unusual question - magnifying weak patterns on...

[tester]

I'm approaching unusual question, but something tells me there are some answers that may help.

Is it possible to distinguish two layers from a mixed signal in following conditions:

1) both layers are noise-like (irregular, quasi-periodic), full spectra signals,
2) both layers represent different "patternicity" so to speak,
3) one layer has let 120dB smaller amplitudes.

So basically, the only difference between them - are amplitude levels, because both are full spectra. But there is maybe one thing that makes difference. Scaling. If signals are relatively clean, and first layer has large amplitudes (100dB difference), then subtle variations typical to that first layer - will be larger. If second layer has small amplitudes, then these amplitude variations will be much smaller. But because both signals are full spectra and irregular - these features will interweave, producing one complex evolving shape.

My question is - is it possible to "split" it somehow (and to what degree)?