I'm using CC2541 in network processor mode connected via SPI to an application processor. I have enabled POWER_SAVING and I'm following the protocol described in the source code: enable chip select, send a dummy byte (is this even required), disable chip select, wait until the device ready is asserted, then enable chip select and send my message. Currently with BLE-1.3.1 release I'm seeing a 700us delay between when the master asserts the wake-up signal and when the slave asserts the data ready signal.
Is this delay something that is intrinsic to the processing capabilities of the CC2541, is it limited by the available clocks (I read somewhere that the 32kHz oscillator needs 400us to get ready), or what else?
Is it possible, by employing additional external clocks or other circuitry to drive that wake-up time down?