I read your blog. It’s great.
I look forward to seeing your power consumption reports when using the nPM1300 (Sleep 0.5uA) and your sketch posts.
’ The XIAO nRF54LM20A runs on nrf54-arduino-core - #11 by msfujino ’
’ Sleep Current of the XIAO nRF54LM20A Using the Built-in nPM1300 ’
Hi there,
SO,
Hold the sauce on that info… C’mon man! 
(espn) Steve
Blaming PlatformIO when you 
let Claude AI take the wheel on brand-new, bleeding-edge silicon? That’s like blaming the wrench because the apprentice tried using it upside down! 
The nRF54L architecture doesn’t have an official, packaged board manifest in PlatformIO’s central registry yet. When you ask an LLM to cook up a custom platformio.ini for pre-release hardware, it’s just going to hallucinate generic nRF52/Zephyr directives, toolchain paths, and upload flags, which is why it wouldn’t flash or boot cleanly.
The phantom wakeups were just early bugs in the underlying core library, which is why pulling the official v1.0.17 update 
in the Arduino Board Manager smoothed everything right out.
Glad you got the sleep cycle, GRTC, and TX power dialed in, though—that 50s sleep / 0.5s wake profile is where the LM20A really starts pulling away from the C6.
LQQK’ing forward to the final power numbers!
GL 
PJ 
In my rough experiment, under conditions where the LM20A advertised BME280 weather data five times at a transmission power of 8 dBm over a 0.5-second period and then slept for 59.5 seconds in the nPM1300’s hibernate mode, the average current over 60 seconds was 35 µA. It looks like it could run for over a month even on a 100 mAh battery.
For reference
Claude (who I like to think is named after the famous lion tamer, Claude Bottom ) actually found the officially recommended Seeed toolchain with a download from a Seeed website, so it was a fairly reasonable choice. Admittedly, some of his other choices have not been so great. He sometimes fantasises complex reasons for relatively simple bugs. He’s great for productivity but you do have to check everything and don’t let him code until you get him to feed back his understanding of requirements and proposed plan before giving a final go ahead. In that respect, he is exactly like some of the junior programmers who worked for me years ago. Maybe having had that experience helps?
Based on experience with my debug model, I’m estimated a possible 50 week battery life but I’ll be happy with 3 months.
I have to confess that I’m not using the battery monitor as there wasn’t room for battery status in the advertisement payload. My website will tell me after an hour when the battery has run out and it’s simple to recharge the device, so I decided that a major restructuring of the payload wasn’t worth the effort.