Everything in the industrial environment revolves around safety, productivity and durability - processes and systems are subject to numerous certifications, changes are planned and tested over a long period of time and nobody is easily dazzled by the advertised possibilities of a new technology. Especially as skepticism towards new things is part of human nature anyway. And the optimization of processes and workflows is best planned with known technology.
On the other hand, the question arises as to when a new technology such as USB3.0 is ready for industrial use. When is there no longer a risk of falling for a technological dud? Where would the automation world be today if it did not embrace new technologies?

Ready for the change

With USB 3.0, the performance of modern CMOS sensors can be fully utilized.

© IDS

Since the first cameras with USB 3.0 technology were presented around five years ago, the camera interface has had plenty of time to 'grow up'. The first test of strength with the tried and tested digital camera interfaces in industrial image processing left a clear picture that can be described in a few words: USB3.0 is fast, suitable for industrial use, easy to handle and can be used across the board; initial teething troubles have been eliminated. This means that a mature technology is now available for future-oriented image processing. A complete system changeover will not take place overnight, but will follow one step after another.

There are a number of good reasons for switching to USB3.0: As USB2.0 is approaching the end of its product life cycle, it is becoming increasingly difficult to obtain hardware and support. In 2012, Intel introduced the 'Ivy Bridge' processors. This made USB 3.0 a mainstream interface. Current chipsets such as Intel's 'Sunrise Point' (Skylake architecture) mark another milestone. The USB 3.0 controller interface (xHCI) combines all USB speed classes for the first time: Lowspeed (1.5 Mbit/s), Fullspeed (12 Mbit/s), Highspeed (480 Mbit/s) and Superspeed (5 Gbit/s). As a consequence, USB 2.0 high-speed controllers (eHCI) are no longer installed. The consumer market is thus inevitably driving the new mainstream technology into the industrial environment.

Image processors are hesitant

However, USB 3.0 is still viewed with caution as an interface for industrial cameras - even though it is well known that the technology not only has a significantly higher data throughput. By limiting the maximum sensor pixel clock, the sensor performance can be adapted to the bandwidth of the interface used. If the frame rate drops too much, operation with the corresponding interface no longer makes sense. Camera manufacturers must therefore decide which interface to combine new sensors with - taking into account what the market requires, of course.

A comparison of modern high-performance sensors: USB 2.0 and GigE only work with reduced sensor bandwidth or are not used at all.

© IDS

A camera application determines the camera system and not vice versa. This primarily refers to the image sensor, whose capabilities must match the application. The necessary data interface is then automatically determined by the data transfer requirements. Modern sensors therefore do not find their way into cameras with an 'old' interface such as USB 2.0.

A direct comparison of two cameras with identical megapixel sensors (for example the 'e2v EV76C560' from IDS), but with different interfaces, shows clear differences in the maximum data rate. By changing the interface to a USB 3.0 camera, the sensor can show what it is really capable of: by using the maximum pixel clock, the sensor bandwidth increases to 86 MP/s. USB 3.0 therefore allows the sensor's maximum frame rate (60 fps) to be transferred and used unthrottled.

Thanks to the asynchronous communication of USB3.0, in contrast to the tried and tested 'polling' under USB2.0, the host controller does not have to constantly ask for new data. This is reflected, among other things, in the significantly lower CPU utilization of the host PC. In a test with the 'e2v' sensor, the CPU load of the USB 2.0 camera was three times higher than that of the USB 3.0 variant, with identical camera parameters.

The quality of the system parts

PCIe x4 Rev.1 cards can be used to transfer parallel data streams from four USB cameras, each with maximum USB 3.0 bandwidth, via four separate USB host controllers.

© IDS

USB 3.0 multiplies the provision of image data. But data transmission in the GHz range also requires tested and reliable quality of all associated system components. USB 3.0 strives for a data speed that must be supported by all components in the transmission chain. This means that due to the huge leap in performance of USB technology, all system components must be of consistently high quality.

