A low-noise, radiation-hardened charge-sensitive amplifier for Silicon carbide neutron detector

Silicon carbide(SiC) neutron detectors can be applied to extreme environments with high temperature and strong radiation. The total noise factor is most significantly influenced by the noise factor of the first stage in a cascade amplifier. Hence, the preamplifier must exhibit exceptional noise performance. The readout circuit is inevitably exposed to the influence of radiation, which may result in temporary or even permanent failure of electronic systems. The connection method between the SiC neutron detector and the readout circuit is typically DC-coupled. When the sensor is working, the leakage current generated by the sensor will enter the input end of the CSA and change the baseline position of the CSA. In this paper, a low-noise CSA with leakage current compensation circuit(LCC) is presented. The circuit’s voltage-to-charge amplification is 2.98mV/fC, and the error due to nonlinearity is below 0.01%. Without considering the input parasitic capacitance, ENC is 43.12e-. In addition, the circuit also has good radiation resistance and leakage current compensation function. It is implemented and simulated in the DB Hitek 0.18μm CMOS process model.

current research focus is on enhancing the noise characteristics of the CSA to raise the sensitivity of the readout circuit.
In addition, the readout circuit is inevitably exposed to the influence of radiation.The most critical effect is Single Event Transient effect(SET) and it may cause temporary fluctuations in the circuit [2].If the process of transient change is captured by the readout circuit, radiation effects will directly result in signal distortion.CSA is the most susceptible block in the readout circuit to SET.Some manufacturing process factories have introduced radiation-hardened processes, but due to their complexity and high costs, it is necessary to enhance the circuit's radiation tolerance through circuit or layout design improvements.Figure 2. the model of readout circuit Due to the DC-coupling within the sensor and the sensing circuit, the leakage current generated by the detector will enter the readout circuit, affecting the baseline position of the CSA output terminal, thereby affecting gain and linearity.Within the literature of [3], [4], the investigators proposed a pulsecharge amplifier that can compensate for leakage current.Nevertheless, the noise characteristic is comparatively low due to the application of a common gate configuration in the literature.The setup described in [5] employs a current drain to offset the input leakage current and deliver discharge current.Nevertheless, it can exclusively offset leakage current in a single direction by injecting current into the CSA, and lacks adjustability.
This paper studies the noise performance of CSA and proposes measures to optimize noise performance from structural and parameters.In addition, this paper has conducted radiation-hardening for the CSA.Moreover, a leakage current compensation circuit was designed based on Krummenacher feedback.

2.Circuit Topology And Analysis
The readout circuit (see Fig2) primarily consists of charge sensitive amplifier, pole-zero cancellation circuit, a shaping circuit, and peaking detect hold(PDH) circuit.By reading the output signal of PDH through analog-to-digital converter(ADC), the time and energy information of radiation can be obtained.SiC neutron detectors use a conversion layer for nuclear reactions to generate charged particles.These particles enter the detector, creating electron-hole pairs.These pairs undergo drift motion under an electric field, followed by collection, leading to a weak current pulse.

