This solution to avoid having len(delta_s) = len(forest) is less time…

… consuming

bin/merge_spectra.py

@@ -305,8 +305,7 @@ def main():
                     mmm = np.where((wav_rf<constant.lya) & (wav_rf>constant.lylimit))[0]
                     if len(mmm) > 0:
                         nz = 256
-                        while (nz < len(wav_tmp)) : nz *= 2
-                        nz *= 2  # make sure that the size of delta_s is larger than the size of the forest, to avoid spurious correlations
+                        while (nz < len(wav_tmp)+10) : nz *= 2  # +10 is to avoid correlations from edge to edge of the forest
                         delta_s = np.random.normal(size=nz)   # latter, produce directly in k space
                         delta_sk = fft.rfftn(delta_s, threads=ncpu)
                         k = np.fft.rfftfreq(nz) * 2 * k_ny