Internet Explorer is not a supported browser for TI.com. For the best experience, please use a different browser.
Video Player is loading.
Current Time 0:00
Duration 10:22
Loaded: 1.60%
Stream Type LIVE
Remaining Time 10:22
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected

[AUDIO LOGO]

Welcome to TI Precision Labs. In this video, I will be answering the question, what is 50 Gigabit Ethernet? 50 Gigabit Ethernet is defined by the IEEE 802.3cd standard, which was approved by the Standards Board in December 2018. Let's break down what that means. IEEE stands for Institute of Electrical and Electronics Engineers. This is an organization which defines global standards in a variety of electronics applications.

IEEE 802.3 is a set of standards and protocols which defines the physical and media access control layers of wired Ethernet networks. IEEE 802.3cd is the particular standard which defines 50 Gigabit Ethernet. This standard consists of Clauses 131 through 140 and Annexes 135A to 136D, which define the electrical characteristics, encoding schemes, and test fixtures of 50 Gig Ethernet transmitted across many different mediums and attachment interfaces.

50 Gig Ethernet was preceded by 25 Gig Ethernet, which was approved by the Board in June 2016. Let's take a look at some of the key differences between 25 and 50 Gigabit Ethernet. The biggest difference between 25 and 50 Gigabit Ethernet is the modulation type. 25-gig Ethernet uses Non-Return-to-Zero, or NRZ, modulation, while 50 Gig Ethernet uses Pulse Amplitude Modulation with 4 levels, or PAM-4. I'll explain the differences in greater detail.

But the important thing to note here is that PAM-4 doubles the information throughput without doubling the baud rate. These two versions of Ethernet also utilize different encoding schemes. 25 Gigabit Ethernet implements 64b/66b encoding, meaning data is split into 64-bit words and 2 bits of overhead are added to each word. When the original data speed of 25 gigabits per second is multiplied by the encoding overhead, this results in a bit rate and baud rate of 25.78125.

50 Gigabit Ethernet uses a combination of two different encoding schemes. 256b/257b encoding improves the overhead of 64b/66b encoding and only adds one bit per each 256-bit word. Reed-Solomon Forward Error Correction, or RS-FEC, for short, adds 30 bits of overhead per each 514-bit word. The addition of these encoding overheads increases the original data rate of 50 gigabits per second to 53.125 gigabits per second.

Since 50 Gig Ethernet uses PAM-4 modulation, the baud rate is one half the bit rate, or 26.5625 gigabaud. What are the differences between NRZ and PAM-4 modulation? NRZ defines two distinct voltage levels to represent binary data where high maps to 1 and low maps to 0. The eye diagram shows the clear distinction between voltage levels, as well as the transition crossing points.

The plot below shows how a 10-bit binary value is transmitted serially with NRZ modulation. PAM-4 differs from NRZ because it defines four distinct voltage levels. This means each symbol represents 2 bits and the bit rate of transmission is doubled while maintaining the same sampling rate. Gray coding helps to minimize bit errors by ensuring only one bit value changes with each defined voltage level.

PAM-4 with Gray coding maps the following binary values to each voltage level-- from lowest to highest, 00, 01, 11, and 10. Therefore, if a PAM-4 receiver misses the intended voltage level of a symbol, only one of the two bits will be in error. The biggest drawback of PAM-4 modulation is that it increases the difficulty to maintain good eye openings.

As you can see in the eye diagram, there are three eye openings and many transmission crossing points. The biggest advantage of PAM-4 is demonstrated in the plot below. The same 10-bit binary value is transmitted at the same baud rate in half the time. RS-FEC is a block-based, error-correction encoding scheme. Data is arranged into s-bit symbols and n-symbol codewords, where k symbols make up the original message and n-minus-k symbols are added for parity.

The RS-FEC encoding scheme allows for t symbol errors to be corrected per codeword, where the number of parity symbols is 2 times t. It's important to note that a symbol error occurs if one bit or all s bits are corrupted. So RS-FEC is particularly good at correcting single-bit errors and burst errors. In 50 Gigabit Ethernet, RS-FEC is implemented with 10-bit symbols and 544-symbol codewords, where 514 symbols make up the original message.

With these parameters, 15 symbol errors can be corrected per 544-symbol codeword. RS-FEC is implemented in 50 Gig Ethernet because the inclusion of PAM-4 modulation reduces the signal-to-noise ratio. 50 Gig Ethernet can allow a higher bit error rate threshold than previous generations of Ethernet because RS-FEC can correct many errors. Now that we've covered some of the differentiating factors of 50 Gigabit Ethernet, let's go over some notable clauses of the IEEE 802.3cd standard.

Clause 135, also known as 50GBASE-R, generally defines the physical medium attachment sublayer of 50 Gig Ethernet without any specific medium. Here, 50G represents the data rate, BASE stands for baseband signal, and R stands for scrambled coding. Clauses 136 through 139 define the physical medium dependent sublayer of 50 Gig Ethernet across various mediums.

Clause 136 applies to copper cable mediums, represented by the C in 50GBASE-CR. Clause 137 applies to backplane mediums, represented by the K in 50GBASE-KR. It's important to note that both of these specifications require autonegotiation, which is a feedback protocol used to find common transmission parameters between two devices. Clause 138 applies to multimode fiber, while Clause 139 applies to single-mode fiber.

Clause 139 is split into three different specs. 50GBASE-FR supports reach up to 2 kilometers, LR supports reach up to 10 kilometers, and ER supports 40 kilometers. All 50 Gig Ethernet across fiber mediums does not support autonegotiation. Annex 135F defines the chip-to-chip attachment unit interface, while Annex 135G defines the chip-to-module attachment unit interface.

Annex 136B covers test fixtures which are relevant to the 50GBASE-CR and 50GAUI-1 C2M specs. Let's review how these clauses define different sections of a typical 50 Gigabit Ethernet application. As you can see, 50G-KR covers the backplane connection between the [? MAC ?] on the switch fabric and the [? Fi ?] on the line card. If a retimer or redriver is used on the line card, then the 50GAUI-1 C2C spec defines the connection between the [? Fi ?] and the signal conditioner.

A few different types of cables can be connected to the front port. If a passive copper cable is used, then 50G-CR defines the connection. If an active electrical cable is used, then 50GAUI-1 C2M defines the connection. If an active optical cable is used, then the electrical board connection is defined by 50GAUI-1 C2M, and the optical fiber portion is defined by 50GBASE-SR, FR, LR, or ER depending on the mode in reach.

Let's go through a short quiz to check your understanding. What is the baud rate of 50 Gigabit Ethernet? The answer is 26.5625 gigabaud. How many error symbols can be corrected per codeword by RS-FEC with n equals 544 and k equals 514? The answer is 15 symbols. Which IEEE 802.3 clause defines 50 Gigabit Ethernet with a backplane medium?

The answer is 50GBASE-KR. Thank you for taking the time to watch this video. I hope you learned a bit more about 50 Gigabit Ethernet.