Feasibility demonstration of continuous signal-based neutron noise measurements by experiments and simulations

This paper demonstrates through simulations and experiments at two research reactors that continuous-signal neutron noise analysis, utilizing pulse-shape deconvolution or detector pairs, effectively overcomes the dead-time and pile-up limitations of traditional pulse-counting to provide unbiased kinetic parameter estimation at high detection rates.

The Gist

The Big Problem: The "Too Fast to Count" Dilemma

Imagine you are trying to count raindrops falling on a roof.

• The Traditional Method (Pulse Counting): You stand there with a bucket and a tally counter. Every time a drop hits, you click the counter. This works great when it's drizzling.
• The Problem: When it starts pouring, the drops hit so fast that they overlap. You can't tell where one drop ends and the next begins. Your counter gets "confused" (this is called dead time and pile-up). You start missing counts, and your data becomes useless.

In nuclear reactors, scientists use similar methods to count neutrons (tiny particles) to understand how the reactor is behaving. When the reactor is powerful or the neutrons are moving very fast, the "rain" of neutrons is so heavy that traditional counters break down. They miss the fast, important details needed to keep the reactor safe and efficient.

The New Solution: Listening to the "Hum" Instead of Counting Drops

This paper proposes a clever workaround. Instead of trying to count individual raindrops, imagine you are listening to the sound of the rain hitting the roof.

• Continuous Signal: Instead of a clicker, you use a microphone that records the continuous hum or vibration of the roof. Even if the drops are overlapping, the sound wave still carries information about how hard and how fast the rain is falling.
• The Goal: The scientists want to use this "hum" (the continuous electrical signal from a detector) to figure out the same things they used to figure out by counting clicks.

How They Tested It: Simulations and Real Experiments

The researchers didn't just guess; they tested this idea in two ways:

1. Computer Simulations (The Virtual Lab):
   They built a virtual nuclear reactor on a computer. They simulated a "storm" of neutrons and compared the old method (counting clicks) against the new method (listening to the hum).

   • Result: When the "storm" got too heavy, the clicker stopped working. But the "hum" method kept working perfectly, even when the rain was incredibly heavy. They could also detect "faster" types of rain (higher energy neutrons) that the clicker couldn't see at all.

2. Real Experiments (The Real World):
   They took this idea to two actual research reactors: one in Japan (KUCA) and one in Hungary (BME TR).

   • They hooked up special microphones (fission chambers) to record the continuous electrical signal.
   • They ran the reactors at different power levels, from very quiet to quite loud.
   • Result: In the quiet settings, both the old clicker and the new hum method agreed. But in the louder settings, the clicker failed (it missed too many counts), while the hum method kept giving accurate results.

The "Noise" Problem and the "Magic Filter"

There was a catch. Just like a microphone picks up wind noise or electrical static, the continuous signal had some "junk" in it caused by the electronics and the shape of the signal itself. This made the "hum" look a bit distorted, like a voice speaking through a bad phone connection.

To fix this, the scientists used a digital trick called Deconvolution.

• The Analogy: Imagine you hear a song played through a room with bad acoustics (echoes and muffled sounds). You know exactly what the original song should sound like. You can use a computer to mathematically "undo" the bad room acoustics and restore the original song.
• The Result: By using this "magic filter" (specifically a Wiener filter), they cleaned up the signal. This allowed them to get clear results even from a single detector, without needing a second one to help cancel out the noise.

Key Takeaways

• The Old Way: Counting individual neutrons works well when things are slow, but fails when things get fast or intense.
• The New Way: Analyzing the continuous electrical "hum" works even when things are fast and intense. It doesn't get confused by overlapping signals.
• The Fix: If the signal gets distorted by electronics or the detector's own shape, they can use math to clean it up (deconvolution).
• The Verdict: This method is a reliable, "dead-time-free" way to listen to nuclear reactors. It allows scientists to measure things that were previously impossible to see because the signals were too fast or too crowded.

What the paper does NOT claim:
The paper does not claim this method can be used to treat cancer, generate electricity for cities, or predict earthquakes. It strictly focuses on improving how scientists measure and diagnose the behavior of research reactors, specifically by overcoming the limitations of counting neutrons when they are moving too fast.

Technical Summary

Technical Summary: Feasibility Demonstration of Continuous Signal-Based Neutron Noise Measurements

Problem Statement
Traditional neutron noise techniques, such as Rossi- and Feynman-alpha methods, rely on registering discrete detector pulses to determine kinetic parameters like the prompt neutron decay constant ($\alpha$). However, these pulse-counting methods suffer from significant limitations at high detection rates. High flux environments lead to pulse pile-up and dead-time effects, causing lost counts and distorting the statistical moments required for noise analysis. While dead-time corrections can recover mean count rates, they fail to accurately correct higher-order moments (variance, covariance) essential for noise methods. Furthermore, dead-time effects limit the measurable magnitude of $\alpha$, preventing the accurate characterization of fast reactor systems, deeply subcritical states, or higher-order kinetic modes where $\alpha$ values are significantly larger than in thermal systems.

