5 The present invention relates to a method for correcting

inefficiencies at high rates of single photon counting detector systems and single photon counting detector systems having a high rate inefficiency correction mechanism. 0 Current state of the art (for single photon counting)

Hybrid pixel detectors consist of a X-ray sensitive layer (silicon sensor) and a readout chip, both divided into corresponding pixels. Each pixel in the sensor is directly5 connected (bump bonding or flip chip bonding) to the

corresponding pixel in the readout chip. The pixel size is therefore limited by the pixel size in the readout chip and the number of electronic components per pixel in the readout chip is very limited and additional electronics to extend0 the rate capabilities of the detector system are therefore hardly possible to implement. A generic pixel detector of this type is generally disclosed in the European Patent Application 1 581 971 Al . 5 The readout chip contains an array of n x m independently working channels (pixels) . Each channel has a charge

sensitive preamplifier with tunable gain, a signal shaper with tunable shaping time, a comparator and a counter with simple pixel control and readout logic. A photon impinging a0 sensor pixel generates electron-hole pairs. These electron- hole pairs are separated by an electric field generating a charge pulse. This charge signal from the sensor is

amplified and filtered by the low noise preamplifier and shaper in the corresponding pixel cell (in the readout chip) . The shaped signal is fed to a comparator with a global reference voltage and an on-pixel trim DAC. An incoming signal exceeding this threshold will toggle the comparator state. If the chip is in Expose mode (counting the photons), the comparator pulse increments the digital counter by one. During the Readout phase the pixel counter states are transferred to the chip periphery, where they are readout via dedicated readout logic. The known state of the art according to the EP 1 581 971 Al has several limitations for single photon counting: i) A general problem of readout pixel chips is related to the pixel size. Known pixel detector have a pixel size in the range of about 172 μπι ² . A smaller pixel size can

increase the image resolution, but limits also the number of transistors, and hence the functionality, that can be put on the pixel itself. This is specifically the case for

radiation hard designs where the transistors are larger as compared to standard transistors. ii) For fast frame rates the readout time (dead time) is very significant and limits the frame rate. Many

measurements are currently limited by the frame rate. The pixel detector known in the prior art having 256 x 256 pixels at a data depth of 12 bit requires at a read-out rate of 200 MHz about 4 to 6 ms for its readout. iii) For pump and probe measurements a sample is excited (pumped) and then after a selectable time the counting is enabled for a short period (probe) . This is then repeated as often as required to gain statistics, accumulating the images. Where the conditions are not constant it is

necessary to make (at least) 2 simultaneous measurements (usually pumped and un-pumped) . This is currently not possible since the counts can only be accumulated in one internal counter. iv) Further, this detector system basically suffers from inefficiencies at high rates of incoming photons. In order to cope with this problem a rate correction mechanism is provided to correct at high rates (hundreds of kHz) the efficiencies. These efficiencies at high incoming rates are predominantly caused by the overlap of analogue signals of the pre-amplifier from temporally adjacent photons that are counted as a single event instead of independent events. The size of the overlap window is a function of the temporal shaping and the signal discriminator threshold of the detector. Thus, the probability for overlapping events, moreover the probability to miss events, increases as the probability of adjacent events arriving with shorter temporal spacing increases, i.e. at higher photon fluxes. Current methods correct for this inefficiency by applying a flux dependent correction factor to the average flux over the complete exposure interval. Typically, corrections are based on the following analytic model,

Nobs = o exp( -N ₀ *T)

where N _0bs stands for the detected rate, N ₀ for the true incident photon rate and τ is the dead-time, ie. The minimum required time between two temporally adjacent photons such that both are counted. Usually an off-line rate correction in software is applied using the equation above by creating a look-up table which maps the recorded counts ( N _0bs (t) ) to the corresponding incident counts ( N ₀ (t) ) . This table has to be recreated for a specific dead-time and a specific exposure time, when either of them is changed.

