I wasn't involved in these lights, but certainly I've seen big variations from MCU to MCU, like +/- 15C (maybe more). That's a 30C+ range which is huge.
So if they aren't individually calibrated, roll the dice. What we really should do is force the use of sensor calibration, but the spec sheet says from a 1 point measurement, you will only see accuracy to +/- 10C. They recommend 2 point calibration, meaning 2 temp readings that are somewhat far apart.
The thermal algorithm in Anduril looks at the current temperature, the rate of change, and a sum of several recent readings, and makes a decision based on where it thinks the temperature will be in the near future. So if the light is heating up very quickly, it’ll predict a very high number and regulate downward. This is done to allow it to “steer into turns”, so to speak, on hot-rod lights like the Emisar D4.
Similarly, if the temperature is dropping quickly, it tries to adjust upward to keep the overall temperature close to the limit.
On top of that, it also has an extra cautionary clause on lights which heat up quickly. With this, at the slightest hint of overheating, it drops down to about 200% of the level which was found to be sustainable during testing… and doesn’t ever attempt to go back up above that level. It does normal regulation below that level though. So we end up with three ramp sections:
Lights with more thermal mass (like the Q8 and the SP36), or with less power (like the Emisar D1S) don’t have that top-most section… they do normal regulation all the way to the top. But it sounds like you’re seeing the quick ramp-down from turbo to high, which is triggered mostly by heating up quickly.
At some point I want to completely rewrite it, but it’s going to be a long, tedious, and frustrating process so I’ve been avoiding it.
IIRC, maukka received one with NarsilM, which uses a completely different thermal algorithm: Every N seconds, check current sensor value. If it’s higher than T, step down by a large amount. So maukka’s results here won’t be the same because his had different firmware.
Do you maybe know how to turn OFF blue leds?
I tried with 7 quick clicks but nothing happens.
Do I always have to unwind tailcap, is this the only solution? 
Please help.
Thank you.
My ROT66 219B is unfortunately already a month in transit.
But Jack did add a spare lens, swapped the driver for a 13*7135 one and turned the aux leds off with a screwdriver all for free which is nice.
Which batteries are the most suitable to power it? I’d rather not fry the leds by too high current. Are the VTC6 safe or is the NCRGA better?
Both are safe. The VTC6 should give higher measured lumens than the Panny GA 3500, if that’s what you want.
All packages from China have a good chance of delay from Nov-Feb due to high sale volumes and holidays.
It bothers me too, and I’ve advised against having any LEDs which can’t be turned off, but … that’s how it was made.
They’re hard-wired to power. To turn them off, unsolder the wires and optionally connect them to a different part of the driver.
I suppose you could remove the switch boot and then insert a circle of black paper into the top of the boot. the Switch may still light, but the paper should block the light so the button won’t light up.
Not good news. I did not know that blue leds are always ON. This is incredible. Realy bad decision from FF. 
I do not use flashligh everyday. Do you maybe know how much power consume these leds?
Now I will have to unwind tailcap each time I will stop using this FL. Crazy, such a nice FL with big problem (in my opinion).
Kind regards
I believe toykeeper did the math before on the aux leds, and said it would take more than a year or two to drain the batteries.
She got the parasitic drain down pretty low, and just having the 2 of 4 switch LED's on doesn't take much juice. No surprise, but someone could actually measure the drain and post, if not done so already. I'd guess the results would be a year or two. Of course a higher capacity cell would be the best.
EDIT: For a PL47, just measured 0.08 mA (Fluke 175 model measures to 1/100th's of mA), which is 80 uA. So for a 4000 mAh cell, 4000/0.08 = 50000, which is in hours, 2083 days, 5.7 years. Not sure how accurate this is, it's not my usual meter, but it could be in the ballpark. If the drain is the same on a ROT66, triple that to 17 years for 3 4000 mAh cells.
If you want to do some sanity test of your own, let it sit for 10 days or so and measure the cell before/after. Swap cells if you are gonna use it.
Are we talking about the 4 switch leds only or also the 12 main auxiliary leds in the front? I would imagine 16 leds total would be a pretty good drain.
Of course they also act like a battery indicator and can also shut themselves off if the battery voltage gets too low. Still, it seems like a twist for mechanical lockout is a good idea. I always mechanically lock out my lights when I’m not using them for more than a couple hours.
Oooops! On a PL47, 7 clicks, followed by 7 clicks turns off 2 of the 4 switch LED's and all the front AUX LED's - think the ROT66 works the same? My ROT66 was before the added AUX LED's.
Updated prior post for a PL47...
My ROT66 with AUX LED’s, cant switch off the AUX LED on the emitters.
Only difference 7-clicks does, is switch between two or four switch LED’s ON
I think it’s because of the low voltage warning the main auxiliary leds can give that they are wired up directly to the battery. The side switch lights are wired to the mcu which has regulated voltage (regulated on the 2S and 4S drivers, I’m not sure about the 3v drivers).
If you don’t mind losing the low voltage warning, I believe you can wire the aux led board to the side switch lights in parallel. Then the 7 clicks will control them both. I think 2 of the 4 side switch lights are still wired direct. Those can be disconnected or wired in to the others.
Just for the hell of it, I just measured my PL47 and some others. And hit exactly the same current. on the PL47 
PL47:
AUX ON: 0.566mA (4 switch leds on)
AUX OFF: 0.08mA (2 switch leds on)
ROT66:
AUX LOW: 0.28mA
AUX HI: 0.575mA
D4S:
AUX OFF: 0.04mA
AUX LOW: 0.085mA
AUX HI: 1,24mA
So for each of these 3 lights, 3 different AUX LED options, and on the ROT66, can't turn them off at the head? Oh boy...
You can always turn them off with a twist. Mechanical style. All e-switch lights have some amount of parasitic drain. The Thorfire S70S has a crazy high 4,430 micro amp drain and it only has a single green led switch light. Compare that to the Q8 at 123 microamps. So it’s best to always cut the power.
I don’t think the MCU is regulated.
Here’s a previous comment by Toykeeper about making it so you can shut off the aux LED board and all 4 button LED’s:
One possible option for turning off aux LED’s is a component called a bistable or latching reed switch. Some of those I’m finding spec sheets for are very small - about the size of a 1/4W resistor - but I’m not sure where to buy them.
It might be possible to wire one in series with the aux LED board, and if it can be located close to the lens, trigger it with a neodymium magnet.
Normal reed switches are momentary. Latching reed switches switch on or off depending on the direction of the magnetic field.
On Fireflies lights so far, the buttons have 4 LEDs:
… and the aux LED boards have more variety:
The 2 controllable button LEDs could have low and high modes, but this option isn’t compiled into the firmware because Fireflies didn’t want it. It doesn’t really make much sense when the other 2 are always on.
The aux LED boards typically have potentiometers to manually adjust the brightness, and the adjustment range tends to go from ~200 uA to ~800 uA. However, the pots are physically inaccessible on some lights, like the PL47.
The aux LED boards also typically have their own built-in LVP:
Because of the way this LVP function is implemented, and the way it attempts to regulate aux LED brightness, the aux LED boards cannot do low and high modes controlled by the MCU.
There is no LVP on the 2 hardwired button LEDs. Also, there is not yet any LVP in Anduril while it’s asleep, so if voltage drops too far during standby it won’t respond until after the user wakes it up. But the aux LED board LVP is always active, which helps a lot since it draws about 98% of the power during standby.