Accelerate Video Development on FPGAs with Video Kits

With the advent of next-generation video compression standards, the industry has shifted from basic video processing to more complex integrated processing solutions, which has led to system requirements that exceed the video performance of independent DSPs. FPGAs provide more than 20GMACs of DSP performance at a price of less than $30, thus filling this gap for cost-sensitive military, automotive, medical, consumer, industrial and security applications. Only FPGAs can provide logic, embedded processing, OS support and drivers for a complete end-to-end video solution.

What prevents developers from using FPGAs for video applications is not their lack of understanding of the performance advantages of FPGAs, but their lack of experience in using their design flow, especially for traditional DSP programmers who are accustomed to programming in C language.

Developers can take advantage of the performance of FPGAs by using the flexibility of configuring a hardware architecture optimized for a specific application. This flexibility adds freedom to the development process, but also promotes its complexity.

The XtremeDSP Video Starter Kit (VSK) provides a complete and easy-to-use design environment. This development kit includes application examples and fully supports standard tool flows, which helps accelerate the design process while still enabling differentiation of the final product.

Develop video applications using the basic platform

An embedded system called a base platform provides a framework from which you can develop video applications using the VSK. The base platform is an embedded system created using the Base System Builder (BSB) of Xilinx Platform Studio and includes a MicroBlaze embedded processor.

This framework can provide a starting point for new designs or a convenient migration path for existing applications developed on processor-based systems. On the MicroBlaze processor, any C code can be easily recompiled for an external processor; once a high-performance video chain is connected, it can be migrated from software to the FPGA fabric.

To assist in this migration, VSK includes a library of custom peripheral IPs that can be easily added to the base system using Platform Studio. Custom peripherals can also connect to the video interface, manage data frames, and perform memory access and basic video processing. These custom peripherals include:

• DVI input
• DVI output
• Camera
• Video Frame Buffer Controller (VFBC)
• Video Processing Pipeline

This VFBC is ideal for video applications that require hardware control of two-dimensional data for real-time operation.

Jump-start your development process with VSK reference designs

VSK provides three reference designs to jump-start the development of video applications running on FPGAs. Each reference design is built on a base platform and uses custom peripherals from VSK's IP library. Table 1 lists the reference designs and the video processing and connectivity functions they demonstrate. These reference designs are intended to provide a starting point from which further development can be built. Figure 1 shows how the DVI pass-through port reference design can be connected to a base system.

Figure 1 Basic system with video pipeline

Table 1 VSK reference design overview

[page]

Creating Video Applications with Model-Based Design

Accelerating video applications on FPGAs requires porting performance-critical operations from software running on a processor to hardware. VSK supports a variety of hardware design flows, including those using VHDL/Verilog that leverage a solid hardware design background, and those that require little or no hardware design experience using more abstract modeling environments including C, MATLAB, and Simulink.

The MathWorks Simulink is a model-based design environment that can be used to develop algorithmic models of video systems. The MathWorks provides an optional video and imaging blockset for Simulink, which includes a rich set of video building blocks that can be used to easily process streaming video and display the results at each step in the model.

You can start by building an abstraction for the video processing algorithm itself using floating-point data types and high-level video and imaging blocks, and then optimize the algorithm in a way that the designer believes balances complexity, system cost, and performance.