# Can white Gaussian noise improve signal level in non-linear device?

In the system shown below, can white Gaussian noise amplify or attenuate the input signal?

Most of the time we check signal level proportional to noise level (such as noise figure) not singly.

I have obtained PSD of my desire system's output and it is a combination of

• a) Dc term

• b) reaction of input noise to itself

• c) reaction of input signal to itself

• d) reaction of input signal and input noise.

At system's output at frequency $\omega$ should I assume terms (c) and (d) as the signal level? Or c is enough?

• I'm not sure exactly what you're asking, but it sounds like you're asking if there are situations where you might add noise to a signal before it is input to a nonlinear device. The answer is yes; this technique is called dithering, and can be used to turn what would be deterministic quantization errors in an ADC (resulting in discrete spectral artifacts) into random errors, flattening the quantization noise spectrum. – Jason R Aug 31 '16 at 11:21
• Dithering is a specific example of a more general idea - see stochastic resonance – David Aug 31 '16 at 12:55