MVD digitiser

1
MVD digitiser
Christina Anna Dritsa IPHC, Strasbourg / GSI,
Darmstadt

• Outline
• Motivation
• Model
• Clustering
• Comparison with data
• Preliminary results

CBM collaboration meeting 15/10/08
2
Why a digitiser?

• Define the MicroVertex Detector properties
• What is the optimal pixel size for the CBM-MVD?
• What is the occupancy for a given collision
  energy?
• What is the collision pile-up that the CBM-MVD
  can handle?
• What is the maximum beam intensity for CBM-MVD?
• Optimize the detector by means of realistic
  simulation.

3
The operation principle of MAPS
Particle trajectory
P-Well
Epi-layer
Substrate
Preamplifier (one per pixel)?
Diffusing free electrons
20µm
4
Digitisation model simple description
Derived from the CMS and ILD simulation software.

• Digitisation model for non-depeleted detector
• Particle trajectory divided in segments
• Energy deposited in each segment is translated
  into charge
• Charge is spread at the surface according to a
  gauss distribution
• s

sensitive volume

• Need to adapt the models parameters (s) in order
  to reproduce experimental data.

5
How a digitiser changes simulations

• Before
• The result of a particle passing through detector
  was a point like hit.
• No pixels simulated (no charge on pixels)?
• No threshold on charge of pixel for reading out a
  pixel
• No cluster of pixels
• No pile-up of clusters
• Now/Soon
• The pixel structure of the MVD is represented.
• There is charge distributed on pixels.
• Possibility to apply thresholds for cluster
  reconstruction.
• Possibility to simulate an ADC with up to 12 bits
  for pixel read-out.

6
Simulation VS real data (1)?

• Q How to check the quality of the digitisation
  model?
• A Compare real to simulated data!
• In reality
• Beam test _at_ CERN-SPS pions 120 GeV Mimosa17
  (30µm pitch, 14µm epi) measured angles 0-80
  degrees wrt the detector plane no magnetic
  field.
• In simulation
• BoxGenerator-gt shoot pions of 120 GeV, 0-80
  degrees, no magnetic field.

7
Simulation VS real data (2) Real data
Simulated
Arbitrary color scales
s 1 µm
s 50 µm
8
Comments on each parametrisation

• Parametrisation A The present model can
  reproduce the average shape of the cluster
• BUT for a limited number of neighbors
• Parametrisation B The model can reproduce the
  charge spread
• BUT not the cluster shape at large angles
  (asymmetry in XY- axis)?
• Not correct approach if we want to guess the
  track
• inclination.
• For the studies shown next, parametrisation B is
  used. Reason More conservative estimation of the
  occupancy, no need for studying track inclination
  yet.

9
Quantitative comparison of data
of pixels above threshold
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Pixel index in the central line
In the following example the threshold is 45
electrons 3noise
Histo of Simulation
1
Histo of Beam Data
10
Beam Data
Simulation
Beam/Sim
0
45
60
11
Cluster Finding AlgorithmAcknowledgments to
M.Deveaux for his contribution

• Q How to reconstruct the track position from the
  cluster?
• 1) Define charge threshold above which a pixel is
  a seed
• 2) Identify seeds on the detector plane
• 3) Define charge threshold above which a pixel is
  a neighbor
• 4) Seek for neighbors around the pixel
• 5) Flag pixels already used
• 6) When no neighbors are found anymore then save
  the cluster in a 7x7 array.
• 7) Hit position is the center of gravity of the
  charge

12
Cluster Finding AlgorithmAcknowledgments to
M.Deveaux for his contribution

• Example

Seed Pixel
Neighbor Pixel
13
Cluster Finding AlgorithmAcknowledgments to
M.Deveaux for his contribution

• Example

Seed Pixel
Neighbor Pixel
14
Cluster Finding AlgorithmAcknowledgments to
M.Deveaux for his contribution

• Example

Seed Pixel
Neighbor Pixel
15
Cluster Finding AlgorithmAcknowledgments to
M.Deveaux for his contribution

• Example

Seed Pixel
Neighbor Pixel
16
Some results (Preliminary)reconstruction
efficiency
1 AuAu central collision _at_ 25 AGeV MVD station _at_
5cm Pixel selection threshold 3noise Pixel
pitch 30µm 284 reconstructed over 296 true
hits 96 reconstruction efficiency Efficiency
loss because of selection threshold and cluster
merging
y cm
x cm
17
Some results (Preliminary)?
25AGeV MVD station _at_ 5cm 1 AuAu central
collision with delta electrons from 100 Au
Ions 30µm pixel pitch 10 of reconstructed
clusters are merged
25AGeV MVD station _at_ 5cm 1 AuAu central
collision 2 AuAu mbias with delta electrons
from 300 Au Ions 30 of reconstructed clusters
are merged
18
25AGeV MVD station _at_ 5cm 1 AuAu central
collision 2 AuAu mbias with delta electrons
from 300 Au Ions 3252 reconstructed hits 30
of reconstructed clusters are merged
19
Some results (Preliminary)?
30µm pixels
(Pile-Up)?
20
Digitiser model summary

• A digitisation model for MAPS was implemented
• Qualitative comparison between reality and
  simulation shows the limitation of the existing
  model in generating both the cluster shape and
  charge spread.
• For a given parametrisation Quantitative
  comparison shows good match with reality the
  percentage of pixels of the central line (and
  column) above threshold is similar to the one for
  real data.
• Next steps
• Try different algorithm for the digitiser so that
  both the cluster shape and charge spread are
  reconstructed.

21
Performance and limits for cluster finding

• All possible cluster shapes can be identified.
• Not easy to do hit-matching (more than one track
  contributing per cluster).
• Are all the reconstructed hits real hits?
• Almost 95-99 of hits are reconstructed,
  depending on the thresholds.
• Not easy to calculate uncertainty on hit
  position set uncertainty to 5µm for all hits
• Code is not speed optimised (20s/evt)?

22
Performance and limits for cluster finding

• From preliminary low statistics simulations it
  seems that cluster merging is not negligible
• Next steps
• Study cluster merging with smaller pixels (10µm)?
• Improve the cluster finding algorithm implement
  pattern recognition
• Further simulations with different MVD geometries
  and pile-up.