Smoothing

The smoothing function is typically applied to peaks in sequential mass spectral scans in order to average out the quantitative error in the source.   While it takes multiple forms, in general they all result in averaging each peak subsequent scans for both amplitude.  The Filter Width (FW) adjustment defines over how many scans the individual single scan data point will be averaged.  

Supposedly the smoothing is beneficial to accurately locating the apex and its amplitude.  The user must also be aware that the smoothing function will change the amplitude, and RT (and the m/z if it too is smoothed) for all peaks.  For big peaks the differences introduced are relatively minor (8.3% in the illustration below).  However, for smaller peaks the smoothing more strongly alters the data  (26.3% in the illustration below).   Another way to look at this is to realize that larger peaks tend not to broaden as they are more strongly smoothed as much as smaller peaks.

When peaks are very narrow any FW larger than the width of the peak will have  strong consequences and may result in loss of the peak altogether as a usable datapoint.

Note:  There are many different ways to implement a smoothing routine but they all likely suffer these effects.