IMX6ULL mem/idle power usage

[michiel_tjampen]

I've used the latest BSP to generate a bootloader and kernel, then based on the DTS from the evk and the default one from the manufacturer of the carrier board (attached) I made a DTS (attached) to suit my needs which are:

• one ethernet
• as many as possible UART
• two i²c

I think I disabled pretty much everything else. But power consumption is nowhere as low as i hoped...

• idle (eth0 'okay' but no cable attached, usb eth gadget) 953mW
• idle but eth0 and usb0 down  750mW
• standby  580mW (and 'mem' doesn't seem to make a noticable difference)

When using the default dtb from the carrier manufacturer (single uart,no i2c, both eth available)

• idle (both eth okay but no cable attached, usb eth gadget, single uart ok)  is 815mW
• idle both eth down, usb0 down 335mW

According to the power consumption app note i should use 

echo 1 > /sys/bus/platform/drivers/imx_busfreq/soc\:busfreq/enable

But this doesn't make a difference in the above scenario's.

So before i try to dig even deeper:

• is it normal that UART's make such a big difference to the 'idle' state?
• what would be the reason standby and mem don't make a difference?
• is there a document/table somewhere that shows things like 'if you enable this module, then this will be the result of on the overall current draw?

I've attached imx6ull-das which is the dtb for the uarts etc and imx6ull-seeed with is the default i'm referring to.

[igorpadykov]

Hi Michiel

• is it normal that UART's make such a big difference to the 'idle' state?

yes, if (for example) uart pins are connected to some external load like pull resistor.

• what would be the reason standby and mem don't make a difference?

reason may be wrong power measuring procedure which is thoroughly defined in NXP

documentation, like sect.3.7. Measuring points on the MX6ULL EVK platform

6ULL Power Consumption Application Note

Measuring Current in i.MX Applications

• is there a document/table somewhere that shows things like 'if you enable this module, then this will be the result of on the overall current draw?

sorry such documents are not available. In general some figures can be found in sect.4.1.5 Maximum Supply Currents

i.MX 6ULL Applications Processors for Consumer Products

Best regards
igor
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------

[michiel_tjampen]

According to the schematic 3 UART's are normally connected for another ethernet chip (for ENET2) but don't have pull ups, UART4 is the only one that has pull ups. But I might be using internal ones (don't understand dts file that well yet), so i'll check that.

Regarding power consumption in standby/mem. How can 'measuring wrong' be a cause? I measure the total current drawn by the board (which equals the 6ull package with ram,emmc and phy hardware)  either way i should see 'some' difference between mem and standby. Could something prevent it going into mem?

I've read the referenced application notes (before i made this thread) that's why i'm so surprised about what i measure. In it, the 'heavy load usecase' it's video playback over ethernet and the total consumption is about 0,5W... I already measure more than that idle! So where is my mistake/error?

[igorpadykov]

> How can 'measuring wrong' be a cause? I measure the total current drawn by the board 

this is wrong measuring, sorry. Measuring total board current does not make sense as

excessive current may be caused by external board components. So you should make

additional steps to put them in state preventing additional current.

Please reread carefully sect.3.7. Measuring points on the MX6ULL EVK platform

6ULL Power Consumption Application Note

Measuring Current in i.MX Applications

There are special testing points on EVK allowing to measure and verify processor current in

low power states. Note, other board components may be left in active state, though processor

is in low power state, causing overall board current to be not so small as expected.

>I've read the referenced application notes (before i made this thread) that's

>why i'm so surprised about what i measure.

May be suggested to test it on i.MX6ULL EVK NXP reference board.

MCIMX6ULL-EVK|i.MX6ULL Evaluation Kit | NXP 

If you are measuring on some third party board suggest to apply to its tech support

for additional help.

Best regards
igor

[michiel_tjampen]

Ok, so it was partially as you predicted. The repurposed pins are also connected to an ethernet PHY including the enable pins... and because all UART pins seem to have a default 100k keeper, they got pulled high.

Removing the chip fixed it.

Disabling the keeper's didn't seem to make any noticable difference (and nothing else is connected to the pins) but now i'm at about 250mW idle (eth's down). But still there's the issue of the mem/deep sleep.

echo mem > /sys/power/state increases total system to 315mW

echo standby > /sys/power/state drops it down to 180mW

What could cause echo mem to not work? Like i mentioned, i use the latest BSP. 

Furthermore, is there a document with suggested keeper/slewrate etc settings for the various modules?

[igorpadykov]

>What could cause echo mem to not work?

may be suggested to measure current on "correct" testing points as it is

made on EVK, not measuring total board current. This will allow to find

what part of board, chips or modules consume eccessive current.

Best regards
igor

[michiel_tjampen]

Ok, nevermind then, the board i'm using has all those covered with a shield, and i'm not planning on removing that.