Waveform 1 shows an incoming 7076kHz CW signal converted to a 2kHz IQ signal (only the real part is shown in the figure) with a receiver tuned to 7074kHz.
The LO is adjusted to generate a 2kHz signal, and Waveform 2 shows the multiplier output after decimation. (The real part in red, the imaginary part in green, and the absolute value in white.) The signal does not become a pure DC signal, because there is always a minor discrepancy everywhere.
The multiplier output is downsampled with the decimation ratio of 960 (= 192,000 / 200) to allow only the signals very close to DC.
After decimation, the (almost) DC signal is 40 times upsampled to obtain 8k samples per second and is multiplied with the output of a 400Hz side tone oscillator to finally generate the audible signal with the sample rate of 8kHz.
In this way, if there is another CW signal with the frequency difference of +f Hz, you will also hear one more audible signal at 400+f Hz.
This figure is for the case when the LO is detuned downwards, and you will hear a slightly higher pitch sidetone.
And this figure for real CW signals. Are you now happy with QRMs and QSBs?