The current method comprises a number of disadvantages:

a) the rate correction is averaged over the complete exposure interval, and therefore, is accurate only when the intensity of the events on the detector is constant

throughout the complete acquisition time. Thus, current methods do not treat variations in the intensity caused by changes in the sample or temporal instabilities in the x-ray sources during the acquisition time. Unfortunately, changes in the sample condition, today, is specifically a problem in protein crystallography where a continuous rotation of a single crystal can cause the intensity of Bragg peak to vary from 0 to 100% and back to 0 during just one single exposure interval. Diffraction experiments with sample under non- ambient conditions (like temperature scans, application of external stress etc.) also have the potential for dramatic changes in intensity during an exposure interval. b) the rate correction factors that correspond to the variable complete exposure time are a function of the complete exposure time and thus must be recalculated based on the actual exposure time of each individual acquisition.

It is therefore an objective of the present invention to provide a single photon-counting pixel detector (chip or multi chip assembly) and a method for correction of high rate inefficiencies of a single photon counting detector system and single photon counting detector system having this property in order to be more precise in terms of photon counting . This objective is achieved according to the present

invention by a single photon counting pixel detector chip or multi chip assembly, comprising:

a) a layer of photosensitive material;

b) an N x M array of photo-detector diodes arranged in said layer of photosensitive material; each of said photo-detector diodes having a diode output interface; c) a N x M array of readout unit cells, one readout unit cell for each photo-detector diode;

d) said readout unit cell comprising an input interface connected to said diode output interface, a high-gain charge to voltage amplifying means and a pixel counter being connected to an output of the high-gain voltage amplifying means,

e) a calculation means for integrating the total number of incident photons over a predetermined total exposure time, wherein the total exposure time is split into sub- exposure intervals, wherein distinct rate corrections are applied to the sub-counts of the sub-exposure intervals and then the number of incident photons is calculated by summing the sub-counts which are individually corrected.

With respect to the method this objective is achieved according to the present invention by a method for

correcting a total number of photons being counted over a predetermined total exposure time by a single photon counting pixel detector chip or multi chip assembly, comprising the steps of:

a) providing the single photon counting pixel detector chip comprising:

i) a layer of photosensitive material;

ii) an N x M array of photo-detector diodes arranged in said layer of photosensitive material; each of said photo-detector diodes having a diode output interface; iii) a N x M array of readout unit cells, one readout unit cell for each photo-detector diode; and

iv) said readout unit cell comprising an input interface connected to said diode output interface, a high-gain charge to voltage amplifying means and a pixel counter being connected to an output of the high-gain voltage amplifying means,

b) integrating the total number of incident photons over the predetermined total exposure time, wherein the total exposure time is split into sub-exposure intervals that are summed together,

c) applying distinct rate corrections to the counts of the respective sub-exposure intervals; and

d) calculating the number of incident photons by summing the corrected counts of all sub-exposure intervals over the total exposure time.

This detector chip or multi chip assemblyand method allows the correction of incident photons coming in at high frequencies. By the definition of sub-exposures intervals, two main advantages can be achieved: 1) The rate corrections on the very short sub-exposures correspond to more

instantaneous, less time averaged rate corrections,

therefore, leading to a more precise overall correction. 2) equal length sub-exposure intervals lead to a single discrete list of rate correction factors.

In a preferred embodiment of the present invention, the total exposure time may be divided into equal acquisition time sub-exposure intervals. The length of the sub-exposure interval (few ms) is chosen to balance a shorter, more accurate (less averaged) sub-exposure time; with statistics in the sub-events and the inefficiency arising from

unavoidable readout related pauses between sub-events.

In a further preferred embodiment of the present invention, the length of the sub-exposure interval is fixed and independent of the incoming flux, thus the rate correction factors are only a function of preamp shaping (dead time τ ) . It is, therefore, possible to provide a single discrete list of correction factors that can be applied. I.e., the rate correction factors are defined for a fixed sub exposure time t _SUb and must in this case not be recalculated in order to correspond to the variable total exposure time

since t _SUb can be held constant and only n changes to change ttotal ·

Further preferred embodiments of the present invention can be taken from the remaining depending claims.

