USB 3.x - Feeling the need for speed -

USB 3.x – Feeling the need for speed


In today’s increasingly busy world, knowledge is everything and the time in which you can expect to receive such knowledge is critical. As such, data transmission in engineering terms has continued to evolve, so that large quantities of information can be transferred at a faster pace and carried significantly further than ever before. From early smoke signals (not the frying PCB kind) to modern superfast fibre optic cabling, every application has looked for new and innovative ways to improve communication capabilities. It must be agreed that there is unlikely to ever be any real change in this fundamental dynamic.

One technology that has seen relentless progress in order for it to remain relevant in modern high-speed systems is universal serial bus (USB). Originally unveiled back in the mid 1990’s, USB 1.0 was a 1.5Mbit/s slow speed serial solution that was predominantly aimed at replacing traditional PC data interfaces, such as RS232 and PS/2. The technology was robust enough to allow reasonable transfer rates over distances of up to 5m. This meant that it was suitable for a wide variety of PC peripherals – especially mice, keyboards, printers, etc.

Further improvements came a few years later, with the introduction of USB 1.1. This enabled full speed data rates of 12Mbit/s to be achieved. It was at this point that USB really started to take off and before long it would effectively become ubiquitous within the computing domain. What really helped propel the success of USB at this time was the tight adherence it had to standards. Standard connectors, and standard cable lengths would be pivotal in ensuring that it was relatively easy to swap peripherals in and out – especially as, in addition to data, the standard cables also made provision for the supplying power to peripherals. Another key aspect to the standards was the development of particular device classes, thereby permitting standard drivers to be accessible to multiple USB devices from different vendors without the cost and inconvenience associated with developing and subsequently installing numerous bespoke drivers.

By the turn of the century USB was well established and looking to grow into new markets and so another upgrade was warranted. The advent of USB 2.0 pushed the supported speed up considerably compared to the previous generations, reaching 480Mbit/s. Over time the power delivery dimension came to have greater value. This resulted in USB’s proliferation into the portable consumer electronics sector – becoming the established route for charging in digital cameras, MP3 players and smartphones. Now every airport and an increasing number of hotels will have USB-based charging points. 

For secure, lossless data transmission at USB 1.x and USB 2.0 speeds, a simple bi-directional, differential signal over twisted pair wiring has proved to be more than sufficient to ensure low noise and crosstalk. As the next chapter in the USB story begins, upholding of signal integrity needs to be accordingly dealt with via more sophisticated mechanisms.

SuperSpeed USB 3.0 (also referred to as USB 3.1 Gen1) has been responsible for an even more significant leap in performance. It offers data rates that are an order of magnitude greater – hitting the 4.8Gbit/s mark. This has opened up a whole new array of applications that USB can now comfortably target, such as faster, larger memory devices and the transfer of high definition (HD) video or imaging data. Thanks to this new level of operational throughput, USB technology has been able to maintain its relevance (while other interface technologies have become obsolete). It is now even offering a viable alternative to Ethernet, in scenarios where cable distances are short but speed and power delivery are both of importance.

Figure 1: The Speeds Corresponding to Each USB Generation (Source: FTDI)

A key factor responsible for aiding the ongoing evolution of USB is that this technology has retained backward compatibility to older legacy systems. This means that the new technology can still be adopted by applications currently unable to take full advantage of the greater speeds that USB 3.x possesses, but will be adequately future proofed for when the need arises. This more cautious approach to design and development has many engineering advantages over a complete wholesale change to some form of technology that has previously been unproven in an application area. However, that does not necessarily mean that design for USB 3.x hasn’t got its own challenges to be taken into consideration. The higher data rates enforce far tighter rules on PCB layout. Traces should be shorter to avoid timing issues, traces should be contained on one PCB layer where possible and superior grounding is required.

Also included with the new USB 3.x specification upgrades there has been a pressing need for new standardised connectors. To facilitate backward compatibility, many solutions will include both a USB 3.0 and a USB 2.0 PHY, with separate data signalling. The USB 2.0 PHY retains the original single bi-directional differential D+/D- signals, while the new faster USB 3.0 PHY requires differential signals for TX+/TX- and RX+/RX-. Creating a full duplex system removes any delays in bus turnaround times, thereby ramping up the overall system transfer speeds. Meanwhile the differential signalling mitigates the effects of noise and reduces the number of packet losses or errors. The upshot is that fast and reliable communication can be assured.

Incorporating these additional data signals as well as power has resulted in connectors such as the USB 3.0 Type B and Micro B, which essentially have pins to handle both USB 3.x and USB 2.0 separately, being developed. Using these new standard connectors not only makes it much easy to swap in/out standard cables, but it also means that the geometries of the pins and the pin spacing are of greater importance – especially for the higher speed signalling.

The long term future for USB also seems secure, with USB 3.1 Gen 2 being designed to offer even faster speeds than are currently available – 10Gbit/s already being outlined by the USB Implementers Forum (USB-IF). This extension to the specification will undoubtedly continue to consolidate advancements into larger memory devices and higher definition video/imaging tasks, as well as presenting further application areas for engineers to explore.

When selecting a suitable USB component, whether it is a PHY, a bridge IC, a connector or a cable, it is important that a diligent approach is taken. Choose wisely and, wherever possible, select a component certified by the USB-IF compliance program. This will ensure the component meets the specification requirements and will be fully functional once implemented, correctly interfacing with any other certified parts. It is also important to look at other elements of the overall system when contemplating a high speed interface. There would be little benefit in employing a very fast interface if the rest of the system is not in a position to be able to keep up. As such it is important that other system elements match the performance of the data transfer bus, otherwise ‘logjams’ will occur that lead to significant delays or call for large (often expensive) buffering to be included to alleviate the problem.

How semiconductor manufacturers are addressing next generation USB standards has not in reality been as effective as it could have been and this, to some extent, has limited the scope of its adoption so far. Whilst many USB 3.x SuperSpeed options currently on the market have directly focussed on specific applications (such as memory devices), other areas where it could have huge potential (such as industrial data logging and HD inspection of items on factory production lines) have been overlooked. In reality it is exactly these sort of applications that could gain the most from such technology though – as SuperSpeed-supporting imaging equipment will be much more convenient to deploy in most scenarios than alternative based on CamerLink, HDMI, Gbit Ethernet, etc. It eliminates the need for separate cabling for power and data, thereby saving space and reducing overall bill of materials costs. In addition, its represent a smaller software stack with much shorter associated development times as a result. Another benefit that should be taken into account is the familiarity that engineers tend to already have with USB.

Figure 2: Comparison of Pinouts for USB High/Full Speed (USB 2.0) and USB SuperSpeed (USB 3.0) (Source: FTDI)

FTDI Chip has decided to take a more generic approach. The company’s FT600 and FT601 USB 3.x products offer vendor class device data transfer functions over a bulk interface endpoint to either 16-bit or 32-bit wide parallel FIFO interfaces. Although increasing the pin count required to connect the system compared to a totally serial solution, this methodology makes USB SuperSpeed bridging more accessible to a wider range of devices (including FPGAs and microcontrollers), thus expanding the breadth of applications that can be undertaken. Converting the serial USB to a parallel interface also allows the system to maintain high data rates without the need to clock data in or out of the chosen USB interface IC as fast as would otherwise be required for a serial architecture when operating at the same overall data rate.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.