TimePix device

The hybrid silicon pixel device TimePix was developed at CERN by Medipix collaboration. It is based on its predecesor Medipix2. The device consists of a semiconductor detector chip (300 um thick silicon) bump-bonded to a readout chip. The detector chip is equipped with a single common backside electrode and a front side matrix of electrodes (256 x 256 square pixels with pitch of 55 um). Each element of the matrix (pixel) is connected to its respective preamplifier, discriminator and digital counter integrated on the readout chip. The noise of analog circuitry is about 650 electrons. Each TimePix pixel can work in one of three modes:

  1. Medipix mode - Counter counts incoming particles.
  2. TimePix mode - Counter works as a timer and measures time of the particle detection.
  3. Time over threshold (TOT) mode - Counter is used as Wilkinson type ADC allowing direct energy measurement in each pixel.

Each individual pixel of the TimePix device in TOT mode is connected to its own analog circuitry and AD converter. Thus the device contains 65&nbsp536 independent ADCs to be calibrated to energy.

Timepix device
Pixel detector Timepix. Device consists of two chips connected by bomp-bonding technique. The upper chip is pixelated semiconductor detector (ussualy Silicon). The bottom chip is ASIC read-out containing matrix of 256 x 256 of preamplifiers comparators and counters.

Charge sharing effect - clusters

A single particle often creates signal in a cluster of adjacent pixels. It is because the charge created by the particle is spreading out during the charge collection process and it can be finally collected by several adjacent pixels forming the cluster. The charge collected by each pixel in the cluster can be measured with the TimePix device. The total charge can be revealed by summation of all these fractional charges i.e. by determination of the cluster volume. As the charge collection speed depends on applied bias voltage the cluster size (number of pixels in the cluster) also depends on that voltage.

Occurence of different cluster sizes in dependence on photon energy for sensor bias voltage of 23 V. Measured with radioactive sources (55Fe and 241Am) and samples emiting characteristic X-ray fluorescence (Cu, Zr, In).
Occurence of different cluster sizes in dependence on photon energy for sensor bias voltage of 100 V. Measured with radioactive sources (55Fe and 241Am) and samples emiting characteristic X-ray fluorescence (Cu, Zr, In).

If the computation of a cluster volume is done without pixel calibration then resulting spectrum is distorted as shown in following picture (measured with 241Am source).

Distorted spectrum of 241Am Categorized spectrum of 241Am
Spectrum of cluster volumes measured with 241Am source and not calibrated TimePix device in the TOT mode. The spectrum is distorted (we cannot identify both dominant peaks). The cluster volume spectra of 241Am for different cluster sizes. Both peaks are already identified. The broad continuum in low energy region is caused by many overlapping lines of characteristic fluorescent X-rays emitted by the holder made of stainless steel.

Calibration method

As we need to characterize each individual pixel (independently on its neighbors) we have to use just single pixel clusters for calibration. For energy calibration we used gamma radiation from radioactive sources (55Fe: 5.9 keV, 241Am: 59.5 keV) and characteristic X-rays emitted by fluorescent materials (26Fe: 6.4 keV, 29Cu: 8.0 keV, 40Zr: 15.8 keV, 42Mo: 17.5 keV, 48Cd: 23.2 keV, 49In: 24.2 keV). The fluorescence was initiated by tungsten X-ray tube.

The cluster volume spectrum was generated selecting just single pixel clusters in measured data for each source. The appropriate peak was identified in the spectrum and fit by Gaussian. The resulting calibration is shown in following charts. This experimental calibration curve can be described by a nonlinear surrogate function f which is shown in the same chart. The function f depends on four parameters: a,b,c and t. Parameters a and describe linear region of the curve (for high energies), parameter t is connected with threshold level and parameter c affects curvature. Values of these parameters are determined by the least-squares fit. The calibration was performed for each pixel determining the appropriate parameters a,b,c and t.

TimePix energy calibration Calibration curve measured with test pulses
"Time over threshold" dependence on particle energy measured with X-ray fluorescent materials and radioactive sources. The dependence is modeled by “surrogate” function f.  "Time over threshold" dependence on particle energy measured with test pulses is marked by yellow triangles. Red squares denote 5.9 keV line from 55Fe and 6.4 keV Ka line  from 26Fe. The “surrogate” function f is the black line with violet circles.

Calibration verification

Energy resolution of per-pixel calibrated Timepix device was tested using the same set of radioactive sources and flourescent materials. Acquired TOT frames were calibrated per pixel to energy using inverted form of surrogate function f, then cluster volume spectra were generated as shown in following charts.

Inverted surrogate function f-1 transforming TOT (y) to energy (x):

Calibration function TOT->Energy

The measured energy resolution is 2.8 keV at 15.8 keV (Ka line of Zr), 3.2 keV at 24.2 keV (Ka line of Indium), and about 35 keV at 5.5 MeV (alpha particles from 241Am).

Calibrated 241Am spectrum Calibrated spectrum of Zr
Calibrated spectrum of In
Cluster volume (energy) spectra of 241Am measured with calibrated TimePix device in TOT mode. The peak positions are aligned for all cluster sizes. The cluster volume spectra of Ka line of Zr for different cluster size (Click to enlarge). The cluster volume spectra of In line for different cluster sizes (Click to enlarge).

Energy spectrum of alpha particles from Pu+Am source measured by calibrated TimePix device in the air.

The calibration procedure is described also in this article. Improved calibration method is described in this article.
The software tool for calibration computation (spectra fitting) is available. Please contact author.