noise analysis
CSA requires low noise to ensure accurate energy measurements, enabling it to amplify weak charge signals and distinguish their subtle differences.The essence of CSA is an integrator circuit and it converts the charge of input current pulses into a voltage step.The charge-to-voltage conversion gain ACQ can be expressed as: (1) As depicted in Fig3, a model for an equivalent voltage noise source and an equivalent current noise sources is established at the input.Assuming that the gain of CSA is significantly large, the voltage noise primarily concentrates at the input transistor [6], [7], and it can be expressed as: Current noise can be expressed as: where kthe Boltzmann constant, T -temperature, γn -2/3(in saturation region), gm-the input transistor transconductance, Kfcoefficient of low-frequency noise, Coxgate oxide capacitance per unit area, W and Lthe input transistor dimensions, ffrequency, Rjthe resistance at the junction between the detector and the input of the frontend electronics, qthe electron charge, Idthe detector leakage current, RFthe feedback resistance, RBiasthe detector bias resistance.
The simplified noise model of CSA The noise power spectral density at the CSA output is represented as: where CG represents the CSA input transistor gate capacitance.
In front-end readout systems, Equivalent Noise Charge (ENC) is commonly used as the index for circuit noise performance.Presuming the transfer function of the shaping circuit to be H(s), then multiplying equation( 4) by the square of H(s), performing a frequency domain integration, and dividing by the maximum value of H(t), where H(t) is equal to the inverse Laplace transform of H(s).This can be defined as ENC. 1 where tpthe shaper peaking time, ap, aw, afthe noise coefficient associated with the transfer function of the shaping circuit.In addition, differential amplifiers offer better power supply rejection ratios, but they tend to have poorer noise performance and involve relatively complex circuitry [8], [9].Furthermore, substrate noise is also present.PMOS transistors are positioned in N-well, and they exhibit lower substrate noise when compared to NMOS transistors constructed on a P-substrate [10].
Based on the analysis above, increasing Cf, RF and the input transistor transconductance can be beneficial for optimizing the noise in the CSA block.Moreover, using a PMOS transistor as input transistor can help reduce noise.Nevertheless, increasing Cf and RF will result in extending the reset time, which can affect the count rate.In the end, the input transistor transconductance was increased to reduce noise, while Cf and RF are adjusted as a tradeoff based on the requirements of the count rate.According to the analysis above, the core operational amplifier adopts a single-input, single-output structure and uses a PMOS transistor as input transistor.Fig4 shows the structure of the CSA. .reset block A telescopic cascode amplifier consists of M1-M4 is the core operational amplifier due to its significant gain, excellent linearity and operational bandwidth.The disadvantage is that the output swing is small.However, considering the weak output charge of the neutron detector, and the relatively low output magnitude of CSA, it will not constrain its performance.The input reference noise of telescopic cascode is [11]: M8 and M9 supply extra current to the input transistors, thereby increasing gm1 while maintaining gm4 unchanged, and this can also lead to noise reduction.Finally, the dimension of input transistor is W/L=3000μm/500nm and its drain current=400μA.Additionally, the current in the extra brunch is 350μA.The small-signal gain of the CSA can be expressed as: A g g r r g r r g r r g g r r g r r (7) The output of the core amplifier links to two source followers formed by M6-M7 and M14-M17.The reset block, LCC and feedback capacitor link to the output of one of the source follower, while the pole-zero cancellation circuit(PZC) is connected to the output of the other.The method effectively eliminates the impact of the shaper circuit on the CSA's output signal.
During the operation of the CSA, charge accumulates at the terminals of the feedback capacitor, leading to signal pile-up.Consequently, it is customary to parallel a large resistor named reset block across the terminals of the feedback capacitor to discharge the accumulated charge.The reset block in this article adopts the topology shown in Figure 5.The equivalent resistance of this circuit remains relatively constant within a certain range of input differential voltage, allowing for a substantial equivalent resistance value [12].In comparison to regular resistors, this circuit also permits the adjustment of the output baseline position.The equivalent resistance Rreset is given by: where VBL -baseline voltage, ΔIthe drain current difference between M6 and M14.

radiation-hardening circuit
Radiation-hardening is required for both the bias circuit and the core amplifier of CSA.Each node in the bias circuit has sensitivity, and the drain terminal connection node of PMOS and NMOS is the most susceptible to SET [13].The bias circuit can be protected by multiple sensitive node active charge cancellation(M-SNACC), and the circuit structure is shown in Figure 6.The transistors M10 and M11 are mentioned in Fig. 4. When their source regions are subjected to SET impacts, there will be transient voltage variations.If MH3 and MH4 also collect charge, ideally, an equal amount of current will occur through MH1 and MH2, resulting in no change at the output node.The gate of MH4 is connected to ground, and it remains non-conductive during the normal operation of the bias circuit.It does not generate any additional power consumption but does occupy a certain amount of area.The same method is used to enhance radiation resistance in M5.
In contrast to noise analysis, it was discovered through simulation that each transistor in the CSA core amplifier requires radiation hardening.Therefore, the M-SNACC method is not suitable.In this case, the node splitting hardening technique is chosen, dividing the circuit into two branches, as illustrated in Figure 7. Since SET events are random, when one branch is impacted by particle strikes, the other branch can continue to operate normally.The probability of both branches being struck simultaneously is very low.As a result, the node splitting method can reduce the overall circuit's sensitivity to SET events.

leakage current compensation circuit
The circuit structure of the LCC is shown in Fig 8 .The compensation principle is based on Krummenacher feedback [14], [15].In the described CSA, it is possible to differentiate between two independent feedback paths.The first one is established through the connection of transistors M1, M2, and Cf, forming a structure that can be considered as a resistor in parallel with Cf.Due to the fact that M1 and M2 have same dimensions, the equivalent resistance is RLCC=2/gm1.Transistor M3 and compensation capacitor Cc, via the drain of transistor M2, create the second feedback pathway.The drain current of M2 is integrated into capacitor C1, thus controlling the gate of M3.When viewed from CSA input, this branch can be treated as an inductive path connected in parallel with Cf, the equivalent inductance is LLCC=2Cc/gm1gm3.Equivalent circuit model of CSA Leakage current is the direct current present between the positive and negative plates of the detector during its regular operation.Therefore, this leakage current is continuous and is not dependent on the signal magnitude.Most of the leakage current flows into the inductive branch rather than the first branch, thus achieving self-adjusting compensation within the scope of conception.The gate of M1 is connected to the baseline voltage，allowing the LCC regulates the output baseline position.In addition, to prevent circuit oscillations, it is essential to use a large compensation capacitor.Furthermore, in order to increase the counting rate, the designed value of τ= CfRreset is relatively small.Therefore, RLCC is significantly greater than Rreset.The reduction of LCC bias current can bring about this effect, and it will also lead to a decrease in the noise contribution from M1 and M2.
The CSA structure diagram with LCC is shown in Figure 9.The calculation of the transfer function can be performed and simplified as:

