UCC28070A: Current peaks

Hello Oscar, 

Please include the test conditions under which this waveform was captured.   Without it, I have to make assumptions and "educated" guesses. 
Also, a list of the design targets and a schematic diagram of the PFC section (with AC input) will be very helpful. 

At 25A/div scale, I see ~45A peaks which corresponds to ~31.8Arms.  For 3.3kW output (I assume that is the loading), assuming ~90% efficiency, that indicates input voltage to be about 115Vrms 50Hz.  Please correct me if my assumptions are wrong.

I disagree about the markings: I believe that the green circles indicate where the AC line voltage crosses 0V, and the red circles indicate where the UCC28070A controller takes some actions.  I believe that the green points are 0-A flat spots in the input current just before the diode bridge, but these flat spots are phase-shifted and level-shifted by a significant amount of reactive current flowing into the EMI filter X-caps.   Please verify this interpretation. 

The red-circled points, appear to be step-increases of input current which is consistent with a gain change in the internal multiplier block due to a reduction of input rms voltage. 
A description of the Qvff gain system in the UCC28070A multiplier can be found in the datasheet.  
When the input voltage is falling (even if very slowly as appear to be the case here) the VINAC peak voltage will eventually cross a Qvff level boundary threshold (in the falling direction, which is 95% of the rising threshold).  When the boundary is crosses, the kVff gain factor for the new level will be immediately applied.  
Falling voltages require higher input currents, hence the kVff factor (in the denomination of the IMO equation) will drop and the Vimo reference for the current loops will step up. 
This generates the current step in the falling side of the AC current half-cycle.  

The UCC28070A interprets the input "RMS" level by peak-detecting the voltage on VINAC each half-cycle. When the input voltage is falling, it can only determine that the immediate peak has decreased from the previous peak by comparing the new vs. old peaks when there is a zero crossing.  The UCC28070A defines the zero-crossing to be when VINAC falls below 0.7V.  This occurs repeatedly at the same phase-angle in every half-cycle, when the voltage is relatively constant. 
In this case, for ~115Vrms input it is occurring at about 8ms after a new half-cycle rises from 0V.  

All that said, we see that the first two current steps increase the amplitude of the input current, and this indicates that the VINAC signal voltage is falling two half-cycles in a row. 
The third red circle shows the current suddenly dropping to zero for about ~2ms.  This indicates to me a shut down of the GDA and GDB gate drives due to output overvoltage.  

I am not really sure of what happens to the input current after the OVP shutdown clears. Obviously some dynamic disturbances; but it looks like the current rises to a peak limit point (due to PKLMT setting).  Then the current drops to a very short segment of a sine wave for 1ms, then drops again to another sine segment for ~3ms, then shutoff again apparently due to OVP again.  After that, the OBC switches off the system operation. 

I do believe that the 1ms sine segment followed by a drop to a lower-amplitude 3ms sine-segment is indicative of the VINAC voltage rising up above the previous half-cycles peak by a large amount, resulting in a large drop of IMO gain (kVff factor increases significantly). 

So I think there is something happening with the input voltage just before this shutdown.  It seems to start to decrease, and the subsequent gain changes set off an output overvoltage situation, and the dynamic current changes may be provoking some resonance in the EMI filter which affects the voltage point that the VINAC input is sensing and those affects are making the problem worse.  Then the OBC shutdown happens before anything can settle to a new steady state.     

Please check my scenario with the actual system operation to see if it explains (most of) the behavior.  
Other waveforms will need to be captured: the actual AC input voltage, the voltage after the diode bridge, PFC output voltage, VINAC, and Vimo, for example. 

If there is a disturbance on the AC voltage, it may be increased by unwanted resonances in the EMI filter, for example.
But possibilities need to be identified first, before trying to solve them.

Regards,
Ulrich