Now THAT was funny! 
Wight- I’ll dig up that page and post it.
Cute, and AA some micron heat sinks into the new vent holes.
Link to data sheets see 7.2, Layout considerations. Most of this is Greek to me but enough soaks through to suggest it might be more likely to make a difference if we physically separate the two functional halves of the board.
OK, these are all considerations for what happens on the top side of the board - all within the buck circuit itself. I see no considerations which affect your suggestion of using 2 boards.
On the other hand, I do see one thing that differentiates the layout Mattaus put together from the layout example in section 7.2 / 7.5. I don’t know if it’s a problem though. Section 7.2 specifically mentions not connecting the analog ground pin and power ground pin. I don’t understand the text of their description (“underneath the output capacitor”). If you look at their layout the power ground goes directly to the fill where GND hooks up. The analog ground pin connects to the exposed pad, which is electrically the same but signal wise it’s a longer distance - for a signal to get between the two pins it must go through the backside of the board and come back out. I’d say it’s over 1/4in extra distance. Doesn’t seem like much, but they do specifically point it out.
Is it just me, or is DigiKey kind of strange to order from?
I knew I was getting 12 more of the Knucklehead boards. I knew I ordered 10 more of the Coilcraft inductors, direct from Coilcraft. I brought up the shopping cart from DigiKey with the parts on it, and thought I upped the numbers to match the boards. I did manage to get 25 more ATiny13A and 25 diodes for the BLF17DD boards. But somehow the order came in with my previous numbers of 3 of each of everything else.
That place confuses me!
So now I need to try again, after this storm blows over, as when it rains I lose my internet connection and we’re between thunderstorms at the moment. 
If it’s not one thing, it’s another.
Oops, found another problem. This is much worse I think, because we can’t just tweak the layout to fix it.
Section 2, Maximum ratings
Table 2. Absolute maximum ratings
Ptot / Power dissipation at TA < 60 °C / 2W
Well, that explains it. We’re way beyond that trying to drive an XM-L2 at 4A.
An XM-L2 needs 3.6v at 4A. That’s 14.4w of output. Input voltage to output voltage delta will make a big difference in efficiency, but I’ll estimate our efficiency using TaskLED’s 3A chart for the b3flex. That puts us at 85% efficiency or worse. I would say approximately 80, but we’ll use 85. At 85% efficiency your input is about 118% of your output. Thus your dissipation is about 18% of your output wattage. 14.4w * 0.18 = ~2.6w That’s bad.
Actual dissipation with the MT-G2 depends strongly on efficiency.
4A * 6.6v = 26.4w
at 90% effic that gives 2.9W
at 95% effic that gives 1.39W
~93% efficiency is the crossover point, at least for the absolute maximum rating of 2W
The cart linked in the OP has extra parts to meet minimums. Everything with qty 10 is just there to meet a minimum.
When I ordered before I got 3 of everything as I had 3 boards coming (with the exception of those minimums). When I re-ordered the other day I used my history and duplicated the order. I would have sworn that I changed each line item to 10 ea as I have 12 new boards coming. But it kept the 3 ea from the previous order. Each line, when changed, blinks over and saves the new quantity, one line at a time, I KNOW I went through that before. But only got 3 of each instead of the 10 I was expecting. At any rate, I just ordered 10 more of everything except the ATiny13A’s as I now have about 35 of the ATiny13A’s and 10 of the ATiny13’s on hand. It’s only money, right? 
I also added up the per ea. cost to get a total on a complete driver, board included, and it’s just under $10 ea.
Edit: And now I know we’ll be able to run these easily at a power level of about 135mA. Woohoo!
I took a look at DBCstm’s numbers from post #257 in order to eyeball our efficiency with a single XM-L (not XM-L2 I think?) LED. He did not post input voltage, but we can assume that it was 3.7-3.9v per cell under load for now. I’ll call it 7.6v for the battery voltage. The higher it was, the worse our results would be for the purpose of this calculation.
input:
1.75A * ~7.6V = 13.3W
output:
3.64A * 3.09V = ~11.2W
That’s about 84% efficiency. Due to the lower overall wattage of the XM-L vs an XM-L2 we aren’t looking at quite the terrible situation an XM-L2 would be, but it’s still around 2.128W.
