Modeling digital signal processing to confirm frequency response and performance
summary
A Julia-based text API and GitHub repo models DSP blocks, computes exact transfer functions for composed multi-rate chains, and extracts frequency response from time-domain runs to validate performance. It implements a fifth-order IIR low-pass made from two parallel allpass sections (coefficients from the paper), converts to a fixed-point coefficient quantization implementation under ~200 lines, measures SNR/IMD/THD via weighted least squares, and shows time-domain results matching the analytic transfer function down to parts-per-million ripple.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:23Defining DSP problem
- 3:50Composing processors and loops
- 5:37Low-pass filter example
- 9:53Fixed-point implementation
- 12:05Measuring frequency response
- 14:53Transfer function vs time results
- 15:57Additional examples and flexibility
- 16:19Summary and Julia details
- 16:41Q&A
related talks
37mBuilding a Zero-Copy DSP Offload Framework in Linux Using RPM...
29mpyadi-jif: JESD204 tools for mere mortals
27mEnabling Performance Portability of GNU Radio on Heterogeneous Systems
20mNew DSP Extensions to the Embench Benchmark Suite
23mCedarEDA for open source silicon
16mRISC-V DSP (P) Extension Proposal
24mMeet the Family: RFNoC Blocks in UHD
29mMega Hertz, Mega Samples, Mega bits, Mega Confusing