The CCD camera for X-ray detection

The first CCD camera was developed in 1969, Bell laboratory, USA. Due to the wide range of advantages over other types of detectors CCD cameras became a breakthrough in astronomy also for X-rays detection [102].

For efficient X-ray photon detection in diffraction imaging experiments the following characteristics of CCD are the most important:

• High dynamic rangeis an important feature since the central area of a diffraction pattern normally has few orders of magnitude more photon counts than distant areas with higher scattering angles. High dynamic range ensures that the diffraction im-age is not overexposed in the center and at the same time each photon on the weakly illuminated border areas of the CCD chip is detected.

• Quantum efficiency(QE) is the probability of a detector to respond to the photon signal and convert it to the measurable signal. To assure the best QE parameter the CCD camera has to be designed for specific energy range that is correlated with the thickness of the doped region.

• Signal-to-noise ratio(SNR) characterizes the ultimate performance of a detection system representing the ratio of the measured X-ray light signal to the noise which mostly consist of dark noise and read out noise. Dark noise is related to thermal current that is a result of spontaneous generation of electrons and holes not depend-ing on whether the detector was exposed by X-rays or not. Read out noise is the result of converting CCD charge carriers into voltage signal and mostly comes form the on-chip preamplifiers.

Fast in vacuum CCD camera at MAXYMUS

Features PnCCD is a full frame fully depleted silicon detector for direct X-ray detec-tion. This camera has image area12.7×12.7 mm²and pixel size 48µmwhich can act as energy spectrometer or photon counter depending on operation mode. In comparison with other commercially available CCDs suitable for in-vacuum operation in soft X-ray range (Princeton Instruments, Andor, Hamamatsu, etc.) the current camera has comparatively big pixel size and small area. However taking into account sampling criteria and optical geometry of the set up it is sufficient for high numerical aperture and oversampling in the q-space. Implementation of frame store operation and column parallel readout, when all pixels in a row are processed in parallel, allows detector frame rate up to 930 Hz. The optimal operation frequency used for diffraction pattern capturing with sufficient statistics

Figure 4.9: Illustration of the frame store operation for high speed read out [103]. Image area is split into two halves. Each channel ends with readout anode and JFET for first amplification, which are bonded to CAMEX chips for further amplification and filtering.

is about 440 Hz corresponding to 440 frames per second (fps). Normally, operation fre-quencies of the cameras with comparable size, read out noise and efficiency, are an order of magnitude smaller.

In order to provide high-speed operation it works in split frame transfer mode (figure 4.9). Each half image with 132 lines and 264 channels is transferred within 50 µsto the corresponding storage region which is shielded from X-rays [103]. The high speed operation of the CCD dramatically reduces time for data acquisition during ptychographic measurements.

The main advantage of this CCD camera is a high dynamic range in combination with a high readout rate. Commonly double exposure technique is used to increase the dynamic range of detector. It combines long and short exposures in one diffraction frame for each scanning point [50]. However using this approach noticeably increases data acquisition and calculation time. The high performance of the CCD and application of the dynamic stacking of frames allow to use only single exposure diffraction data set. Due to this approach very fast read out rate do not reduce dynamic range of the registered patterns.

The optimal settings for the experiment can be found as a combination of read out rate and gain to ensure noise level as low as possible and at the same time accurate registration of desired diffraction features.

The pnCCD has large detection volume with the thickness of 450µmthat helps to achieve high quantum efficiency for soft X-rays. As it is shown in figure 4.10 the quan-tum efficiency is above 95%in the range from 3 to 10 keV. The CCD camera used at

MAXYMUS has 150 nm optical light stop layer made of Al that corresponds to red curve on the graph. For soft X-rays the most efficient part of80−90%stays in the energy range of 700-2000 eV. The main characteristics of the pnCCD camera are summarized in the table 4.3 [103].

Parameter Value

Number of pixels 264×264

Pixel size 48µm

Image area 12.7×12.7mm²

Frame rate up to 1 kHz

Quantum efficiency 95%at 3-10 keV

Read out speed 28 MPixel/s

Read out noise (RMS) <3e⁻/pixel Full well capacitance 10⁵e⁻ Charge transfer efficiency 0.99995

Table 4.3: Specification of in-vacuum pnCCD detector

Cooling To reduce the dark current the CCD camera is operated under the optimized temperature−25^◦C. At this low temperature and frame rate 440 Hz dark current noise contribution is less than 1e⁻/pixelRMS. Cooling is produced by thermoelectric modules with 31W cooling power each, which are connected to a copper heat exchanger cooled by water with a temperature of19^◦C. The cooling water for the thermoelectric cooler comes from the external cooling circuit through two flexible vacuum hoses (figure 4.11).

