Access Official Technical PDF

Share Summary:LinkedInEmail

Short-range radar systems must often handle two different sensing needs: detecting objects at a distance and resolving nearby targets with greater detail. The AWR1642 reference design TIDEP-0092 demonstrates one way to address both requirements in a single FMCW radar application. It combines a short-range radar (SRR) mode with an ultra-short-range radar (USRR) mode, processing detections to estimate object position and relative velocity and, in the SRR mode, track selected targets.

The design is intended as a starting point for automotive applications such as lane change assist, blind spot detection, cross-traffic alert and autonomous parking. Its architecture also offers useful lessons for engineers evaluating radar sensing, embedded processing and the trade-offs involved in configuring a multi-mode system.

Why use FMCW radar for short-range sensing?

Frequency-modulated continuous-wave radar transmits signals known as chirps and analyzes their reflections. From those returns, the system can estimate target range and relative velocity. Radar can operate in darkness and is comparatively resistant to environmental conditions such as rain, dust and smoke. Unlike camera-based sensing, it does not depend on visible light; compared with ultrasound, radar can support longer-range sensing.

These characteristics make radar relevant wherever a vehicle needs to detect nearby obstacles or monitor traffic around it. The TIDEP-0092 design uses the AWR1642, a single-chip FMCW sensor operating in the 76–81 GHz band, with four receive channels and two transmit channels. The device also integrates processing resources used by the application, including an ARM Cortex-R4F processor and a C674x DSP.

How the two radar modes complement each other

The reference design creates two subframes with different sensing priorities. The SRR subframe targets detection out to approximately 80 m, while the USRR subframe provides more detailed sensing over an effective range of approximately 20 m. In practical terms, the system can monitor farther objects while also producing a richer point cloud for closer targets.

The modes make different compromises between range, resolution and velocity. The published system specifications list SRR range resolution at approximately 36.6 cm and USRR range resolution at 4.3 cm. Maximum range is specified as 80 m for SRR and 20 m for USRR. The corresponding native maximum velocities are listed as 90 kph and 36 kph. These figures describe the design’s configured operating modes, not universal performance for every installation or target.

Range resolution describes the ability to distinguish targets at different distances along the same bearing. Improving that resolution generally requires a different chirp configuration, and the longer-range mode has to account for limits including available intermediate-frequency bandwidth and memory. Using separate subframes allows the design to tune each mode to its own sensing objective rather than forcing one profile to serve both.

Chirp and antenna configuration

The SRR and USRR subframes also use different transmit patterns. SRR uses one transmitter. USRR alternates chirps between the two transmitters using time-division multiplexing. Combining the transmitter and receiver measurements in the USRR mode forms an eight-element virtual receive array, compared with the four physical receive antennas. This increases angular resolution; the design describes it as approximately 14.3 degrees, against a field of view of about ±60 degrees.

For SRR, the design uses groups of fast and slow chirps. Their different repetition periods provide separate velocity estimates. When the estimates are inconsistent because a target’s velocity exceeds a chirp pattern’s native unambiguous range, signal processing can compare velocity hypotheses across the two patterns to resolve ambiguity. The reference design describes testing up to 90 kph for this enhanced SRR velocity measurement.

From radar samples to object tracks

After the radar front end captures reflected signals, the DSP processes samples through a sequence of steps. A one-dimensional FFT extracts range information from each chirp. A two-dimensional FFT then contributes velocity information, producing a range-velocity representation that can be searched for target detections.

Detection uses constant false alarm rate (CFAR) processing, followed by peak grouping and azimuth estimation to help locate objects in the X-Y plane. The application also applies additional pruning to limit effects such as ground clutter. Detected objects can then be grouped into clusters using the dBScan algorithm. For SRR tracking, the strongest object in a cluster serves as a representative measurement, reducing the number of inputs passed to the tracker and helping limit switching between nearby reflectors.

The SRR tracker uses an extended Kalman filter with position and velocity states. Its measurements include range, relative velocity and azimuth information. Track initialization and data association depend on assumptions about object movement. For example, the reference design initializes velocity under the assumption that motion is along the radar’s longitudinal axis. That assumption suits some traffic scenarios better than cross-traffic, so system integrators should treat tracking behavior as an application-specific design consideration.

Embedded processing and implementation considerations

Signal processing is only part of the engineering challenge: radar data must also move efficiently through the device’s memory. The design uses enhanced direct memory access (eDMA) to transfer data between buffers and memories while the DSP continues processing. Ping-pong buffering lets one buffer receive data while another is used for computation. This approach helps coordinate data movement with FFT, detection and angle-estimation workloads.

The two subframes are processed separately, and many scratch-buffer locations can be reused between them. This makes memory allocation an important part of the design. The reference design also separates responsibilities between the master subsystem, which configures and starts the radar, and the DSP subsystem, which processes radar signals. Because the chirp configuration is fixed in this arrangement, changing it requires corresponding software updates rather than an unrestricted live adjustment.

What the reference design demonstrates

The design was evaluated with field tests involving a small vehicle and a motorcycle. The documentation reports continuous visibility of a small car to 80 m and detection of a motorcycle up to 50 m. It also reports a highway test in which a vehicle traveling at 100 kph was tracked, while distinguishing this test observation from the stated 90 kph SRR system specification.

For engineering teams, the main value of TIDEP-0092 is its end-to-end example: configurable chirps feed signal-processing pipelines, detections become clusters, and selected clusters support tracking. Its dual-mode approach illustrates how a radar can balance long-range awareness with higher-resolution near-field sensing, while exposing practical constraints around bandwidth, velocity ambiguity, memory and motion assumptions.

As a reference design, it is a foundation for adaptation, not a substitute for application-specific validation. Teams developing automotive radar functions should assess whether the configured range and resolution meet their use case, review the assumptions built into tracking and data association, and test the complete system in the intended operating environment.

Access Official Technical PDF

About The Author