### SET operations: Post-Processing

After the neutron data is downloaded from NMDB/NEST, some post-processing is applied to make the time series data more useful for supsequent operations:

- The Efficiency and Pressure-corrected data (including a filtered time series), the Pressure-corrected data, and the atmospheric pressure are assembled into one data file. In nearly all cases, the Efficiency/Pressure-corrected data and Pressure-corrected data are identical (the JUNG data are the exception). It has been recommended to use just the Efficiency/Pressure-corrected data.
- Missing data are interpolated, resulting in monotonically increasing time, i.e., there are no data gaps (large gaps will be poorly estimated however). The IDL routine INTERPOL is used, with the LSQUADRATIC option
- The data are smoothed with a Butterworth filter to impove signal-to-noise, and a time series of hourly means using boxcar smoothing.

Extreme oulier rejection is supposedly performed by the operating station, so it is not done in SET operations.

### About Neutron Data: Smoothing with a Butterworth Filter vs. a Boxcar Average:

The Butterworth Filter is a type of signal processing filter in the frequency domain designed to have a frequency response as flat as possible in the passband. It is particularly useful for extracting the signal out of time series that has a very large high-frequency component (noise). It is a commonly used in electronic resonant circuits.

Below are two comparisons, showing how data with a one-hour boxcar average compares to a Butterworth filtered time series.

Note the differences between the two methods of smoothing the data: The Butterworth filter results in better end-point estimates, and there is an improvement in reducing high-frequency noise. However, in all cases tested, the overall differences between the filtered and mean time series were on the order of 1%.