Jake Carpenter: : Research Opportunities: Center for Reliable & Trusted Electronics: Indiana University

Bio

James Carpenter, commonly referred to as Jake, is currently in his Ph.D. candidacy in Intelligent Systems at Indiana University at Bloomington (IUB), having previously earned his B.S. and M.S. degrees in Electrical Engineering from the University of Tennessee at Chattanooga (UTC). During his tenure as a Master’s student at UTC, Jake specialized in the design and execution of radiation experiments, with a particular focus on advanced signal processing techniques. Presently, as a Ph.D. student researcher at IUB, he is deeply immersed in the complex domain of radiation hardening for machine learning hardware. Jake’s research underscores his commitment to integrating his extensive radiation testing experiences into the realm of AI hardware, strategically addressing formidable challenges within the radiation environment.

Research

Hierarchical Framework for Classifying Analog Single-Event Transients

Hierarchical clustering for classifying and sorting Analog Single-Event Transients (ASETs) based on similarities in waveform shape is demonstrated. Rather than solely relying on measures such as pulse height and width, the approach organizes ASETs into structured families based on shape that are correlated to sensitive regions within the device, facilitating the identification of underlying circuit response mechanisms. Dynamic Time Warping (DTW) is used to develop a distance metric within the hierarchy to align and compare transient shapes. Using pulsed-laser experiments conducted at the U.S. Naval Research Laboratory (NRL), spatial maps of ASETs in the LM124 operational amplifier are generated and clustered to reveal regions of vulnerability and distinct charge collection behaviors. The method is further validated using heavy-ion data from the Lawrence Berkeley National Laboratory’s (LBNL) 88" Cyclotron, demonstrating its robustness across different radiation sources.

DTW-based clusters superimposed on the LM124 photomicrograph

Full set of waveform families corresponding to data in (top image) organized into 10 DTW-based clusters. Fifty waveforms are superimposed within each subplot, emphasizing the general ASET shapes. Data were obtained from pulsed-laser QBB testing of the LM124.

Analysis of Single Event Transients in Arbitrary Waveforms Using Statistical Window Analysis

Window or taper functions are commonly used in data processing to detect transient events or for time-averaging of frequency spectra. A generalized window function is demonstrated using the ionizing radiation effects spectroscopy (IRES) technique to enhance the measurement of transient anomalies within arbitrary waveforms. The IRES filter is used to convolve time data with a sliding window consisting of a moment-generating function. The resulting time-dependent statistical moments can be used to eliminate any steady-state signatures, including noise, and extract transient behaviors. The IRES filter is used to analyze data from heavy-ion exposures of commercial off-the-shelf (COTS) operational amplifiers (Op-Amps), laser-induced transients in CMOS phase-locked loops (PLLs), and simulated transients in digital and analog circuits. The performance of the IRES filter in noisy environments shows that transients can be measured with higher fidelity than standard amplitude thresholding. This statistical window analysis technique may remove the need for complex triggering mechanisms on instrumentation and does not require a priori knowledge of transient characteristics. Potential applications of IRES include real-time measurement, in situ data analysis, and machine learning (ML).

Example IRES spectrogram (d) for visualizing various statistical moments of a signal’s behavior (a) pre-, (b) during, and (c) post-strike. Probability density functions are estimated within the sliding windows, and several statistical moments are calculated. Here, the time evolution of the mean of the signal’s frequency [μ(f)], the mean of the signal’s phase [μ(ϕ)], the variance (σ2), standard deviation (σ), kurtosis (κ) and skewness (γ) of the phase are illustrated.

(a) is an analog SETs in the LM124 (b) is that same analog SETs in the LM124 with random WGN added to the transient for an SNR of −17 dB. (c) and (d) Corresponding spectrograms following one application of the IRES filter (IRES1). (e) and (f) Corresponding spectrograms following two sequential applications of the IRES filter (IRES2). (We can eliminate noise by run the signal 2 through IRES)

Radiation Testing

Jake Carpenter has (200+ hours) of hands-on experience conducting radiation testing across a range of facilities. His experimental background includes short-pulse gamma testing at the Naval Surface Warfare Center (NSWC) Crane Bumblebee facility, Radiation Test Solutions’ (RTS) Colorado Springs facility Pulserad 112A flash X-ray source, proton irradiation testing at ProNova Solutions, and all heavy-ion testing facilities in the USA, the NASA Space Radiation Laboratory (NSRL), the FRIB Single Event Effects Facility (FSEE) and the K500 Chip Testing Facility (KSEE) at MSU, the Lawrence Berkeley National Laboratory (LBNL) 88″ Cyclotron, the Texas A&M University Cyclotron Institute. Across these campaigns, he has supported test planning, hardware integration, beam operations, data acquisition, post-irradiation analysis, and communication to the sponsors, with a primary focus on circuit-level radiation response.