What you'll learn:
- Why are retimers are better option than redrivers for USB4?
- What are the myths surrounding cost and complexity of retimers?
- Myths regarding retimer size and latency.
The Universal Serial Bus (USB) is an industry-standard specification for cables, connectors, and protocols that link computers and peripherals. USB-compliant devices are widely used for communications and power delivery between them.
Since it was introduced in 1996, the USB spec has changed with the addition of higher rates and more complex protocols. In recent years, design engineers have typically used redriver devices in their USB solutions.
That’s changing with the advent of USB4 due to several disadvantages of analog redrivers. These include amplifying both the signal and their internal noise, the inability to fully clean intersymbol interference (ISI), and the inability to restore the eye width and associated jitter. The combination of these factors can lead to errors and poor performance. Those three disadvantages are ending the era of redrivers in USB and ushering in retimers for high-speed data channels.
Retimers are nothing new. Even so, myths exist about them. Digital retimers have been used since the 1960s for telecom T1 and E1 digital carrier systems to carry multiple channels of voice circuits over twisted pairs with a digital retimer installed every few thousand feet.
1. Retimers aren’t necessary.
It used to be that USB needed neither redrivers nor retimers—then speeds increased. Redrivers could fill the bill for a few generations. USB4 has eliminated the use of redrivers because of signal-integrity problems caused by its higher rates, as described above.
2. Retimers burn much more power than redrivers.
The linear amplifiers used in redrivers make them consume an amount of power similar to retimers. At least one redriver consumes more power than one retimer on the market.
3. Retimers add complexity.
While retimers add more complexity in software, redrivers add much more complexity in signal-integrity analysis. Many more engineers are good at software than at signal-integrity analysis. As an example, effective use of retimers has the effect of making each USB4 port in a system have the same performance because the signal is refreshed by a retimer located next to each port.
Redrivers can leave the port in the far corner of the PCB with inferior properties due to loss of signal integrity caused by the long PCB reach leading to that port bleeding through the redriver. This can lead to headaches for both system vendors and users.
4. Retimers are for experts.
While retimers do add some complexity, retimer vendors are trying to make it easy by solving the key problems, such as how to best combine a retimer with a power-delivery (PD) device to deliver the most optimized USB-C solution.
5. Retimers are new.
Digital retimers, developed by Bell Labs, were used in the 1960s for telecom digital carrier systems T1 and E1. They carried multiple channels of voice circuits over a shielded twisted pair with a digital retimer installed every few thousand feet. They were mass produced for T1/E1 systems in the 1970s. What’s new is their use in USB links.
6. Retimers are only for the Ethernet/OIF ecosystem.
The Ethernet/OIF ecosystem has indeed used retimers for many years. That ecosystem has long put bandwidth at a premium. USB systems used to be able to get by with redrivers, given their lower speeds. The increased speeds of USB 3.1 Gen2 and now USB4 have thus closed the door on the use of redrivers.
Retimers are required in all but the smallest USB4 systems to clean up those high-rate signals. That’s because for a USB4 port to achieve its full reach, the loss budget can’t be used crossing a PCB from the microprocessor to the USB-C connector. In most host scenarios for USB4, a retimer is needed, particularly when a system doesn’t use the most expensive types of PCB materials.
7. Retimers are too expensive.
Retimers are indeed a percentage more expensive than redrivers. That said, there’s no choice but to use retimers in USB4 systems because of the signal-integrity considerations mentioned above.
However, a protocol-aware retimer solution offers the signal-integrity performance for USB4 applications, as well as a development path and a cost-effective system solution to meet consumer demand.
8. Retimers add too much latency.
Redrivers add only a small analog latency. Well-designed retimers add only a modest additional latency, on the order of 30 ns for USB4 Gen3.
9. Retimers are too big.
Retimers also come in small packages. Retimers are available in a 4x4mm high density interconnect (HDI)package and in a 5x5mm non-HDI package.
10. Retimers are single-sourced.
This is a temporary issue. The marketplace for redrivers has long been robust and the same should be true for retimers. A few available retimer solutions support USB4, DisplayPort, and Thunderbolt protocols, allow for direction reversal, support connector flipping and external PD controllers, and include extensive test and measurement capabilities. Kandou, for example, offers one such solution.
11. The specification beyond USB4 is the one that will require retimers.
This is wishful thinking. USB4 presents challenging signal-integrity problems that can only be solved with retimers. Redrivers simply won’t work for USB4.
The time is now for USB4 retimers. They are important for interconnect designs and essential for the successful operation of USB4 and the other high-speed protocols defined for USB-C. To put it succinctly, USB4 is the end of the road for redrivers.
Editor’s Note: Kandou recently announced production sampling of its USB-C multiprotocol retimer with USB4 support.