Modify overall normalisation in NormaliseToMonitor

[Russell Taylor]

I will follow Paolo's suggestion for what to do:

n our chain of events, normalisation to monitor typically comes just before focussing, at least at present (see below). Normalisation to monitor only makes sense in wavelength units – there is no reason to do anything else. In fact, this is the case for all diffraction-type instruments (inc. the SANS and LSS ones). For us, the next step is always a conversion to q, followed by focusing, which implies rebinning. For this reason, it is imperative to maintain extensive (i.e., histogram) character – at least if we want to use the present focussing routine.

If we call m_i the counts in the monitor at bin i, s_i the counts in a generic spectrum and Dlam_i the width if the bin i, the simplest normalisation formula that fulfils our criterion of preserving histogram character is:

Norm(s_i)=s_i/m_i*Dlam_i

The problem with this formula is that it yields a histogram in strange units (Angstrom). In my view, the best formula is:

Norm(si)=s_i/m_i*Dlam_i*Sum(m_i)/Sum(Dlam_i)

Sum(m_i) is the sum of all monitor counts in all bins, whereas Sum(Dlam_i) is the sum of all bin width i.e., the total wavelength range. The units of Norm(si) are counts, so we have not changed units . The other advantage of this formula is that it is completely reversible, i.e., you can go back to the original data without storing anything else. In fact the total monitor counts are preserved:

Sum(Norm(m_i))= Sum(m_i/m_i*Dlam_i)*Sum(m_i)/Sum(Dlam_i)=Sum(m_i)

Another advantage is that this formula preserves the extensive character in counting time, so we can still sum two normalised runs.

---

Also need to deal with infinities when monitor count is zero.

[Russell Taylor]

(In [1482]) Modify overall normalisation factor. Re #269.

[Russell Taylor]

(In [1484]) Adjust overall normalisation by integrated count. Re #269.

[Stuart Campbell]

This ticket has been transferred to github issue ​1117