Think Copper for High Speed

Too many engineers automatically think first of fiber optics when they think about high-speed data transfer. They should also give a thought to copper. Why? Primarily because it's less expensive.

Of course, many engineers believe that copper can't achieve the lengths required by systems at the faster data rates of the future. But they were wrong!

Originally, fiber philes thought that copper could not reach longer than a couple of meters at 2.5 Gbps (giga bits per second). But then, passive equalization was applied to copper and it got up to as long and far as 20m lengths at 2.5 Gbps.

Another successful technology is the XAUI protocol. This takes a serial data stream, perhaps at 10.0 Gbps, and splits it into four parallel data streams, each operating at one fourth of the 10.0 Gbps data rate. The copper achieves long lengths because each stream is operating at a slower data rate. At the end of the long-length copper interconnect, the four streams are recombined back into the original 10.0 Gbps serial data rate.

Parallel with the XAUI effort, there are copper technology improvements that yield long copper lengths where each serial lane operates at fast data rates: These breakthroughs include 20m passive at 2.5 Gbps; 50m using active assemblies at 2.5 Gbps; 15m passive at 3.125 to 4.25 Gbps; 30 to 40m using active at 3.125 to 4.25 Gbps; 10m passive at 6.25 Gbps, longer active; and 15 to 20m active at 10.0 to 12.5 Gbps.

What applications can use this fast copper?

• Fiber Channel (storage area networks) currently up to 4.25 Gbps serial, 10.0 Gbps using XAUI protocol (10GFC)

• Gigabit Ethernet (telecommunications) 10.0 Gbps serial using X2, XFP, XENPACK, etc. 10.0 Gbps using XAUI protocol (CX4)

• InfiniBand 2.5 Gbps. Working on 5.0 and 10.0 Gbps using XAUI, and active assemblies for serial

• PCIExpress implementing 2.5 and 6.25 Gbps using passive and active serial, multiple parallel (up to 32X, or 16 lanes each way) using XAUI type protocol

The next big jump in copper technology is active cable assemblies. I call it "the copper miracle."

Currently, applications need to get from a SERDES on the transmit circuit board, along a length of circuit board trace to an I/O port, then down a long length of copper cable, then onto another length of circuit board trace at the receive I/O port, finally to the receive SERDES. The transmit SERDES must launch a large enough signal (typically 0.8V minimum) to end up with at least the minimum signal required for the receive SERDES to function without error (typically 0.2V minimum).

The copper miracle is to supply 2.2 or 3.3V dc power and return on two pins at the cable assembly I/O port. Existing high-speed connectors have had these pins designed in, in anticipation of active connector technology coming on line. An active cable assembly needs only 0.05V signal at the I/O port. The cable will re-launch a large voltage (0.8V minimum) signal into the cable length. Only a 0.05V signal arrives at the receive end of the assembly. Then, the cable assembly at the receive I/O port will again launch a large voltage (0.8V minimum) signal onto the receive circuit board in order to achieve a 0.2V signal at the receive SERDES. The technology is available up to 12.5 Gbps. Copper should have a bright future if engineers give it a chance.