Vitis HLS: Constant False Alarm Rate detector implementation

The Constant False Alarm Rate (CFAR) detector is a crucial component in radar signal processing, enabling the reliable detection of targets in the presence of noise and clutter. Here I would like to introduct the implementation of CFAR detection algorithms, including Cell-Averaging (CA-CFAR) and Ordered-Statistic (OS-CFAR), using Vitis High-Level Synthesis (HLS).

What is CFAR?

CFAR detectors dynamically adapt detection thresholds based on the surrounding noise environment, ensuring a constant rate of false alarms across varying signal and noise levels. They are widely used in radar systems for:

• Target Detection: Identifying potential targets amidst noise and clutter.
• Clutter Rejection: Reducing the impact of environmental variations on detection accuracy.
• Noise Adaptation: Adjusting thresholds in real time for robust performance.

Architecture Overview :

The CFAR HLS IP core consists of the following functional blocks:

Input Handler (Read Axis):

• Processes incoming radar data via the AXI streaming interface.

Preprocessing:

• Computes the absolute value, logarithm, or square of the input signal, preparing it for thresholding.

Shift Register:

• Maintains a sliding window of signal samples, enabling the analysis of guard and reference cells around the cell under test (CUT).

CFAR Processor:

Implements the CFAR algorithm, including:

• Cell-Averaging CFAR (CA-CFAR): Calculates the detection threshold by averaging reference cells.
• Ordered-Statistic CFAR (OS-CFAR): Determines the threshold by ranking reference cells for improved performance in cluttered environments.

Scale and Compare:

• Scales the calculated threshold and compares it against the CUT to determine detections based on scaling factor set via AXI-light interface

Output Handler (Write Axis):

• Outputs detection results via the AXI streaming interface..

Performance and Resource Utilization :

CA-CFAR Implementation:

• Pipelined design with a latency of 2089 cycles.
• Efficiently balances throughput and resource usage, using 254 DSPs and 9933 LUTs for a 1024-point signal.

OS-CFAR Implementation:

• Slightly higher latency of 2205 cycles due to ranking operations.
• Provides better performance in non-uniform clutter at the cost of increased resource usage (354 DSPs and 10801 LUTs).

Interfaces :

• AXI Stream: Handles data input and output for radar signal streams.
• AXI-Lite: Provides configuration registers for controlling the CFAR processor and adjusting algorithm parameters (e.g., scaling factor).

Hardware Architecture :

The system architecture is designed to integrate programmable logic and processing system components for seamless operation:

Programmable Logic (PL):

HLS CACFAR (Cell-Averaging CFAR):

• Implements CFAR detection based on the average of reference cells.

HLS OSCFAR (Ordered-Statistic CFAR):

• Implements CFAR detection using a ranked statistic of reference cells for improved clutter handling.

AXI Stream Switches:

• Facilitates flexible data routing between the CFAR modules.

AXI DMA v7.1:

• Provides efficient data transfer between the programmable logic and the processing system.

Processing System (PS):

Powered by the Arm Cortex-A9, the processing system manages:

• Configuration and runtime control of CFAR IP cores.
• Data transfer and post-processing of detection results.

DDR Memory:

• Shared memory for input data, CFAR threshold results, and detection outcomes.

The following plot illustrates the performance of the CFAR detectors — Cell-Averaging CFAR (CACFAR) and Ordered-Statistic CFAR (OSCFAR) —for detecting targets in a radar signal with varying Signal-to-Noise Ratios (SNR) of -2, -4, -8, and -10 dB

Key Plot Elements :

Targets:

• Detected at bins 10, 256, 600, and 1018 including input samples edges.
• The peaks represent the presence of targets in the noisy signal.

SNR Levels:

• Lower SNR values result in targets that are less distinguishable from the noise, demonstrating the importance of CFAR for adapting detection thresholds.

Detection Algorithms:

• ABS (Blue Line): Represents the raw magnitude of the input signal.

CACFAR (Green Markers):

• Employs cell-averaging to set thresholds.
• Performs well in uniform noise conditions but is susceptible to masking by clutter or strong neighboring targets.

OSCFAR (Red Markers):

• Uses a ranking-based approach to handle non-uniform noise and clutter.
• Provides superior detection accuracy in challenging conditions, as evident from clear detection of all targets.

For more details about CFAR detector IP core implementation using Vitis HLS visit please my website.