Looking at data with the diffuser in, the width of the residual distribution at TOF1 looks pretty broad (100 mm RMS). Obviously the diffuser makes a huge amount of scattering. But this means that the track matching cut is far too tight.

Consider attached plots - I show the residuals distribution for

• run 8681, which has no diffuser (3-140 beamline, 2016/04 1.2 cooling channel with 3 T in SSU).
• run 8645, which has the full diffuser (10-140 beamline, 2016/04 1.2 cooling channel with 3 T in SSU).

To generate these plots, I opened the TOF1 position cut right up when running global recon.

---

---

---

The red histogram are events that I consider to be "good", i.e. TOF and TKU are consistent as muons, chi2 in TKU is good and there is only evidence of one particle in TKU and TOF0/1.

The difference in width of the residuals is pretty clear. By default track matching uses a cut at 60 mm, which rejects many "good" tracks when diffuser is in. In the first instance, I propose opening the track matching cut right up so that the algorithm can be used at all when diffuser is in the beamline. Then analysts can make a run-by-run decision on which tracks to accept in post-processing.

One could imagine, in the future, propagating the uncertainties. That is probably a bigger job (though many elements of the code are in place, it will require some reasonably substantive changes).

Thoughts? Comments?