Summary
Gigabit Multimedia Serial Link ™ (GMSL ™ ) and Gigabit Ethernet (GigE) are two popular link technologies in camera applications, commonly found in different end markets. This article provides a comparative analysis of the system architecture, key features, and limitations of the two technologies. This will help explain the basic principles of the two technologies and provide insight into why GMSL cameras are a compelling alternative to GigE Vision ® cameras.
Background knowledge
GigE Vision is a network camera interface standard based on Ethernet infrastructure and protocols. It is widely used in the industrial field. ADI's GMSL is a point-to-point serial link technology specifically designed for video data transmission, originally designed for automotive camera and display applications.
Both technologies are designed to extend the distance that image sensor video data can be transmitted, but each solution has its own characteristics. Over the years, we have seen an increasing adoption of GMSL cameras in areas outside of automotive, often as an alternative to GigE Vision cameras.
Typical system architecture
Image sensor connection
The signal chain of a GigE Vision camera (shown in Figure 1) typically consists of three main elements: image sensor, processor, and Ethernet PHY. The processor converts the raw image data from the image sensor into Ethernet frames, a process that typically involves image processing and compression or frame buffering to fit the data rate into the bandwidth supported by Ethernet.
Figure 1. Key signal chain elements on the sensor side of a GigE Vision camera
The signal chain for GMSL cameras (shown in Figure 2) is typically much simpler, consisting of just the image sensor and a serializer. In a typical application, the serializer converts the raw data from the image sensor and then sends it over the link in its native format. These cameras do not require a processor, making their design simpler and more suitable for applications that require a small camera size and low power consumption.
Figure 2. Key signal chain elements on the sensor side of a GMSL camera.
Host processor connection
GigE Vision cameras are widely recognized in the industry for their compatibility with a wide range of host devices. Gigabit Ethernet ports are almost standard on personal computers (PCs) or embedded platforms. Some GigE Vision cameras can use universal drivers, providing a true plug-and-play experience.
GMSL cameras require a deserializer on the host side . In most use cases, the host device is a custom embedded platform with one or more deserializers, like: eyecloud.ai's ECFG product .
The deserializer transmits the image data in the raw format of the image sensor's MIPI output through its MIPI transmitter. For such cameras, each custom camera design requires a matching driver, just like any other MIPI camera. However, if the driver for the image sensor already exists, a pair of SerDes only requires a few preset registers or register write operations to transfer the video stream from the camera to the SoC.
When only one camera is used, GigE Vision may have some advantages over GMSL in terms of system complexity, as it can be directly connected to a PC or embedded platform with an Ethernet port. However, when multiple GigE cameras are used, an Ethernet switch is required. It can be a dedicated Ethernet switch device, a network interface card (NIC) with multiple Ethernet ports , or an Ethernet switching IC between multiple Ethernet ports and the SoC. In some cases, this will result in a reduction in the maximum total data rate, or worse, it will introduce unpredictable delays, depending on the interface between the camera and the end device. See Figure 5.
Figure 5. Typical GigE Vision Network
In a GMSL camera system, one deserializer can connect up to four links with a MIPI C-PHY or D-PHY transmitter supporting the aggregate bandwidth of all four cameras. As long as the SoC can handle the aggregated data rate, using one or more GMSL devices does not affect bandwidth or add too much system complexity.
Figure 6. Typical GMSL camera to host connection
Feature Comparison
Sensor interface
GMSL serializers only support parallel LVDS (GMSL1) and MIPI (GMSL2/GMSL3) sensor interfaces. MIPI is a widely used image sensor interface for consumer electronics and automotive cameras, so GMSL cameras can support a wide variety of image sensors. However, due to the use of processors inside GigE Vision cameras, they are more flexible in terms of sensor interfaces.
Video Specifications
Working principle
Figure 7 shows an example of a timing diagram for the transfer of data from an image sensor to a GMSL link or GigE network in a continuous video stream.
Figure 7. Video transmission timing diagram
In each frame of the video stream, the image sensor emits data immediately after the exposure period and then enters an idle state before the next frame begins. The example diagram better illustrates the situation for a global shutter sensor. For a rolling shutter sensor, the exposure and readout are controlled individually for each row, so the exposure and readout periods on the frame level will overlap.
The GMSL serializer on the sensor side serializes the data from the image sensor and then immediately transmits the data to the link via its proprietary protocol.
The processor in a GigE Vision camera buffers and typically processes the data from the image sensor before arranging the video data in Ethernet frames and sending it to the network.
Link rate
The link rate specifies the theoretical maximum speed at which data can be transferred over the link. Link rate is often the key metric when comparing different data link technologies. GMSL2, GMSL3, and GigE Vision all use discrete, fixed link rates.
