2.1.1 Detection

As shown in Fig. 1, the detection function consists of ²³⁵U FC, as well as the preamplifier. Because of the wide range and low uncertainty of the reaction cross section, ²³⁵U FC has no limitations on the energy of the incident particle, making it applicable to the incident neutron with primary energy varying from 0 to 200 MeV. The fission reactions between neutrons and ²³⁵U in FC generate two fission fragments with kinetic energy between 40 and 110 MeV. The fragments are multiply charged heavy ions and considerably easy to ionize, generating strong current pulses in ²³⁵U FC. This is accomplished by calculating the amount of fragments the flux of the input neutron can obtain. Because current pulses induced by fission fragments are considerably stronger than the current induced by the γ-ray, ²³⁵U FC provides great advantages on background γ suppression.

To minimize the noise induced by the cable between the preamplifier and FC, and to improve the output signal to noise ratio, the broadband current preamplifier is applied. Moreover, in order to allow the output voltage of a single neutron signal to range between 0.5 and 1 V, the total amplification factor of the preamplifier is set according to the quantity of electric charge induced by a single neutron in FC.

2.1.2 Discrimination

Figure 2 indicates the structure of the discrimination function, including the signal processing unit (SPU) and host system.

Fig. 2Full-screen View

(Color online) SPU Structure

An essential part of the TMSR-NFM system is SPU, which processes the input signal to obtain real time neutron counts. The output signal of the preamplifier is connected to SPU through a long cable. The analog signal processing circuit first adjusts the input amplitude and then converts the single-ended signal into differential ones in order to satisfy the requirement of the 250-MHz, 12-bit ADC. Thereafter, the digitized data stream is directly transferred to FPGA for further data processing.

The NFM algorithms and flow control are all realized in FPGA. In this process, Altera Cyclone III EP3C16Q240C8 [15] is applied in SPU to perform the following functions.

A. Calculate the input baseline offset. In the self-checking mode, the input baseline offset is calculated and recorded, then deducted from the ADC measurement as a means to obtain actual input signals during the NFM operation.

B. The implementation of the count and Campbell algorithms is realized. Both algorithms operate in parallel and results are simultaneously uploaded to the host.

C. Multiple functions for the interaction between the host and SPU, including the SPU self-check, trigger processing, data upload, and realization of CAN protocol, as well as system instructions and settings were implemented.

The communication between the host and SPU is via the CAN bus, which is mainly used for system settings and upload of neutron signals from FC. The application of the CAN bus guarantees the quality of data communication by taking advantage of its superior real time performance, long transmission range, and high electromagnetic immunity.

2.2.1 Count Algorithm

When the neutron flux is relatively low, neutron pulses are discrete, and their amplitudes are considerably stronger than other signals are. Thus, by setting a proper noise threshold and discriminating the pulses with amplitudes beyond that threshold, an accurate number of neutron pulses can be obtained. This simple method remains available when the input neutron count is lower than 4 × 10⁵counts per second (cps). However, when the input flux exceeds the given value, miss counting would occur because pulses pile up, and the bias would increase when the flux intensifies.

2.2.2 Campbell Algorithm

Under a high neutron flux situation, we applied the Campbell algorithm to obtain the neutron pulse count from the output neutron waveform of FC. The Campbell theory states that the output of a nuclear detector is subjected to the Poisson distribution through time. Furthermore, when it undergoes a linear time-invariant system with an impulse response of h(t), the variance of the output voltage at time t is as follows:

where v is the output voltage at time t, Q is the current impact strength, h(t) is the impulse response of a linear time-invariant system, and $\lambda$ is the average output pulse number of the detector.

For the same fission reactor and detector, $\overline{{\mathcal{Q}}^2}$ is a constant, whereas for the same preamplifier, $\underset{0}{\overset{\infty }{{\displaystyle \int }}}{h}^2\left(t\right)\text{d}t$ is a constant. From Eq. (1), the variance of the continuous output voltage is proportional to the neutron flux; thus the neutron pulse number can be calculated. The Campbell algorithm is employed when the neutron count is greater than 4 × 10⁴cps. When the flux is lower, irrelevant signals and noises gain larger proportions in the input signal, potentially affecting the accuracy of the algorithm.

2.2.3 Normalization

Figure 3 depicts the result of a simulation test of the algorithms. Simulated neutron pulses are generated by Keysight 81160A Pulse Function Generator, and the neutron pulse numbers individually calculated by the two algorithms are plotted as ordinates. In the figure, yellow dots represent the calculation result of the count algorithm, whereas blue dots represent that of the Campbell algorithm. The figure also shows that the available measuring range of the count algorithm is between 0 and 4×10⁵cps, whereas that of the Campbell algorithm is between 4 × 10⁴and 10⁸ cps. An overlap between the two regions exists, where both algorithms can work properly, so that the results of the two algorithms should be linear, as follows:

Fig. 3Full-screen View

(Color online) Simulation test of count and campbell algorithms

Utilizing linear fitting on Eq. (2), the values of a and b can be obtained so that all calculated Campbell results can be converted to an actual count value. The count waveform can be obtained by unitizing results from the two algorithms and intact neutron. The count algorithm results are applied when the flux is below 10⁵cps, and the Campbell algorithm results are applied when the flux is above 10⁵cps. The combination of the two methods made it possible to overcome the respective limitations of each algorithm on the applicable neutron flux range, which considerably broadens the total measurement range of the system.

3.2.1 Reactor background.

When all the control rods were dropped, and the reactor was not operating, we conducted the neutron background count test, which lasted for 14.283 h. The neutron count value was obtained every second, and a total of 51 418 measurement results are plotted in Fig. 7, revealing the fluctuation of the background neutron counts of the reactor. The experiment demonstrates the appreciable real time performance and stability of the NFM system.

Fig. 7Full-screen View

The measurement of reactor background

3.2.1 Reactor operation

As shown in Fig. 8, for the first 143 s, all control rods remained in full insertion. From 144 to 882 s, the control rods were extracted at certain speeds, and the neutron flux increased. A total time of 341 s was measured for the sampled neutron counts, which increased from 5 × 10³ to 5 × 10⁷cps. At 754 s, the sampled neutron count reached the upper limit of the system of 1.08 × 10⁸cps. Because of the output scale of the preamplifier, the output voltage of the preamplifier reached its upper limit. This limited the measurement range of SPU when the neutron count from FC exceeds 1.08 × 10⁸cps. However, a slight waveform drop could be observed. Because the input signal exceeds the measurement range of the preamplifier at that time, the transistor quiescent operation point would shift, and the output gain of the preamplifier would decrease because of the transistor’s operating characteristics. From 882 to 1488 s, the control rods were inserted, and 343 s elapsed for the neutron count to decrease from 5 × 10⁷to 5 × 10³cps. From 1488 to 1517 s, the control rods remained still. From 1517 to 2170 s, the rods were extracted for a second time. The time that elapsed for the neutron count to increase from 5 × 10³to 5 × 10⁷cps was 348 s. The time spent during each period was approximately equal, because the insertion and extraction of the control rods had the same velocity.

Fig. 8Full-screen View

(Color online) Measurements during reactor operation

Because the insertion and extraction speeds of the control rods were fixed, the neutron flux in the reactor developed exponentially [20–21]. As shown in Fig. 9, sampled neutron counts during the extraction of the second control rod were exponentially fitted. From 1566 to 1941 s, the sampled neutron count increased from 2 × 10³to 5 × 10⁷cps, during which the measurement value fitted well with the exponential function model (R²= 0.9897).

Fig. 9Full-screen View

(Color online) Exponential fit during the extraction of control rods