Signal Tables: An Extensible Exchange Format for Simulation Data

Abstract

This article introduces Signal Tables as a format to exchange data associated with simulations based on dictionaries and multi-dimensional arrays. Typically, simulation results, as well as model parameters, reference signals, table-based input signals, measurement data, look-up tables, etc., can be represented by a Signal Table. Applications can extend the format to add additional data and metadata/attributes, for example, as needed for a credible simulation process. The format follows a logical view based on a few data structures that can be directly mapped to data structures available in programming languages such as Julia, Python, and Matlab. These data structures can be conveniently used for pre- and post-processing in these languages. A Signal Table can be stored on file by mapping the logical view to available textual or binary persistent file formats, for example, JSON, HDF5, BSON, and MessagePack. A subset of a Signal Table can be imported in traditional tables, for example, in Excel, CSV, pandas, or DataFrames.jl, by flattening multi-dimensional arrays and not storing parameters. The format has been developed and evaluated with the Open Source Julia packages SignalTables.jl and Modia.jl.