Many people seem to confuse bandwidth with "bit rate capacity." Bandwidth is measured in Hertz. Over and over again...
we see bandwidth measured in bps. I guess we are to assume one bit per Hz of capacity. Why not more than one bit per Hz? The FCC tells us that we must get more than 4.5 bps per Hz if we want to license a microwave transmitter. Following this thinking, we should easily place 1.544 Mbps digital baseband bit stream on a transmission medium that only has capacity for 1.544 MHz at the conventional definition of bandwidth, one bit per Hz. We wish those folks lots of luck. They'll do it if the pulse risetime is zero seconds. Physicists will tell us that that is against the laws of nature. If we do not provide sufficient bandwidth for a signal, the signal will be corrupted and that corruption can be measured in error rate performance.
I stick with the IEEE Dictionary, now in seventh edition. They don't put their foot in their mouth very often. This edition of the dictionary has 17 definitions for bandwidth, none are measured in bps. Bandwidth is a premium commodity in the world of transmission. We want to get as many bits in one Hz of bandwidth as possible. We pay good money for it based on its "quality." How is the quality of bandwidth measured? Most of us just assume the bandwidth definition is where the signal level drops 3-dB at either end. There should also be some measure of noise across that bandwidth. Now there are four types of garden variety of noise. We find most people only think of thermal noise. That's a good start. There is also impulse noise, crosstalk and intermodulation noise. If we really want to get into it, a signal has phase noise around its skirts -- and so on.
The good technologist will provide a "bandwidth" which should be defined in Hz. That same person will want to impress so many bits per second down that bandwidth pipe. The signal must suffer some corruption at the receive end of the pipe. The amount of corruption is generally measured in error rate. It is against the laws of nature to have a completely error-free signal. There are important way-points for bit error rate values we should respect. For example, 8-bit PCM telephony requires better than 1 error in 1000 bits received, which is stated as a BER of 1X10-3. This value is dictated by the supervisory signaling on the circuit, not on the PCM itself. Intelligibility of PCM is not lost until the BER of the signal degrades from 1X10-2 value. To save money on circuit quality, a BER of 1X10-3 is all I should be required to provide. I'll be drawn and quartered at dawn by the computer people if that is the best I can do. Many commercial common carriers provide a raw BER of 1X10-12. By "raw" I mean without the help of such stratagems as FEC (forward error correction). How long would it take me to measure the error rate if I transmitted 1000 bps over the circuit? The answer is "a very long time," 109 seconds. Before applying such BER values, remember that you may, at some time, be asked to measure it. The bottom line of this little essay is that bandwidth is measured in Hz and bps is a measure of bit rate capacity.
About the author:
Roger Freeman has worked in telecommunications since 1946 when he joined the Navy and became an aviation radioman. Later, Roger served as a radio officer in the merchant marine for nearly 10 years. He then held several positions with ITT assigned to their Spanish Standard Electrica subsidiary. He also served the International Telecommunication Union as Regional Planning Expert for Northern Latin America based in Quito, Ecuador. Roger is bilingual. His last employee position was principal engineer with the Raytheon Company, Marlboro, MA where he took early retirement in 1991 to establish Roger Freeman Associates, Independent Consultants in Telecommunications. He has been giving seminars in telecommunication disciplines at the University of Wisconsin, Madison for nearly 20 years. Roger has been writing books on various telecommunication subjects for John Wiley & Sons since 1973. There are seven titles which he keeps current including the two-volume work, Reference Manual for Telecommunication Engineers, now in 3rd edition. He holds two degrees from NYU. His Web site is www.rogerfreeman.com and his e-mail address is firstname.lastname@example.org.