The 2 cells started out measuring 8.16V when the light was first turned on (active reading during test) and fell to 7.82V at the point by the time the light started blinking and shifted down.
The static measurement of the 2 cells was 8.31V. The emitter was reading 3.71A about an inch from the emitter on the negative lead, with tail amp measuring 1.86A. Sorry, actual emitter amperage was indeed 3.71A, but it fell to 1.86A when the driver stepped down…that wasn’t a tail reading but a step down reading. Tail reading was flickering between 1.74 and 1.75A.
Good work, but that’s bad news for this driver. That puts your setup at approximately 82% efficiency. Which should mean that approximately 2.4W was being dissipated through the buck circuit. EDIT: OK so we don’t know the battery voltage under 1.75A load, no point speculating further. It’s bad either way.
With an XM-L2 output voltage would be a fair amount higher, this would have a positive effect on efficiency. Maybe 1-2% or more, I dunno. I think 3A on an XM-L2 may keep dissipation at about 2W. I’m not sure that “about 2W” is where we wanna be though, considering that that’s an absolute maximum rating.
Ok, so now, only getting low mode running twin XM-L’s would you say that’s probably in the ATiny13A? Might have damaged it removing it then re-installing it on the board? I can pull it and flash a new one with this new code of yours, try again?
Short PWM and VCC on the ATTINY (do not short PWM and BAT+ or risk frying the ATTINY). If that puts it in High then it’s a problem with the ATTINY, reflash. If it puts it in OFF then I got it backwards and you need to short PWM to GND on the ATTINY. Do not run either way over a period of time, but you can do it w/ a jumper wire for a few seconds at a time w/out fear of damage I think.
EDIT: There’s no reason not to re-flash, but this is just a faster way to test. If you want to reflash that’s fine too.
Tried that, with both LED’s lit up quite low I bridged pin 6 to pin 8…nothing. I then bridged pin 6 to ground and it shut down the lights.
Edit: Measuring from the negative lead connection on one LED to the positive lead connection on the second LED, my DMM shows 7.78V.
Something on the buck side is zapped. One Million Dollars™ says that it’s the LED2001.
EDIT: by “nothing” I assume that you mean brightness remained the same. That is to say, brightness was low and when you bridged 6 and 8 it remained low.
I pulled off the suspect one that I soldered the heat sink on top of, put a new one on, hand soldered the ATiny13A back on along with the other components necessary on that side of the board, wired 2 Led’s to it and it would only work dim right off the top.
Correct, bridging 6 to 8 made no difference in the output, not even a flicker.
- Do you have modes?
- Were you careful to place the correct components as usual?
- Measure MCU VCC to ground please.
EDIT: also what 2 LEDs do you have hooked up, just 2 random XM-L ones? Maybe total Vf is too high, but I really doubt it.
Mind you, some of this stuff is brand new to me, like hooking up 2 leds in series. Series being the way batteries are run, with the negative and positive together in the middle, the negative LED is connected at positive to the driver, the positive LED is connected at negative to the driver. Just like batteries would be. Right?
Driver () to LED () LED () to LED (-) LED () to driver (+)
Geesh, hold on, I got a pic…
I think the forum messed up your notation there. It’s just like batteries. You have it correct.
VCC to ground measures 3.31V. No modes, only the dim output. Seems to me like something must be shorting somewhere.
Edit: Freakin blind now! I removed one of the LED’s from the circuit, the driver works fine, 3 modes, high kicked my ass.
Edit II: I have a loose MTG2 on Noctigon, brand new, got it yesterday…
Edit III: I also have an XM-L2 on a Noctigon mounted in an HD2010 pill backed by copper that is a test emitter, would I fare better using that?