UCC28180: About sudden changes in load

Hello Ishiwata-san, 

Thank you for your interest in the UCC28180 PFC controller. 

All PFC controllers necessarily have very slow transient response in order to achieve the current-shaping function.  
The UCC28180 can tolerate a 100% load step, and the slew rates up and down do not really matter.

For 100kHz switching frequency (for example) the switching period is 10us, and a load step current from 0 to 100% can have a rise times from 1us to 100us (for example). To the PFC voltage loop, these steps are virtually instantaneous.   The loop will have very little reaction to the change in output voltage over the ~10 switching cycles since the start of the step. Increasing the voltage loop bandwidth for faster response will result in high distortion to the input current.    

Figure 19 on page 18 of the UCC28180 EVM User Guide ( https://www.ti.com/lit/pdf/sluuat3 ) shows a typical PFC response to a repeating 100% load-step. 
Unfortunately the time sweep (500ms/div) is too slow to clearly show the time interval from the start of the step to the point where output voltage stops falling and begins to rise. 
Zooming in to the figure, it looks (to me) like 10~20ms.  

This is a worst-case situation because the V-loop control signal (VCOMP) is still saturated at its lowest level because the output has not fully recovered to regulation from the previous unload overshoot.  An isolated load-step during steady-state regulation at 0% load should have a faster response and less drop in voltage. 

The total recovery time back to regulation appears to be about 600~700ms.  This time can be made faster (if necessary) by over-sizing the output power capability so that more extra power is available to recharge the capacitor even while delivering full rated load power.  

Concerning the large input filter coil:  I don't know who designed this EVM, but I suspect that his interest was concentrated more on the PFC section and less involved with optimizing the EMI filter.  This coil was chosen to reduce conducted-EMI noise from the 120kHz switching frequency and its harmonics; possibly only a limited set of coils were available to choose from.  Unfortunately no data is provided to show how effective it is.  Maybe the EVM designer is not an expert in EMI filter design. 

Regards,
Ulrich

Hello Ishiwata-san,

Because of the very-slow loop response, it is very unlikely that the PFC output will resonate due to a load current fluctuation. 

Please consider the repetitive step-load waveform of Figure 19 in the User Guide.  The timing between load transitions (about 2 seconds) allows the output to attain regulation when loaded, but not fully recover from over-voltage when unloaded.  In the worst-case, if this 0-100% load-step period is decreased from 4s to 400ms to 40ms to 4ms, you can imagine that the output voltage will not fully recover regulation before the unload happens. Then the overshoot will be about the same jump, but starting from a lower point.  When the load is reapplied, Vout starts from a higher point.  

This indicates that the peak positive and negative voltages due to the worst-case 100% load current fluctuations are bounded and not expected to increase beyond what is shown in the waveform already. 
Basically, the V-loop cannot keep up with the load changes and the output capacitor handles all of the filtering of the current fluctuations.  The capacitor is normally sized to limit the full-load 100-Hz peak-peak ripple voltage to ~5% of average Vout.  The V-loop compensation is designed to keep small-signal perturbations under stable control.  Large-signal perturbations are handled by the output capacitor and 100% load-steps are about as large of a perturbation signal as there is.  The size of the output capacitor fully handles these steps within the peak-peak limits seen in the step waveforms, so the PFC is inherently "stable" even under these conditions.   

For the typical 390Vdc PFC output, using 400-V rated MOSFETs (and diodes) does not provide enough margin for long-term reliability.  There are not many 500-V MOSFETs to choose from, so the 600-V and higher ratings of MOSFETs and diodes have wide range of selection.  These provide plenty of margin for reliability and can withstand occasional line surges as well.  There is no danger of exceeding the MOSFET's breakdown rating. 
 
Lastly, even if the small-signal loop compensation was poorly designed with very little phase margin, if some perturbations did start to induce output oscillation, the over-voltage protection (OVP) shutdown feature of the PFC controllers would stop switching immediately and interrupt those oscillations before they could build to excessive voltage levels.    

I think the PFC topology is probably the least likely system to break into an oscillation.  
A more likely danger (although rare in occurrence) is open feedback loop where the output feedback signal is interrupted by a broken component or track. 
Most controller can detect an open feedback path and shut down.  The worst-case of this kind of failure is the "stuck-at" open-loop, where the feedback voltage pin is stuck at some voltage lower than the regulation level.  This will drive the PFC to continuously pump out more power to attempt to bring the feedback level up to the reference level, but the output voltage rises out of control until something fails (usually venting the output cap). 

Some controllers have a second independent voltage-sense input strictly for OVP to avoid the single-point of failure that the feedback signal path represents.  The UCC28180 does not have a second OVP (due to low pin count), but the UCC2806x family of PFC controllers (for example) does have this feature.

Regards,
Ulrich