mmWave · FMCW Radar · Radar Signal Processing

From IWR6843 Configuration to 3D People-Tracking Output

2026-09-04 · updated 2026-09-04 · Hyeongrok Ryu

A system-integration case spanning IWR6843 EVM setup, DSS processing, dual UART, and Visualizer output.

Series
FMCW Radar Signal Processing · 6
Type / level
project-log · advanced
Tools
IWR6843ISK, IWR6843AOP, TI Radar Toolbox, Code Composer Studio, SYS/BIOS, UART

Test setup

I mounted the IWR6843-family EVM on a stand and aligned its height and orientation with the direction of movement. Mounting pose and the observation region affect whether points leave the display or floor clutter dominates, so I adjusted the physical setup while watching the GUI output.

IWR6843 EVM on a stand beside the PC Visualizer environment
The test setup combines the IWR6843 EVM, interface board, and Visualizer.

Radar processing

The retained DSS project shows range and Doppler processing followed by 2D Capon/MVDR beamforming and dynamic/static CFAR. Those stages produce point-cloud and target-list outputs with angle, Doppler, and SNR information. I mapped each GUI point back to the sequence from chirps through detection instead of treating the screen as a standalone sensor output.

DSS-to-MSS handoff

The C674x DSS build and SYS/BIOS configuration handle radar processing. Results then move through shared memory and a mailbox toward the MSS data path. The retained package contains an MSS binary but not its matching source, so I focused on the interface between DSS output and the serial packet consumed by the GUI.

Two UART links

I separated the CLI and data ports during connection. The CLI uses 115200 bps for sensor configuration, while the data port uses 921600 bps for point-cloud and target packets. A swapped port or mismatched baud rate can let configuration pass while the screen stalls—or leave incoming data undecodable. This distinction helped isolate hardware, firmware, and PC-application issues.

CLI UART  : 115200 bps  -> sensor configuration
Data UART : 921600 bps  -> point cloud and target packets

Visualizer output

After changing sensor position, FOV, range, boundary, and detection thresholds, I compared target IDs and bounding boxes over time in the 3D tracking view. I also inspected the range profile, point cloud, and vital-sign display.

People Tracking runtime view with a target ID and 3D bounding box
One frame from the run, with the target ID and position updated together.

Current scope

Recorded work: EVM setup, DSS build analysis, dual-UART integration, and Visualizer output using TI examples. The 18.17-second run contains a moving 3D target. The vital-sign view records UI output; clinical, CAN/LIN, and environmental/EMC evaluation are separate stages.

Sources used

Publication first-page preview