1. Equalization of Medipix2 imaging detector energy thresholds using measurement of polychromatic X-ray beam attenuation Josef Uher a,b, Jan Jakubek c a CSIRO PSE, Lucas Heights, NSW, Australia, b CSIRO MDU, NRF, Australia c IEAP-CTU, Prague, Czech Republic

2. CSIRO. Outline Medipix detector description Energy threshold equalization using noise floor Energy threshold equalization using fits of X-ray attenuation curves Examples of application Microfocus X-ray tube Sample Imaging detector X-ray transmission was used in measurements

3. CSIRO. Detectors of Medipix family Images from www.utef.cvut.cz Medipix detectors: array of 256 x 256 pixels pixel size 55 x 55  m total area 1.5 x 1.5 cm 2 Planar 300  m thick silicon pixel detector (GaAs and CdTe also available) Bump-bonded to a readout chip containing amplifier, discriminator (low and high threshold) and counter for each pixel 3-bit fine threshold adjustment in each pixel

4. Threshold equalization 3-bit fine threshold adjustment in each pixel values minimizing the threshold level spread has to be found the common technique is based on finding the noise floor in each pixel threshold can be set just above the noise floor CSIRO. Adjustment bits = 0 Adjustment bits = 7 Optimized adjustment bits

5. Signal-to-thickness calibration (STC) CSIRO. Replacement of the flat-field correction STC converts the measured number of counts into equivalent thickness of the calibrating material (Al) Measured signal Equivalent thickness X-ray attenuation in Al foils is measured

6. Signal-to-thickness calibration (STC) CSIRO. X-ray tube running at 40kVp, energy threshold set to ~13 keV i.e. above k-edges of Ni and Cu and below k-edges of Ag and Sn STC raw imagecorrected image sample photo Al Ni Cu Ag Sn 13 keV threshold Sn Ag Ni Cu

7. STC and threshold close to k-edge of Sn CSIRO. Aluminum ~3 mm thick Tin foil 25µm thick STC corrected imageImage histogramDetected X-ray spectrum X-ray tube running at 40kVp, energy threshold set to ~27 keV (Sn k-edge is at 29.2 keV) Can we further reduce the threshold spread?? Note: The spectra do not include the detector response.

8. Better threshold equalization CSIRO. Equalization using monochromatic X-rays: + reduces the total threshold spread by 30% － requires change of the X-ray setup arrangement － count rates are low => timely procedure Equalization using polychromatic X-rays measurement fit Analytical function calculating the detected signal: Requires: mass attenuation coefficients X-ray spectrum Medipix detector response matrix Attenuation curves at different thresholds Free parameters of the fit: amplitude thresho ld 6 keV 12 keV 20 keV 30 keV

9. X-ray spectrum CSIRO. Measured using Silicon Drift Detector: AmpTek SDD X-123D sensor thickness 450 µm Be window 12.5 µm thick collimator ∅ 1 mm X-ray tube: Oxford Instruments Monoblock XRS-75-200D W target Inherent filtration @75kVp: 1.5 mm Al The measured spectrum was corrected on the SDD detector response Precise knowledge of the spectrum is essential for this technique!

10. A function that calculates the charge sharing in the detector was developed. Its parameters (charge spread) are obtained from fit of experimental data. Needs to be constructed only once for the given detector bias. Medipix detector response matrix CSIRO. Photon Pixel cells The electric charge deposited in one pixel spreads into adjacent pixels. The hit pixel “sees” less charge! Pd k α Pd k β The charge sharing produces a large tail in the measured spectrum. IEEE NSS 2010, Knoxville, Tennessee, USA

11. Noise optimized adjustment bits Equalization using the X-ray attenuation curves CSIRO. Adjustment bits = 0 Adjustment bits = 7 STC optimized adjustment bits X-ray attenuation in Al is measured for both extreme values of THL adjustment Measured attenuation curves are fitted in individual pixels and threshold value in keV is determined Optimal THL adjustment value is determined FWHM: 1.4 keV → 1.1 keV FWTM: 2.7 keV → 1.6 keV Note: The upper threshold can be equalized relative to the lower threshold

