1      Introduction

GMSL2 and GMSL3 provide high-speed video and bidirectional control communication over coaxial cabling. With Power over Coax (PoC) circuitry implemented on Connect Tech products, the same cable can also supply power to the camera. Although these technologies share many characteristics, differences in the underlying hardware and software can influence camera integration and platform selection.

This application note focuses on the JCB022 and JCB009 camera platforms, though the same selection considerations apply to other Connect Tech products available in GMSL2 and GMSL3 configurations, including Anvil-RX (ESG635).

JCB022 is intended for GMSL2 camera applications, including stereo and multi-stream configurations that benefit from the additional video pipes available with the MAX9296A deserializer.

JCB009 is intended for GMSL3 camera applications and takes advantage of the increased link bandwidth available with the MAX96792A deserializer.

The MAX96792A also supports 3 Gbps and 6 Gbps GMSL2 operation and can communicate with compatible GMSL2 serializers. This provides interoperability between generations, although support for the serial link does not, by itself, establish compatibility with a complete camera implementation. The serializer, video-stream configuration, available video pipes and software support must also be considered.

2      From GMSL2 to GMSL3

One of the primary changes introduced with GMSL3 is an increase in forward-link bandwidth from up to 6 Gbps with GMSL2 to 12 Gbps with GMSL3. Rather than doubling the signaling frequency, GMSL3 achieves the higher data rate through PAM4 (four-level pulse-amplitude modulation).

GMSL2 uses NRZ signaling, where each symbol represents one bit using two signal levels. PAM4 uses four signal levels, allowing each symbol to represent two bits. A 12 Gbps GMSL3 link therefore operates at approximately the same 6 Gbaud symbol rate, and the same 3 GHz fundamental/Nyquist frequency, as a 6 Gbps GMSL2 link. This allows GMSL3 to approximately double the transmitted data rate without requiring the channel to operate at twice the fundamental frequency.

PAM4 has smaller voltage margins between its four signaling levels than NRZ, so GMSL3 incorporates forward error correction (FEC) to improve link robustness. ADI specifies FEC as part of 12 Gbps GMSL3 operation; the associated overhead means that the full 12 Gbps line rate is not available entirely as application payload.

The pipe counts above are characteristics of the MAX9296A and MAX96792A IC’s used by these respective architectures, rather than a general limitation of every GMSL2 or GMSL3 device. The MAX9296A provides four video pipes, while the MAX96792A provides two, one associated with each GMSL input.

3      GMSL2 Compatibility on JCB009 and Anvil-RX (GMSL3 Configurations)

The MAX96792A provides backward-compatible GMSL2 operation at 3 Gbps and 6 Gbps. This allows JCB009 to establish a GMSL link with compatible GMSL2 serializers and can be useful for applications transitioning between GMSL generations.

Camera integration, however, involves several layers beyond establishing the serial link. Camera drivers frequently configure the serializer and deserializer in addition to the image sensor itself. This can include link initialization, serializer/deserializer register programming, stream IDs, data-type and virtual-channel routing, GPIOs, synchronization, and CSI-2 output configuration.

For this reason, an existing GMSL2 camera implementation may require software or configuration changes when used with the MAX96792A architecture. ADI similarly identifies configuration and register-setting differences when moving between GMSL2 and GMSL3 operations.

The difference in available video pipes is particularly relevant for cameras that produce multiple streams. The MAX9296A provides four video pipes and notes that HDR or other cameras carrying multiple data types may use more than one pipe. The MAX96792A provides two video pipes, with the total number of pipes used by connected serializers limited to two.

As a result, some GMSL2 stereo, multi-imager, HDR, or multi-stream camera implementations may not map directly to the JCB009 architecture, even where the underlying GMSL2 link can be established. These applications are generally better aligned with JCB022 and its native GMSL2 architecture.

4      Video Pipe Allocation

A video pipe is an internal path used to receive and process video data within the serializer or deserializer. Pipes are distinct from physical camera connectors, CSI-2 output ports and CSI-2 data lanes.

A common configuration uses one pipe per sensor stream. However, the actual allocation depends on how the serializer packages the camera data, including the selected transport mode and whether multiple streams share a pipe. The number of sensors or virtual channels alone does not necessarily establish the required pipe count.

The accompanying infographic illustrates stereo cameras that each transmit two image streams using two serializer pipes.

On the MAX9296A, the serialized video data received over Links A and B can be routed into four video pipes. In this example, the two image streams received over Link A are routed to Pipes X and Y, while those received over Link B are routed to Pipes Z and U.

On the MAX96792A, the two image streams received over Link A are routed to Pipes Y and Z. This allocates both available video pipes to a single stereo camera, leaving no video pipe available to process additional video data from Link B in this configuration.

These examples show allocation within one deserializer, rather than the complete camera platform. JCB022 and JCB009 each contain four dual-input deserializers, so pipe availability must be reviewed within the relevant pair of camera inputs.

The infographic illustrates pipe usage only. Successful integration also depends on driver support, serializer configuration, synchronization, link bandwidth and the shared CSI-2 path to the host.

5      Platform Selection

When selecting a camera platform, the camera’s native GMSL generation, serializer architecture and available software support should be the primary considerations. Link bandwidth, video-pipe requirements and the number of simultaneously active streams should then be reviewed for the intended configuration.

For new designs, Connect Tech generally recommends matching the camera platform to the corresponding native SerDes architecture:

If there is any uncertainty around platform selection, Connect Tech Technical Support can help review the camera architecture and integration requirements to determine the most appropriate platform and assess compatibility for a specific camera configuration.

6      Contact Information