|
|
|
|
|
|
 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Killer Apps Driving Adoption of USB 3.1 Gen 2By Corigine, Inc. Abstract  Design-in of USB 3.1 Gen 2 has become much more common in semiconductor devices. This adoption is being driven by a broad range of applications, most involving large file sizes or high-resolution video streams, that demand high bandwidth. Neither earlier generations of USB nor other connectivity solutions such as wireless Ethernet or Bluetooth can meet these requirements. The 10 Gbit/second bandwidth of USB 3.1 Gen 2 satisfies all current needs, with room for future growth. This white paper discusses the alignment of applications and Gen 2, showing how alternatives fail to make the grade. 1. Introduction The electronics industry has an astounding range of technologies and standards available for inclusion in new designs, and more are introduced every day. For a team developing an advanced semiconductor device such as a system on a chip (SoC), selecting the most appropriate technologies from the array at their fingertips is challenging, yet making the right choices is often critical for eventual product success. In most cases, the alignment between product requirements and appropriate technologies is closely tied to applications. In fact, it is common to hear the phrase “killer app” to describe the application space where a given new technology or update to an existing technology is not only appropriate, but required. This establishes an interesting reciprocal relationship. A product design must include the right technologies to be successful, while at least one killer app must be found for each new technology for it to be successful. High-speed Internet access and streaming video is one oft-cited example of a technology and its killer app. One relatively new connectivity standard is currently a hot topic in the industry because multiple killer apps are converging to drive rapid adoption. Revision 3.1 Gen 2 of the standard for the Universal Serial Bus (USB) was released in 2013, but many designers were initially slow to adopt it for their end products because earlier versions sufficed. This situation has changed dramatically over the past year or so as advanced applications require the much greater performance provided by the latest USB standard. 2. Applications Driving USB Adoption When USB was named with word “Universal” included, its inventors showed considerable foresight in anticipating how ubiquitous the standard would become. It is hard to imagine any class of peripheral device that can’t be connected to multiple computing platforms with USB. Every new version of the standard has included new features and offered more bandwidth. Table 1 shows the growth in the achievable bandwidth. The second generation (Gen 2) of USB 3.1 has been out for several years. Commercial IP and chip solutions are readily available, allowing adoption of the latest USB version without having to design, verify, and certify an implementation from scratch. The current 10 Gbit/second bandwidth provides a level of performance almost unimaginable in the early days of the standard. There are several categories of killer apps that require this higher speed and are therefore driving adoption of USB 3.1 Gen 2.
Table 1: History of USB Releases Any application with large data files and challenging performance requirements will benefit from having USB with 10 Gbit/second bandwidth on all devices that must process the large amount of data. For example, storing and retrieving larger database and video files is putting pressure on hard drives and flash drives to be faster, so many are being upgraded. Traditional backup drives with older USB versions can take days to save and restore the contents of an internal hard drive, so these devices are also ideal candidates for USB 3.1 Gen 2. TransferJet is a close proximity wireless transfer technology whereby data can be exchanged between two touching (or very close together) electronic devices. Although the first version achieved only modest speeds, TransferJet X (IEEE 802.15.3e standard) was recently defined to enable data transfer speeds of 13.1 Gbit/sec and above. USB 3.1 Gen 2 will be required to support TransferJet X. Another motivation for the adoption of USB 3.1 is the new Type-C connector. It supports the high bandwidth of Gen 2, but it has additional benefits. The connector is smaller, so it can be used in smaller and thinner devices. It is reversible, so it is much