Hi,
I am trying to optimize power consumption on some machine learning applications with the SparkFun Edge board and I am facing higher current consumption figures than expected.
Board is powered from a 3.30V power source with integrated current probe, connected between VCC and GND pins of the programming header (programmer disconnected).
Here are my measurements when running the LowPower Arduino sketch (figures are similar outside Arduino as well):
As you can see I’m getting around 1mA more than expected in every situation. I’m probably overlooking something but I doubt this 1mA floor is only due to microphone and accelerometer. I’d expect at most a couple hundred micro-amps in deep sleep mode even with accelerometer and microphone amp still powered.
Where can this high current consumption come from?
As a comparison, I can easily get an STM32L476RG on a Nucleo-L476RG board down to 9µA in Stop 1 mode.
Quite off-topic but I also observe that the Nucleo-L476RG is able to run similar code around 30% faster at the same core frequency. This is somewhat confirmed with 2.479 Coremark/MHz on Ambiq Apollo3 (when running the benchmark from Ambiq SDK examples) vs. 3.42 Coremark/MHz on STM32L476RG (from the datasheet); so I really wanted to be able to capitalize on the SparkFun Edge lower power consumption to save energy overall when running on battery power.
I’m also surprised the Ambiq Apollo3 increase in performance from activating Burst Mode is only of about 25% when running memory stressing and intensive floating-point operations but that might also be an issue on my side.
Thanks,