12. STC and threshold close to k-edge – new equalization CSIRO. Aluminum ~3 mm thick Tin foil 25µm thick STC corrected imageImage histogramDetected X-ray spectrum X-ray tube running at 40kVp, energy threshold set to ~27 keV (Sn k-edge is at 29.2 keV) Aluminum ~3 mm thick Tin foil 25µm thick STC corrected imageImage histogramDetected X-ray spectrum FWHM: 723 → 634 µm FWTM: 1671 → 1099 µm old new

13. New equalization – raw images before the STC CSIRO. Aluminum ~3 mm thick Tin foil 25µm thick Raw image with noise equalization Image histogram X-ray tube running at 40kVp, energy threshold set to ~27 keV (Sn k-edge is at 29.2 keV) FWHM: 6647 → 4058µm FWTM: 12690 → 7587 µm Raw image with attenuation fit eq. FWHM: 5146 → 3368 µm FWTM: 9220 → 6298 µm

14. No Medipix response matrix? CSIRO. The attenuation curve shape should be dominated by the material of sample, sensor, the X-ray spectrum and threshold. Can we omit the response matrix? Attenuation curve fit Fit residuals THL adj bit set to 7 pixel 111,87 THL adj bit set to 0 pixel 111,87 measurement fit with no resp. matrix fit with resp. matrix measurement fit with no resp. matrix fit with resp. matrix threshold from fit: 19.1 keV vs 17.5 keVthreshold from fit: 27.0 keV vs 26.2 keV

15. THL adjustment with and without the response matrix CSIRO. Distributions of STC calibrated values for Sn foil 25 µm thick Can we omit the response matrix? Distribution of differences of the adjustment with and without the response matrix adj. without resp. matrix adj. with resp. matrix

16. k-edge imaging - examples CSIRO. Sn Ag Pd Mo Sn Ag Pd Zr Cd Sn Ag Cu Ni X-ray tube: 50kVp, 500µA Threshold scan: 7÷40 keV Energy step: ~1.4 keV Total measurement time: 15 min K-edges identified by relative differences of the measured spectrum and a spectrum under Al. Mo 25-75µm k = 20.0keV Pd 25-75µm k = 24.3keV Ag 10-30µm k = 25.5keV Sn 25-75µm k = 29.2keV Ni 20-60µm k = 8.3keV Cu 20-60µm k = 9.0keV Ag 10-30µm k = 25.5keV Sn 25-75µm k = 29.2keV Zr k = 18.0keV Pd k = 24.3keV Ag k = 25.5keV Cd k = 26.7keV Sn k = 29.2keV Thicknesses: 10 – 200µm Mo Sn Pd Ag Ni Sn Cu Ag

17. Conclusion Advantages of the presented equalization method: better energy threshold equalization compared to the noise based technique comparable performance to the monochromatic source based equalization simple to implement easy to scale to large area detectors easy automatization of this method rough threshold energy calibration signal-to-thickness calibration for raw image correction precise calculation of the detector signal can be used: to calculate signal-to-thickness calibration for different materials other than the one used in the measured calibration better beam-hardening correction in iterative CT reconstruction techniques (the Expectation Maximization method) CSIRO.

18. Contact Us Phone: 1300 363 400 or +61 3 9545 2176 Email: enquiries@csiro.au Web: www.csiro.au Thank you Process Science and Engineering On-Line Analysis & Control Josef Uher Phone: +61 2 9710 6720 Email: Josef.Uher@csiro.au

19. Backup slides CSIRO.

20. Detection of k-edge in the spectrum CSIRO. Spectrum behind Al Spectrum behind Sn ∆E I Al /I Sn =1 I Al /I Sn >1 Spectrum of k-edges is generated by running the window across the measured spectrum. Only values above and below the edge can be measured if the elements in the sample can be anticipated.