like the 105C mods and the explanation behind the AMC7135 source limits of 6Vs (7V absolute max from the manufacturer).
We seem to be operating in a niche environment where this may just be correct.

I haven’t seen the output curve for these 7135 configurations as yet driving a 6V MT-G2. I suspect the regulation will be much like the one in this video:

That is why I put the caveat on this post where regulation is a requirement. I would also hope that efficiency is thereby also preserved.
I prefer using a 3SnP configuration as long as the driver can manage the losses without heroic measures. To me, this just makes for a nice flashlight body.

As to open source, I can already see this evaporating on this forum of all places.
Fortunately, I have friends outside the forum that are also very talented and have nothing to protect.
I am hoping we can entertain this concept of true regulation circuits from a collaborative development sense rather than pure profit motives.
After all, we can reverse engineer any of the Chinese buck drivers and tweak them to fit a single solution consistently.
Its just a crude switching power supply with a current regulation circuit, right?

I went through this a couple of decades ago when MOSFET technology became “the bomb!” for RC cars. Same issues; different game.
One little device that would make or break winners.

6

7135s aren't anything to do with a buck driver, and wouldn't survive 12.6v input with a 6v MTG2. It's not that they'd be inefficient, they wouldn't survive long enough for it to matter, they'd flat out burn up. Forget about 7135s. A buck (properly designed) driver shouldn't be under any stress with a high input voltage and low output voltage, after all that's what a buck driver is for. Takes a high voltage low current and turns it into a lower voltage higher current.

7

This will take some of the mystery out of the buck circuit:

8

I think you misunderstand the hurdles here. We can design relatively low dropout buck circuits. It’s not a lack of headroom - if it was we’d be feeding 6v MT-G2 parts from a 3s supply and be done. Mostly we’re concentrating on small, high current solutions. Big ones and low current ones are already available.

I’m actually not really sure what you are looking for. The Knucklehead V3 can probably do what you want with a 6v part and can certainly do what you want for a 9v part on 4s. The HX-1175b can also do what you want. There may be several other buck options depending on the specific application.

Where exactly is the problem?

@RBD - actually both projects are straight rips of the application circuit from the controller datasheet!

9

I have worked quite a bit with the Maxim MAX16820 and I would highly recommend it for this application. You can’t use it for single-cell because of the 4.5V minimum, but for a 6V output it would be perfect. 2MHz switching can get your inductor pretty small

10

Thanks. That looks like a very straight forward solution. Really would prefer no sense R in the circuit but if the value comes out to be is low enough, this might be acceptable.

11

Thank Wight. I am simply after a good 3A current regulated driver that will hold up to 6.2V if the source remains above 3V/cell.

Size is less of an issue since I don’t need it to fit on a 17mm diameter PCB.
Again, I am measuring performance as constant output while the input varies from 12.6V to ~9V.

Seeing a test similar to the linked videos with the 6V MT-G2 at 3 amps to know if or when regulation fails would be very informative. Obviously the current meter would have to be between the emitter and the driver to see if it remains steady throughout the measured Vin range. I am just not getting that kind of feedback from the many MT-G2 posts.

12

Your OP seemed to frame the situation a little differently. It really didn’t make it clear that you are looking for… that kind of exhaustive data. I’m certainly not making any videos!

If all you want is regulation you can use a zener modded 7135 driver and 2s. You’ll stay in regulation with <200mV of headroom, that meets your goals right away. (I forget the precise dropout voltage.) EDIT: sorry, I forgot the comment that I was replying to mentioned 12v. To me the OP made it sound as if your ultimate goal would be to regulate a 6v emitter on 2s or a 9v emitter on 3s, but that no longer seems to be the case?

If you specifically want a buck circuit (for a higher Vin maybe), why don’t you compare the items from the OP in MRsDNF’s handy MT-G2 driver thread with the list of drivers reviewed by HKJ. There’s bound to be a couple of overlapping items. Actually I don’t think there are. The T6 driver HKJ tested is specifically what i had in mind, but the one comfychair used is different. FWIW I think all of those drivers are actually just different implementations of the QX9920 driver, so they should be in the same ballpark performance wise. At a glance the HX-1175b looks nearly identical in component selection to this driver which HKJ has done a comprehensive review on (the “T6 driver” I was referring to earlier). Note that HJK’s review is done with a “normal” voltage SST-90, not the higher voltage MT-G2. I don’t expect that to make a difference in the quality of the regulation.

