I still don't have the delta-x numbers to correct the alignment.
3-layer XFT is gone, and that leaves our Level-1 rate too high.
Suppose we make new mezzanine cards to put more restrictive cuts on the
trigger. I use a Z sample with 9719 BMU muons. I'm interested in what
happens when we use both a delta-T cut on pairs of hits and demand that
there be two pairs of hits. A 50-nsec deviation between hits in a pair
seems to fit nicely.
|9660||have Pt < 200 GeV/c|
|3095||have Pt < 10 GeV/c low Pt sample|
|6565||have 10 < Pt < 200 good sample|
|3685||of the good sample have 4 BMU hits|
|3671||of the good sample with 4 BMU hits have max ΔT < 400 nsec silver sample|
|942||of the low Pt sample have 4 BMU hits and max ΔT < 400 nsec lead sample|
|1976||of the silver sample lie in the same stack|
|1707||of the above (86%) have max ΔT < 50 nsec|
|1695||of the silver sample cross stacks|
|1460||of the above (86%) have max ΔT < 50 nsec|
|477||of the lead sample lie in the same stack|
|174||of the above (36%) have max ΔT < 50 nsec|
|465||of the lead sample cross stacks|
|160||of the above (34%) have max ΔT < 50 nsec|
So if a trigger requires two pairs of hits, both differing by less than 50 nsec, in either the same stack or the adjoining stack, we retain 3167 out of 6565 in the good sample, or 48%, and retain 334 out of 3095 in the low Pt sample, or 11%.
This doesn't include the edge effects, of course: we'll lose another 4% of both samples because half a stub is in one TDC and half in another, so these should be more like 46% and 10% respectively.
This needs to be checked against minimum bias samples, but if we can cut our trigger rate by a factor of about 9 or 10 we might be able to get back in the game.
A little closer look: The sample I'm running on is a Z sample, enriched
in BMU muons. It has
|43345||Have at least one pair of BMU hits||47%|
|14129||Have 2 pair in a 3-cell range, 12 stack blocking||15.47%|
|14385||Have 2 pair in a 3-cell range, 24 stack blocking||15.75%|
|14061||Have 2 pair in an interleaved range, 12 stack blocking||15.39%|
|14314||Have 2 pair in an interleaved range, 24 stack blocking||15.67%|
|9719||Number of BMU stubs found|
For a 3-cell range I look for a pair in cells 0 and 2 and another pair in cells 1 and 3 in the same or next neighboring stacks. This algorithm has the advantage of being symmetric East and West, but it lets a few fakes leak through. An interleaved range looks for the matching 1/3 pair only in the same or the interleaved stack, which can be a higher or lower stack number depending on whether this is on the west or east. Notice that the difference between the two is not great: about 0.5% excess. We can easily live with that, so a 3-cell range is the way to go for programming simplicity.
The blocking reflects the edge effects we get. A TDC has 24 stacks, and so the mezzannine card will miss interleaved pairs that wind up on a different TDC. However, I don't yet know from the card design whether or not the stacks are handled by two identical chips or not. If they are, then each chip handles 12 stacks, and we have more losses due to edge effects.
The next step is to run on a min-bias sample.
Modified 25-May-2006 at 13:34