nRF52DK 50uA bursts every 20 seconds when advertising??

[essele]

I'm running some power tests on the nRF52 DK and, after fiddling with some of the power settings, I'm managing to get an idle current of about 1.8uA on the coin cell, and a few more uA if I power with USB ... which is great.

However, roughly every 18-20 seconds, I get a burst of ~50uA which lasts for about 8 seconds or so. (This is using a multimeter with 12 NPLC averaging - its much more bursty in reality and I'm not convinced I'm getting a good picture, but the PPKII is out of stock everywhere!)

I'm using ble_app_template, but with an APP_ADV_INTERVAL of 1200 and a zero advertising timeout. I also only use slow advertising, so it's not a transition from slow to fast. Changing the ADV_INTERVAL does seem to impact the frequency and intensity ... is there something that happens every X advertising packets?

I can't see any timers or anything else in the code that might explain it ... and it only happens if I enable advertising, and also doesn't happen if there's a connection (so it does seem advertising related.)

This is day 1 for me with this device, so I may well have done something stupid. Any help would be greatly appreciated.

(Image attached)

Thanks,

Lee.

[Stian Røed Hafskjold]

Hi,

Interesting. Never seen that before. I have a theory though:

I counted the dots between the time marks, and it seems like the sampling interval is 750ms? And, an advertising interval of 1200 corresponds to 750ms as well. So what you are seeing is due to aliasing.

First it samples exactly each advertising event every 750 ms interval for 8 seconds, then it drifts a little further and misses every advertising event for 12 seconds, and then it starts to sample the advertising events again, and so on.

Try to change the advertising interval and I would guess that you will see a different pattern.

[essele]

Hi Stian,

Thanks for the response ... I think you may be right. 12 NPLC is every 240ms in the UK, so I guess this could be close enough to cause the problem ... and I do see different results with different sampling times.

I will hook up a scope tomorrow to see if I can get better results ... plus I have now found a source for the power profiler, and I think that’s likely to give me the most usable measurements (although will be a while in shipping.)

Cheers,

Lee.

[essele]

Further analysis ... I had thought about this further and was starting to question the theory given that the meter effectively averages over the 240ms time period, however I then realised that auto-zero was on, which means that some of the time it's not sampling which could absolutely aggravate the situation.

Turning auto-zero off I get a much more sensible graph, with peaks at advertising intervals and average numbers that I can believe.

Thanks for your help.

Lee.