PPK II reliability and interaction with DC/DC converters

[ecasado_iz]

Hi,

We've been using the PPK II for basically every IoT project we've carried out in the last few years, and it has been an incredibly useful tool.

Our standard use case is using it in source mode to replace a battery during development for IoT products and we're really happy with the results. However, during some measurements for projects that have a DC/DC converter after the battery (be it buck or buck-boost), we observed some weird behaviors and inconsistencies.

Examples of behaviors that we're seeing when measuring boards with DC/DC converters:

• High (> 0.8A) "phantom" current spikes. After investigating in the Devzone, it seems they're caused because of the switching logic inside of the PPK. We can live with them because they doesn't affect the average current too much, but it's worth mentioning that we observe them way less using PPK version 4.1.2 than the latest one (as seen in this post). Also, if we try to measure after the DC/DC (for example in the 3V3 output), these spikes are not there.

• Variable "noise floor", that significantly increases the average current sometimes. We've tried to place a capacitor right at the output of the PPK as suggested in another Devzone post, and we observe that it helps alleviating the issue as you can see in these screenshots. Left one is without capacitor, right one with it.

Zooming on the "floor" area betweek peaks you can observe the difference too (scale is a bit different, sorry about that)

What is the reason for this behavior? Which one of the measurements should we take as the most correct one?

• Also, trying different boards with the exact same hardware and exact same firmware, we observe differences in consumption. I don't have screenshots of this, but one recent example was one board measuring 150 uA and another measuring 400 uA. We thought there was some hardware issue and thoroughly reviewed everything, until we plugged both of the boards to a Monsoon Power Monitor and in both of them we read 400 uA. On paper, Power Monitor has a way lower sampling frequency (5 kHz) so aliasing from switching regulators should be affecting even more; but with it we don't observe the previous phenomena. Again, measuring at the DC/DC output (bypassing the regulator), gives more consistent results but we can't always do this.

Are we doing something wrong, or are there some limitations of the PPK2 that we're not considering?

Best regards and thank you,

Enrique

[Asbjørn]

Hello Enrique,

High (> 0.8A) "phantom" current spikes. After investigating in the Devzone, it seems they're caused because of the switching logic inside of the PPK. We can live with them because they doesn't affect the average current too much, but it's worth mentioning that we observe them way less using PPK version 4.1.2 than the latest one (as seen in this post). Also, if we try to measure after the DC/DC (for example in the 3V3 output), these spikes are not there.

This is good to hear. The version you are referring to have a built in routine that discards "unexpected" results within reason. Intention is to remove the phantom spikes, and show a more accurate result in the use case where the average current is "high". The ppk will also behave as a terminal print with this version and print out the number of discarded results.

What is the reason for this behavior? Which one of the measurements should we take as the most correct one?

The reason is the same as with the high current case, there's a mode switch that distorts the both the peak and the "floor" area. The one with capacitance should be closer to the correct. When you zoom in between the peak, the floor is correct, but you miss out from the peaks in the window view. If you find average with peaks and without. It's tricky to know exactly how much the peak overshoot without longer control measurements with more advanced setups, but somewhere between average with peaks and average without would be the "right" results.

Also, trying different boards with the exact same hardware and exact same firmware, we observe differences in consumption. I don't have screenshots of this, but one recent example was one board measuring 150 uA and another measuring 400 uA. We thought there was some hardware issue and thoroughly reviewed everything, until we plugged both of the boards to a Monsoon Power Monitor and in both of them we read 400 uA. On paper, Power Monitor has a way lower sampling frequency (5 kHz) so aliasing from switching regulators should be affecting even more; but with it we don't observe the previous phenomena. Again, measuring at the DC/DC output (bypassing the regulator), gives more consistent results but we can't always do this

My guess here is that with one of these boards the ppk goes above the threshold of discarding result and the other one doesn't. This should be visible with connect to the ppk with a serial terminal and see the print out from the capture. I would expect that the board showing 150uA to have discarded more results during the test compared to the other one. The differences might just be very close to the trigger for discarding and you end up seeing it on one and not the other.

Best regards

Asbjørn

[ecasado_iz]

Hi Asbjørn,

Sorry for the late reply.

I really appreciate the insights about this, because they help us deepen our understanding aboyut how the PPK works and confirms we're not getting crazy with the different results.

We will test with the serial terminal to check the discarded samples as suggested to confirm the behaviors.

For now, to get the most accurate results, do you recommend to add an input capacitor (around 10-22 uF?) when the PCB has a spiky DC/DC right after the battery to "smooth" the waveform and get more reliable results?

Best regards,

Enrique

[Kenneth]

Hi Enrique,

Asbjørn is away, but I would say yes to the question of adding input capacitance can help.

You might also try this firmware for test to see if that improve your measurements:
RE: Power Profiler Kit II measurement error 

Kenneth