Los Alamos National LaboratoryCenter for Space and Earth Science
Part of the National Security Education Center

Los Alamos Radiation Effects Summer School

June 3 – July 26, 2019

Contacts  

Mentors and projects

Each student will work on challenging research projects with a LANL scientist or engineer as their mentor. Projects are related to current research topics in radiation effects with access to Los Alamos data, accelerators, and computation facilities.

Students are highly encouraged to contact potential mentors before applying to discuss the mentor's suggested projects and your own project ideas.

2019 Projects

Mentor

Project

Mike Holloway

mholloway@lanl.gov

Title: “Single-event effects testing in high-powered GaN HEMTs”

Description: Irradiate high-powered, high-electron mobility transistors (HEMTs) with neutrons to characterize single-event effects response and single-event gate rupture. This work would support the qualification of HEMTs for an “accelerator in space” project.

Required student skills: Familiarity with RF and analog amplifier concepts and terminology.

John Michel

jmm1@lanl.gov

 

Title: “Resilience of SHERLOC cosmic ray removal algorithm for spectral data analysis”

Description: Characterize errors in cosmic ray removal vs. total ionizing dose of spectroscopic analysis software running on gamma-irradiated processors. On missions such as Mars 2020 or Europa Clipper, planetary exploration rovers use CCDs to collect spectral data of their environments. Spectral data can be processed on-board as part of the instrument’s automated decision process to return to regions of interest. Removal of cosmic rays that affect the CCDs is a key step to having consistent spectral data and has been incorporated into the SHERLOC instrument’s software that is planned for the Mars 2020 rover. It may be possible for the software to also detect and remove anomalous data cause by radiation effects of the system’s memory. Although the code will run on the rad-hard LEON3 processor in the flight instrument, testing will be done using commercial grade memory and an FPGA with an instantiated LEON3 processor.

Required student skills: C and python.

Sam Gutierrez

samuel@lanl.gov

 

Title: “Resilience of Legion for scheduling tasks in parallel codes”

Description: Compare the resilience of a parallel processing algorithm, such as a multi-grid or conjugate gradient solver, implemented in Legion, MPI, MPI+, and OpenMP running on 1) a CPU under irradiation by neutrons with 2) an unirradiated CPU. Resilience will be characterized through output errors, number of crashes, time of convergence, and memory footprint.

Large scale, high performance computing facilities are subject to single-event effects from atmospheric neutrons and cosmic rays. Quantification of single-event effects will assist LANL scientists and engineers in assessing the fidelity of computationally intensive simulations running on multiple nodes at these facilities.

Required student skills: C++ and parallel programming experience.

Bob Robey

brobey@lanl.gov

Title: “Resilience of single and double precision mini-apps running on processors with and without ECC”

Description: Compare the resilience of single and double precision implementations of CLAMR, a fluid dynamics adaptive mesh code, as it runs on processors built 1) with ECC and 2) without ECC as they are irradiated with neutrons. Resilience will be characterized through the relative error rates of the four test conditions: 1) single precision/no ECC, 2) double precision/no ECC, 3) single precision/ECC, and 4) double precision/ECC.

Required student skills: C++ and experience in data analysis and statistics.

Bob Robey

brobey@lanl.gov

 

and

 

Laura Monroe

lmonroe@lanl.gov

Title: “Resilience of single and double precision mini-apps running on CPUs vs. GPUs”

Description: Compare the resilience of single and double precision implementations of CLAMR, a fluid dynamics adaptive mesh code, as it runs on CPUs and GPUs as they are irradiated with neutrons. Resilience will be characterized through the relative error rates of the four test conditions: 1) single precision/CPU, 2) double precision/CPU, 3) single precision/GPU, and 4) double precision/GPU.

Required student skills: C++ and experience in data analysis and statistics.

Lissa Moore

lissa@lanl.gov

 

and

 

Laura Monroe

lmonroe@lanl.gov

Title: “Comparative Resilience of Neural and Traditional Machine Learning Models in Training and Inference Stages”

Description: Compare the resilience of a neural net application on a particular dataset and task to linear regression and random forest models on the same dataset and task when single-event effects are incurred during the training and/or inference stages. Single-event effects will be introduced through neutron irradiation of processors running each machine learning model. Resilience will be characterized by the relative number of classification errors observed for four test conditions on each model: 1) no irradiation during either the training or inference stages, 2) irradiation during the training stage only, 3) irradiation during the inference stage only, and 4) irradiation during both the training and inference stages.

Required student skills: Python and machine learning or statistics experience.

Brett Neuman

bneuman@lanl.gov

 

Andy Dubois

ajd@lanl.gov

 

and

 

Laura Monroe

lmonroe@lanl.gov

Title: “Resilience of half, single and double precision algorithms”

Description: Compare the resilience of half, single and double precision implementations of an eigenvalue solver running on a processor being irradiated with neutrons. Resilience will be characterized through the relative cross-sections of errors which are 1) detected but corrected, 2) detected but not corrected, or 3) neither detected nor corrected (silent data corruption).

Required resources: half, single and double precision implementations of an eigenvalue solver,  time at LANSCE, test boards, FPGAs or GPUs

Required student skills: C++ along with VHDL/Verilog for FPGAs or CUDA/ Open CL for GPUs.

