How to automatically distinguish a "good" from a "noisy" signal?

In the example below, you see on the left side a "good" signal, which in my case would be characterized by little fluctuation (especially during the baseline period, seconds 1 to 10) and then sometime after second 10 (where a stimulus happens) a large peak. I don't want to make any assumptions about the shape, just assign some kind of qualitative measure so that I can find the optimal curve of many of these...

I think my best bet so far would be find the highest z-score between ~11-20s relative to baseline...

[Jan]

You have very few data for the automatic estimation of a quality. Even for a human eye it is not trivial to distinguish noise from the signal in your diagrams.