AntSDR T510 AI is a new integrated FPGA-based RF platform built to connect sensing, processing, understanding, and decision-making. If you are looking for a wireless platform that brings together the real-time signal processing capability of an RFSoC and the intelligent computing power of a GPU, T510 AI is unlike any other option on the market.

T510 AI tightly integrates RFSoC and GPU resources to create a high-performance edge platform for next-generation intelligent wireless applications. The RFSoC is optimized for high-speed, low-latency, deterministic RF data paths, including acquisition, buffering, digital front-end processing, and multi-channel synchronization. The GPU provides massive parallel computing capability for spectrum analysis, signal classification, wireless environment perception, and AI inference. Through this heterogeneous architecture, T510 AI goes beyond RF transmit and receive: it enables an integrated workflow from wireless sensing and local understanding to rapid response.

Heterogeneous Computing That Goes Beyond RF Processing

By pairing a powerful FPGA with a GPU, T510 AI unlocks new wireless processing possibilities. With massive MIMO onboard, next-generation wireless prototyping is within your grasp, along with phased-array radar and many other scenarios demanding the highest performance. Test, measure, and dial in your next wireless designs with one powerful device.

T510 AI is based on the AMD Zynq UltraScale+ RFSoC ZU47DR. This single-chip architecture deeply integrates high-speed RF data converters, programmable logic, and embedded processing. The platform covers a 1 MHz to 6 GHz RF range and integrates 14-bit ADCs up to 5 GSPS and 14-bit DACs up to 9.85 GSPS, supporting an 8T8R architecture. It supports synchronization across 8 ADC channels and 8 DAC channels, as well as multi-board synchronization for expansion to 16, 32, or more channels. The platform also supports digital up/down conversion, interpolation, decimation, and other digital front-end functions, with up to 2 GHz baseband bandwidth per channel.

T510 AI also integrates a high-performance GPU module, providing large-scale parallel computing and AI acceleration. The GPU is well suited for deep-learning inference, matrix operations, and high-throughput data processing, enabling efficient signal understanding, feature extraction, and acceleration of complex algorithms. In the T510 AI architecture, the RFSoC handles deterministic, low-latency RF and baseband processing, while the GPU performs dense intelligent computing. Together, they form a tightly coupled heterogeneous computing platform that efficiently bridges real-time signal acquisition and intelligent processing.

Features and Specifications

• XCZU47DR RFSoC for high-performance RF processing
• Direct RF input coverage from 1 MHz to 6 GHz
• 8-channel 14-bit ADCs with sampling rates up to 5 GSPS
• 8-channel 14-bit DACs with sampling rates up to 9.85 GSPS
• 8T8R architecture with multi-channel ADC/DAC synchronization
• Digital mixing, DDC/DUC, interpolation, and decimation with up to 2 GHz baseband bandwidth
• NVIDIA Jetson for real-time AI signal processing and intelligent spectrum analysis

Time Synchronization

• Clock reference options:

  • External clock input
  • Onboard OCXO/TCXO
  • GPS clock

• Supports external PPS trigger input

• Supports clock output

• Clock system supports multi-board synchronization

Connectivity

• 100 G QSFP28 high-speed optical interface (RFSoC)
• SSMC RF interfaces for multi-channel RF input/output
• Gigabit Ethernet (RJ45) for system control and data
• USB 3.0 Type-A for peripherals
• USB Type-C for UART and JTAG debugging
• NVIDIA Jetson interfaces including USB, HDMI, Ethernet, and M.2 SSD expansion

Storage

• PS 4 GB DDR4
• PL 2 GB DDR4
• 32GB Onboard eMMC
• QSPI Flash: two 64 MB chips, 128 MB total
• SD card slot for storage expansion
• NVIDIA Jetson M.2

Physical

• Dimensions: 205 × 160 × 45 mm (8.07 x 6.29 x 1.77 in)
• Operating temperature: 0 to 55 °C
• Power input: 12 V DC (6-pin ATX connector)
• Typical power consumption: 45 W (8-channel full operation + Jetson NX @15W)

System Block Diagram

Software

• NVIDIA Jetson platform with Ubuntu 22.04:

  • Ubuntu 22.04 and CUDA are pre-configured and validated for AI, RF, and signal-processing development

• WaveSight multi-channel RF viewer:

  • Visualize, monitor, record, and replay up to eight synchronized RF channels in real time from a host PC or the integrated Jetson platform

• GNU Radio and SoapySDR compatibility:

  • Quickly migrate existing GNU Radio workflows with native SoapySDR support and custom GNU Radio integration

• Open IQTAXI driver framework:

  • Open-source drivers and APIs for RFSoC control and IQ data stream acquisition.

• Flexible development environment:

  • Build applications in Python or C++, use GPU acceleration, and access RF hardware through high-level APIs without FPGA or HDL development

Example Applications

• SignalLab AI demonstration:

  • A GPU-accelerated demo for real-time Wi-Fi, Bluetooth, and modulation classification using RF data captured by the platform

Open Source

T510 AI is designed to be a developer-accessible platform, and we plan to publish the key hardware and software resources needed to study, customize, and build on the system. All related files will be available in the MicroPhase/T510-AI GitHub repository. The public repository will include hardware reference materials, RFSoC firmware sources, host-side streaming examples, bare-metal demos, and Jetson-based AI workflow examples.

Planned open-source resources include:

Hardware reference files

• A public, abridged version of the T510 AI schematics for system-level review and integration
• DXF mechanical files to support enclosure design, mounting, and system integration

RFSoC Linux firmware source

• Source code and build resources for creating T510 AI Linux firmware without the NVIDIA Jetson software stack

High-speed host data-transfer examples

• A 100G Ethernet example for high-throughput IQ data transfer between the RFSoC and a host system
• An Aurora-based example using an FPGA front-end card to transfer data between the T510 AI RFSoC and a host system

Bare-metal examples

• Low-level demos for RFSoC bring-up, peripheral control, data-path validation, and board-level testing

NVIDIA Jetson integration examples

• Example projects showing how to combine RF acquisition on the RFSoC with GPU-accelerated processing, AI inference, and local application logic on the Jetson module

Interested in This Project?

You can sign up at the top of this page to be notified when the campaign launches and to receive other updates. We only send out relevant content, and you can unsubscribe at any time.