simplified quick non gaussian fit function so that it is more robust

src/mdevaluate/utils.py

@@ -357,14 +357,15 @@ def quick1etau(t: ArrayLike, C: ArrayLike, n: int = 7) -> float:
     return tau_est
 
 
-def quicknongaussfit(t, C, width=4):
+def quicknongaussfit(t, C, width=2):
     """
     Estimates the time and height of the peak in the non-Gaussian function.
     C is C(t) the correlation function
     """
-    # TODO this is a very experimental interpolation, can fail
-    def ppoly(x,a,b,c,d,e,A,mu,sig):
-        return A*np.exp(-(x - mu)**2 / (2 * sig**2))+a+(b*x+e)*1/(1+np.exp(c*(x-d)))
+    def ffunc(t,y0,A_main,log_tau_main,sig_main):
+        main_peak = A_main*np.exp(-(t - log_tau_main)**2 / (2 * sig_main**2))
+        return y0 + main_peak
+    
     # first rough estimate, the closest time. This is returned if the interpolation fails!
     tau_est = t[np.argmax(C)]
     nG_max = np.amax(C)
@@ -375,13 +376,11 @@ def quicknongaussfit(t, C, width=4):
             tau = time[np.argmax(corr)]
             mask = (time>tau-width/2) & (time<tau+width/2)
             time = time[mask] ; corr = corr[mask]
-            guess = [0.001,-0.001,5,tau-0.5,1.0,nG_max, tau, 1.5]
-            popt = curve_fit(ppoly, time, corr, p0=guess, maxfev=10000)[0]
-            # TODO instead use some root or solve
-            xspace = np.linspace(*time[[0,-1]], 10000)
-            y = ppoly(xspace, *popt)
-            tau_est = xspace[np.argmax(y)]
-            nG_max = np.amax(y)
+            nG_min = C[t > 0].min()
+            guess = [nG_min, nG_max-nG_min, tau, 0.6]
+            popt = curve_fit(ffunc, time, corr, p0=guess, maxfev=10000)[0]
+            tau_est = 10**popt[-2]
+            nG_max = popt[0] + popt[1]
     except:
         pass
     if np.isnan(tau_est):