easier to plug in. It also supports the USB Power Delivery (PD) specification, enabling charging of large devices such as laptops. Upgrading to the Type-C connector is an ideal time to consider replacing an older USB implementation with a 3.1 Gen 2 core. The broadest category of killer apps demanding 10 Gbit/second connectivity are those using high-quality video. On many types of computing platforms and networks, video applications place some of the highest demands in terms of memory usage, computing power, and bandwidth across various interfaces. Consumers may find that streaming video drains their smartphone batteries faster than other applications. PC, tablet, and laptop users may see their system performance slowing during video playback. Sufficient bandwidth is required whenever high-quality video images must be captured, viewed, edited, or preserved. The most obvious applications are those in which the video stream must be analyzed carefully by a human, video processing systems using artificial intelligence (AI), or both. Security scans at locations such as airports are a common example. In addition to the metal detectors and full-body scanners, checkpoints may have multiple cameras aimed at the person in question. In addition to viewing by trained security personnel, videos can be matched against image databases to identify known or potential threats. Matching technology has made major advances in recent years. In such a case, the better the video is, the more accurate the assessment. This is also true for machine learning, possibly by combining human input with the image analysis. Traditional black-and-white or low-resolution security cameras will not suffice. There is no substitute for real-time bandwidth when picture quality matters. Although a security video may be saved for later review, much analysis such as image-matching is performed on one frame at a time. For this reason, frame grabbers are a natural fit for USB 3.1 Gen 2. They are also commonly used in industrial inspection applications where images are analyzed to check that manufactured parts meet specifications. As with security cameras, machine vision and industrial imaging are most successful with high-quality images grabbed from an ultra-high-definition video stream. Medical applications also benefit from the highest possible resolution. A surgeon watching an in internal or external video feed during a procedure surely would demand high-speed image transmission. Similarly, designers of control rooms are upgrading to be able to display much more readable information on every screen. Video conferencing and other telepresence systems, set-top box (STB) video storage and retrieval, and video streaming from laptops to monitors via docking stations also benefit from better resolution and demand more speed. In any of these numerous applications, all cameras, video equipment, and monitors must be able to handle high bandwidth for maximum flexibility. It turns out that video transmission is at an interesting inflexion point. Many parts of the industry have adopted the 4K UHD ultra-high-definition video standard and many more are moving rapidly toward adoption. The next two sections calculate the bandwidth requirements for 4K UHD and show how it is another major factor driving the industry to USB 3.1 Gen2. A final point must be addressed: the length of USB cables. USB 3.1 Gen 2 cables are limited to one meter by the specification, although some tests with longer cables have been successful. One meter will suffice for many applications, but if a longer distance is required several manufacturers provide “active” extender cables that embed repeaters (one-port USB hubs) at regular intervals. Advanced transfer speeds are not needed for many types of IoT devices and wireless Ethernet may be the best solution for those cases. 3. Video Bandwidth Requirements Both professional and consumer applications are demanding more video quality, with transmission requirements for 4K UHD crossing the 5 Gbit/second limit of USB 3.1 Gen 1 and requiring Gen 2. The bandwidth required for video transmission depends heavily upon the quality desired. In general, the more pixels in each frame and the faster the frame rate, the better the quality. There is a wide range of video formats defined by various standards organizations as well as numerous proprietary and domain-specific formats. Table 2 shows several of the better-known formats and the rather confusing names used to define them.
