E-mail

In the fall of 2000, vision and camera experts established the Camera Link standard to address the needs of high-speed digital cameras. This standard defines low-voltage differential signaling (LVDS) communications that transmit serial pixel data from a camera to a computer. Recent changes to the Camera Link spec include a method that lets manufacturers provide power to a camera through a Camera Link cable. Sprucing up the Camera Link spec may seem like a good idea, but the vision industry should also concentrate on developing a new approach to camera-to-PC communications.

This development kit from Avnet includes a camera module, an Analog Devices’ Blackfin digital signal processing (DSP) board, and software tools. Designers use this type of kit to develop and test smart-camera algorithms.  Courtesy of Avnet.

I see several paths that designers of cameras and vision equipment can consider.

First, look to a new serial-data standard. One possibility is USB 3.0, which Intel demonstrated about nine months ago. Reports say that Intel demonstrated USB 3.0 with a fiber-optic connection operating at 5 Gbps, but Intel has released no hard information about the final physical interface. Could USB 3.0 use a regular USB connector and an electrical bus? Who knows? At this point, implementation of USB 3.0 in vision equipment seems far off.

Second, consider whether recent changes to FireWire make the bus appropriate for your application. The IEEE 1394c-2006 standard, which codifies the FireWire S800T link, should lead to equipment that operates at 800 Mbps over inexpensive Cat-5e cable and uses standard RJ45 connectors. In late 2007, the 1394 Trade Association announced FireWire S3200, which will increase data rates to 3.2 Gbps. But in my opinion, FireWire comes on as a weak contender. It uses a local-access protocol rather than a network protocol, so it’s not great for networked cameras. And who knows when S3200-based hardware will arrive?

Third, build smarter cameras. This approach puts vision-processing and vision-analysis tasks where they belong—close to the sensor. Why stream masses of raw data through a bottleneck, er interface, only to process it in a PC and discard most of the collected pixels? Texas Instrument’s Open Multimedia Application Platform (OMAP) ICs and the Blackfin signal-processing devices from Analog Devices, for example, will help developers quickly extract useful information from raw image data. Low-power field-programmable gate arrays (FPGAs) also let engineers put custom image-analysis algorithms within a camera. These types of proven chips and associated design tools already exist. Take advantage of them.