AMD Versal RF Series: Adaptive SoCs With RF Applications

With integrated direct RF-sampling converters, AMD’s Versal RF Series Adaptive SoCs offer the highest computation in a single-chip device in the industry.

AMD Versal RF Series

Broadband Spectrum. enormous DSP computation. Just one chip.

Summary

Wideband-Spectrum, Accurate Observability

Allows for quick, accurate, and adaptable signal analysis throughout a broad visible spectrum for applications such as spectral analysis, radar, and test and measurement.

Large-scale DSP computation

Certain, frequently utilised DSP operations are implemented in hard IP blocks, which greatly lowers power consumption and programmable logic utilisation while speeding up processing time.

SWaP-Enhanced

Enables quick data processing while optimizing size, weight, and power (SWaP), which is essential for applications involving sophisticated signal processing.

Boost RF Signal Processing

Utilize a single-chip adaptive SoC with programmable logic to access the highest-resolution integrated RF sampling converters.

In a single chip that satisfies SWaP standards, the monolithic AMD Versal RF adaptive SoC offers high-performance computing for the demanding DSP workloads of the contemporary aerospace and defence as well as sophisticated test and measurement applications.

Benefits of the Product

Facilitating Large-Scale DSP Computing at the RF Edge

With the addition of 18 GHz direct RF-ADCs and RF-DACs for wideband-spectrum observability and dedicated hard IP cores in a monolithic die, the AMD Versal RF Series builds on the advantages of the Versal architecture, including NoC, AI Engines, programmable logic, and flexible DDR, to meet demanding SWaP requirements for A&D and test and measurement applications. Four GSPS FFT/iFFT, channelisers, a polyphase arbitrary resampler, and an upgraded LDPC decoder that supports DVB-S/S2x codes are among the hard IP blocks.

Important Features

Summary of the feature combinations that are available for each bundle or device.

RF-DACs and direct RF-ADCs

14-bit RF-ADCs and RF-DACs with calibration that function up to an input and output frequency of 18 GHz are examples of Versal RF Series products. The RF-ADC offers a wide-observable spectrum and supports both 8 GSPS and 32 GSPS setups, whilst the RF-DAC samples up to 16 GSPS. Integrated coarse and fine mixers and adjustable decimation/interpolation enable frequency agility. Scalable solutions are made possible by the support for a number of RF-DAC and RF-ADC combinations, including 16 RF-DACs and 16 RF-ADCs @ 8 GSPS, 8 RF-DACs and 4 RF-ADCs @ 32 GSPS, and 16 RF-DACS and 8 RF-ADCs @ 32 GSPS in compatible packages.

Signal Processing Hard IP Blocks

Compared to an AMD soft logic implementation2, implementing frequently used functions in hard IP blocks can cut dynamic power by up to 80% and take up less physical space, which lowers overall size, weight, and power (SWaP). Programmable logic (PL) and DSP engines required in a typical design may be significantly reduced by the specialised IP contained in the Versal RF Series, which is adaptable for common use cases.

The following hard IP blocks are present in many occurrences in the Versal RF Series:

FFT/iFFT

Fast spectrum analysis of the input signals is made possible by a programmable 8-pt to 4K-pt, 4 GSPS hard FFT/iFFT block. Multiple instances of the hard FFT/iFFT block can be paired with the PL if higher point size FFTs are required.

The Channeliser

By dividing the wideband input spectrum into tiny subchannels, a polyphase filter bank channeliser makes it possible to quickly identify and characterise input signals. Cascaded channeliser IP instances can improve signal characterisation accuracy and resolution. It is also possible to utilise the hard channeliser IP block as a synthesiser in reverse. Furthermore, if the channeliser function (16-point FFT/iFFT) is not utilised, channeliser IP can be applied as a general filter.

LDPC Decoder

Versal RF Series devices use an upgraded SD-FEC seen in Zynq RFSoC Gen 1 and Gen 3. This SD-FEC variation includes the DVB-S2/S2X satellite codes for space applications, although it only supports LDPC decoding. When this hard block is used, the LUTs for systems that use LDPC forward error correction are much reduced, which leads to a much lower power consumption. Note: Although they are available as soft IP cores, turbo codes and LDPC encoding are not supported by the dedicated IP.

Arbitrary Polyphase Resampler

For test and measurement applications, a hard polyphase arbitrary resampler IP block (1 ≤ R ≤ 2) is offered. General filtering is another purpose for this hard IP restriction.

System of Processing

To accommodate a variety of application requirements, the Versal RF Series has a dual-core Arm Cortex-A72 application processor and a dual-core Arm Cortex-R5F real-time processor. programs that require low latency, determinism, and real-time control are best suited for the real-time processing unit, whereas sophisticated programs backed by an operating system are best suited for the application processing unit. System boot, security, and debug are controlled by a third, independent platform management controller.

I/O & Programmable Logic

Custom computational blocks for dynamic algorithms may be created to programmable logic. The programmable logic, which is tightly connected with programmable I/O and packed with a wide range of memory components, enables customers to build potent accelerators for any application.

High-performance X5IO that supports MIPI D-PHY at up to 4.5 Gb/s and C-PHY at up to 10 Gb/s offers future adaptability, flexibility, and real-time processing.

Engines for DSP

New operations and data formats, such as complicated 18×18 operations and single and half-precision floating-point operations, are supported by enhanced DSP engines. Users may either change their designs to take use of the enhanced computing performance made possible by this new architecture, or they can continue to utilise the libraries and intellectual property that are already in place for AMD UltraScale+ device designs.

Multipurpose AI Engines

Versal RF Series adaptive SoCs include a first-generation AI engine with predictable performance that is tuned for real-time DSP calculation.

LPDDR5 and DDR5

Hard DDR memory controllers offer up to 136.5 GB/s of memory bandwidth and support DDR5-6400 and LPDDR5X-8533 memory types with new inline encryption capability.

Industries & Applications

Enhancing Electromagnetic Spectrum Operations for A&D

As the spectrum is more crowded and contested, Electromagnetic Spectrum Operations (EMSO) has grown more difficult. EMSO capabilities are addressed by the Versal RF Series in a highly optimised monolithic SWaP-efficient chip. The Versal RF Series combines the high-DSP count of the Versal Premium Series with the capabilities of a Zynq UltraScale+ RFSoC into a single Versal RF Series device.

AMD Versal RF Series Adaptive SoCs for Wideband Signal Analysis Solutions

A wide range of test and measurement applications in semiconductor ATE, aerospace and defence, benchtop instruments, and 6G are supported by the AMD Versal RF Series, which has monolithically integrated, sophisticated DSP computation and 14-bit data converters sampling 32 GSPS to 18 GHz.

For Every Developer

For developing with adaptable SoCs and FPGAs, AMD offers a top-tier software development environment that comprises tools (compilers, simulators, etc.), intellectual property, and solutions.

Developers may attain great performance per watt in this setting while cutting down on development time. With the use of these technologies, developers of all stripes may utilize AMD adaptive computing solutions, including application and algorithm engineers, embedded software developers, and traditional hardware developers.

Product Specifications

RF Data Converters

AI Engines and Hard IP Functions

Processing System Features

Programable Logic Features

Memory, Transceivers, and Interface