GMSL2 supports data rates of 3 Gbps and 6 Gbps. GMSL3 supports data rates of 12 Gbps, and all GMSL3 devices are backward compatible with GMSL2 devices using the GMSL2 protocol.
GigE Vision follows the Ethernet standard. GigE, 2.5 GigE, 5 GigE, and 10 GigE Vision cameras are often seen in common applications. As the names suggest, they support link rates from 1 Gbps up to 10 Gbps, respectively. Advanced GigE Vision cameras will support 100 GigE at 100 Gbps link rates. 1 With GigE Vision, all high-speed protocols will backwards support lower-speed protocols.
Although link rate is closely related to video resolution, frame rate, and latency, it is difficult to make a direct comparison between the two technologies based on link rate alone.
Effective video data rate
In data communications, the effective data rate describes the data rate capacity excluding protocol overhead. This concept also applies to video data communications. Typically, the amount of effective video data transmitted in a packet or frame is: pixel bit depth × number of pixels. Figure 8 illustrates the relationship between effective video data and overhead.
Figure 8. Payload and overhead in a data frame/packet
GMSL transmits video data in the form of packets. GMSL2 and GMSL3 devices use fixed packet sizes, so the effective video data rate is also well defined. Take a GMSL2 device as an example. When the link is set to 6 Gbps, it is recommended to use no more than 5.2 Gbps of video bandwidth. However, because the link also carries some overhead and blanking time from the sensor MIPI interface, the 5.2 Gbps reflects the aggregate data rate of all input MIPI data lanes, not 5.2 Gb per second of video data.
Ethernet transmits data in frames . GigE Vision does not have a standard frame size, and it is often used as a trade-off in software solutions to increase efficiency ( advantage of long frames ) or reduce latency ( advantage of short frames ). For these cameras, the overhead is typically no more than 5%. Higher speed Ethernet reduces the risk of using long frames to achieve a better effective video data rate.
Both technologies transmit data in bursts. Therefore, the average data rate over a longer period (one video frame or longer) may even be lower than the effective video data rate during the transmission. For GMSL cameras, the burst time depends only on the readout time of the image sensor, and in practical applications the burst ratio may reach 100% to support the full effective video data rate. GigE Vision cameras may be used in more complex and unpredictable network environments, in which case the burst ratio is typically lower to avoid data collisions. See Figure 9 for an example.
Figure 9. Data traffic of GMSL and GigE Vision networks
Resolution and frame rate
Resolution and frame rate are two critical specifications for cameras and they are the key drivers for increasing link rates. For these specifications, both technologies have their own advantages and disadvantages.
GMSL devices do not provide frame buffering and processing. Resolution and frame rate all depend on what the image sensor or sensor-side ISP can support within the link bandwidth, which is usually a simple trade-off between resolution, frame rate, and pixel bit depth.
GigE Vision has a more complex model. Although its available link rate is slower than GMSL in many cases, it can take advantage of additional buffering and compression to support higher resolutions and/or higher frame rates. However, this comes at the expense of increased latency and power consumption, and requires expensive components on both sides of the camera system. In some less common use cases, such cameras also transmit raw image data at lower frame rates.
Delay
Latency is another critical specification for cameras, especially in applications where data is processed and decisions are made in real time.
From the serializer input/sensor output to the deserializer output/receiver SoC input.
due to internal processing and more complex network traffic . However, these latencies do not always result in higher system-level latencies, especially when camera- side processing is part of the system's image pipeline and is more dedicated and efficient.
Other Features
Transmission distance
The GMSL serializer and deserializer are designed to transmit data up to 15 meters over coaxial cable in a passenger car. However, the transmission distance is not limited to 15 meters as long as the camera hardware system meets the GMSL channel specification .
Using the Ethernet protocol, GigE Vision can transmit data up to 100 meters using copper cables, and even further using fiber optics, though it may lose some features, such as Power over Ethernet (PoE).
PoC and PoE/ PoDL
Both technologies are able to transmit power and data over the same cable. GMSL uses Power over Coax (PoC), GigE Vision uses PoE for 4-pair Ethernet and Power over Data Lines ( PoDL ) for Single Pair Ethernet (SPE ). Most GigE Vision cameras use traditional 4-pair wiring and PoE.
PoC is simple and is often the default for camera applications that use a coax configuration. In this configuration, power and data on the link come from a single wire, and the PoC circuit requires only a few passive components.
PoE circuits that support data rates of 1 Gbps or higher require dedicated circuitry, with active components on both the camera and host (or switch) sides. This makes PoE functionality more expensive and less readily available. GigE Vision cameras that support PoE also typically have a local external powering option.
