TIMEPIX AND TOT MODE 1 Taken from Llopart’s 2007 thesis: Paper X and J. Jakubek, Nucl. Instr. and Meth. A (2010), doi:10.1016/j.nima.2010.06.183 Ryan Badman, July 15th OUTLINE Introduction to Timepix Calibration with charge (test pulse) Calibration with energy 2 IMAGE OF TIMEPIX Fig. 1 3 TIMEPIX LAYOUT Analog part has one band-gap circuit to internally generate a stable reference voltage to be used by the 13 on-chip global DACs Reference voltage has sensitivity of -0.22 mV/ deg C and a power supply of less than 1mV/V Has eight 8-bit current DACs and four 8-bit voltage DACs, and a single 14-bit voltage DAC Digital part has all input/output control logic, the IO wire-bonding pads, and a 24-bit fused blown registry for unique chip identification 36 million transistors total in the entire chip, and a read-out time of less than 300us 4 CHARGE COLLECTION IN A SINGLE CHANNEL Electronics for each individual pixel Any charge collected by the pixel anode is integrated and compared to the global threshold. If the preamplifier voltage crosses the threshold then the output of the discriminator generates a pulse whose width corresponds to the length of the time the preamplifier output voltage remains over threshold Minimum detectable charge is expected to be 650 e- and the noise is 100 e-rms 5 TIMEPIX ABILITIES Timepix uses an external clock with frequency up to 100MHz Each pixel has its own preamplifier, discminator with hysteresis, and a 14-bit counter Each pixel has 4 modes: Masked mode: pixel is off Medipix mode: incoming particles are counted TOT mode: Counter is incremented continuously as long as signal is over TH, used to measure particle energy. (TOT = time over threshold) Timepix mode: counter is continuously incremented from the time the fist hit arrives until the end of the shutter Fig. 11 shows a full matrix response to 10 test pulses of about 2.3 ke- 6 TOT CALIBRATION TOT vs. Qin (test pulse) In TOT mode the energy resolution is better than 5% if the input charge is greater than 1 keV above threshold Fig. 6, TOT measurements on the right, and the measured energy resolution with 1% and 5% markers on the right. Fig 1 (above, another Timepix schematic. 7 TIME-WALK Defined as the difference between the time measured from an input charge that is 1ke- over threshold and an infinite input charge. The faster the preamp peaking time, the better the time-walk value One method of compensating the time walk is to arrange the pixel matrix in a chess pattern where pixels alternate between TOT mode and arrival mode, and then use knowledge of the input charge in the neighboring pixels. Measured time-walk per pixel is less than 50ns. Time-walk of 1 pixel at the matrix center for five different thresholds. The thicker lines represent Preamp 255 (1.8uA) and the thin with Preamp 127 (900nA). Y axis is time of arrival. 8 TIMEPIX EXAMPLE RESULTS Fig. 12 shows TOT measurements of a single cosmic background particle on the left after interacting in the gas of a GEM detector, and on the right the arrival time measurements of the cosmic particle obtained with a Micromegas gas gain grid coupled to the Timepix chip. 9 TOT MODE ENERGY CALIBRATION Each pixel needs to be calibrated for proper energy measurement. This calibration is nonlinear in lower energy ranges, and linear in higher. The calibration can be described by a surrogate function containing four parameters. Fig. 2 (a is gain, b the y-intercept, and t relates to threshold) 10 HOW THE FIT PARAMETERS VARY PER PIXEL (MY RESULTS, TH400) Note: These our for test pulse, not for Jakubek’s energy calibration Results for (col 0-40) x 256 pixel 11 COMPARE TO RICHARD’S (SAME BINNING) TH300, Col (0-40) x 256 12 PROBLEM WITH ENERGY CALIBRATION Calibration requires measurement of at least 4 spectral lines and at minimum 5 least-squares fits per pixel. However, fitting spectral peaks with Gaussians in the non-linear portion close to threshold gives incorrectly shifted results. 13 IMPROVEMENTS TO ENERGY CALIBRATION The solution to the bad peak shape is to combine a Gaussian with the inverse of the surrogate function for the nonlinear region (called model M); and just a Gaussian in the linear region. Fig. 3. The TOT spectrum of 55Fe (the same spectrum as shown in Fig. 2 on the previous slide) fitted with the model M. 14 IRRADIATION WITH MULTI-ENERGY CALIBRATION SOURCE. Jakubek irradiated the Timepix chip to obtain TOT spectra for each pixel in single pixel clusters. The threshold was just above the noise edge, sensor bias voltage was 100V, the clock frequency was 10 MHz, and Ikrum was 1. 15 CALIBRATED GLOBAL SPECTRUM The calibration quality was tested without the Fe layer and the following global spectrum was observed. Fig. 6. Calibrated spectra of dual energy source (241Am+In) for different cluster sizes (the cluster size is denoted by label). Peaks for different cluster sizes are well aligned. Peaks are fitted with Gaussians. The energy resolution (sigma of Gaussian) is 2.3keV for both peaks. 16 DISTORTION CORRECTION Since the cluster volume spectra is distorted (its supposed to be independent of bias voltage), the calibration function is not accurate for high ionization charges. The difference between estimated and measured cluster heights is a linear function so the calibration function can be corrected above 0.9MeV (up to 1.2MeV). 17 Fig. 12. Cluster volume spectra of 5.5MeV alpha particles. The distorted response of pixels was corrected above 0.9MeV and the spectrum shape restored (compare with Fig. 10). CLUSTER HEIGHT VS BIAS VOLTAGE Fig. 8. Maximum signal seen by a pixel in the cluster (cluster height) as a function of the bias voltage for 5.5MeV alphas. Fig. 9. Dependence of the cluster height on the bias voltage for 5.5MeV alpha particles. When the pixel response reaches about 0.9MeV the cluster height starts to grow unexpectedly. 18 EXAMPLE OF DISTORTION Fig. 10. Cluster volume spectra for sensor bias voltage of 39V (top) and 72V (bottom) showing distortion caused by the deviation of pixel response from the calibration function f 19 SUMMARY Timepix offers more control over individual pixels than its predecessor Medipix2. The small periphery will cause no loss in coverage if the chip is chosen for the VELO upgrade. The power consumption, radiation tolerance, pixelated format, small pitch, and convenient programmability are also favorable for the VELO upgrade. Jakubek’s correction successfully corrects the cluster height measurements for higher pixel responses up to 1.2 MeV. We will come up with a better energy calibration? 20