3.1 LaBr₃ Scintillator

The usage of LaBr₃ scintillators in high-energy experiments could be traced back to the beginning of this century, and it became an excellent choice for X-ray and γ-ray measurements in a short time [8]. The high light yield and fast decay time of LaBr₃ offer both higher energy resolution for X-ray photons at 30 keV and very small probability of signal pile-up during high-intensity flares. An assessment of radiation tolerance shows that there is no measureable degradation of LaBr₃ exposed to simulated solar protons with intensity up to 10¹² protons/cm² [9], larger than the expected orbit intensity of ∼10¹⁰ protons/cm² in a lifetime of four years.

The LaBr₃ crystals used in the spectrometer, with a diameter of 25 mm and a height of 25 mm as shown in Fig. 3, were produced by the Beijing Glass Research Institute (BGRI). Owing to the deliquescence of LaBr₃, each crystal is packaged in an aluminum shield with a head-on quartz window. Additional Teflon film and optical coupling glue are packed into the shield, to improve the uniformity of light yielded, and to provide sufficient mechanical strength against launch vibrations.

Fig. 3.Full-screen View

LaBr₃ in X-ray spectrometer: section view and picture

The thickness of the aluminum of the X-ray entrance window of the scintillator must be carefully controlled, as there is no movable attenuator similar to that used in RHESSI and STIX [5, 6]. X-ray attenuation in aluminum with different thicknesses has been calculated with X-ray flux data of a "typical large flare" from RHESSI [10]. The result in Fig. 4 shows that a compromise could be reached in the dilemma of attenuating the flux of X-ray photons below 20 keV without losing too many photons above 30 keV by choosing the thickness of aluminum as 2.5 mm.

Fig. 4.Full-screen View

X-ray attenuation in aluminum (with total cross-section)

An onboard radioactive source is required to calibrate the gain degradation of PMTs and the dependence of FEEs on temperature. Accordingly, ¹³³Ba is chosen for its multiple characteristic X-rays in the energy range 30–383 keV. The intensity is ∼30 Bq for balancing between the aim of reducing the intrinsic background of the spectrometer, which limits the sensitivity to weak flares, and the requirement of acquiring calibration spectra in ∼10 min with sufficient statistics.

3.2 Photomultiplier Tube

The PMTs used in the spectrometer are produced by Hamamatsu with a super bialkali cathode to achieve a good energy resolution of X-ray photons at ∼30 keV. These R1924A-100-01 PMTs, fortified in structure with the criterion adopted by FERMI in a private proposal, could sustain the vibration at launch and the temperature fluctuation in orbit. Fig. 5 shows the structure of a PMT installed in a polyimide socket with a permalloy shield to reduce the magnetic field from the earth.

Fig. 5.Full-screen View

PMT with socket and base circuits: section view and picture

The base circuit of the PMT is limited to an area of 45 × 45 mm², owing to the small installation space in the spectrometer. In the design of the base circuit, a critical issue is to guarantee the linearity and gain stability at high event rate [11]. The circuit is based on Hamamatsu socket E2924-05 [12], with the values of voltage divider resistors and couple capacitors being reduced, as shown in Fig. 6. An LED test described in Sect. 4 shows that this circuit provides good linearity and gain stability at an event rate up to 100 kHz.

Fig. 6.Full-screen View

PMT base circuit: initial (upper) and modified (lower)

3.3 Front-end Electronics (FEEs)

The architecture of the FEEs is shown in Fig. 7. The signals from all 99 detectors will be connected to eight sets of identical read-out circuits, each collecting charge pulses from 12 or 13 PMTs with a 16-channel charge measurement ASIC. The analog outputs of ASIC will be digitalized using analog-to-digital converters (ADCs) under control of the FPGA, and accumulated to form a spectrum. The spectra, together with some other information (such as time code and trigger status), will be packaged by the FPGA, and then sent to the electronics control box by an LVDS link with a clock frequency of 25 MHz. The status of the FEEs, and commands from the electronics control box, are both transmitted in a half-duplex RS422-based channel with a user-defined UART protocol. The same protocol is also used in the DArk Matter Particle Explorer (DAMPE) mission [13]. An extra synchronization pulse is sent to all FEEs by the electronics control box simultaneously, to synchronize their beginning and end times. The pulse interval can be adjusted according to the level of solar activity, to reduce data volume during solar quiet periods.

Fig. 7.Full-screen View

Architecture of FEEs for the spectrometer

The critical element of the FEEs is the charge measurement ASIC, i.e., IDE3380 from IDEAS, which was originally designed for SiPM, but could also be used for PMTs [14]. This ASIC contains 16 channels, each having a charge-sensitive amplifier, trigger comparator, and CR-RC shaper with track-hold circuit, as shown in Fig. 8. An improved version named IDE3381 is under development with optimizations in the dynamic range and readout speed, and will be used in the qualification model of the HXI.

Fig. 8.Full-screen View

Block diagram of charge measurement ASIC IDE3380 more details in [14]

There are three types of data produced by FEEs simultaneously: i. Spectrum data: accumulated from the digitalized output of the ASIC directly, and could be used for indirect imaging; ii. Event data: digitalized output with trigger information, and could be used for in-orbit trigger threshold calibration; iii. Counting data: obtained from trigger counters in the ASIC, determined by four configurable charge comparators to achieve event counting in different energy ranges, and could be used for flux correction of spectrum and coarse indirect imaging in high-intensity flares. Owing to the limitation of IDE3380, the counting data will only be available in the qualification model of the HXI with the new IDE3381 ASIC.