Rodney Howeedy

rodneyh@lanl.gov

 

Title: “Resilience of database files in Flash and SDRAM memory”

Description: Compare the resilience of Flash and SDRAM memory to have corrupted data recovered through block reconstruction. Four test conditions will be evaluated: Flash memory with corruption introduced 1) as an intentionally corrupt dataset and 2) as initially uncorrupted data that is corrupted through neutron irradiation of the Flash memory, and SDRAM memory with corruption introduced 1) as an intentionally corrupt dataset and 2) as initially uncorrupted data that is corrupted through irradiation of the SDRAM memory. Four types of data errors will be characterized for each test condition: 1) media read errors, 2) indeterminate data errors, 3) database block level reconstruction errors, and 4) errors putting data in tables.

Required resources: RAID 1+0 hardware, corrupted and uncorrupted test data sets, error detection software, known database (e.g., MySQL), time at LANSCE

Required student skills: Python or C plus familiarity with SQL and relational databases.

Tom Fairbanks

fairbanks@lanl.gov

 

and

 

Josh Pritts

Title: “Identification of single-event induced transients in analog-to-digital converter (ADC) data”

Description: Identify single-event transients in the output signal of an analog-to-digital converter (ADC) under irradiation by neutrons. The single bit upset (SBU) and multiple bit upset (MBU) cross-sections will be identified from output data collected using an FPGA. Identify transients in realistic signals from ADCs to DACs to support future work on test standards for ADCs.

Required student skills: Strong coder with data analysis experience. VHDL experience would be ideal.

Paul Graham

grahamp@lanl.gov

 

and

 

Heather Quinn

hquinn@lanl.gov

 

Title: “Effect of operating supply voltage on single-event response of input-output blocks (IOBs) in multiprocessor system-on-a-chip (MPSoC) FPGAs”

Description: Characterize the effects of operating supply voltage on the single-event response exhibited by IOBs in MPSoC FPGAs. Single-event upset (SEU) and single-event transient (SET) cross-sections will be generated for neutron-irradiated MPSoC FPGAs with different operating supply voltages (e.g., 1.8 V, 2.5 V and 3.3 V).

Required student skills: C for data analysis. VHDL for FPGA operation.

Lowell Wofford

lowell@lanl.gov

Title: “Resilience of Linux kernel critical bits”

Description: Compare the resilience of two Linux kernels (e.g., 3.10 and 4.9) to identify critical bits that would benefit from radiation hardening strategies. Benchmark software that stresses memory and cache will be run on test machines installed with either the 3.10 or 4.9 Linux kernel as the DIMMs of each machine are 1) irradiated with neutrons and 2) perturbed by fault injection software. A backtrace will be generated each time the operating system crashes. The backtraces will be analyzed to identify the system call responsible for the crash, and critical bits will be inferred from a statistical analysis of system calls. Two crash metrics will be evaluated: 1) the total number of errors attributed to a system call, and 2) the ratio of errors attributed to a system call to the number of times the system call was made. This work supports efforts to make an operating system resilient for space applications.

Required student skills: C required. Experience with Linux kernels, hardware, and computer engineering would be ideal

Paolo Rech

frinhard@gmail.com

Title: “SEE characterization of object detection neural networks”

Description: Characterize and compare the single-event effect response of object detection neural network phases as they execute on GPUs or FPGAs irradiated with neutrons.  The effects of single-event transients and single-event upsets on classification errors will be investigated.

Required student skills: C or python, CUDA or VHDL, and a basic understanding of neural networks.

Paolo Rech

frinhard@gmail.com

Title: “TID Resilience of object detection neural networks”

Description: Characterize and compare the resilience of object detection neural network phases as they execute embedded GPUs irradiated with gamma rays. Classification errors will be quantified as a function of total ionizing dose to identify the most vulnerable phases of the neural network.

Required student skills: C or python, CUDA, and a basic understanding of neural networks.

Josh Pritts

Title: “Characterization of a COTS embedded general purpose graphics processing unit (GPGPU) for CubeSat Missions”

Description: Radiation effects characterization to include neutron SEE and cobalt-60 TID. Develop in-situ data retrieval tools and novel radiation benchmarks for accelerated testing of NVIDIA embedded GPGPUs. Develop data reduction and post analysis tools.

Required student skills: Python, C, and CUDA. Experience with Linux-4-Tegra would be ideal.

Josh Pritts

Title: “SPA LASER SEE Characterization of a COTS embedded general purpose graphics processing unit (GPGPU)”

Description: SPA LASER SEE characterization of an NVIDIA embedded GPGPU. Develop in-situ testing and data retrieval tools. Develop data reduction and post analysis tools.

Required student skills: Python, C, and CUDA. Experience with Linux-4-Tegra and MSP430 would be ideal.

Jeff George

jsgeorge@lanl.gov

Title: “Characterization of neutron-induced effects in DC-DC converters”

Description: Characterize destructive single-event effects as a function of input bias for neutron-irradiated DC-DC converters. DC-DC converters of different vendors and geometries will be compared.

Required student skills: Python plus general lab experience with oscilloscopes and other test & measurement equipment.