i.MX6UL rev 1.0 - VDD_SNVS_IN 400µA in Power-Off Mode

[d_lohse]

We measured the current drawn from VDD_SNVS_IN (3.0V coin cell CR2032) and got the following measurements:

Multimeter: Fluke 8846A:

Supply (3.3V) Off, RTC in Reset (not adjusted):     +209uA

Supply (3.3V) On, RTC adjusted, CPU running:      +11uA (because coin cell voltage is about 3.2V and LDORTC is 3.0V)

Supply (3.3V) On, Sleep Mode (suspend to RAM): +16uA

Supply (3.3V) Off, RTC adjusted:                             +439uA (this should be lower than 25uA)

In the IMX6ULCEC.pdf Rev. C

Page 32:

-> Current consumption 4uA + 1uA + 1.5uA = 6.5uA

Why did we get 439uA current drawn from the coin cell in power-off mode? This reduces backup time rapidly to some days/weeks.

The SNVS is clocked from an external 32kHz crystal (not the 40kHz ring oscillator).

We using Linux Kernel 4.3.0, started from u-boot. Could there be an configuration mistake (e.g. tamper detection enabled) or is this an hardware issue? I don't think we need tamper detection, but may be it's activated somewhere.

[jamesbone]

Hello Dirk,

In order to try to reproduced your issue, can you please point me out the exact steps that you are using.,  I am going to test it using the i.MX6ULEVK board, under linux.

[d_lohse]

1. connect the ampere-meter between the coin-cell and the vdd_snvs_in - Pin. If there is no backup capacitor the rtc should be in reset state when there is no voltage on this pin for several seconds.

While the board is powered off, you should now measure the 210 uA.

2. boot the cpu. While it is powered on you should now measure only a low current (11 uA), because the snvs is powered from the supply. Now, the rtc should stay at 1970. Set the clock (e.g. using rdate and hwclk command).

3. you can put cpu in sleep (suspend to ram), i use a tool called pm_sleep. Other tools like rtcwake also should work, or see here:

https://www.kernel.org/doc/Documentation/power/states.txt

While in sleep you should be able to measure near the same current like in power-on state. But this is not relevant for us.

4. Now turn off power-supply. The coin-cell should keep the rtc running. I can measure the frequency (32kHz) at the external crystal. The current now is much higher than in step 1. I get 439 uA.

As described in datasheet we should have not more than 25uA. This is much in comparison to an rtc like the ds1339, but would be acceptable. 450uA instead is too much.

[jamesbone]

Hello Dirk,

I try myself using the  NXP BSP  Linux 3.0.53  and I do not get the same measurements.  But when I try the Kernel 4.3.0 which it is not release yet from our side,  I get the wrong measurements.  So I think you may have wrong the sleep driver. and it is not really getting into the Deep Sleep mode. of the device.