Yes, the LED2001 has a large exposed pad area on the bottom. Have you run any of the TaskLED drivers at high power?
EDIT: here is the datasheet which shows pads etc:
http://www.st.com/web/en/resource/technical/document/datasheet/DM00079085.pdf
It has a pad almost the full size of the IC. I made a heat sink for it from a strip of copper, flat to go on top of the regulator and coiled on either side. Glued it on with Arctic Alumina Thermal Adhesive. On high the adhesive let go at about a minute. Well, the light stepped down from 3.6A to 1.76A at 36 seconds. Then again stepped down to .88A at about 45 seconds. And again a step down to .42 at around 55 seconds. Stabilized back up to .54 and ran as long as I’d leave it on.
This repeated in a second attempt, almost exactly. With it settling at .55A for as long as I cared to test it. The heat sink fell off, again. Too hot for the adhesive to hold it on if trying to run it on high.
Edit: Measurements on the negative lead about an inch from the emitter.
Edit II: The light flashed like 5 times before stepping down, each time.
Edit III: The heat sink I made is the width of the long side of the IC (1 to 4 pins) and it’s about 2 1/4 inches long. I clamped hemostats on it in the middle to keep that flat, then coiled the 2 ends to make it narrow enough to possibly fit into a light. The flat almost exactly matches the profile of the IC, with the 2 coils rising up above it. But if it won’t stay on… Might have to figure out how to assist the thermal adhesive with a more active brace…
I think that just indicates a bad bond; I’m pretty sure the adhesive should hold. Try clamping a heatsink?
I remove this adhesive by holding the iron against it, when it’s hot enough it lets go.
I clamped the heat sink onto the IC with stats, waited til the adhesive left on the wax paper was solidified, then turned the light on to heat it up a bit and speed the process. Should’ve been good, maybe I rushed it. 
I looked at the Oshpark board and the data sheet and yes it does have a thermal pad. If this pad area is too small Matt might need to shift some components around to make it bigger and /or accessible so that more sinking can be added. It looks like this chip is designed to dump heat so maybe it’s not a firmware or component issue but board design. No offense intended Matt just looking for anything that might help.
The datasheet does not discuss flashing. Sounds like an LVP issue.
Dale, are you soldering to the large center pad under the IC or just the pins?
I compiled that firmware with a lower ADC threshold to accommodate the 100k/7k divider. It’s possible that change was enough to cause an issue with the floating ADC that RMM would not have run into previously.
EDIT: If you want just change ADC_LOW and ADC_CRIT to 0 and recompile. That should prevent LVC from kicking in.
Dale, it sounds like you are hitting the low voltage threshold on the firmware itself by the stepdown. Go into the C file then comment out (add two slashes "//" before the voltmon section) then recompile and flash.
Even better, thanks.
Scott, yes, it’s soldered at all points via reflow. I make sure the solder is liquified, move the component with tweezers to make sure it’s got that wicking action going on, and then remove it from the plate.
The Voltage of the 2 cells is at 8.31V in series. Your’e telling me that you think these 2 cells are falling below 3V? These 2 cells are functioning quite nicely, would shock your heart back from a heart attack. I can use newer hotter cells if y’all would feel better about it. I can go with 3 or 4 cells if need be, I’ve only got a bit over a hundred Li-ions.
Nobody is saying that the cells are falling even close to 3V each, just that the attiny thinks they are.
The disconnected pin is giving an improper reading. Changing batteries will not help.
wight, speaking of the floating ADC pin...I just remembered that I usually just remove one resistor to disable monitoring and have never run without both! Doh!!!
So should I connect pin 7 and 8?
I just monitored the battery while running the light. With a start value of 8.16V it fell to 7.82V by the time the light flashed and dropped a level.
Dale, don't bridge anything!
Don’t use the single resistor technique either! Voltage is too high for that here.
Yes, I’ve reviewed and I’m recommending bridging 7 and 8 if you do not want to recompile.
??
To Bridge or not to Bridge, That is the question…
For to bridge shall surely cause one to carry on,
While not to bridge might guarantee a falter
So do we give this horse the reins?
Or pull back sharply on it’s halter?
Make it rein?