zoul/remote-revb: initialize unused pins to minimize current draw (WIP) #1920
Conversation
alignan
changed the title
|
note to self: Scheduling a |
|
Yes, a tickless implementation would help (as previously discussed). |
| #define CLEAR(x, y, z) GPIO_SOFTWARE_CONTROL(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); \ | ||
| ioc_set_over(x, y, z); \ | ||
| GPIO_SET_OUTPUT(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); \ | ||
| GPIO_CLR_PIN(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); |
There was a problem hiding this comment.
do/while(0) or a GCC compound statement ({}) would be better.
| GPIO_SET_OUTPUT(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); \ | ||
| GPIO_SET_PIN(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); | ||
|
|
||
| #define INPUT(x, y, z) GPIO_SOFTWARE_CONTROL(GPIO_PORT_TO_BASE(x), GPIO_PIN_MASK(y)); \ |
There was a problem hiding this comment.
I mean that z should be used with ioc_set_over(), not that it should be removed.
|
|
||
| INPUT(GPIO_D_NUM, 0x03, IOC_OVERRIDE_DIS); /* RTCC interrupt */ | ||
| CLEAR(GPIO_D_NUM, 0x04, IOC_OVERRIDE_DIS); /* LED */ | ||
| CLEAR(GPIO_D_NUM, 0x05, IOC_OVERRIDE_DIS); |
There was a problem hiding this comment.
IOC_OVERRIDE_DIS is fine for the outputs, but it's not for the inputs (besides the issue with the last parameter of INPUT()), unless the pin is always driven externally, whether it is by the output of another chip or by an external pull resistor, etc.
Except for PC0-3, all the pins are set as GPIO inputs with internal pull-up resistors by the CC2538 after reset, which is already optimal for power consumption, unless there are external components forcing a low level on the pin or enabled by a high level on an input of theirs.
Also, all the components of the board could be switched off by default, and only enabled when needed (dynamically), but this sometimes lengthens the access times because of an additional wake-up time.
The sleep modes could be better used.
The µDMA and interrupts should be used instead of polling where possible.
The system frequencies could be tuned according to the application requirements, even dynamically.
There was a problem hiding this comment.
Thanks for the pointers!, I will fix the INPUT macro left-over
gitignore: remove Cooja tmp files
This is a work in progress to reduce the current draw from pins not used, please feel free to drop any comment or suggestion!
Testing with the
examples/zolertia/zoul/zoul-demowith the following changes:NETSTACK_MAC.off(1);in L128 commented outPROCESS_YIELD();in L145 changed toPROCESS_WAIT_EVENT_UNTIL(0)NullRDCcommented out in project-conf.hSo far in PM2 the current drops from 2.73mA to 0.672mA, but some work is still required. There is around 500uA consumed by the XTAL oscillator (32KHz), as gatting the pads reduces the current consumption to 50uA. Dropping to PM3 yields around 32uA.
The values are measured on the battery input of the RE-Mote.