Lisa Bechtold

Berk-Tek Inc.

There has been endless confusion about what defines the speed in high-speed networks. Here we will explain the difference between megahertz and megabits, as well as define the relationship between the two.

Bit rate does not necessarily equal frequency. They match closely for some systems; for others, the disparity is great. What, then, is the relationship among frequency, hertz, cycles and bits?

The bit, or binary digit, is the smallest piece of information that can be processed by a computer. In many systems, such as the American Standard Code for Information Interchange, it can take 8 bits, or 1 byte, to make one character--a letter, numeral or symbol. A bit is either a 1 or 0, a "yes" or "no," or an "on" or "off."

The frequency of a signal voltage is measured in cycles per second. One hertz is one complete cycle per second. While higher frequency can mean a faster system, a truer measurement of communication speed is bit rate.

Most data communications systems operate at millions of cycles per second, or megahertz. In high frequencies, such as values in the MHz range, the time the cycle requires is measured in minute fractions of a second.

If one cycle of signal carries 1 bit of information, then the frequency of the system (in hertz) equals its speed (in bits per second). However, there is no reason why a single cycle cannot carry more than 1 bit of information. Therefore, increasing the speed of a system without changing its frequency is possible. With the premium placed on speed in the world of data communications, it is not surprising that several encoding schemes have been developed that accomplish just that.

Encoding schemes increase speed

For example, fiber distributed data interface uses the non-return to zero, inverted digital encoding scheme. This scheme represents the 1s and 0s in digital transmission using alternating low and high voltages. Any change in voltage represents a digital 1, and no change represents a digital 0.

Because each change is recognized by the receiver as a bit, nonreturn to zero, inverted can generate 2 bits per cycle. In this case, the bit rate is twice the nominal signal frequency.

To increase the bit rate or "speed" of the signal in the example above, we would have to increase the frequency. The system still sends out 2 bits per cycle, but does it in shorter cycles.

Although this may seem to be a suitable solution in the search for higher communication speeds, there is a problem. Increasing the frequency of transmission can greatly elevate radiation or electromagnetic interference emissions from the system, which violates Federal Communications Commission regulations. The cable effectively becomes a transmitter that sends signals into the air.

To illustrate an example of increased frequency in a real-life application, the FDDI standard permits highly repetitive bit patterns. As we have already seen, non-return to zero, inverted represents a logical 1 as a change in voltage level. A long string of 1s, then, would necessitate a constant changing of the voltage level. Because this change takes the form of a sine wave that moves from positive to negative voltage and back, it follows that as the speed of voltage level alternation increases, so does the frequency of the signal.

Stated in another way, signal frequency can vary in any transmission system dependent upon the content of the information being sent at any given time. Peaks and lulls will occur where the frequency of the voltage increases and decreases. People who design encoding schemes and transmission systems must be concerned with peak frequencies. They must also be concerned with how much of the energy used in the scheme falls at higher frequencies because of the radiation problems mentioned earlier.

Multilevel encoding more efficient

Bandwidth-efficient encoding schemes, then, are designed to transmit more bits of information using lower frequencies. MLT-3, for example, is a scheme for 100-megabit-per-second FDDI over copper. It uses a multilevel threshold approach; three levels of voltage change are used rather than the two levels used for non-return to zero, inverted.

If the fiber-based non-return to zero, inverted scheme is translated to MLT-3, which runs over unshielded twisted-pair copper, then every time non-return to zero, inverted changes its logic level, MLT-3 must do the same. Using three levels instead of two, however, places the maximum fundamental frequency of MLT-3 at half that of non-return to zero, inverted. Much of the frequency range used is less than 30 MHz, within the limit imposed by the FCC for UTP data transmission. The 100-Mbit/sec FDDI signal runs at 31.25 MHz over UTP. A 155-Mbit/sec signal, such as that proposed for asynchronous transfer mode, can run at less than 50 MHz.

More bandwidth-efficient encoding schemes than MLT-3 are possible, and can transmit 9 or 10 bits per cycle. This technology is being applied in state-of-the-art modems.

Multilevel encoding schemes are capable of transmitting larger packets of information--that is, more bits--in efficient patterns at lower frequencies than the most commonly used encoding schemes allow. The non-return to zero, inverted code was described here because it is used in FDDI transmission, and also because it provides a simple example of how digital information can be translated into a signal for transmission.

Most of the familiar codes used today are not as efficient as non-return to zero, inverted, which can only be used for certain timing-tolerant applications. Differential Manchester encoding, used for l0Base-T and token ring, is only half as efficient as non-return to zero, inverted. One cycle can represent only 1 bit of information, and bit rate does appear to match frequency. For example, 10-Mbit/sec l0Base-T does run at 10 MHz.

Increasing the frequency to increase the number of bits transmitted does not always answer the need for more speed. Increased frequencies produce increased emissions, making their use impractical in the real world. Bandwidth-efficient encoding schemes are designed for real-life applications, such as 100-Mbit/sec twisted pair-physical media dependent and 155-Mbit/sec ATM, where higher data rates are required but the systems must perform at usable frequency levels.

The expanded number of logic levels in bandwidth-efficient encoding schemes makes them more susceptible to noise, as well as more frequency-efficient. With more possible signal levels being driven more quickly by the system, there is less room for errant signal noise.

The critical parameter for proper transmission of bandwidth-efficient encoding schemes over UTP cable is the attenuation-to-crosstalk ratio. This ratio is a measure of the difference between the desired signal and undesired interference and loss. It defines the usable bandwidth of the cable. Cables that offer very low crosstalk to prevent this internal interference, the primary source of noise in UTP transmission systems, must be chosen. Cables with high crosstalk squeeze the bandwidth and reduce their information-carrying capacity.

High-performance UTP cables that offer enhanced crosstalk values will be able to carry the high-speed network signals currently on the market or under development. Efficient encoding schemes will keep frequencies and emissions within tolerable levels.

Lisa Bechtold is an applications engineering manager at Berk-Tek Inc., New Holland, PA.