串行/解串器与铜质千兆连接：串行/解串器如何通过铜质千兆连接技术取代光学数据连接(

　　Copper cable data links have been evolving for many years. They have transitioned from very low data rate links to links running at multiple gigabits per second. With advances in serdes /gigabit transceiver technology (serializer and de-serializer), we are likely to see optical data link implementations relegated to its traditional optical networking arenas such as SONET, optical GbE and Fibre-Channel standards compliant applications. Copper cable-based data links are likely to replace optical implementations especially for short distance (<10 meters), high-speed links within proprietary-natured applications that tend to be non-standards based.

Historical Perspective:
For early gigabit link adopters, the lack of semiconductor physical layer device advances prevented copper-based gigabit data link implementation. At the time, optical was the only choice available for achieving gigabit serial link performance. Additionally, optical provided system designers with easy solutions to solve difficult problems, i.e., gigabit serial data communications. This solution came at the expense of higher implementation costs. The high cost of optical modules, fiber, test equipment and the hard-to-find expertise often made end-products more expensive to build, test and support. For these reasons, designers typically avoided optical serial implementations and used familiar methods such as wider-parallel copper buses running at slower, but more manageable line rates.

This brute-force method using numerous parallel links, however, had limitations. As the aggregate data rate moves higher, the designer must add more parallel lines in order to scale with the data rate. At some point, depending on the design, the large number of parallel links becomes unmanageable. In addition, these implementations demand valuable board space and consume vast amounts of power. Beyond a certain number of parallel links, designers also have to carefully mange the channel skew between the copper lines to maintain signal integrity. In today’s low-power, low-cost and small form- factor driven markets, optical and parallel gigabit copper solutions do not meet a designer’s requirements who is "pushing the envelope" in gigabit data rate applications.

The birth of Serdes
With no optimal solution available, systems designers began to demand answers from semiconductors vendors. At the time, few companies could make a solid business case for investing in the serialization/de-serialization (Serdes) and clock data recovery (CDR) technologies addressing these issues. However, with communications equipment, consumer electronics and industrial automation equipment explosion, applications requiring the serdes technology increased. Thus, semiconductor vendors could justify developing the serdes technology that would eventually foster the gigabit serdes. Even with the advent of the new gigabit serdes technology, many systems designers still questioned its "robustness" for transmitting gigabit-per-second payloads over copper media while maintaining target bit error rates (BER). Over time, as the engineering community understood the underlying concepts of CDR/PLL technology, gigabit serdes became a powerful tool.

How fast, how far?
Typically a system design engineer wants to understand what data rate can be achieved by a gigabit serdes for a given transmission length of a particular copper media type. First, it is important to select a serdes that is designed for driving copper media. Some serdes are specifically designed to drive optical modules, for example. In this case, these serdes do not have the transmitter drive-strength or the receiver-sensitivity to drive copper media. Some serdes solutions are specifically well-suited for driving copper media as well as optical modules. In addition, serdes typically support a specific data-rate range so engineers need to select a serdes that meets their desired data rate. For example, one gigabit serdes may feature a data-rate range from 1.5 Gbps to 2.5Gbps while the another supports a data-rate range from 600mbps to 1.5Gbps.

Another key factor in developing a gigabit link is the copper media selection. Usually the media type is chosen based on numerous criteria such as system cost, weight and performance among other requirements. A popular media-type used in today’s gigabit copper links is Category-5 (CAT-5) twisted-pair cable. Ready availability from multiple vendors, its known performance characteristics as well as cost-effectiveness makes Cat-5 (and other Cat-5x flavors) popular with system designers. Since gigabit links are typically comprised of differential signals, twisted-pair cable media is almost always selected for multi-gigabit links. Other considerations, such as the connector selection, play an important role in defining the link as well.

How is the performance?
Often the performance and robustness of gigabit links are gauged based on a composite of several factors. Some of the most important measurements are:

• Jitter: A measure of the signal’s integrity; the short-term variations of the digital signal’s significant instants in time from its ideal position.

• Vod: Differential output voltage; the amplitude of a signal being driven into the channel. Typically as the signal traverses the media signal, amplitude is "lost" due to media attenuation.

• Data Eye Diagram: An oscilloscope display in which a receiver’s PRBS pattern is repetitively sampled and applied to the vertical input, while the data rate triggers the horizontal sweep. Link performance information is determined by analyzing the eye pattern. An open-eye pattern corresponds to minimal signal distortion. Distortion of the signal waveform due to inter-symbol interference and noise appears as closure of the eye pattern. (ATIS Definition)

• BER: The number of erroneous bits divided by the total number of bits transmitted, received or processed over some stipulated period. Examples of bit-error ratio are: (a) transmission BER, i.e., the number of erroneous bits received divided by the total number of bits transmitted; and (b) information BER, i.e., the number of erroneous decoded (corrected) bits divided by the total number of decoded (corrected) bits. The BER is usually expressed as a coefficient and a power of 10; for example, 2.5 erroneous bits out of 100,000 bits transmitted would be 2.5 out of 105 or 2.5 - 10^-5. (ATIS Definition)

Once an engineer selects the serdes, cable-type and other key components, typically the information is collected relating to the above measurements (Jitter BER, Vod) to determine if the link will be "robust" enough for the target applications. For example, using TI’s TLK2501 Gigabit Serdes, Cat-5 cable media, and SMA connectors the following results were obtained.

 

                                                                     Figure 1. Eye Diagram at 2.0-Gbps Data Rate

For complete test results, please refer to www.focus.ti.com/lit, "High Speed Gigabit Data Transmission Over Various Cable Media" (document number SLLA091).

The results show that a 2Gbps link over five meters of Cat-5 is very achievable. Different cable, serdes and connector combinations will yield different results. In addition, system "noise" sources such as power supply, cable and trace mis-match are just a few of the issues that can degrade a link’s performance.

With this perspective on serdes-based copper gigabit links, let’s look at a typical case of applying a copper-based gigabit link resulting in a more cost-effective application. In this case, let’s assume the two system processing nodes that need to be interconnected are 5m apart. Let’s further assume that the total data throughput required for the link is 2Gbps. In this case, there are two obvious choices. This link can be implemented using an optical module such as an SFP and Fibre or using gigabit serdes and copper cable such as Cat-5.

The optical case:
The added costs for the optical option are as follows with prices depending on volume.

 

The optical link’s major components cost $71. This total does not include optical testing and measuring equipment that may be needed for debugging the link. In most cases, even an optical solution will need serdes-type functions for clock data recovery as well as data serialization and de-serialization. These device types are not included in the above cost number.

The copper case:

 

The copper link’s components cost $26. In the case with copper, the same basic equipment that is used to test other electrical ICs can be used to debug. high-speed scopes may be needed for both cases, however. Please note that this cost-analysis only looks at the large-cost components that go into making each link type.

In this case, the copper gigabit solution costs approximately 60 percent less than the optical one. This is a substantial savings, especially if multiple gigabit links will be used within an implementation. As technology progresses, both the optical and silicon technology costs will move lower making both optical and copper links more robust and cost-effective than they are today. However, silicon technology innovations such as fixed and adaptive receive equalization, multi-level signaling and improved receiver sensitivity will allow copper gigabit links to maintain, and even increase, their cost advantage over comparable optical solutions.

Conclusion:
There will always be a need for optical-based links in areas such as telecommunications (SONET, Long-Haul, Metro) and extremely high bit rate >10Gbps serial implementations. However, for short-haul gigabit links (<10 meters up to 6Gbps), copper cable-based implementations using gigabit serdes technology is likely to become the industry preference due its cost-effectiveness, robustness and ease-of-implementation.