1. Switching ripple noise
Ripple is always the dominant noise in a switching regulator because the peak-to-peak voltage amplitude is typically a few mV to tens of mV, depending on the switching regulator topology and basic operation. It should be seen as a periodic and predictable signal. If operating at a fixed switching frequency, it can be easily identified and measured by an oscilloscope in the time domain, or by Fourier decomposition in the frequency domain.
2. Broadband noise
Broadband noise in switching regulators is random amplitude noise on the output voltage. It can be expressed in terms of the noise density over the entire frequency range, in V/√Hz, or in V rms, which is inseparable from the density over the frequency range. Due to the limitations of silicon process and reference filter design, broadband noise is mainly located in the 10 Hz to 1 MHz frequency range of the switching regulator, and it is difficult to reduce it by adding filters in the low frequency range.
Typical buck regulator broadband noise peak-to-peak amplitude voltage is about 100μV to 1000μV, which is much lower than switching ripple noise. If additional filters are used to reduce switching ripple noise, broadband noise can become the dominant noise in the output voltage of the switching regulator.
3. High frequency spikes and ringing
The third type of noise is high frequency spikes and ringing noise because the output voltage is generated by the switching regulator turn-on or turn-off transients. Consider parasitic inductance and capacitance in silicon circuits and PCB traces; for buck regulators, fast current transients will cause high frequency voltage spikes and ringing at the SW node. Spike and ringing noise increases with current load. Figure 8 shows how a buck regulator forms a spike. Depending on the on/off slew rate of the switching regulator, the highest peaking and ringing frequencies will be in the 20 MHz to 300 MHz range, and the output LC filter may not be very effective in rejection due to parasitic inductance and capacitance. Compared to all the above discussion about conduction paths, the worst is the radiated noise from the SW and VIN nodes, the output voltage and other analog circuits suffer due to its very high frequency.