Methodology
This study investigates the feasibility of using the raw continuous voltage signal from fission chambers as an alternative to discrete pulse counting. The approach is grounded in the stochastic model of fission chamber signals developed by Pál, Pázsit, and Elter (2014), which treats the detector current as a convolution of a deterministic average pulse shape $f(t)$ and random detection events.

The methodology involves two primary avenues of analysis:

1. Theoretical Derivation: Extending the stochastic model to derive continuous-signal analogues of the Rossi-alpha (autocovariance) and Feynman-alpha (variance-to-mean) formulas. These derivations account for the finite decay constant of the detector pulse ($\alpha_e$), which introduces distortion terms in the correlation functions.
2. Signal Processing Techniques: To mitigate pulse-shape distortions and dead-time effects, the study employs:
   • Detector Pairs: Using cross-covariance between two detectors to suppress self-correlation artifacts caused by finite pulse widths.
   • Deconvolution: Applying inverse Fourier and Wiener filtering to deconvolve the average pulse shape from the continuous signal. This aims to recover an ideal sequence of Dirac delta pulses, effectively removing the detector's time constant from the analysis.
3. Validation: The feasibility was assessed through:
   • Simulations: A 1-speed point Monte Carlo model of a subcritical multiplying system (based on the BME Training Reactor) was used to generate detector signals under varying source intensities and $\alpha$ values (up to $10^5$ s$^{-1}$).
   • Experiments: Measurements were conducted at the Kyoto University Critical Assembly (KUCA) and the BME Training Reactor (BME TR). Data acquisition utilized high-speed FPGA digitizers (125 MHz sampling) to record continuous signals, alongside traditional pulse-timing chains.

Key Contributions and Results

• Simulation Results:

  • High-Rate Performance: Simulations demonstrated that continuous-signal analysis remains accurate at count rates where traditional pulse counting fails due to pile-up (several $10^5$ events/s/detector).
  • High $\alpha$ Measurement: The continuous method successfully estimated significantly higher $\alpha$ values (up to $10^5$ s$^{-1}$) compared to pulse-based methods, which degrade rapidly as $\alpha$ increases.
  • Deconvolution Efficacy: Deconvolution of the pulse shape significantly improved single-detector evaluations, allowing for accurate $\alpha$ estimation even when the detector pulse decay constant was comparable to the system's $\alpha$. Wiener filtering was shown to be robust against electronic noise (up to 2.5% Noise-to-Signal Ratio), whereas simple inverse Fourier deconvolution was more sensitive.

• Experimental Results:

  • KUCA Measurements: In subcritical and low-power critical configurations, continuous signal analysis yielded $\alpha$ values consistent with traditional pulse counting. However, at higher power levels (CR-1, CR-2), continuous signals recorded in a non-compressed format exhibited artifacts due to the nonlinear frequency transfer of the data acquisition (DAQ) chain. Compressed recording (storing only signal excursions) mitigated low-frequency noise and DAQ artifacts, enabling successful analysis.
  • Deconvolution Application: At KUCA, the CR-2 configuration (higher power) could not be analyzed via standard pulse counting due to pile-up. By applying Wiener deconvolution to the continuous signal, a pulse-based signal with reduced dead time was reconstructed, allowing for successful Rossi- and Feynman-alpha evaluations that matched lower-power results.
  • BME TR Measurements: Experiments confirmed the method's viability but highlighted limitations imposed by low detection efficiency (detectors placed outside the core). A secondary decay constant ($\sim 1500$ s$^{-1}$) was observed in the data, which OpenMC simulations attributed to neutrons scattering back from the surrounding moderator/reflector materials rather than the core kinetics.
  • Artifact Mitigation: The study identified that the DAQ chain's frequency transfer characteristics can introduce artificial short-lag correlations. Compressed recording and specific smoothing techniques were effective in suppressing these artifacts.

Significance and Claims
The paper concludes that continuous-signal neutron noise analysis is a robust and unbiased alternative to traditional pulse counting, particularly for high-rate reactor diagnostics. The primary significance lies in extending the operational range of noise methods to higher count rates and higher $\alpha$ values, which are inaccessible to pulse-based techniques due to dead-time limitations.

The authors claim that:

1. Dead-Time Independence: The continuous method inherently avoids the dead-time limitations that plague pulse counting, provided the signal is processed correctly.
2. Access to Higher Modes: The ability to measure higher $\alpha$ values makes the technique suitable for fast reactor systems, deeply subcritical states, and the characterization of higher-order kinetic modes.
3. Practical Feasibility: While electronic noise and DAQ frequency transfer characteristics present challenges, these can be managed through compressed recording, detector pairing, and Wiener deconvolution.

The authors remain modest regarding the current limitations, noting that the precision of the experimental results is currently constrained by data acquisition hardware and detector placement efficiency rather than the theoretical methodology itself. Future improvements in electronics and the use of higher-efficiency detectors are identified as necessary steps to fully realize the technique's potential.