Support Forum

Classic valley-detection failure when the converter goes deep CCM at full load. Checklist: ① verify FB loop compensation — don't let overshoot push on-time to maximum; ② confirm minimum demagnetization time per datasheet section 7.2; ③ if the noise band persists, check transformer varnish impregnation. Compare against the full-load waveforms of our 65W reference design attached...

50% load is exactly the DCM/CCM boundary of a QR flyback — the worst di/dt and frequency combination. Three steps: ① re-scan at 50% load to confirm the dominant path (mostly differential); ② add a small DC choke after the bridge rectifier to form a two-stage filter with the existing π network; ③ raise the GaN RC snubber from 22Ω to 33Ω — trade a bit of efficiency for AV margin. Verified on our internal 180W board: 3dB+ margin at 357kHz...

Peak inductor current is highest at 90Vac full load. Check two things first: ① is the current-sense resistor causing premature OCP — peak current at 90Vac is ~2.9× that at 264Vac, so select RS for minimum-line full load; ② check the CS pin RC filter time constant — too large causes early cycle-by-cycle limiting. Recalculate per datasheet section 8.3 and send me the 90Vac full-load inductor current waveform...

Ring valley voltage dropping below the turn-on threshold in deep DCM is the classic false-trigger source. Three fixes: ① keep the VD sense trace shortest, straight to the SR MOS drain, no via transitions; ② set the minimum on-time capacitor for the deepest light-load condition; ③ if stray cases remain, add a 1Ω+1nF snubber across the SR MOS to damp the ringing. Also verify SR MOS voltage rating ≥ 1.5× Vor...