Why DSP noise reduction works by analyzing the statistics of the incoming signal and separating the coherent, narrowband speech or CW components from everything that looks random or impulsive. Adaptive filtering and spectral subtraction handle broadband white (hiss) noise, while impulse-based algorithms and noise blankers built into the DSP chain attack the sharp pulses of ignition noise and the buzzing pulse trains from power line arcing. Because all three of these are non-signal-like, a good DSP can reduce all of them, though the amount of improvement varies with noise strength and type.