LED drivers and Accessories you want, but don’t exist

True. It really depends on the LED you want to use. Buck-boost is good for lights with single high-Vf LEDs and a single Li-Ion cell. Buck is good for low-Vf LEDs. Linear is good for medium-Vf LEDs.

Some high-Vf LEDs: Cree XP-G2 (especially the S4 2B), Cree XM-L2, Cree XP-L, Samsung LH351D and the Osram Black Flat.

Great but, are you planning in somehow making this driver available to the DIY community? If so, let me throw in a few comments (bear in mind I've barely read anything in this thread LoL):

  • First of all, hope to see the driver in fully assembled form. From a DIY standpoint I can't see myself having a lot of fun having to pan reflow one of the board sides to later use a hot air gun for the other side while watching for components not to fall from side one over the table. Of course, I may have dramatized a bit, it is just to stress the kind of scarecustomer challenge it can be.
  • I am seeing large inductor pads. This means having to apply massive amounts of heat over the board in case the inductor is to be replaced. Can live with it but can't say I fully like that. Hope it helps with driver heat management.
  • Hope you are using a tiny sensing voltage, tiny enough for sense resistor heat to never be problematic. Improves efficiency too. Did you tried sensing voltage drop at the switching MOSFET? :-)

Cheers ^:)

A single cell buck/boost might work, with 2A maximum. We could put 2 of those in parallel to decrease conduction losses. Current-Sense Amplifiers that could be used for this are available.

Buck converters could be possible with low-Vf LEDs and an “oversized” Buck-IC (which is made for more current).

When you solder a board like that do you do one side with high melting point solder and then use lower temperature solder on the other side?

How close is this board to being the final version?

Yes, making it available for DIY is the plan, and it always was! I will try to make it like the TA-drivers.

Building the driver is not easy. But not in the way you describe it. The parts won’t fall off, you really would need a lot of time heating up the board to melt the other side, and even then, the surface tension of the solder should keep everything in place, except for the inductor.
The most painful part is to place the solder paste and the parts. It is a nightmare, because I don’t even use solderpaste stencils, I do everything with a syringe and a toothpick!

With the inductor, I may actually change the plan and use just one for both 6V and 12V LEDs, but let’s see how it turns out. You can somewhat “shield” the other parts with some aluminum foil or kapton tape while heating the inductor (the aluminum foil works pretty well with an IR solder station (which I was able to use once, and its way better than hot air)).

I’m using a 10mΩ current sense resistor, that’s 10mV/A. At 6A, that is 0.36W of heat. Sensing voltage on the switching FETs is possible, but not a good way. The resistance changes too much with the temperature (a bad characteristic from Silicon), and the next problem is that their switching on and off. Sensing at the inductor (with an RC filter) is the better idea, until it changes from PWM control at high loads to discontinuous mode at light loads. Then you can’t sense the current anymore. And the inductor current changes with the cell voltage.

Pretty close (I hope). The plan is that I will reflow one this weekend and see how it works.

But why shouldn’t 3A work? Zebralight and Armytek managed to do it (for years now).

Thanks. I sort of know that last part you say about the toothpick… and related time consuming stuff.

Here's something that will help keeping the tiny parts onboard:

Better grab another toothpick. ;-)

Cheers :-)

3A is possible as well, I just read the datasheet value. But now I see the 2A is for Vin >=2.5V
It depends on the input and output voltage, how much current is possible.

3A might be just on the edge with a 3.5V output, 3V input.
Certainly not easy with 2.7V minimum input like they say on their product page. They then go over the “minimum maximum switch current”, and use the typical max switch current instead (but it seems so work, has somebody heard of SC62s dying?)

Where do you folks publish schematics to these various designs? I’ve been messing with various synchronous switchers so it would be great to see what other folks are doing. So far I’ve only found the board layouts, and a few unfinished schematics elsewhere in the forums.

