Here’s what I’m basing my recommendation on. Take a look at section 14.6.2 of the ATtiny13 datasheet.. It indicates that “Single ended channels that exceed VREF will result in codes close to 0x3FF”. Honestly I’m not really sure what “single ended” means, but I assume it applies here. The absolute maximum ratings section rates all IO pins as <VCC+0.5v. And I’ve done lots of bad things to Atmel chips without damage 0:). 0x3FF is the maximum 10-bit reading, this should apply for our reading as well.
While y’all were debating the merits of bridging pins 7 and 8, I created a strong attachment for the little heat sink. I twisted 4 strands of silver coated wire from 20ga teflon in order to create a strong “support cable” for the sink. I soldered one end into a ground via on one side of the board, protected a capacitor with a sleeve of the teflon coating and pulled the strand tight across the heat sink while soldering to the ground pad on the other side of the board. Then I soldered the “cable” to the heat sink itself. It’s solid and isn’t going anywhere. And it pulls heat from the soldering iron FAST! lol
Now this is quick and dirty, both heat-sink design and implementation as well as the photography, so bear with me…
Function over form, and all that, right? 
I’ll do better next time, promise! 
I defer to wight! (I think he's done far more research on the subject than I have!)
Not to confuse things but doesn’t running 2 cells to one led result in poor efficiency? If it’s as low as the Bflex then it might be only 85% which would mean close to 2W losses and the data sheets quote 40C/W above ambient. 2 cells to 2 LEDs(or 1 mt-g2) should result in better efficiency and lower ic temp. Just going by the numbers again for the Bflex losses for 2s-2s dropped to ~1.6W. I know you have a heat sink on top of the chip but even better would be a sink directly to the ground plane. A 17mm board should fit pretty well on a 1/2” copper pipe cap or coupler and help some.
That’s a pretty nifty heat sink. I think one end of R1 is tied to the pad under the IC and on the next iteration I would tie a fat copper wire to that. Actually, I’d solder the wires to the heat sink first, then glue down the sink, then solder to the board I’ve had bad luck soldering things already attached to the package.
I was actually thinking of slotting a post made from 5/16” brass rod to make a cylindrical heat sink.
Difficult to attach the inside of the board (inductor side) to anything flat like a cap or solid copper disc as the inductor is significantly taller than the regulator. One could always get creative of course…
108 C or F?
We’uns down roun these heuh pahrts use ’Merican!
The heat gun is trying to sense an area about 1” in diameter, so it’s averaging the board…can’t read the single IC where the heat’s coming from. Would need a thermal connection to do that.
The via directly below the sense resistor is directly associated with the large pad under the regulator. A copper transit could be brought up out of that via and tied to the heat sink to further the sink’s abilities to remove the problem.
Nobody’s come up with a good reason why I shouldn’t bridge 7 and 8 so here goes…
I meant the ground ring to the ring of the cap not the flat but the closer you can get to that pad under the chip the better and I think R1 is the closest. If there is a redesign of the board it might be well to shift R1 over towards L- and allow more of the pad to be exposed and extended towards C2 but I think it’s too early for that.
So F it is. Still, if it’s too hot to touch on the outside then it going to be much hotter inside the chip. One thing at a time though and sorry for the distraction but you might try adding a second series led to lower the losses in the IC.
Seems I’m now dead in the water again. lol
Except for the fact that the very reason I was in need of a buck driver was for a light that has one emitter and uses 2 cells. If I can’t make it work for my primary purpose, anything else is not really relative.
I probably fried the regulator when I soldered the wire across the heat sink. And now the problem is that I’ll have to de-populate the mcu side of the board to get the regulator off via re-flow.
Are we having fun yet?
Rufusbduck - We lose some efficiency (5-7%?) BUT having a large delta between battery pack voltage and LED string voltage allows us to maintain regulation throughout the discharge.
If bridging 7/8 made it stop operating but didn’t result in any sizzling, just unbridge them.
If you are hot-air reflowing do you really need to depopulate the other side before doing it?
EDIT: have you considered stripping off all changes to see if it still works as it was before? Meaning remove the wire you added, the heatsink, and the bridge on 7/8.
Yeah, that makes sense. I think it can be made to work with 2 cells to one led but I’m just not sure it’s up to that yet and I feel like we need to start with a set up that is less strain on the buck IC and work toward that goal. I’m just grasping at straws really, trying to help find some answers. As soon as RMM gets these in I can get some(with back up parts) and join in the fun.
There is absolutely no point in trying to make this run with one cell, it's two cells+ or bust.
I totally agree and I’m not suggesting that. Just suggesting we start with 2 cells to 2 LEDs to improve efficiency and ease the strain on the buck IC while we work things out. I could be way off base but the other high power buck drivers I’ve seen(Sharkbuck, Bflex, etc,)implemented some form of heat sinking dedicated to the buck IC solder point and this one seems to need help in that department.
I think Matt originally had the mtg with it’s higher Vf in mind for this board but I wouldn’t dream of suggesting the use of a $20 led for testing when 2 cheaper ones will do.
I understand your point RBD. Adding an LED or switching to an MT-G2 or just lowering the drive current would all be ways to reduce the strain on the LED2001. I guess none of us really want to admit that the LED2001 may not be up to the job!
On the other hand, most of these failures have something other than stress that we can point to for the failure. Bypassing the sense resistor, too much heat while reflowing, too high PWM.
Exactly, I just want it to survive so we can find out what needs tweaking to get the most out of it and that’s hard(and expensive)if it blows before we find out why. If we can do lower power tests that show us something without destroying it each time then we should be able to ask questions and eliminate possibilities a bit more frugally. I think we all want the same end result but maybe if we back it down a notch or two we might actually get there more quickly.
On the bright side, we’re learning what NOT to do!
I don’t have a hot-air re-flow station, I use a 6” dia. 3/16” thick 304 SS plate on a radiant heat glass stove top.
At your expense for now. I was kind of hoping a gb would spread this around and as soon as Rich has the parts it may yet do that.
If we do any testing without the at tiny maybe for that purpose only a board could be modded with a fat copper wire through the board to the buck IC pad since it should be possible to get at least a 10-12 awg wire(or copper nail) to fit the footprint w/out touching B+. If that alleviates the heat issues then it might help direct any board design changes. I’m still confident the chip can deliver 4A(it says so right on the box)but that much power means every I dotted and T crossed