[Blog] High-Speed Data Transfer: The Future of Embedded Vision

The demand for high-speed data transfer is growing rapidly. With advancements in smart devices, datacenter systems, and software, organizations need data to move quickly through their technology infrastructures while staying adaptable, scalable, and secure. This increase in real-time data transfer puts more strain on systems, leading to higher infrastructural needs.

From enhancing cell phone video quality and aiding autonomous vehicles in avoiding crashes to activating smart home security devices and monitoring industrial quality, the need for efficient data transfer is paramount. In this blog, we will explore the hardware, software, and interface protocol requirements essential for high-speed data transfer in common embedded vision use cases.

Key Protocols and Components for High-Speed Data Transfer

Common infrastructure components and protocols for high-speed data transfer can be categorized into short range and long distance media protocols:

• Short Range Media Protocols
  • Mobile Industry Processor Interface (MIPI): MIPI is a standardized interface for connecting peripherals and sensors to a device’s embedded processor that is common in mobile and embedded vision applications. It can also support high-speed data transmission between various sensors, processors, and displays.
  • Gigabit Multimedia Serial Link (GMSL): GMSL is a highly configurable SERDES interconnect solution used for the transmission of real-time data, control, and power signals in high-speed and high-resolution video and display applications, all over a single wire. /li>
  • Display Serial Interface (DSI): DSI is a high-speed serial interface that enables the transmission of data between a host processor and a display. Essentially, the DSI helps facilitate the communication of processed video or image data to a screen, making it an important component of many modern embedded vision applications.
  • Peripheral Component Interconnect Express (PCIe®): PCIe is a high performance, scalable, and well-defined interface standard used to connect a system’s processing unit to peripheral devices, such as cameras, sensors, and more.

• Long Distance Media Protocols
  • Ethernet Cable: Ethernet is a wired connection often leveraged in LAN (GigE Vision, Ethernet AVB / TSN, IPMX, Dante, and NDI) or WAN (SMPTE 2022, SMPTE 2110) transport.
  • Coaxial Cable: Coaxial cable (or coax cable) is an electrical cable that transmits radio frequency (RF) signals from one point to another (typically 100m or less), commonly used with CoaXPress, SDI, and HDBaseT protocols.

Each of these protocols and components plays an important role in supporting vision applications. That said, they still require complementary hardware and software to support their most sought-after capabilities, and Field-Programmable Gate Arrays (FPGAs) are often chosen for their flexibility and advanced capabilities that enable the seamless integration of these embedded vision protocols for developers to leverage in powerful ways.

Empowering Embedded Vision with Lattice FPGA Solutions

Lattice offers hardware and software solutions that integrate these protocols and components for vision applications.

Hardware Solutions

Lattice FPGA families built on the Lattice Nexus™ and Lattice Avant™ FPGA platform offer embedded flash, high I/O, and class-leading power efficiency. These FPGAs support high-speed applications through optimized connectivity features, low power consumption, and small package size.

Software Solutions

The Lattice mVision™ solution stack includes a range of tools to help embedded vision system designers quickly develop and deploy FPGA-based applications from machine vision, robotics, ADAS, video surveillance, and drones. It provides designers with:

• Reference designs and demos for everything from sensor bridging and aggregation to display interfacing and image processing to serve as examples or pre-built starting points when building embedded vision systems.
• Software tools such as Lattice Radiant™ and Lattice Diamond™, which offer powerful design capabilities to ensure ease of FPGA integration in embedded vision designs.
• IP cores such as Modular MIPI/D-PHY and USB3/GigE Vision, which can act as building blocks for scalable designs.

In embedded vision application development, bridging/queuing and channel aggregation are critical functions used to set priorities and convert from one media/format to another. In the example below, video could come in over many different physical media, and then be aggregated, prioritized, or optionally pre-processed before being packetized and sent over the network.

Below is another example, but instead of bridging over a network for transmission to a distant location, the media can go over PCIe or Ethernet into a host PC or SoC for compute.

Mastering High-Speed, Low Power Data Transfer with FPGAs

Embedded vision applications require high computing power without sacrificing efficiency or data quality. Leveraging a Lattice FPGA-based system that supports critical components and protocols can enable developers to create high-speed and low power systems.

Several examples of FPGA-based embedded vision applications were showcased at the most recent Lattice Developers Conference including:

Enabling Edge AI with GSML to Holoscan Sensor Bridge: Using the Lattice Holoscan Sensor Bridge with NVIDIA®, developers can easily pair GMSL-enabled cameras with developer kits and displays using MIPI. This reference design enables hardware developers and system architects to use GMSL components with systems that do not have a direct MIPI interface, all while maintaining the bitrate and connectivity required for embedded and Edge use cases. Watch the full video demo here.

Bridging MIPI DSI to DisplayPort: Using Lattice FPGAs and IP cores, developers can seamlessly bridge the gap between mobile and high-definition displays. By leveraging the Lattice CertusPro™-NX, developers can transmit video data from a DSI to a DisplayPort through the MIPI protocol. This solution meets the flexible connectivity requirements of embedded vision across various industries, including applications in the Automotive sector. It helps to ensure high quality video output while demanding less processing power from components like graphics processing units (GPUs). Learn more about the details of this architecture here, or watch the full video demo.

In addition to these examples, Lattice FPGA-based systems that leverage dynamic interfaces have the potential to enable evolving robotics, AI-enhanced Industrial inspection tools, Automotive vision systems, and more. The combination of powerful hardware and adaptable connectivity interfaces enable developers to support embedded vision systems that require efficient controls and high-speed data transmission.

To learn more about how Lattice FPGAs can enable your high-speed and low power embedded vision applications, reach out to our team today.