4.1 Energy Linearity Test for LaBr₃ with PMT

The energy linearity of the detectors is a critical issue to ensure the reliability of on-orbit observation results. As an absolute necessity, on-ground calibrations of the energy linearity of the detectors have been performed using both X-ray beams and an embedded radiative source. The X-ray beam test has been performed in the hard X-ray calibration facility (HXCF)–an adjustable X-ray beam with a double-crystal monochromator shown in Fig. 9, designed for the Chinese Hard X-ray Modulation Telescope satellite by the Division of Ionizing Radiation Metrology, National Institute of Metrology [15]. This facility could generate X-ray photons covering the energy range 15–100 keV with a monochromaticity better than 0.1%. The LaBr₃ detectors are placed in the test platform with a calibrated high-purity germanium detector.

Fig. 9.Full-screen View

Schematic diagram of the HXCF [15]

For energy greater than 150 keV, an embedded ¹³³Ba source is used to provide several characteristic lines from 30 to 383 keV. The test result is illustrated in Fig. 10 where error bars are defined by one sigma of a Gaussian distribution, which shows that the integral nonlinearity of LaBr₃ with PMT in the energy range 30–383 keV is smaller than 2%.

Fig. 10.Full-screen View

Energy linearity test results

4.2 Gain Stability Test for PMT Base Circuit

Another important milestone in the HXI is to measure the gain stability of PMTs for different event rates. As shown in Fig. 11, an LED test platform driven by a pulse generator has been built in place of the LaBr₃ scintillator to simulate the fluorescence induced by X-ray photons. The periodic pulse width has been set to 25 ns, which is same as the decay time of LaBr₃. Further, the pulse height has been calibrated by the FEEs, ensuring the light from the LED generates the same amount of charges in the PMT anode as those generated by an X-ray photon with specified energy.

Fig. 11.Full-screen View

Block diagram of LED test platform

Fig.12 shows the gain stability results obtained with the modified base circuit (Fig.6) as a function of X-ray event rates. For X-ray photons at 30 keV, the event rate with stable gain could reach up to 100 kHz. Higher event rate will reduce the gain of PMT owing to the saturation in the dynode and anode. With an increase in the energy of X-ray photons, the upper limitation of event rate with stable gain decreases. For a typical solar X-ray spectrum [10], there is a fast exponential decay shape, which indicates that the bottleneck of observation with stable gain in a large flare is the huge flux of X-ray photons in low energy. An attenuation for such huge flux is critical in the design of the spectrometer. In the future, a beam test will be performed to verify the effect of an aluminum shield as the attenuator.

Fig. 12.Full-screen View

Results of PMT gain stability test

4.3 Linearity and Noise Test for ASIC

For monitoring the linearity of ASIC, the same calibration circuit as that used in DAMPE [13] has been added to the FEEs. As shown in Fig. 14, the calibration circuit consists of an analog switch and a coupling capacitor. The step pulse from the analog switch will generate a negative charge on the capacitor similar to the output of the PMT. A digital-to-analog converter and an amplifier are used to change the voltage of the step pulse for different charge volumes. As shown in Fig. 13, the maximum integral nonlinearity is smaller than 2% with a charge input range of 66–495 pC. The error bars are defined by one sigma of a Gaussian distribution of the ADC channels.

Fig. 14.Full-screen View

Block diagram of calibration circuit

Fig. 13.Full-screen View

Linearity performance of ASIC

To measure the noise, we feed in a step signal through a 33-pF capacitor and utilize a synchronized external trigger to collect data from the ASIC. The equivalent noise charge (ENC) defined by one sigma of a Gaussian distribution of the ADC data is shown in Fig. 15. With the gain curve measured in Fig. 13, the ENC of ASIC could be estimated as ∼120 fC, which indicates that the FEEs have no effect on the energy resolution of the spectrometer [16]. Fig. 15 also reveals a constant output for signals smaller than ∼10 pC, which defines the lower limitation of the dynamic range of the FEEs to ∼10 keV, with an energy-charge conversion factor given by LaBr₃ and PMT at ∼1 pC/keV, smaller than the performance requirement of 30 keV.

Fig. 15.Full-screen View

Noise performance of ASIC

4.4 Spectrum Performance

The energy resolution of the entire system has been tested using ¹³³Ba and ²⁴¹Am. As shown in Fig. 16, the full width at half maximum (FWHM) resolution of the spectrometer system is approximately 13.6% at 59.5 keV and 11.9% at 81 keV. The spectrum peak at 300–350 ADC channels is derived from a combination of 31 keV Cs_Kα X-ray in the decay of ¹³³Ba, and 32 keV Ba_Kα X-ray with ∼4 keV Auger electrons in the decay of ¹³⁸La inside the scintillator. As the proportion of activity between ¹³³Ba and ¹³⁸La is already known (∼8:1), this peak could be fitted with a sum of two Gaussians. The fitting result shows an energy resolution of 19.7% at 31 keV, satisfying the requirement in Table 1.

Fig. 16.Full-screen View

Spectrum performance (Gaussian fits in red): (a) ¹³³Ba radioactive source, (b) ²⁴¹Am radioactive source