3.Results and Discussion
The circuit is implemented and simulated in the DB Hitek 0.18μm CMOS process model.It is driven by a pair of positive and negative power supply(±2.5V)and the baseline voltage is designed as ground.
In addition, feedback capacitor is set to 320fF.Within the SiC trench-type neutron detector, the rise time of output current pulses is less than 2ns, while the fall time is approximately 20ns.The minimum amplitude is approximately 200nA, the maximum amplitude is about 4μA, and the output charge ranges from 2 to 20fC.The reverse leakage current does not exceed 100nA.The time gap between the two pulses is approximately 80ns.For a transient simulation, the input current pulse was varied in the range of 200nA to 4μA, resulting in an approximate input charge of 2fC to 20.5fC. Figure 8 displays the output waveform of the CSA.The peak time is 7 ns.The charge-to-voltage conversion gain of the CSA can be calculated as 2.98mV/fC while the theoretical calculation is 3.13 mV/fC.This is due to the fact that during charge accumulation, a portion of the charge is concurrently discharged through the reset module.A four-parameter linear simulation is conducted to assess the relationship between input charge and output voltage, with a non-linear error of less than 0.01%(shown in fig11).The noise performance of the CSA is shown in Figure12.When the detection capacitance is zero, the ENC is 43.12 e -, and the noise slope is 1.81 e -/pF.To simulate the radiation hardening performance, a dual-exponential pulse current source is employed for emulating the current pulses associated with SET.Its simulation accuracy is not high, but the simulation speed is relatively fast, suitable for processes of 90nm and above [16].Its model is: where I0-peak value of the current pulse, τα -charge collection time constant, τβthe time constant of the initial generated particle trajectory.Figure 13 exhibits the current pulse waveform, and adjustments to both the peak value and pulse width can be made for the current pulse source.Given an input current pulse of 4μA, and the corresponding input charge is 20.5fC.Then injecting each MOS transistor drain with a dual-exponential current source results in significant fluctuations in both the output voltage amplitude and pulse width.
Figure 1.detector structureFigure2.the model of readout circuit Due to the DC-coupling within the sensor and the sensing circuit, the leakage current generated by the detector will enter the readout circuit, affecting the baseline position of the CSA output terminal, thereby affecting gain and linearity.Within the literature of[3],[4], the investigators proposed a pulsecharge amplifier that can compensate for leakage current.Nevertheless, the noise characteristic is comparatively low due to the application of a common gate configuration in the literature.The setup described in[5]  employs a current drain to offset the input leakage current and deliver discharge current.Nevertheless, it can exclusively offset leakage current in a single direction by injecting current into the CSA, and lacks adjustability.This paper studies the noise performance of CSA and proposes measures to optimize noise performance from structural and parameters.In addition, this paper has conducted radiation-hardening for the CSA.Moreover, a leakage current compensation circuit was designed based on Krummenacher feedback.

Figure 4 .
Figure 4. Schematic of the core amplifier Figure 5. reset block A telescopic cascode amplifier consists of M1-M4 is the core operational amplifier due to its significant gain, excellent linearity and operational bandwidth.The disadvantage is that the output swing is small.However, considering the weak output charge of the neutron detector, and the relatively low output magnitude of CSA, it will not constrain its performance.The input reference noise of telescopic cascode is[11]:

Figure 8 .
Figure 8. Schematic of LCC Figure 9.Equivalent circuit model of CSA Leakage current is the direct current present between the positive and negative plates of the detector during its regular operation.Therefore, this leakage current is continuous and is not dependent on the signal magnitude.Most of the leakage current flows into the inductive branch rather than the first branch, thus achieving self-adjusting compensation within the scope of conception.The gate of M1 is connected to the baseline voltage，allowing the LCC regulates the output baseline position.In

Figure 10 .
Figure 10.output of transient simulation Figure 11.linear simulation In a simulated circuit, by varying the parasitic capacitance of detectors, the root mean square value of the output noise voltage is equivalent to the input.Equation (10) is utilized to convert it into ENC: rms CQ V ENC Aq

Table 1 .
Table 1 provides specific information regarding the peak magnitudes and durations of voltage transient alterations.transient changes in the output( "*" represents without radiation-hardening) node