 |
|
Copper wire shown to be competitive with fiber optic cable for LANS
December 8, 2003
Penn State engineers have developed and simulation tested a copper wire transmission scheme for distributing a broadband signal over local area networks (LANS) with a lower average bit error rate than fiber optic cable that is 10 times more expensive.
Dr. Mohsen Kavehrad, the W. L. Weiss professor of electrical engineering and director of the Center for Information and Communications Technology Research who led the study, says, "Using copper wire is much cheaper than fiber optic cable and, often, the wire is already in place. Our approach can improve the capability of existing local area networks and shows that copper is a competitor for new installations in the niche LAN market."
Kavehrad presented the Penn State team's results in a paper, 10Gbps Transmission over Standard Category-5, 5E, 6 Copper Cables, at the IEEE GLOBCOM Conference in San Francisco, Calif., Thursday, Dec. 4. His co-authors are Dr. John F. Doherty, associate professor of electrical engineering, Jun Ho Jeong, doctoral candidate in electrical engineering, Arnab Roy, a master's candidate in electrical engineering, and Gaurav Malhotra, a master's candidate in electrical engineering.
The Penn State approach responds to the IEEE challenge to specify a signaling scheme for a next generation broadband copper Ethernet network capable of carrying broadband signals of 10 gigabits per second. Currently, the IEEE standard carries one gigabit over 100 meters of category 5 copper wire which has four twisted pairs of wire in each cable.
"In the existing copper gigabit systems, each pair of wires carries 250 megabits per second. For a 10 gigabit system, each pair will have to carry 2.5 gigabits per sec," Kavehrad explains. "At these higher speeds, some energy penetrates into the other wires and produces crosstalk."
The Penn State scheme eliminates crosstalk by using a new error correction method they developed that jointly codes and decodes the signal and, in decoding, corrects the errors.
Kavehrad says, "Conventional wisdom says you should deal with the wire pairs one pair at a time but we look at them jointly. We use the fact that we know what signal is causing the crosstalk interference because it is the strongest signal on one of the wires." The Penn State approach also takes account of the reduction or loss of signal energy between one end of the cable and the other that can become severe in 100 meter copper systems.
"We jointly code and decode the signals in an iterative fashion and, at the same time, we equalize the signals," adds the Penn State researcher. "The new error correction approach acts like a vacuum cleaner where you first go over the rough spots and then go back again to pick up more particles."
A MATLAB simulation has shown that the scheme is possible and can achieve an average bit error rate of 10 to the minus 12 bits per second. Fiber optic cable typically achieves 10 to the minus nine. The work is continuing.
Penn State