Memos

The EPIC memo series documents the engineering and commissioning work underlying the published papers. It includes profiling results, analyses of failure modes and design trade-offs.

No.	Title	Authors	Date	Content
001	A Roadmap for Efficient Direct Imaging with Large Radio Interferometer Arrays	EPIC Collaboration	2019	Astro2020 white paper: computational scaling of direct imaging; development roadmap covering calibration, wide-field correction, FPGA/ASIC implementations and hybrid architectures
002	Searching for Crab Giant Pulses with EPIC	C. Whitton, A. Beardsley	Oct 2019	Data artefacts in early observations; an interference filter based on reference pixels; the requirement to flag antennas before imaging
003	Romein Optimization	H. Krishnan	Jan 2020	Restructuring of the gridding kernel to one thread per antenna; threefold speed-up
004	1D Omniscope/EPIC Hybrid	A. Beardsley	Mar 2020	Hierarchical-grid FFTs applied to EPIC; computational cost and grid fitting
005	Cross-Correlator Module / Romein Bug Fix	H. Krishnan	Apr 2020	A cross-polarization module supporting grids up to 256²; corrections to the gridding kernel
006	FRB Detectability with EPIC	H. Krishnan	Oct 2020	System equivalent flux density and sensitivity of LWA-SV; dispersion and scattering at 10–88 MHz
007	EPIC Data Assessment	H. Krishnan	Aug 2021	Station multicast configuration; 81.92 ms data products; effects of commensal retuning
008	EPIC vs Beam: Data Comparison	H. Krishnan	Nov 2021	Comparison of solar dynamic spectra with beamformed data; saturation in the beamformer
009	EPIC Code Optimizations (Part I)	K. Reddy	Feb 2023	Profiling of the Bifrost implementation; design of the fused cuFFTDx kernel; GPU requirements
010	EPIC Imager Data Management	K. Reddy	Apr 2023	Data rates; storage and transport of images and pixel data
011	Float vs Half Precision Accumulation for EPIC Images	—	2023	Error budget of half-precision imaging
012	EPIC Imaging & Source Localization Comparisons	C. Taylor, J. Dowell	Nov 2023	Validation against the LWA-SV FX imager; correction of autocorrelation removal, pixel scale and orientation; residual rotation
013	Searching for FRBs with EPIC	S. Michaud, A. Beardsley	Aug 2024	Search of database spectrograms for CHIME/FRB-triggered events; systematic effects in the spectrograms

Recommended reading

The memos may be read in the following order:

1. Memo 001, for the motivation and the scope of the direct-imaging programme.
2. Memos 007, 008 and 012, for the characteristics and validation of EPIC data.
3. Memo 009, for the design of the GPU implementation.
4. Memos 010 and 013, for the data management and its use in transient searches.

Open issues

The memos and papers identify the following unresolved issues, which bear on the interpretation of EPIC data:

• Calibration. No real-time antenna gain or bandpass calibration is applied, and image values are in arbitrary units (see Calibration).
• Astrometry. A residual image rotation of approximately 1°–3°, increasing toward the horizon, was measured relative to the FX imager (Memo 012).
• Interference. No online flagging is performed. Interference near the horizon couples into the full image through the sidelobes of the point-spread function (Memos 002, 013).
• Commensal tuning. The observing frequency is set by the station schedule, and imaging sessions terminate on retuning (Memo 007).
• Transient processing. Dedispersion, automated detection and bandpass normalization are not part of the real-time system (Memo 013).