The CCD camera should be fully cooled in order to switch on the voltages. The cooling can be started when vacuum chamber of STXM is under sufficiently low pressure, in a range10⁻⁵mbar. It prevents condensation of residual gases from the chamber on the surface of the CCD chip. To vent the chamber for sample change all the procedure has to be done in reversed order.

Data acquisition To provide the voltages, which are necessary for pnCCD camera op-eration, a control unit with programmable power supplies is introduced and placed out-side of the microscope vacuum chamber. The sequencer that works as a pattern genera-tor produces pulses for charge transfer from one pixel to another, controlling clocks for CAMEX amplification and 4 analog-digital converters (ADC). In case of the power failure at BESSY the uninterruptible power supply (UPS) unit ensures back up to avoid camera damage. Transmission of data from ADC with approximate speed of 450 Mbit/s is made by two high-throughput fiber connection, which are stored in the Linux-based computer.

Figure 4.10: Quantum efficiency of the pnCCDs as a function of energy for a 450µmthick pnCCD detector with different light blocking filters. Red curve Al coating of 150 nm, green curve -Al coating of 100 nm, blue curve - -Al coating of 50 nm, black curve - a thinSiO2 andSi3N4

passivation layer [104].

The fast data acquisition, filtering and transfer to the storage PC allows to perform data processing in real time.

Dynamic stacking For ptychography data acquisition the single diffraction pattern is created by stacking the multiple frames during the dwell time for each scanning point.

Stacking the frames is a way to enhance the image contrast by reducing noise and increas-ing the dynamic range. Since the noise of the separate frames changes randomly from one to another, summing them together decreases the overall amount of background noise. At the same time the signal from the diffraction speckles is similar for all frames that results in enhancing of the real signal. The correct gain mode and slits opening gap are chosen in the way that the diffraction image is not overexposed at the region with high illumination and at the same time it is possible to distinguish dim speckles. Stacking helps to increase the number of possible digitized values linearly with the number of stacked images. In this case the dynamic range is improved by accumulating the signal from the dim regions which in case of one frame image would be drowned in the noise. The number of stacked

Figure 4.11: Photo of pnCCD camera without front cover. On the back side two flexible hoses for water cooling are attached.

frames is determined by the dwell time for each scanning position and frame rate of the CCD. The process is done automatically using CCD software.

Figure 4.12 shows radial profiles of the diffraction patterns obtained from the magnetic labyrinth domain sample (see Chapter 5.3) with dwell times 100 ms, 300 ms and 500 ms. The profiles demonstrate that the most of the photons concentrated in the first order diffraction of FZP that is equivalent to zero order light in regard to scattering. At the higher spatial frequencies we see drastic drop of the counts with variations in background signal depending on the dwell time. The shorter the dwell time the higher the noise, for instance, 100 ms dwell time has the noise about 2 times higher then signal at 500 ms. As a result using longer dwell time in diffraction imaging helps to improve contrast of the reconstructed image due to increase of SNR value.

Diffraction pattern correction

The CCD camera doesn’t have a global shutter, so it is still sensitive during the signal transfer towards the frame store areas. Therefore diffraction data need to be processed in order to reduce the artifacts produced by the signal from the highly illuminated first order of FZP, which is located in the center of the CCD chip. The noise creates long streaks propagating to the very edges of the diffraction image (figure 4.13). The CCD consist of two halves, and signal acquisition is done by transfer of the electrons as they are shifted from the center to both sides of the split CCD. The correction is done by subtraction of the median array obtained from the first and last 30 columns from the corresponding left and right halves of diffraction pattern. As a result it helps to avoid unwanted artifacts in

Figure 4.12: Radial profiles of diffraction images obtained from the low scattering sample using different dwell times: 100 ms, 300 ms and 500 ms.

reconstruction images, that normally would result in vertical periodic stripes and reduction of contrast.

Figure 4.13: Correction of diffraction pattern: a) raw diffraction image and b) the same image with read out noise filter, both images are in logarithmic scale.