Using a test pulse split into two channels of the DRS4 Evaluation Board v5, I looked at the time resolution using a leading edge threshold. The voltage and timing calibration was performed. One method (1) is to linearly interpolate between two points of the raw waveform that is above and below the threshold (this is exactly the algorithm given in read_binary.c in the drs4 source distribution); and another (2) is to use a cubic spline interpolation of the raw waveform. The results I obtained are:

Method 1: dt = 1.298 ns +/- 7.22 ps

Method 2: dt = 1.293 ns +/- 15.48 ps

I am really puzzled why the time resolution of the spline interpolation is about a factor 2 worse than the simple linear interpolation. Has anyone studied the time resolution using similar or other interpolation methods?