More than 175 years ago, Samuel Morse sent the first telegraphic message from Washington DC to Baltimore on copper wires.
About 150 years ago the first stable transatlantic cable was laid and messages were sent between Ireland and Newfoundland over copper. The data rate on this cable was not even 1 bit/sec, it was 0.1 bit/sec. Concerns over the cable reliability and the slow data rate led some to believe copper was doomed as a communication media for these long cables.
It was Oliver Heaviside, among others, who realized one of the limitations to the bandwidth of the long-haul cables was the high series resistance of the cable compared to the inductive reactance, making the cable an RC network and not an LC, low loss network. It took another 30 years for inductive load cells to be added to cables to increase the transmission rate to 10 bits/sec.
Fast forward to 75 years ago, when the TAT-1, the first transatlantic copper cable system, was implemented. The lower loss dielectric in the cable and vacuum tube repeaters every 69 km increased the initial bandwidth to 144 kHz for analog voice signals.
The world of communications changed 45 years ago, in 1977, when the first fiber optic cable communications system was deployed by AT&T in Chicago. The introduction of low loss fibers and modulated lasers heralded a new era in communications. It hinted at the death of copper as a transmission media.
Just three years later, in 1980, my first assignment at Bell Labs, as a Ph.D. physicist fresh out of graduate school, was to investigate the use of optical interconnects and integrated optics at the circuit board level. I had joined a PCB manufacturing technology group. They were hearing of all the work other groups at Bell Labs were doing on fiber optic communications, getting ready to deploy the first fiber optic undersea cable, scheduled for introduction in 1988.
As the token physicist in a group of chemists and chemical engineers, they wanted my assessment of copper circuit boards, their bread and butter, would be made obsolete by optical backplanes. Did copper have a future when optical communications was about to proliferate?
My naïve assessment was that at the then bleeding edge data rates of 10 Mbps, copper in backplanes had at least two generations to go before it might reach a limit due to material constraints or silicon. Printed circuit boards (PCBs) were a good bet for a few more years, but optical backplanes were sure to come.
For the last 40 years, while I have been involved in the interconnect technology industry, I have heard the same refrain over and over again. With every generation of yet higher data rate, copper will reach a practical limit in one or two generations, and optical interconnects will be necessary. Optics will spell the death of copper.
At conferences, I’ve heard, “maybe we can get to 2.5 Gbps in copper backplanes, but surely, we will need optical backplanes for 10 Gbps.” And then new low loss materials and equalization features in the RX and TX were introduced. Then I heard, “surely we will need optical backplanes for 28 Gbps backplanes,” and then lower loss laminates, smooth copper with better connectors, and good design practices enabled volume deployment at 28 Gbps. But, “surely we can’t do 56 Gbps over copper,” and then PAM4 with FEC was introduced and we have volume deployment at 56 Gbps in copper backplanes.
Of course, for long distance communication, optical interconnects make sense. The right place to transition from electrons to photons is a moving target. The decision factor is not about what is possible, it is about what is cost effective. This includes the total cost of ownership of the system, the power consumption per bit, the density of lanes as well as the bandwidth-length product. The power consumption per bit is flattening for silicon and the cost per bit of optical systems is coming down. The most cost-effective location to switch from electrons to photons is getting closer to the backplane.
As we move toward deploying 112 Gbps single lane data rates, we might say, “maybe we can do the next generation in copper, but surely, this will be the limit and we will have to switch to optical backplanes.” And then some clever engineers surprise us with a combination of powerful technologies like cabled backplanes to extend the life of PCBs to yet another generation.
But, surely we will need optical interconnects for 224 Gbps, won’t we?