The CC2520 consumes an average of 250uA in Low Power Mode 1 (LPM1) with the internal voltage regulator turned on. In LPM2 without the regulator the CC2520 consumes <1uA, but there is no state retention.
What I tried was to supply the CC2520 with 1.8V through the DCOUPL pin while putting it into LPM2. After waking it up again I verified that the state of the configuration registers was intact. The current into device's DCOUPL pin was just below 1uA in LPM2 (now with state retention).
That means that the internal voltage regulator simply burns 250uA for nothing, and that 1uA would be enough for state retention.
Now these are the obvious questions for the TI folks:
Thanks,
Michael
The current you are seeing when you supply power externaly on the dcoupl pin is the leakage through the digital logic. While it may indeed be as low as 1uA, this cannot be relied on for all devices in all temperature conditions (Specifically, it will go up a lot if you operate the device at high temperature as leakage has an exponential relation to temperature. If you know that the device in your application won't operate in high temperature and you can supply 1.8V +- 10% without significant quiscient current in your power supply it would seem to be ok.
The only issue I can think of is that you might break some assumptions we made regarding power-on-reset release sequencing internally, so I can't promise you that you won't break something. (There are also some potential of introducing unwanted noise by fiddling with the power supply, but I guess you will be able to see if this becomes an issue in your application)
The reason the internal voltage regulator spends as much current as it does is because it is designed to supply the device's maximum power expenditure within tolerance from all supported external voltages. This makes the supply less than optimal for stand-by use as you have discovered, however adding multiple internal power supplies would have added cost to the system and it was our intent that most applications shouldn't need to upload a lot of configuration, so the value of retaing configuration should be small. Obviously this was a design consideration and might not apply for your particular application.
Thanks for the fast reply, and for your detailed explanation.
I guess it all depends on factors like speed of the SPI, active current of the MCU, supply voltage, etc. I counted 20 registers that need to be configured for my application, half of which are undocumented test and control registers that need to be updated. I did some more detailed computations and came to the conclusion that the cost of adding an external supply is probably not justified after all.
-Michael