First thing to check is the output capacitor value and ESR. At 1.2 kW the ripple current can be huge—roughly Iload × (1–D). If your C is too small or the ESR too high, the voltage will sag. Try a bulk electrolytic with a low ESR, or add a ceramic in parallel for high‑frequency filtering. Next, look at the duty‑cycle ripple of the inductor. A higher inductance reduces the current ripple, but it also raises the peak current rating. Verify the inductance meets the L = (Vout × (1–D))/(ΔI × fs) requirement for your switching frequency and desired ΔI. If the inductor is underrated, the current will oscillate more than expected. Check the MOSFET gate drive. A weak drive can cause slow transitions, increasing the effective duty cycle and creating more ripple. A stronger gate driver or a snubber network on the drain can shave off that extra ripple. Make sure the input voltage ripple is low. If the input bus isn’t clean, the buck sees more variation than the output stage can handle. Adding an input bulk capacitor or a small‑value ceramic in series can help. Finally, look at your PCB layout. Long traces or a poor ground plane can introduce inductance that looks like a ripple source. Shorten the return path and keep the switching nodes close to the capacitor bank. Once you run the numbers through those checks and tweak the values, the ripple should drop in a predictable, linear fashion. If it still spikes, there’s probably an unaccounted‑for parasitic somewhere—time to open the board and trace it.

That’s the right sequence—start small, tweak one thing, see what happens. Keep a notebook of the numbers; the devil’s usually in the ESR or a tiny trace. Once you lock each variable, the ripple should drop out cleanly. If it still sticks, pull the scope up and watch the edges—you’ll catch that hidden inductance before it gets you into a full‑on crisis.

Sounds good. Grab the multimeter, lock the scope in place, and let’s start with the capacitor. Record the ESR, swap in a low‑ESR part, note the ripple drop. Then hit the inductor, tweak L, record the ΔI, watch the waveform. One change, one readout, one repeat. If the spike still shows up, the hidden parasitic will finally reveal itself. Let's keep the steps straight and the notes tidy.