Take the cable, for example: classic copper-based 'passive' cables are bound by basic physical rules that do not make it easy to meet all mechanical and electrical requirements. A high-frequency data signal such as USB 3.0 (5 GHz) is weakened by the physical cable resistance - the longer the copper cable and the smaller the wire diameter, the greater the weakening. We are talking here about insertion loss. The reduction in cable resistance can be 'bought' by increasing the cable cross-section, i.e. the thicker a data copper cable is within a USB 3.0 cable, the better the transmission properties in the high-frequency range. Conversely, however, this increases the overall thickness of the USB 3.0 cable, which makes it less flexible and more difficult to install the connector.
In short, a lot can go wrong with USB 3.0 cable production.

In particular, cheap and simply produced cables for the consumer market often do not have the necessary processing quality to meet the requirements in an industrial environment. This leads to an increase in USB transmission errors or connection interruptions. The result is a reduction in transmission performance or an unstable connection. Reduced transmission quality can manifest itself in a camera only reporting as a USB 2.0 high-speed device when establishing a connection with the camera driver, despite USB 3.0 hardware.

The cable challenge

Another frequently discussed topic is the cable length. The USB 3.0 specification does not specify a maximum cable length. Instead, it refers to the relationship or compromise between the cable length and relevant high-frequency properties (insertion loss) or the voltage drop. If you manage to keep these factors to a minimum with the cable design, you can also produce longer, functioning cables with a length of 5m or 8m. Active fiber optic cables also transmit USB 3.0 data over even greater distances at a consistently high data rate.

Possible data throughput in the USB 3.0 multi-camera system for direct and distributed use.

© IDS

For example, ready-made USB fiber optic cables (Active Optical Cable, AOC) are available up to 50 m in length. If a longer cable length is required, they can simply be replaced with a passive cable. The electronics required for signal conversion and amplification are integrated directly into the connectors. Despite this, the connectors are only slightly longer and wider than their passive counterparts.

In addition to the fiber optic cable, there is also a copper-based power cable, which gives rise to the term 'hybrid' cable. This can also be used to supply a camera directly with power through the host PC via an AOC cable.

But: The devil is in the detail, as we all know. Therefore, camera manufacturers should take a closer look even at the expressly awarded super-speed USB ports. Front ports inside the PC are connected to the mainboard via cable bridges, which means that the same rules apply as for cable connections between the PC and camera. Nevertheless, cable assemblies with poor shielding at the plug connections or with exposed cable strands are often found here. In addition, several front ports are often connected via the same USB controller. As a result, they share the maximum possible USB 3.0 bandwidth.

The back ports of a PC are permanently soldered to the mainboard. Cable or connector-related difficulties are not to be expected here. However, experience shows that their properties vary greatly depending on the variety of mainboards and operating systems used. A general suitability for high-performance USB 3.0 transfers should therefore not be expected here either. The chipset drivers also contribute to this: As these are now responsible for far more hardware components than just the USB controllers, they have to be maintained and expanded for each new processor generation. Drivers should remain maintainable or manageable in order to avoid systematic software errors. It is therefore obvious that older operating systems are not supplied with updates for new hardware indefinitely.

One solution is to use USB 3.0 PCI Express plug-in cards. With these, the characteristics of the connection can be precisely determined and the specific requirements for the interface can be added. These include, for example, identical USB 3.0 hardware for all systems, stable driver software, a sufficient power supply directly via the PC power supply unit and, last but not least, maximum data bandwidth thanks to a separate USB controller for each USB port.

Knowledge and experience of USB 3.0 technology has grown considerably in the meantime. This is reflected in the further development of cameras and accessories. And when deciding for or against a new camera interface, one thing must not be forgotten: The next evolutionary stage in the form of USB 3.1 with 10 Gbit/s and other innovations is already just around the corner.

Author:
Heiko Seitz is Technical Editor at IDS Imaging in Obersulm.