In my experience, in Advanced mode, the first reading is most of the time inaccurate. I always take the the 2nd or 3rd reading as the actual value. IMO, internal voltage readings will be less accurate from volt meter readings due to factors such as heat. Not sure this is an issue with anduril itself
Either VCC or voltage divider depending on the driver. FET/7135 drivers from VCC with a RPP diode and a Low pass filter (usually 4.7R). Regulated drivers like Emisar/Noctigon or Fireflies Lume1 have a LDO, so they use a voltage divider.
VCC reading on my drivers is always accurate (only a LPF before the MCU, but virtually no drop). Of course if measured just after turbo it can be lower than after a short rest. And indeed the issue than SammysHP mentionned if the memorised mode is high, since the light turns on during the 3C battcheck sequence.
I just check my Sofirn light battery check. Made sure it was in advance mode.
Meter read 4.165V, Battery check read 4.1V. Good enough for me.
Maybe it was due to using the light then checking, then checking the voltage the next day.
I didn’t realize it used a memorized mode to check. My memorized mode is 10/150, so it’s always a low level that shouldn’t affect Batt Check.
I typically go through at least 2 cycles simply because I don’t trust myself 
Did I count that right? Not sure about my SP10 but the other lights, including the TS21, seem to be accurate. When it moves it seems to be border line voltages or load influenced.
This has happened to me since the days of my first SP36. Anduril Batt Check is more a point of reference than a fact.
Edit: So due to this, I definitely do it a couple times to get a better baseline.
Never experienced this until Anduril 2 with power bank feature. A1 read consistently though sometimes not calibrated correctly.
I just experienced that weirdity today. Was charging my GTmicro, figure while I’m doing 14500s in my LK, I might as well check the TS10.
Blinked something like 3.6 (expected), then repeatedly 4.4! Every. Single. Time.
3.6, 4.4, 4.4, 4.4, 4.4.
Charged it after, just now tried it when reading this, and it was 4.1, then 4.2, 4.2, 4.2, etc.
At first I was wondering if maybe it would blink out the voltage, then temp? Didn’t consult the dox, so wasn’t sure.
Again, right before saving the text: 4.1, 4.2, 4.2, 4.2, 4.2, …
By constantly using the battery check, one after another. Will this drop the volts?
Not sure what you mean. Eventually your battery will be empty.
Every time you enter batt-check by pressing 3C you enable the light at ceiling level for a short time (after the second click). This will drop the voltage for a short time and the first readout will report a lower value. After entering batt-check the ADC buffer should be cleared and a new measurement started before voltage is blinked out. Sadly everything happens asynchronously in the background, so it’s not easy to synchronize without many code changes.
Not true. My light output maintains the memorized level at 10/150 until the batt check function begins to blink out the reading. 10/150 means, there are 150 steps (as I understand it ) between the bottom of the floor (lowest output level) and the top (Turbo). My light has a memorized level for turning on at 10 steps above the floor.
Oh yeah, you’re right! It starts with the memorized brightness (from 1C). Ceiling is only entered after the button timeout of 2C.
Does it draws power from the battery to do the battery check? Constantly checking will drain the battery.
14500 batteries can take a few seconds longer to recover from sag from a higher setting then an 18650. Or any single battery from a really high setting. Something entirely different I’ve seen on my DMM is when I first put the leads on any battery whether it’s a lithium ion or car battery. Initially and for a very split second the voltage readout will be slightly higher than what it changes to after a full second of being on the terminals. A Lithium-Ion battery sitting at 4.0v might show 4.25v. I’m pretty sure I’ve seen the same thing on 12 volt car batteries also even with two different meters. But with 14500, unless you were in moonlight or have a very low memory set always let it flash two or three times. Put a 14500 light on turbo for 10 seconds and then shut it off quickly and then quickly do a battery check and with some lights and batteries you’ll see some serious voltage sag recovery.
Yes, the main emitters blink and the microcontroller is fully awake. When the light is off the microcontroller is in sleep mode and requires much less power.
But there’s some load at the other end of the leads? A DMM passes practically no current in voltage measurement mode. If it sill shows a higher value, it is a glitch in the DMM (or in its auto ranging).