Ticket #1225 (closed: fixed)
Interpolation Rebin
| Reported by: | Nick Draper | Owned by: | Steve Williams |
|---|---|---|---|
| Priority: | critical | Milestone: | Iteration 23 |
| Component: | Keywords: | ||
| Cc: | Blocked By: | ||
| Blocking: | Tester: |
Description
I think this boils down to a special interpolating rebin algorithm.
Follow up and ask him for the formulae he mentions and see if we need further info than that.
From Richard Heenan:
I am looking again at normalization using M1 rather than M2 at long sample-detector distances.
There is an issue here in that the time bins for M1 are a little too coarse, converting to wavelength bins of e.g. 0.5481549 in equal steps of 0.41113 angstrom. If the data reduction is asked for in smaller steps than this, or without matching boundaries, then some odd things may happen.
Would it be possible to interpolate M1 and M2 (say using a local cubic interpolation – I can give you the formulae) rather than use the normal rebin?
Also we will need a separate option to use M1 rather than M2 for the transmission calculation please, though this is not so urgent just yet…
Change History
comment:2 Changed 10 years ago by Steve Williams
- Status changed from new to accepted
Writing main algorithm took about two days.
Martyn's estimate on time required to change the Rebin tests for this case: two hours.
It took 4 hours because I found and fixed three mistakes that I'd made in the original algorithm.
comment:4 Changed 10 years ago by Steve Williams
(In [4684]) This allows InterpolatingRebin to deal with histograms whose y-values are all the same even when the x-values are invalid. The new VectorHelper function isConstantValue() returns if all the values in a vector are the same. The SANS scripts and GUI now work with InterpolatingRebin. re #1225
comment:5 Changed 10 years ago by Steve Williams
- Status changed from accepted to testing
- Resolution set to fixed
Richard Heenan said he would look at this functionality
comment:7 Changed 10 years ago by Russell Taylor
- Status changed from testing to closed
Looks OK, though really this needs to be validated by Richard Heenan. If he comes back with any issues they can go into a new ticket.
(In [4560]) Add a new rebin algorithm that inherits from SimpleRebin, also added properties documentation strings to ReBin. The algorithm uses cubic splines for a more accurate interpolation than SimpleRebin. Exactly what you get is a little hard to understand, however. For example most interpolations can be used to give better estimates of integrals while on our data, of course, this isn't the right thing to do. Unit tests to follow re #1225