Where does 1/f noise stop at low frequency?

The integral of a $1/f$ spectrum diverges logarithmically at both ends, so a physical system needs a lower cutoff $f_{\min}$, set in the fluctuator picture by the longest relaxation time $\tau_{\max} \sim \tau_0 \exp(E_{\max}/T)$. Measurements over many decades have not found a flattening. Determine $f_{\min}$ in a specific well-characterized system (metal film or SQUID) and relate it to the fluctuator distribution.

Long, drift-controlled noise records on a stable device showing a flattening of S(f), compared with an independently measured distribution of fluctuator relaxation times.