Extra 240mW draw coming out of LP1

We are working on an application where low power is critical. By turning of display (/sys/class/graphics/fb0/blank) and a bunch of other settings we are able to bring down the idle power consumption on Ixora (without the CAN trancivers) to something around 148mA at 12V.

Once is this state, we successfully put the TK1 into LP1. And after a few second woke it up using a Wake-on-LAN.

The interesting thing we found is that the current draw goes back to up 168mA instead of 148mA once the TK1 was woken up. It almost feels like some peripheral ends up in a different state after waking up from LP1. Any ideas what could be drawing the extra 20mA?

Here is a kernel log (taken from UART1) during a LP1/Wakeup sequence using RTC interrupt.

kern.log

Can you provide kernel log (that includes suspend and resume)?

Added in original post.

Can you also share output from cat /sys/kernel/debug/clock/clock_tree before and after suspend?

Good idea @dominik.tx. Here you go.

before LP1

after LP1

I did a cursory diff, and it seems hda2hdmi is “on” after waking up.

OK. I tried to get the clock_trees again on a freshly re-flashed board (the previous one was from a board that was actively being used for development).

before LP1

after LP1

Now the diff shows a bunch of differences.

Hi @notthetup
Could you double check the last files you commit?
According to the files, the module should consume less power after LP1.