Preferred embodiments of the present invention are

hereinafter discussed in more detail with reference to the following drawings which depicts in:

Figure 1 a schematic view of the design of a photo-detector diode ;

Figure 2 a schematic view of a part of a detector module comprising an array of photo-detector diodes as one of them is shown in Figure 1 ;

Figure 3 a schematic view of a readout unit cell comprising an input interface connected to said diode output interface, a charge sensitive preamp/shaper, discriminator and counter. Figure 4 a plot of the incident intensity on a pixel

(dashed curve), the uncorrected measured prior art distribution (courser-binned distribution) and the uncorrected measured sub-exposures distribution of the present invention (finer-binned distribution);

Figure 5 a plot of the percent error in the rate correction over an exposure time according to the prior art (dashed distribution) and with rate correction for each individual sub-exposure interval according to the present invention (solid line distribution) ;

Figures 1 illustrates schematically the architecture of a photo-detector diode 2 having a doped semiconductor p ⁺ , n ^~ , n ⁺⁺ trespassing section 4. The most commonly used material is a silicon crystal but also germanium, gallium arsenide or cadmium telluride are used.

An incident photon 6 having an energy in the range of 100 eV to several KeV before entering the doped semiconductor p ⁺ , n ^~ , n ⁺⁺ trespassing section 4 passes through a possible cover layer (e.g. aluminum) 8 and causes according to its energy and to the energy needed to create an electron hole pair a respective number of electron hole pairs 10 after x-ray absorption. In the drawings, this number of electron hole pairs is exemplarily shown by three electron-hole pairs 10 being divided by the electrical field generated by a source of bias potential 12. Figure 2 shows a schematic view of a two-dimensional pixel detector 14 having a number of photo-detector diodes 2 arranged in an array of n rows and m columns. The photo detector diodes 2 have a length 1 and a width w of about 200 μπι (or less) and a height of about 300 μπι. Below the plane of these photo-detector diodes 2 a readout chip 16 having a corresponding number of readout unit cells RO is arranged for collecting the charge from the electron hole pairs 10 generated in the respective photo-detector diodes 2. The electrical conjunction between a diode output interface of the photo-detector diodes 2 and an input interface IN of the readout unit cell RO is achieved by bump bonding using for example indium bumps 24. Figure 3 shows a schematic view of a readout cell. In which the charge signal is amplified and filtered by the low noise preamplifier and shaper in the corresponding pixel cell (in the readout chip) . The shaped signal is fed to a comparator with a global reference voltage and an on-pixel trim DAC. An incoming signal exceeding this threshold will toggle the comparator state. If the chip is in Expose mode (counting the photons), the comparator pulse increments the digital counter by one. Figure 4 shows a Gaussian like fluctuation example of the incident intensity on a pixel as a function of time or rotation angle. The upper dashed curves represents the incident intensity impinging on the pixel, while the two lower distributions represent the uncorrected measured intensity distribution in which events that temporally overlap are lost (200 ns dead time assumed) . The lower courser-binned, distribution represents a prior art

measurements, where measurements are taken every 100 ms as "desired" by the experiment. The finer binned distribution represents the uncorrected sub-exposures of the current invention, for which the sub-exposure bins (e.g. 5ms) are first individually rate corrected then combined into the courser "experimentally desired" 100 ms bins. Figure 5 shows the percent error of the measured intensity distribution after rate corrections are applied: a) the dashed upper distribution is the percent error in the prior art's rate correction where the corrections are applied on the "experimentally desired" 100 ms bins, b) the lower solid distribution is the percent error of the current invention's rate correction, in which the 5 ms sub-exposure were first individually rate corrected then summed into the

"experimentally desired" 100 ms bins.

Further, this sub-exposure rate correction can preferably be implemented in hardware (e.g. on an FPGA in the readout system) . This avoids the big overhead in data transfer and data processing that would be necessary if all sub-exposures counts would have to be transferred, stored and post processed. This also makes much shorter sub-exposure intervals possible since the frame rate for the sub

exposures is not limited by the data transfer or storage rate .