Query regarding UCC28633 application
**********************************************
I am trying to apply the UCC28633 in a SMPS circuit with 120ac input, 100V output, and 8mA up to 1A (transient or pulsatile) load. At loads of only ~20VA there is already difficulty in avoiding Vdd(OVP) due to the FET-turnoff/primary-demag spike (width of ~100 to ~200ns, increasing residual energy with increasing load). Because UCC2863x use bias winding to sense output voltage there is no Vdd regulation in IC itself.
Unless energy in spike can be largely dissipated (rather than stored in Vdd holdup capacitance), Vdd(OVP) threshold will be exceeded. Datasheet appears not to mention this issue at all. Datasheet example circuit does not appear to address the issue either.
What does TI recommend to deal with this problem? Tuned RLC circuit in series between bias winding (voltage-swinging terminal and Vdd holdup cap is my preliminary thinking but it seems strange that this whole issue is not mentioned in the datasheet. Please comment and advise if possible. Thanks in advance.
P.S. My intended SMPS app is completely unconventional -- it is not a DC voltage source at all. But at the present stage, for purposes of my query, it can be considered as such.
Further related UCC28633 app questions are:
1. How wideband is the Vdd overvoltage detection? How wideband is the output OVP detection, and is this only based upon sampled (once per switching cycle) value?
2. What is a likely cause of ~2kHz oscillation on output (~5% peak-to-peak) with loads light enough to achieve full output voltage? UCC28633 is cycling between higher and lower switching frequencies (~30kHz and ~55kHz, for example).
3. What is a likely cause of fairly rapid transition (i.e. within a few switching cycles) into light-load mode (i.e. Fsw drops to ~200Hz) during output ramp-up before achieving full output voltage with heavier loads?
I think (2) and (3) are essentially the same problem or have the same root, and I am observing both in my circuit.