Thingy:52 Power Concept

[ZAC]

Hi, the configuration for the 9-axis sensor used in Thingy is interesting where a switch is used to cut the power line from the sensor. From the description: The motion sensor is interfaced through the I2C bus (0x68). To reduce power consumption when the motion sensor is not in use, the power supply and signal lines are connected through an analog switch (U4). This enables the device to be completely cut off from the rest of the circuit when not in use.

How much current would the signal lines MPU_INT, SDA and SCL draw when the sensor is not in use and assuming only the power line is cut, i.e. while the signal lines are still connected?

I am planing to implement a switch, although the planned connection with the BT chip is SPI. For my design, size and power consumption are of big importance. The NX3DV2567 used in Thingy is a four-pole double-throw analog switch. Maybe in the case of SPI, I can live with a smaller one-pole double-throw or (single load switch) for the power supply?

Note: I am 99% sure the answer for SPI power consumption can be found somewhere. So please answer this fully only if you are friendly excited to provide a quick answer, perhaps in case you have experienced the same situation and willing to share your finding on Devzone. Otherwise google is there. Many thanks!

Edit:

[MartinBL]

Hi,

I couldn't find anything about what kind of state the I2C pins are in when the MPU is powered down. This is probably a question for Invensense. If they are high impedance though, the pull up resistors on the bus should pull the lines high and it shouldn't matter. If the MPU of some reason pulls the line down to GND you will have a current flowing through the pull ups. There seems to be some 4.7kOhm pull ups on the bus and the TWI drivers probably also enables the internal 13kOhm pull up resistors.

Electrical spec for the nRF52's SPI can be found here.

Hope this helps. Not sure if I quite understood your situation.

[ZAC]

Thanks for the reply. I have added pics for that. Indeed pulls-ups are there. So should I assume it is not for power saving? Did not find anything about pull-down to GND in the MPU's datasheet: The MPU-9250 always operates as a slave device when communicating to the system processor, which thus acts as the master. SDA and SCL lines typically need pull-up resistors to VDD. The maximum bus speed is 400 kHz.

[ZAC]

I think, if the VDD_MPU pin alone is cut (i.e. a single-pole switch is used), the SDA and SCL will still be connected to VDD through the R22 and R23. So the background why 4-pole switch must have to do with leakage current and therefore energy saving?

[MartinBL]

I talked to the guy who designed the Thingy.

The double throw switch is there because there are multiple slaves on the same I2C bus. If you turn off one of them it is likely that it will mess with the bus. The reason why a NX3DV2567 double-throw switch is used instead of a possibly smaller single throw is simply because the NX3DV2567 was the cheapest option at the time.

[ZAC]

Good (and nice new webpage for devzone!). Do you think that this implementation would bring any benefit in case of an SPI implementation or one can live in that case with a switch for cutting-OFF VDD_MPU only?