I've got a lot more understanding now than I did last year so I would have to revise some of my earlier statements. I had lucked into a few things working, which means that they do work, but it doesn't mean that they were ideal. I definitely don't recommend anyone using a 47uH inductor on one of these unless you want a 1A to 2A output. You want something more like 8uH-15uH for the ~6A I was shooting for.
Some of my initial issues were due to overcurrent, plain and simple. I started playing with these at the same time that the vF on the XM-L2s started to change, lowering the maximum current significantly (You used to be able to get 7A out of most of them---now you are lucky to get a lot of them to survive at 6A).
Output caps: Yes, across the LED leads. This will greatly decrease the output ripple current while maintaining the same inductor ripple current. This is where you always put the capacitor(s) on a hysteretic constant current buck driver like this one, which uses the QX9920 IC, but it also applies to the MAX16820. You have to look not only at the capacity and overall ESR of the capacitor, but you also have to look at the ESR at a particular frequency. Generally, smaller capacitors are more effective at higher frequencies, while bigger capacitors are more effective at lower frequencies. Putting too much capacitance or high ESR capacitors across the terminals can actually increase output ripple (and will also decrease the dimming range). With an MT-G2 I wouldn't worry too much about adding anything, since it is a tough LED to kill. With an XM-L2 you can probably eek a bit more out of one without frying it, but they are still current limited. If you're using ceramics, I'd go with a 0.1uF and 1uF ceramic capacitor (you probably already have some 1uF capacitors laying around). You can parallel more capacitors up to a point, but the gain diminishes quickly.
I'm not sure about the inductor shielding. I'm not sure if foil is a good idea. On mine I was just putting some heat shrink tubing over the inductor to keep the wires physically farther away from it, but that isn't the same thing as really shielding the inductor.
If you are looking at changing inductors, I would use the formula from the TI datasheets or the the Maxim datasheets to determine what your switching frequency ends up at. You have some wiggle room with the factory inductor.
This is still a heck of a beefy driver for the money. It is easy to piggyback onto and can handle a lot of current. You can turn the output down a bit by changing the sense resistors (I=0.25/R). The main downside to this driver is that it is hard to turn into a good momentary driver due to the high quiescent current drain. You couldn't buy the parts to build one of these for the price you get these for, so they're still a sweet deal.
^ Thank you for more great info. Very helpful. I won't pretend to understand everything you said as you went deeper than my electronics knowledge base. I will be googling and trying to learn the stuff I don't understand. Nothing like a real world problem/challenge to encourage learning theory.
I've made about a half dozen lights with this driver and all of them are momentary switches. Definitely have to use discipline locking them out. All the drain appears to be in the up front voltage regulation that feeds the MCU and part of the buck converter. It's a very common system using diodes that seems to be in every buck driver I have encountered. Is that something you have addressed in your buck drivers that you are developing?
I'll look for my notes, but can't promise anything. I know I scribbled one out a long time ago, but I never go around to verifying it. It's hard to see the traces with the large copper traces on the battery side.
I have prepared one in easyeda and it is more or less I can see under the light and by the use of my digital multimeter, so I can share it to work it out together.
Sure. I can PM you my email address. Do you use Eagle? IIRC, I designed a 20mm and maybe a 17mm version of this buck. I designed it to use the parts from the HX-1175, but with a LDO voltage regulator to feed the MCU and Buck Converter. I never verified that the design would work though. I should have the schematics for those in Eagle format. If you want, I can send you those files.
I tried to find my Eagle files on the above boards last night. Apparently, I have changed computers at least once since I last touched them and did not pull over my Eagle files. Fortunately, I still have my previous 2 computers. I just need to dig them up. I will search their drives tonight. I sure hope I find them because I put a lot of work into those boards.
I'm not familiar with the program you mention above. Is there a free version of it available for download? Or at least just a viewer?
EDIT: Googled it. On-line and free. Cool. Take a look at your schematic and get back to you.
EDIT2: Just enough time for a casual observation for now. I don't know if your schematic need to have the same physical lay out as reality. If it does, Pins VDD and CS need to be swapped but connected as you show in your diagam (i.e. CS to R101 and VDD to C2 and the diode labled "S4" that goes to regulated 5v voltage regulation.
Just in case I can't file my Eagle Files. This should be close to how wired (except no MCU in below diagram) the boards I mentioned above.
The LM2936 is rather large. So I just use the TO-92 package. You don't have to have designated pads that way. Just run the legs to pads already on the board. Oriented sideways, it should be no taller than a through-hole inductor. I'm sure there are better voltage regulators, but I'm just comfortable with the LM2936. I have a thread on it here.
