MINOS


UCL MINOS Homepage
Back to my MINOS


1: dE/dX - Left:Truth , Right:After Tracking


2: dE/dX - Truth and Tracked Superimposed

You can see that for the most part there is good agreement. There are a few extraneous hits in the truth just up to plane 60. I'm going to show that these are caused by (photocathode) noise that doesn't get tracked. A bit more energy is seen in the tracked in the high energy dip after the last plane hit (lph) mean. This is caused by crosstalk hits close to the track being included in the track. This should decrease when the crosstalk hits are folded back into tracks in both tracked and truth.

3: Number of Hits - Ratio of Truth/Tracked


4: Close Up

But note: This is a ratio, in fact the number of hits plane>=54 only 0.32% of all hits. The reason can be seen by superimposing lph on this plot...

5: Fit to LPH on number of hits ratio

So the discrepancy is occuring just at the edge of the high energy tail of the distribution. The overwhelming majority of these 1.8 GeV muons have stopped by the time they get here. LPH mean = Plane 48.1 (Click for lph fit to truth)

Table: comparison of tracked to truth
Plane No. MC Tracked MC Truth Ratio Comment
19 9802 9868 1.007 Typical plane
48 6163 6267 1.017 Lph mean here
49 4208 4332 1.03
50 2267 2357 1.04
51 881 931 1.058
52 236 289 1.22
53 73 104 1.42
54 23 40 1.74
55 6 25 4.15
56 3 24 8
57 0 12 -
58 0 12 -
59 0 6 -
60 0 6 -


You can see from this table that this is largely a statistical effect. But I still wanted to know: So I started looking at some events:

6: Overexuberant Tracking!


7: The truth for the above plot

The top plot is tracked. The bottom is truth, weighted by HitBit. A red HitBit in this case means crosstalk. The last two hits in the second and fourth views turn out to be crosstalk!

8: Tracked

9: Truth HitBits

This is an effect we expected to see. It's very hard to reduce because the tracker assumes crosstalk in line with the track is just more track. A cut on energy deposited may help a little.

10: Photocathode noise
(*)This is Truth. Look at the last view - you can see some photocathode noise (from HitBits, above 0.3pe) in plane 57. This is included in the dedx for the truth but falls outside the tracking window.

11: HitBits=0!

This plot is truth in pe. The last hit in the first view has its HitBit set to 0. Chris informs me this is 'noise'. It also doesn't get tracked.

12: HitBits

(*)HitBit breakdown on logarithmic x and y scale. 50 hits in this subset of data has HitBit=0 which corresponds to about 0.24 hits/event across all events in the file. There are very few noise hits, why should this be a problem?

But there seems to be something fishy...

13: Distribution of noise in the detector


14: Y-Profile of Noise Distribution

Shouldn't it be evenly distributed in the detector?
OK, so Ryan rightly points out that if there are any hits on a pmt, then by definition the HitBit can't be zero so this plot is biased against planes that see lots of hits...
However, if this is the case, wouldn't we expect to see the number of noise hits increasing after lph max? And certainly not decreasing from plane 56 as this appears to do.

15: PCN Distribution in detector

(*)
16: Y-Profile of PCN Noise Distribution

(*)This seems correlated with hits. In fact it's almost an inverse of the crosstalk distribution
This issue remains unresolved... but it seems likely that it it's a bug in some part of the generation code.

17: Crosstalk Distribution

This is the distribution of *only* crosstalk in the detector. (i.e. no contribution from track or photocathode noise)


The next step is to fold the crosstalk back into the tracked and truth to try and even out dE/dX a bit.

The MC truth has a 'pixel' word output which (I think) can be used to identify which strip crosstalk comes from...

However I'm NEVER certain which plane it comes from.

As a result, a preliminary plot of 'x-folded' has a saw-tooth fine structure... (not shown)


Conclusion

18: dE/dX - Tracked/Truth

The comparison between Tracked and Truth is good, at least for the vast majority of events that stop roughly where we expect... (48.1 is lph mean with a sigma of 1.9) The peak remaining at plane 53 is mainly caused by higher energy deposits in the tracked compared to the truth due to some crosstalk being tracked accidentally.

19: dE/dX - Tracked/Truth - plane 53+ cut out
If you ignore >plane52, the comparison is already good to 1% without folding back crosstalk.


* (September 2003) The discussion here on Photocathode noise was later found to be erroneous but is left because it did subsequently bring to light a problem with the Monte Carlo simulation. It is not thought to significantly affect the other results presented. Future work may re-evaluate the level of PCN observed.