Table 2: Examples of Standard Video Formats Only formats considered as high definition (HD) or ultra high definition (UHD) are included. Calculating the bandwidth required for a specific video format takes several steps. Since the formats shown in the table are all digital, one place to start is the raw number of bits that must be transmitted in a given period, typically one second. The math for this step is simple enough: (frame size in pixels) X (bits / pixel) X (frames / second) = (bits / second) The frame size and maximum frame rate can be obtained from the table. The number of bits per pixel is less standardized. For many years, most digital systems used 24 bits per pixel (“true color”) for both still images and video. This allows 8 bits (256 values) for each of the three colors in an RGB (red-green-blue) encoding. Some high-end graphics systems used 30 bits per pixel (10 bits for each color) and this “deep color” option has been extended to 36 and even 48 bits per pixel although these are not yet widely used. The most common high-definition format in use today, Full HD, uses a frame (refresh) rate of 30 Hz (29.97 Hz) and uses 16 bits of color information per pixel with a luma/chroma specification rather than RGB. Applying the previous formula to such a video stream calculates the raw bit rate as follows: (2,073,600 pixels) X (16 bits / pixel) X (29.97 frames / second) = (994,332,672 bits / second) Thus, just under 1 Gbit/second of bandwidth is needed to transmit the video information. Of course, there is overhead in packaging the data for transfer, adding synchronization packets, and supporting error detection and correction. Most digital video is transmitted using the HD-SDI standard, which includes the SMPTE 292 standard from the Society of Motion Picture and Television Engineers (SMPTE). Using this scheme, the total bandwidth for sending a Full HD stream over a cable is about 1.485 Gbit/second. Note that this is the bandwidth required for a raw, uncompressed video stream. When HDTV is broadcast over the air, the compression algorithms reduce the bandwidth by roughly a factor of 50, for a 19.3 Mbit/second rate. But the full raw image is required in many pre-broadcast and non-broadcast applications, including transmission from high-resolution cameras to computers and other video editing and processing equipment, and between these pieces of equipment and display monitors. The standard in the middle of the table, UHD-1 (also known as “4K UHD”) is the primary video format considered in this white paper. Many consumers are already familiar with “4K Ultra HDTV” DVDs and it likely that expectations will rise to require this higher resolution in most video applications. Some streaming video services, including Netflix and YouTube, already support 4K UHD or better. Applying the formula to UHD-1 with 10 bits (1024 values) for each luma/chroma value yields the raw bandwidth as follows: (8,294,400 pixels) X (20 bits / pixel) X (29.97 frames / second) = (4,971,663,360 bits / second) This result is about 5 times the bandwidth of the familiar Full HD that still dominates today. There is every possibility that the requirements will grow even higher in the future if deeper color or higher frame rates become more popular. But 4K UHD is clearly the emerging standard today and is likely to dominate in consumer and professional applications for some time. Applying the same ratio used for overhead with Full HD, the total bandwidth required for 4K UHD is 7.425 Gbit/second. 4. Options for 4K UHD Connectivity Knowing that it takes more than 7 Gbit/second to transmit 4K UHD video among cameras, monitors, computers, and video equipment raises the question of how to provide the connectivity with sufficient bandwidth. These days it seems that the default assumption for connecting almost anything is wireless Ethernet. Many computing devices and peripherals ship with no wired Ethernet port at all. They typically have some combination of cellular, Bluetooth, and wireless Ethernet technology. Cellular can be dismissed as an option. Its speed is well below that required for Full HD, let alone UHD. It would also be inefficient and expensive to transfer large amounts of data by uploading from one device to a cellular network and then downloading to another device. Bluetooth, while designed for device-to-device connectivity, lacks the bandwidth for video. Depending upon the version, speeds range from less than 1 Mbit/second to 50 Mbit/second. Zigbee, another device interconnect standard, tops out at 250 Kbit/second. Wireless Ethernet is the default assumption because it has become ubiquitous. It is not unusual to see homes and offices in which all connectivity is wireless beyond the router connecting to the Internet. The IEEE 802.11 standard has come a long way since its first version, with a speed of only 1 Mbit/second. The latest approved revision is 802.11ad, and devices are now available. The theoretical maximum bandwidth is nearly 7 Gbit/second, but real-world results have shown that the best achieved speeds are around 4.6 Gbit/second. This sets up a key inflexion point for the industry. Wireless Ethernet can easily handle HD video streams, but not the 5x faster requirements for 4K UHD. The next revision of the standard, 802.11ax, is being designed to exceed 10 Gbit/second in theoretical bandwidth. However, it