Real Time Evolvable Hardware for Optimal Reconfiguration of Cusp-Like Pulse Shapers

TL;DR

This paper tackles drift in cusp-like digital pulse shaping caused by aging and radiation-induced sensor degradation. It introduces an FPGA-based evolvable cusp-like shaper controlled by a MicroBlaze-driven genetic algorithm that reconfigures four parameters $(k,l,m_1,m_2)$ to reproduce a reference output in real time, guided by fitness functions $F_1$, $F_2$, and $F_3$ (with $F_2$ and $F_3$ preferred). The approach enables rapid auto-calibration, achieving convergence in seconds to minutes and regenerating cusp-shaped outputs for both real CaLMa data and synthetic degenerated signals, with relative peak errors typically below 8% under strong degradation. This work demonstrates a feasible framework for adaptive digital filters in evolvable hardware, offering robust measurements in harsh environments and potential applicability to other cusp-like or trapezoidal shapers. The four-parameter design $(k,l,m_1,m_2)$ expands the configurability to about $2^{40}$ possibilities, and the evaluation pipeline provides a practical path toward self-reconfiguring radiation-tolerant sensing systems.

Abstract

The design of a cusp-like digital pulse shaper for particle energy measurements requires the definition of four parameters whose values are defined based on the nature of the shaper input signal (timing, noise, ...) provided by a sensor. However, after high doses of radiation, sensors degenerate and their output signals do not meet the original characteristics, which may lead to erroneous measurements of the particle energies. We present in this paper an evolvable cusp-like digital shaper, which is able to auto-recalibrate the original hardware implementation into a new design that match the original specifications under the new sensor features.

Figures (6)

• Figure 1: Block diagram of system prototype. REG$_{\{1,2,3\}}$ are register, DELAY$_{\{1,2\}}$ are the delay pipelines, $\Sigma_{\{1,2,3,4\}}$ are adders/subtractors, ACC$_{\{1,2,3\}}$ are accumulators and X$_{\{1,2,3\}}$ are multipliers. $d^k(n)$, $d^1(n)$, $p(n)$, $q(n)$, and $s(n)$ are presented in \ref{['eq:Conformador01']}-\ref{['eq:Conformador05']}. Signal fitness is truncated at the output port of the Fitness Evaluation module. For the sake of clarity, control module and control signals have been obviated in this figure.
• Figure 2: Regenerated response for the three fitness functions, $F_1$, $F_2$, $F_3$. For comparison purposes we also present the golden response. \ref{['fig:fitness_comparison_scratch']} Regenerated response starting from scratch. \ref{['fig:fitness_comparison_degenerated']} Regenerated response after signal degeneration.
• Figure 3: Individual encoding.
• Figure 4: \ref{['fig:ThreeInputs']} Initial and two degenerated events with $\delta = 0.8$ and $\delta = 0.6$, respectively. Two EHW recalibrations of the original shaper: \ref{['fig:ThreeOutputs80']} For the first degenerated event $(k,l,m_1,m_2)_{\mathrm{reg}}=(31,15,68,16)$, and \ref{['fig:ThreeOutputs60']} for the second degenerated event $(k,l,m_1,m_2)_{\mathrm{reg}}=(31,15,89,20)$.
• Figure 5: Histograms of the original and degenerated sensor signals with \ref{['fig:Histogram80']}$\delta=0.8$ and \ref{['fig:Histogram60']}$\delta=0.6$, respectively. Original: is the histogram of the original signal for shaper's parameters $(k,l,m_1,m_2)_{\mathrm{ref}}=(31,15,57,13)$. Damaged is the histogram of the degraded signal for the same parameters. Restored is the histogram of the degraded signal for optimized shapers' parameters $(k,l,m_1,m_2)_{\mathrm{reg}}=(31,15,68,16)$ and $(k,l,m_1,m_2)_{\mathrm{reg}}=(31,15,89,20)$, respectively.