Deep sleep high current on Atom

[oklab-ka]

I'm testing deep sleep current for an Atom project. Atom starts, current is about 50mA. In deep sleep current drops to less than 1mA, which is what I expect. Atom awakes after 30 seconds to do its regular job, then goes to deep sleep again. However, this time (and all subsequent times) deep sleep current is about 11mA. When I press the reset button immediately after power on, the first deep sleep is also 11mA. I can also see a difference in the active current in the order of 11mA. There are no units connected. Whatever I do, disable Serial, I2C, Display, isolate GPIOs or not, there is no effect.
Unfortunately M5Stack doesn't provide schematics for the Atom, just a block diagram which misses out a lot of details.
Any ideas? Should I remove the RGB Led or the IR transmitter. My best guess is it is the FTDI chip coming out of sleep mode due to some fluctuation on the TX/RX lines. But without schematics I cannot do much. What I found on FTDI is something like 8-15mA, depending on the chip type. The only one in SSOP20 package is FT231X. Any help welcome. Below 1mA is good, but 11mA deep sleep current is a nightmare. There are many questions on the deep sleep power issue and I'd be very glad if M5 could do something to improve this ...

[felmue]

Hello @oklab-ka

just curious, are you powering M5Atom from 5 V (e.g. USB) or 3.3 V?

Thanks
Felix

[oklab-ka]

@felmue Tried both. 3.3V has a bit higher current. Strange, if there's just an LDO (guessing - no schematics!).
Also removed IR and RGB LEDs. Tried with Arduino and Espressif IDF deep-sleep example. No success. NodeMCU with Wroom32 is kind of OK (1.6mA), by the way. But mechanically too large ...