I avoid posting anything relevant if I’m not 100% sure it works, that’s why there are just some pictures here. I actually have worked some more on it, but I managed to kill my XHP50.2 within seconds, and now I haven’t ordered another one, and I have other more important stuff to do right now.
If you write me a PM I can send you a picture of a schematic.

Hey Snovotill, I have a page written up for my older GXB17 project here: http://loneoceans.com/labs/gxb17/. Like Schoki, the newer GXB172's page is still in progress :)

Nice update, thanks loneoceans. I see your RC at pin 18 has some unusual values. Is that compensating nicely for the high-side sensing via INA139?

I’m building one of these synchronous boost converters into my Coleman remote-phosphor 1500lm camping lantern. Lanterns seem to be lagging waaaay behind flashlights in terms of technology, probably because the average person has no idea about efficiency and battery life. I’ll be adding current regulation to one of these for that purpose:

“TPS61088 QC3.0 Boost Board”
https://www.ebay.com/itm/TPS61088-QC3-0-Boost-Board-3V-L-5V-9V-12V-98-High-Efficiency/272470324300?hash=item3f707f344c:g:UYoAAOSwZb5bMEKO:rk:3:pf:0

The COMP pin (18) is the output of the internal error amp in the TPS61088, where a few passives can be used for loop optimization. See the datasheet (pg 17/18) on how to choose those values. As is, the GXB17 with this topology works just fine but I prefer the simpler and cheaper implementation in my GXB172 driver which removes the digipot. The TPS61088 breakout board you linked looks nice, good luck with your project!

…but but but, how about in stead of buck and boost in one driver, have buck and linear in one driver?
Buck for fully charged Li-Ion, and go to linear when the battery voltage drops?
Would be nice for modern low Vf LEDs, and i assume simpler than buck boost.

(Just a thought)

But while i’m thinking out loud here, linear boost could be interesting too.
For high Vf LEDs or for LiFePo, CR123A or 2x AA batteries.

I’ve been thinking about that as well.
Resistor~~buck~~>linear->FET. Going from super-low to super-high with good efficiency at any mode.

You could take that one step farther and also eliminate the synchronous boost regulator. The old-school cheapo unregulated LED flashlight circuit shown below could be regulated by varying the DC bias into transistor M6 via the MCU D>A or PWM output pin. No need for a schottky diode so long as your inductance is large and so efficiency is very high. You'd need to sample the flyback frequency and battery voltage to determine actual LED current.The flyback frequency is linked to the hFE of the transistor which varies with operating temperature. hFE of course determines the peak inductor current in each cycle. I know it's not practical due to physical inductor size, but it's an interesting concept which would work for up to 300 lumens or so.

Hey, the drawing you see below is a base-64 encoded image stored directly within this post, so it will still be here a hundred years from now ;o)

Why not just make a buck-boost-linear-fet-dd driver with built in microwave?

a DIY kit for driver flashing

I like the microwave idea …

Jokers….
Once you have a simple low-power buck driver, adding a FET-based linear driver is a matter of like 20 extra components, mostly passives. That’s not a small change. Adds some board space, but not terribly a lot - LD4 is single-sided 17 mm and these components take about half of its board space. May require use of a MCU with more pins than the old SOIC8 ATTinys have. For that effort the payback is much higher regulated output as well as longer regulation.
Once you’re there adding unregulated mode costs nothing and pushes the peak further up.
And then adding a resistor cheap and reduces moonlight to ultra-low level.

Overall: Complex. Expensive. Unlikely to fit on a double-sided 17mm driver. But:

  • peak power is top notch
  • peak regulated power is top notch
  • bottom power is top notch
  • efficiency is top notch

W/out the buck driver, the points 1,2,3 would look the same in a simpler, smaller and cheaper driver. But efficiency at low modes won’t be top notch. So I view that as the high-end choice for 1S 3V lights where it’s not too large….which is quite a lot of lights.