MyFilterRevised11

Notes on Fabric's clustering and filtering.

Limited topology accepted as a cluster:
Highest, 2 nd highest and 3rd highest must form an L with the highest at the knee.
i.e. the second highest and the third highest can never lie on a diagonal with the highest.

This is the spirit behind his routine fLocate16PixCluster
In this routine the pixels of a 4 by 4 cluster are organized into the tLoc matrix where the first index of tLoc is the pixel number 0-15 and the second index is 0 or 1 designating the row or column of the pixel. The ordering of the 4 by 4 array to the 0-15 pixel index is shown in the following figure:

The numbers in this table show the pixel numbering for the first index of the tLoc matrix. The topology has been flopped and rotated such that the pixel amplitudes appear as shown in the 4 by 4 table. The amplitude ordering is not well defined for pixels other than the central 4. Clusters that do not conform to this topology are probably not included, but this needs to be checked.

Definitions of tVar in fCalcClusterFilterVar

tVar0 Highest Pixel Signal to noise
tVar1 Probability that it is noise using ChiSqrd of 2nd and 3rd highest pixel
tVar2 Probability that it is noise using ChiSqrd of 2nd highest, 3rd highest and diagonal
tVar3 Probability that it is noise using ChiSqrd of all but the max amplitude pixel (Fabrice argues that removing the highest pixel removes the bias on this cut from the orginal max pixel cut
tVar4 ChiSqrd of highest, 2nd highest, 3rd highest and diagonal
tVar5 ChiSqrd of all but max amplitude pixel.

Notes, by HW 31-May-2005

Input parameters

in e-

in pitch

in number of noise sigma

Generate impact position

Non-redundant random locations over the pixel

Generate signal for different ionization. From 50 e- to 550 e-

rnorm returns vector with element for each electron with a gaussian distribution
about the x impact point. So this gives the projection onto the x axis

in e-

this distributes the projected number in each x bin into y bins

first index different impact locations, second index different signal amplitude

Generate noise. Only keep the events with a signal > seed threshold on the cental pixel

Suppression factor due to seed threshold

Gaussian noise distribution into each pixel

i counts the different random noise events

keep throwing until over threshold then
store in center

Definitions of tVar in fCalcClusterFilterVariable

tVar0 Highest Pixel Signal to noise
tVar1 Probability that it is noise using ChiSqrd of 2nd and 3rd highest pixel
tVar2 Probability that it is noise using ChiSqrd of 2nd highest, 3rd highest and diagonal
tVar3 Probability that it is noise using ChiSqrd of all but the max amplitude pixel (Fabrice argues that removing the highest pixel removes the bias on this cut from the orginal max pixel cut
tVar4 ChiSqrd of highest, 2nd highest, 3rd highest and diagonal
tVar5 ChiSqrd of all but max amplitude pixel.

Probably returns list of pixel locations for cluster of 16. Haven't worked out topology

returns square cluster only even if non-sq element is larger

tLoc gives each pixel location for the cluser of pixels

locates max adjacent, does not do diag
probably starts with only center 3by3 so actually pixel with max could could have smaller value than that in 2max

Locates pixel with max signal in the
center 3 by 3

First index is pixel number in the cluster
Second index selects row or column

Filter signal and noise