Peripheral control and system connectivity
GMSL is a dedicated camera or display link and is not designed to support a wide variety of peripherals. In a typical GMSL camera application, the link transmits control signals (UART, I2C , and SPI) and communicates only with camera peripherals such as temperature sensors, ambient light sensors, IMUs, LED controllers, etc. Larger systems that use GMSL as a camera interface typically have other lower-speed interfaces such as CAN and Ethernet to communicate with other devices.
GigE Vision cameras generally use their built-in processors to handle camera peripheral control. As a popular connection solution in industrial applications, Industrial Ethernet has a variety of standard protocols to support a variety of machines and equipment. GigE Vision cameras are directly connected to the network through its software and hardware interfaces.
Table 1. Comparison of main features of GMSL and GigE Vision
| GMSL | GigE Vision |
Topology | peer to peer | Point-to-point or via network switch |
Data link rate (Gbps) | 3/6/12, dedicated | 1/2.5/5/10, shared |
Sensor interface from PHY | Yes, MIPI D-PHY/C-PHY | no |
control signal | real time | When the network is idle |
Video Compression | no | yes |
Video Delay | Low and deterministic | High (video processing), uncertain (network conditions) |
Camera Trigger | Bidirectional direct link, µS level latency | Trigger pin (additional hardware), Ethernet packets (latency is not deterministic) |
size | 5 mm × 5 mm (GMSL2 serializer) 4 | ≥5 mm × 5 mm (GigE PHY) 5 excluding processor |
Power consumption | 260 mW (GMSL2 serializer) 4 | > 300 mW (GigE PHY) 6 excluding processor |
Plug and Play | No, MIPI driver required | yes |
Cable power supply | Simple, passive network | Complex, active components |
Standard network synchronization protocol | no | yes |
Transmission distance | ≤15 m (GMSL2, 6 Gbps) *Assuming aging, 105°C LEONI Dacar 302 coaxial cable (–1.1 dB/m) | ≤100 m |
Camera triggering and timestamping
Both the forward and reverse channels of the GMSL link support low-latency GPIO and I2C signal tunneling in microseconds , enabling different camera trigger/synchronization configurations. The trigger signal source in a GMSL camera system can come from the SoC on the deserializer side or from one of the image sensors on the serializer side.
GigE Vision cameras typically offer both hardware and software triggering options, either through dedicated pins/ports or Ethernet trigger/sync packets. In typical applications, hardware triggering is used as the standard method for responsive and accurate synchronization with other cameras or non-camera devices. The main problem with software triggering for these cameras is network latency.
Although there are protocols available to improve synchronization accuracy, they are either not accurate enough (Network Time Protocol (NTP), which synchronizes to the millisecond level2 ) or are not cost-effective (Precision Time Protocol (PTP), which synchronizes to the microsecond level3 , but requires compatible hardware).
When using a synchronization protocol on Ethernet, all devices from the same network (including GigE Vision cameras) will be able to provide timestamps in the same clock domain.
GMSL does not have timestamping capabilities. Some image sensors can provide timestamping via the MIPI embedded header, but this is not usually relevant to other devices on the higher level system. In some system architectures, the GMSL deserializer is connected to the SoC on a PTP network to use a centralized clock. If this functionality is required, use the AD-GMSL2ETH-SL as a reference.
Conclusion
In summary (see Table 1), GMSL is a strong alternative to existing GigE Vision solutions. Compared to GigE Vision cameras, GMSL cameras typically provide equivalent or better link rates and features at lower cost, lower power consumption, simpler system architecture, and smaller system size. In addition, because GMSL was originally designed for automotive applications, it has been proven by automotive engineers for decades in harsh environments. GMSL will provide engineers and system architects with confidence in the development of systems where reliability and functional safety are critical.
References
Kainan Wang. Gigabit Multimedia Serial Link (GMSL) Cameras as an Alternative to GigE Vision Cameras . ADI Analog Dialogue. 2023-12.
Eyecloud.ai is the leading supplier of AI vision appliances and systems, aiming to support tech companies in overcoming the development and production challenges of Edge AI vision products with expertise in advanced hardware design and production, camera and machine vision systems development, image sensor and ISP tuning, and IoT device management. Eyecloud.ai has successfully developed mass-production machine vision solutions for customers around the globe in autonomous driving, electrical vehicles, mobility robots, and surveillance markets. Eyecloud.ai offers engineering services to enable customization and meet unique application requirements in rapid and cost-effective manner. Founded in 2018, Eyecloud.ai has received several industry awards for its insight and innovations in AI vision product deployment.
Kommentare