Chris, could you please reconsider your text equating analog, DC and a PWM signal. Although I do get your reason for saying so.
A quick Fourier analysis will put the lie to a PWM signal being DC. It is a square wave of relatively high frequency with modulated or varying mark space ratio. In fact, in this regard it is analogous to the stretched zero bit that Rob describes for address 0 DC loco operation. In both cases the motor sees this variable mark space ratio square wave but, being highly reactive, filters out all of the high or non-zero frequency components (again Mr Fourier can help here) by being (almost) open circuit to them, leaving only the average DC value to drive the motor.
However, just as the DCC frequencies are high but not infinite, so the motor is not actually open circuit to them but high impedance, so is affected by them to some extent. Hence the buzzing you hear with a DC motor on a DCC track as the motor moves backward and forward slightly to follow the square wave frequency.
PS. Fourier analysis, or frequency domain analysis, doesn’t look at how the signal varies over time, rather it looks at a particular time only and sees what frequencies are contained in the signal. In a square wave, the frequencies present are the fundamental frequency of the wave plus various harmonics of the fundamental frequency with decreasing amplitude as the harmonic number increases.