Current Spikes when in Extended Sleep

Hello,

I have a custom PCB with Dialog 14682. I am running into an issue where, when I enter extended sleep, the CPU appears to wakeup momentarily every 2 or 3 milliseconds. I am using an accurate DMM to see that the sleep current does drop to ~6uA, and that the spikes are in the 1s of mA, so not high enough that the radio is turning on, but rather indicating the CPU is waking up, possibly. This means that my average sleep current is something like ~200uA instead of ~6uA.

This occurs when running simple demo projects from the SDK, like extended_sleep and BLE_adv. This also occurs if I try to enter hibernation rather than extended sleep.

I am powering the Dialog chip off of 1.8V VBAT. This is built in to the system and can't be changed. Proper configuration changes, such as changing the battery configs to be TYPE_NO_RECHARGE, have been made to operate at this VBAT, but maybe I missed something there.

For more context, I have tried this on two separate 14682 PCBs (similar designs) with similar behavior. One board exhibits a 2ms period behavior, while the other exhibits 3ms period. I have some peripheral sensors connected to GPIOs that I have tried to configure correctly, and have cut the traces manually as well, with no change. I have tried playing around with the power domains such as 1V8 and 1V8P (on, off, sleep on, sleep off) and how the FLASH is powered (1V8 or 1V8P), with no effect. I have tried turning off the Watchdog with no luck, as well.

Any help is appreciated! Either any insight into SDK operations (auto-wake up if some power domain decays too much), or any tips on debugging custom hardware. I believe we have followed the layout guidelines to make these boards, but I can't rule out hardware issues, of course.

Thanks,

M