## N-Procedure

The
original N-procedure transforms the experimental spectrum Y(x) according to the
equation

*Y*(x,p,Dx) = *Y*(x)
- p[*Y*(x + Dx) + *Y*(x - Dx)].

Here, *Y* is the
count rate (intensity), x is the number of current “channel” in the spectrum
(running from 1
to N).^{}

p = 2 (-2 - d^{2} + (d^{4} + 5 d^{2}
+ 4)^{1/2})/ d^{2}, where

d
= 2Dx/G,

Dx and G (in channels) are the procedure parameters.

The values of procedure
parameters Dx
and G are selected in order to obtain optimum resolution after the transformation
at an appropriate signal-to-noise ratio. In the principle, one gets the best
results, if the value of G is close to the experimental “half – width” G for the *narrowest* component in the initial
spectrum* *to be distinguished. The Dx value is close to
value of G/2 (or less) while the signal-to-nose ratio is
appropriate. The less is the Dx value, the better is the resolution, but the more we
lose in the signal-to-nose ratio.

As the result one gets a
derived N-spectrum with a better spectral resolution as it consists of narrower
lines, the *line position* and the *relative absorption area* of
every spectral component remaining unchanged.

The very broaden components and various absorption
distributions lose in their amplitudes much more in the N-spectrum, than narrow
spectroscopic lines. It permits to reduce their influence on the narrow
spectroscopic components.

In general, the N-procedure improves the visual
spectral resolution up to about 2-3 times at the expense of signal-to-noise ratio.
The fitting results (for overlapping spectral components) are characterized by
lower global correlation factors, if the N-transformed spectrum is used.