Models are calibrated to public/order-of-magnitude data — not confidential slide plots.
Purple shaded region on the NSD plot = ∫ NSD² df for RMS.
Gaussian noise scaled to integrated RMS in the shaded band — LF/mid “fuzz,” not 2 MHz switching ripple.
Switchers look “noisy” when people focus on MHz switching ripple. Integrated output noise in 10 Hz–100 kHz is a different story — reference, error amp, and loop peaking live there. Drag the integration band, overlay parts, and compare representative NSD models (not bench captures).
LT3045 NSD trace: Vin = 5 V, Vout = 3.3 V, Cout = 22 µF, Cset = 4.7 µF (Noise Analyzer capture, 10 Hz–10 MHz). LT1763 trace: Vout = 3.3 V, Cbyp = 10 nF, Cout = 10 µF (10 Hz–100 kHz plot; above 100 kHz is estimated, not traced).
Compare LTC6226 vs LT6018 with preset 0.1 Hz – 10 Hz (datasheet plot bandwidth). LT6018 advertises slightly higher spot noise (1.2 vs 1.0 nV/√Hz) but wins integrated LF noise because its 1/f corner is lower on the traced curves. Slide the band up to 10 Hz – 100 kHz and LTC6226 can edge ahead. Widen to 10 Hz – 10 MHz and LT6018’s HF rise can flip the ranking again.
Ask: “What bandwidth matters for your ADC / sensor / IF?” Many specs implicitly care about 10 Hz–100 kHz even when the engineer only remembers scope ripple at switching frequency. Narrow the integration band to that range — SS3-class parts win. Widen past 2 MHz and switching ripple spurs appear on the NSD plot (representative 2 mVpp SS3 vs 10 mVpp conventional at 2 MHz, fundamental + harmonics). SS3 is still far quieter, but without HF filtering even a quiet switcher shows a few mV on a scope.
LF/mid NSD (10 Hz–1 MHz) is traced from a slide via the Noise Analyzer; below 10 Hz is extrapolated. Ripple spurs are a teaching model — not a measured bench capture.