is not expected to be released until late 2018, and commercial implementation may lag that date. The evidence is clear that another option is needed for interconnectivity of UHD-based devices. Fortunately, the latest version of the Universal Serial Bus (USB) standard is up to the task. Since its first release more than 20 years ago, the supported speed has increased by almost 2000 times. As discussed previously, USB 3.1 Gen 2 was released in 2013 and is now well established in the market. The 7.425 Gbit/second bandwidth required by 4K UHD video streams is comfortably within the 10 Gbit/second maximum transfer rate of the latest upgrade to USB. USB is also an excellent choice for UHD because an even faster version of the standard was recently released. As mentioned earlier, the demands for UHD bandwidth will likely increase over time. With a maximum rate of 20 Gbit/second, USB 3.2 will easily be able to accommodate a faster screen rate or the 8K version of the standard. Choosing USB 3.1 Gen 2 now will enable easier driver upgrades and a transition to USB 3.2 in several years when IP cores are available. In addition, USB is ideal for streaming video (or audio) because it can operate in isochronous mode. Most interconnect standards and protocols aim for completeness of all transfers; if a packet is corrupted or lost entirely it will be re-transmitted. This behavior is essential in such applications as file transfer, when every bit must be sent and saved correctly. USB supports this “bulk transfer” mode of operation, and in fact many mass storage devices from thumb drives to huge backup disks have USB interfaces. For listening to audio and watching video, transmission of the data must occur at regular intervals so that the stream is as constant as possible. In the case of corruption or loss, it is generally better to move on to the next packet rather than re-transmit. Viewers will notice a single corrupted frame much less than a pause while data is still arriving. The isochronous mode of USB was defined expressly for streams from a microphone or camera, or to a speaker or monitor. 5. The Corigine USB Solution Since USB has been around since 1996, there are many implementations available. Some companies created their own in-house designs and many others have licensed commercially available USB cores. As the standard has been upgraded, many of these implementations have evolved to add new features and support higher bandwidth. But continually tweaking an existing design is not always the best solution, since legacy architectures must carry the baggage necessary for backward compatibility with older versions. Using legacy cores has four consequences: more gates, more power, more memory, and slower drivers. Adding features to an older core requires more gates, and therefore more space on a chip, than a clean design from scratch. More gates mean higher power consumption, a major issue since some classes of USB devices may be battery powered. The amount of memory required also grows since new blocks are simply added to the previous requirements. Finally, control register sets become bloated with legacy support, slowing software drivers. The Corigine USB 3.1 Gen 2 IP core was designed specifically to support the latest versions of the USB standard, and thus carries no legacy baggage. The reduced gate count means that it consumes less power than competitive products. Further, real-world tests have shown that the amount of memory required is nearly halved, reducing the cost for the entire system in which the core is used. Designers who license this core benefit from the industry’s most memory-efficient and power- and cost-sensitive architecture at the fastest USB speed. The Corigine USB 3.1 Gen 2 PC host and device controller IP has been certified by the USB Implementers Forum (USB-IF) as compliant with USB SuperSpeed+ (10 Gbit/second). Corigine provides a complete development environment, including the controllers, verification IP, IP subsystems, IP prototyping, and software development kits. The Corigine solution includes support for USB Dual Role, a significant step up from the USB On-the-Go (OTG) specification. Dual Role makes it possible for a system to be either a USB peripheral (slave) or USB host (master), enabling more interconnection flexibility. For example, a tablet might operate as a slave when uploading photos to a laptop but as a master when a peripheral such as a keyboard or thumb drive is attached to it. USB Dual Role also allows either device to be the power source for the other. 6. Conclusion The Universal Serial Bus (USB) standard has truly become universal. It has continued to evolve, with the current 3.1 Gen 2 version providing 10 Gbit/second transfer speeds and a bevy of features. Adoption of this version has accelerated because a wide range of applications, including those using 4K UHD video, demand the highest speed possible. System designers have USB 3.1 Gen 2 IP cores available commercially now as certified solutions that deliver the most efficient footprint for power, system memory and size. 7. Further Reading
If you wish to download a copy of this white paper, click here
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|
|
Home | Feedback | Register | Site Map |
|
|
|
All material on this site Copyright © 2017 Design And Reuse S.A. All rights reserved. |
