In the fast-moving world of edge computing, the benefits of flexible hardware are manifold. When OEMs and system integrators can customize and repurpose existing designs, they gain major advantages in terms of development time and cost – as well as the ability to enter new markets.
Field-programmable gate arrays (FPGAs) have long delivered in this regard. They can be adapted to nearly any task, particularly parallel signal processing and, more recently, have become an attractive target for AI workloads as these flexible logic devices can implement sophisticated algorithms in hardware.
However, FPGA design has historically been a complex task that requires specialized skills. This has also made building edge devices around FPGAs challenging.
Making FPGAs accessible with SMARC modules
That complexity is being reduced with the introduction of FPGA-based embedded hardware platforms like the SECO SOM-SMARC-ZU. The SOM-SMARC-ZU is a standards-based SMARC module that brings the performance and flexibility of FPGAs to a small, easy-to-deploy form factor. Now, OEMs and system integrators have a powerful tool for delivering highly differentiated hardware without a massive engineering investment.
The SMARC form factor was created to accelerate and simplify the design of small but powerful embedded devices. It pre-integrates the most complex elements of an embedded computer onto a compact module, which is then coupled with a carrier board containing application-specific I/O hardware. This architecture greatly reduces engineering time and effort, while also enabling reuse of core IP across designs.
The SOM-SMARC-ZU is one of the first modules to deliver the capabilities of an AMD Xilinx Zynq UltraScale+ FPGA in this form factor. The popular Zynq UltraScale+ FPGA chip enables engineers to implement virtually any custom feature—including powerful AI and parallel signal processing engines—along with an embedded Arm processor that supports software development environments that embedded product engineers already know.
A modern approach to programmable hardware
The Zynq UltraScale+ represents a significant advancement in the field of programmable logic. Its multiprocessor system-on-chip (MPSoC) architecture includes two Arm processors: a Cortex-A53 application processor with up to four cores and a Cortex-R5F real-time processor with two cores. Together, these processors deliver high performance when it’s needed, low power when it’s not, and excellent responsiveness at all times.
The processors are complemented by an Arm Mali graphics processing unit (GPU), which provides high-performance graphics and video rendering. Together, the processors and GPU form a highly capable basis for edge devices. What’s more, they support popular software stacks and toolchains—meaning that system architects and engineers will have a short learning curve and high productivity.
Meanwhile, the FPGA component can be used to accelerate machine learning and AI. This task is aided by the VITIS AI platform, which provides a comprehensive development environment for creating and deploying AI applications on the Zynq UltraScale+. The platform includes a range of tools and libraries for developing and optimizing AI models, as well as support for popular frameworks such as TensorFlow and PyTorch.
Building out an edge device
To help OEMs and systems integrators accelerate design, SECO offers a breadth of services to support the SOM-SMARC-ZU. For example, we can design the carrier board or even customize the module itself – transforming it into other system-on-module (SOM) standards, custom form factors, or single-board computers (SBCs). These services allow OEMs and systems integrators to focus on software while SECO handles the hardware.
At the same, the company offers FPGA and software design services, including firmware development, operating system customization, and application software design. These services ensure a seamless and integrated experience, from the hardware layer up through the software stack.
Rethinking the value of FPGAs
With its easy-to-use design flexibility, the SOM-SMARC-ZU offers OEMs and systems integrators a unique way to implement complex multiprocessing domain edge designs without a massive engineering investment. The combination of pre-integrated hardware and comprehensive software support enables them to focus on their unique value-add instead of basic processing core design tasks.
Looking to build your own computer vision, sensor fusion, or AI-powered edge device? Contact us today to learn how we can help you deliver value-added solutions and services.
