Testbeam analysis

We assisted in the setup, running and analysis of a test-beam in November 2006.

The first few modules sent to CERN were placed in a support frame (left) cabled up to the data aquisition system and placed in the path of a 400 GeV proton beam (right). Data readout from the detectors was stored on disk for later analysis.

This allowed a test of the full detection chain: silicon sensors, electronics, data acquisition and reconstruction code.

Signal to Noise Analysis

An analysis of the detector response in the absence of beam allows the electronic noise to be evaluated after corrections for common mode and pedestal noise. In the presence of beam, candidate hits in each layer of silicon are selected by requiring significant deviations (>5 sigma) from noise values. The number of candidate hits per beam crossing is shown on the left. There are 12 layers of silicon and the peak about 12 already allows an estimation of the efficiency and purity of the detector. The signal-to-noise distribution when there is exactly one hit in each layer is shown on the right. The peak of the Landau is at 22:1. The table below gives our results for each of the layers of silicon. Full details are in Karol's presentation.

Module	Noise	Sig/Noise	Module	Noise	Sig/Noise
29P	1.8	23.2	29R	2.1	21.5
31P	2.2	26.0	31R	2.1	21.6
37P	1.9	24.2	37R	2.5	21.2
27P	1.9	22.7	27R	2.1	21.1
24P	1.8	25.7	24R	2.0	22.0
28P	1.8	25.3	28R	2.0	22.2

Resolution

The resolution of the detector was investigated by constraining a track to go through the first and last layers and looking at the distance from a hit in another layer to this track. A simple alignment was first performed to ensure everything lined up - the construction of the modules at Liverpool was so good that only small shifts (less than 20microns) were required. Subtracting off the extrapolation uncertainty allowed the resolution to be determined. On the right is plotted this resolution as a function of the radius of the detector. The resolution increases with radius reflecting the increasing strip pitch. Also plotted is the theoretical limit for binary readout (pitch/root(12)). We are doing better! due to the charge sharing between strips. More detail can be found in our presentation.

Crosstalk

To achieve optimal resolution you must know precisely which strips were hit by the particle. Crosstalk between neighbouring channels, either on the physical wafer or in the readout chain, degrades the resolution and can induce biases in the determination of particle position. There are many possible sources of crosstalk: charge-sharing in the silicon detector itself; capacitive coupling on the readout line; synchronisation of the signal information; decoding by the Beetle readout chip etc. Separating out the various sources from the data is a very difficult problem. A first step in this direction is shown in this presentation to the VELO group at CERN . Using test-pulse and beam data, we can clearly identify cross-talk at the level of 10-20% in the readout (and not in the silicon). However, when examined in details, the are many features to the plots that we still don't understand.