Some of us may be more trusting than you. If I establish that I’ve got a functioning current regulated buck driver and then I establish it’s dropout voltage… I’m done! :slight_smile: I might want a couple of test points, but nothing beyond a 1v granularity (8/7/6v for example).

13

I was hoping for that 200mv drop as well, but that doesn’t seem to be the case. If you look at the videos carefully, you will see that at 1amp, the XM-L and the XP-G require a 2.95Vf (~3V) yet the regulation begins to drop off at about 3.5V source. This is a 500mv difference. I am not sure how to account for the discrepancy. And that is also why I am thinking the zener mod 105C will not regulate very long at 2S at a 3A source current.

Of course I am learning as the discussion moves forward. Therefore yes, the most likely candidate is a 6V MT-G2 with a 3S source. And clearly, this requires a buck driver.

I do see a lot of mixed feedback on buck drivers when it comes to comparing single or series emitters. Many seem to work harder when you add more emitters in series. The MT-G2 acts like a series emitter with the higher Vf requirements. So don’t think my overall skepticism is to unfounded.

When drivers cost $2 delivered to my door, I could test and toss whatever failed, but in this range, the cost of development adds up much quicker. I can design the right light from the ground up with the exception of the driver. As stated before, I can package an source a good design easily.

14

I could be wrong, but I assume you’re running into problems with Vdd being too low to (fully) turn the 7135’s on. You can probably simply bypass the reverse protection diode to avoid this problem.

Plus, I thought that you’d established in the OP that you had 600mV to play with. Why would it matter if the driver dropped out at 500mV?

… and also, a Zener modded Nanjg 105c is in exactly the same price range as the drivers you’ve been playing with. This forum is full of positive reports of using them to drive MT-G2 LEDs on 2s. I’m sure you realize that I want good buck drivers as badly as anyone, but you seem to be swimming upstream here.

15

The voltage drop on the Zener mod is largely due to how the 7135 operates. As Vb drops towards Vf the losses diminish(not to zero) and efficiency improves thus increasing time in regulation but won’t work with multi cell to single LEDs other than higher voltages LEDs like the mtg or series low voltage LEDs. Led4powers driver definitely improves on this by eliminating pwm but is limited to 1S input. I only mentioned the possibility of someone working behind the scenes to build one because I think it would sell and maybe someone with the knowledge and skill to design one unaided might think so too. There are also plenty here willing to share in the work on these projects and share the results.

Wight, I hope I didn’t misrepresent the case too badly. If I did, oops.

Right now, the Zener mod is a low cost, small package approach to a difficult problem.

16

RBD, while I think that everything you are saying is accurate, I do not think that you are actually addressing what I was discussing in the first two paragraphs in my post #13.

What NightSpy noticed is that his 105c drivers drop out of regulation much earlier than the AMC7135 datasheet suggests they should. What I’m explaining is that he may not actually be experiencing dropout from the AMC7135’s. In fact I think that he’s running into a problem where the MCU is unable to pull their “Enable” pin (Vdd) high enough. This is due to the combination of the output pins on the MCU being significantly lower than the Vin on the MCU and the fact that Vin is significantly lower than Vbat due to the protection diode.

How does that sound? Maybe we can be on the same page now ;-).

17

I can see that, Wight. That is easy enough to confirm with what I have on hand.
I definitely don’t want to make it more difficult than it needs to be.

18

Thanks wight, that makes sense to me now. Has anyone compared Vbatt to Vpwm at 100% duty cycle? Also, I think the drop across the diode varies with current so the situation gets worse with more chips added.

19

I’ve always use a 0.5V drop across a diode for ballpark. That is just about enough to toggle the 7135 Vdd prematurely if there is an additional drop through the MP. Line regulation for the 7135 requires 3V but Vdd can drop to 2.7V.

In this case, the S4 Schottky Vf typical is 0.38V for 1mA draw where 0.5 is typical at 40mA.

20

Can you elaborate on the part I bolded? It doesn’t make sense to me.