When you calculate uA/MHz for nRF52832, which figures do you use?

[BodyBeacon]

I saw that DA14680 from Dialog comes with a 30uA/MHz processor.

Although the clock can be programmed up to 96MHz, the 30uA/MHz is measures at 16MHz.

How can I get a comparable figure for nRF52832?

[Øyvind Karlsen]

Hi,

When running from RAM the nRF52 also delivers 30µA/MHz(www.nordicsemi.com/.../nRF52-Series-SoC)

The µA/MHz measure does not include the use of the floating point calculations. Floating point calculations using the FPU in an M4F is roughly 20 times faster than the same calculation on the M0. This means lower power consumption. The nRF52 also has DMA to most peripherals, this means that you don't even need to use the CPU for many features, you can run directly from memory. In addition to that the nRF52 has cut peripheral power consumption in half compared to the nRF51.

Best regards,

Øyvind

[BodyBeacon]

So in order to compare chips, you cannot divide TX consumption by clock speed?

[Øyvind Karlsen]

To compare chips you would need to know how often you want to transmit, and how much you need to calculate.

Then you sum the time you use in each state (TX/RX, processing and idle) multiplied with the power consumption of each case.

[FormerMember]

Purely looking at the technical specs with regards to the power consumption, here are my thoughts (aka rant).

The DA14680 has an integerated battery management unit that can directly switch down a Lithium based rechargeable battery to the supply voltage. This is more efficient (and cheaper) than using a regulator to get to a supply voltage and then using the internal regulator. The radio is more efficient at 4.2 / 4.3 mA Tx / Rx at 3 V (assuming 0 dBm tx power).

For me everything else is questionable, especially since we have only access to the banner specs. The 30 uA/MHz for Cortex M0 is kind of same as 38 uA/MHz for Cortex M4 (for non DSP and FP code). When running DSP and using FPU the whole ball game changes. Also what is not mentioned with DA14680 is when is this number achieved. Most probably not when using the 8 Mb flash (I might be wrong). The receiver sensitivity is 3 dBm lower than nrf52, which means the tx power should be a bit higher. Lots of key info such as the different sleep current, wake up time from different sleep state, peripheral power consumption and how autonomous do they operate aren't yet available. So, can't really say too much about it.