Signal to Noise
Standard Report configuration file for Regions allows the inclusion of a
quantity referred to by the heading Noise. The meaning of the Noise column depends
on the manner in which the quantification regions are prepared. Both values
reported for the Noise column require the data are made available in counts. To
estimate the noise in an XPS spectrum, Poisson statistics associated with the
pulse counting system dictates that the noise is proportional to the square
root of the counts per bin.
- The default action is to
compute a metric for the noise in the spectrum for which the region is
defined. The method employed attempts to construct a smooth curve which
passes through the data such that the residual between the spectral data
and the smooth curve measures the distribution of the noise within the
spectrum. Since XPS spectra generally include a substantial background
signal and the noise is proportional to the square root of the counts per
bin, the residual distribution must be normalised by dividing by the
square root of the counts per bin for each bin. The standard deviation for
the normalised residual distribution is computed and reported in the Noise
column of the Standard Report. For a single channeltron
detection system, the standard deviation is typically unity. Multiple
detector systems effectively smooth the data and therefore the value for
the standard deviation tends to be less than unity.
- As an alternative, the signal
to noise for each region can be computed and reported in the Noise column
of the Standard Report. To alter the default action from that described
above, the Tag field in the region must be assigned the string S:N. The Tag field is case
sensitive and there should be no additional spaces between the characters.
Given that the Tag field is set to S:N, the Noise
column report the value (peak-background) / SQRT (peak). Again this
measurement is normalising the signal above background using the counts at
the peak maximum and therefore should be a similar value for similarly
intense peaks regardless of the background height.