Home Networking Basis: Transmission Environments and Wired/Wireless Protocols

Earlier versions of Ethernet, 10Base5, and 10Base2 were built upon interconnected coaxial cables. These coaxial cables are not very flexible. Therefore, the repositioning of a desktop computer in an office might involve efforts of professional installers. 10BaseT is the first version of Ethernet using twisted pair cables. 10BaseT supports a 10-Mbps transmission rate over two pairs of Category 3 or Category 5 twisted pair cable. Because of the convenience of twisted pair wiring in an office environment, 10BaseT and its higher throughput variations have become the norm for most new Ethernet installations. Category 3 or Category 5 twisted pair cable contains four pairs of wires. 10BaseT uses one pair of wires for transmitting data and the other pair for receiving data. 10BaseT is designed for using Category 3 twisted pair cable over a connection distance of 100 m. The connection distance is increased to 150 m if a better quality Category 5 twisted pair cable is used instead.

Although Category 3 twisted pair cable is characterized as voice grade with transmission capabilities similar to those of in-house telephone wiring, the conventional in-house telephone wiring cannot be used for Ethernet application because it usually has only one available pair after the telephone usage and multiple branches normally exist on each pair. The other distinction is that telephone wiring uses a four-position RJ-11 connector while each end of a Category 3 or Category 5 cable for the Ethernet application is terminated with an eight-position RJ-45 connector (Figure 6.4).

Each pair of a Category 3 or Category 5 cable is distinguished with different colors. Solid colors are used for cables with solid conductors, while color stripes are used for those with stranded wires. There are two conventions for connecting each pin of the connector to a particular wire identified by its color. Type A and Type B conventions are listed in Table 6.1 and Table 6.2, respectively.

For a cable with striped wires, wires sharing the same color form a pair. It is interesting to note that wires of the orange pair for Type A and of the green pair for Type B are not connected to adjacent pins. This design allows the wiring to be shared with a telephone connection where the center two pins are normally utilized and the RJ-11 plug can be inserted into an RJ-45 wall connector. When RJ-45 connectors at both ends of a twisted pair cable are connected to colored wires according to the same convention, Type A or Type B, a straight through cable is created. When two different types of conventions are applied at different ends of a twisted pair cable, it becomes a crossover cable.

All 10BaseT connections are point-to-point. This implies that a 10BaseT cable can have a maximum of two Ethernet transceivers, with one at each end of the cable. One end of the cable is typically attached to a 10BaseT hub. The other end is attached directly to a computer station's Network Interface Card. These point-to-point cable connections of 10BaseT result in a star network topology, which consists of a central hub and many radiated NICs. Straight through cables are used to connect a NIC to the hub. An embedded crossover function is implied inside the hub to match the transmitter of a NIC to a receiver of the hub and vice versa (Figure 6.5). Two 10BaseT NICs may be directly attached to each other without a hub using a crossover cable (Figure 6.6). The independent transmit and receive paths of the 10BaseT allow the full-duplex mode of operation to be optionally supported. The full-duplex mode can effectively double the transmission throughput over that particular connection. To support full-duplex mode, both the NIC and the hub must be capable of, and be configured for, full-duplex operation.

The transmission integrity of a particular connection can be verified during both regular data transfer and idle periods. When no data are to be transferred, transceivers still send link pulses to each other periodically to verify the integrity of the connection. 10BaseT transceivers normally provide a link light that remains lit as long as the transceiver receives frames or link pulses from the other end of the segment.

6.2.1 Summary of 10BaseT Ethernet Standards

Standards are usually established to maintain product performances at certain functional levels while not violating existing regulations. Standards are also necessary to make the same product from different vendors interoperable at the required functional level. For these goals, standards sometimes getting into certain technical details; however, standards are generally not specifications of related products. Nor do standards provide all necessary information and means for making a particular product. Ethernet standards for different types usually provide information on transmission environments such as cable channel and crosstalk models, electrical characteristics such as time and frequency domain masks, signaling formats such as line and error correction codings, and set-up procedures such as hand-shake and negotiation sequences.

Some operation details of the 10BaseT Ethernet are specified in Chapter 14 of IEEE 8032.3 standards. Specifically, the Category 3 twisted pair channel model, output differential voltage level, signal time domain masks, output impedance, and timing jitter are defined in Section 14.3.1.2. Near End Crosstalk models of different numbers of disturbers can be found in Section 14.4.2. The transmission characteristics of a 100-m Category 3 twisted pair for the 10BaseT Ethernet application are specifically defined in terms of a circuit model consisting of passive R, L, and C components. The topology and component values of this transmission media model are shown in Figure 6.7. This twisted pair model can be further divided into four second-order T, 100-ohm, input-output impedance matched networks of identical topology but different component values. Four T networks are chosen such that enough degree of freedom can be provided to fit attenuations of this model in the frequency range of interest very close to those of the twisted pair cable. At a very low frequency, the attenuation of each T network is determined by shorting the inductance and opening the capacitance. At a very high frequency, the attenuation of each T network is determined by shorting the capacitance and opening the inductance. The attenuation of the cable model is the sum of these four T networks. Some T network design theories can be useful in selecting each component value during an iterative process.

By defining proper ABCD matrices for different sections of this circuit and after several steps of parallel and serial two-port network combination, the insertion loss of this model is shown in Figure 6.8 together with that obtained according to the propagation model of Chapter 2. Insertion losses are about 9.5 and 15 dB at frequencies of 10 and 20 MHz, respectively.

Typical crosstalk losses of the corresponding twisted pair are specified by

Equation 6.7

Equation 6.8

The frequency is measured in megahertz. Equations 6.7 and 6.8 are for Category 3 cables with 25 and 4 pairs, respectively. Figure 6.9 shows the crosstalk coupling losses corresponding to these expressions as well as one, in the middle, according to the NEXT model of Chapter 2 with 25 disturbers. The specified Category 3 cable crosstalk loss is around 26 and 30 dB at 10 MHz for 25 and 4 pairs, respectively. Although there exists only 1 crosstalk pair in a practical 10BaseT setting, multiple-pair crosstalk models can be used for a better transceiver design. The four-pair crosstalk model has less crosstalk loss and, therefore, produces more crosstalk noises. It has been chosen for following 10BaseT transmission simulation studies.

Information bits are converted to signal levels applied to the twisted pair using the Manchester encoding where a 0 bit is represented by a positive to negative transition and a 1 bit is represented by a negative to positive transition as shown in Figure 6.10. This encoding process can be implemented by switching between two opposite phases of a bit-rate clock that alters between positive and negative amplitude during each bit interval.

The differential voltage level across two wires of a twisted pair with a nominal impedance of 100 ohms is specified to be between 2.2 and 2.8 volts (V). The transmit power is therefore between 48.4 and 78.4 milliwatts (mW) or between 17 and 19 dBm. The normalized Power Spectrum Density of a Manchester encoded signal can be described by

Equation 6.9

where T is the duration of a bit interval. Figure 6.11 shows the PSD of this Manchester encoded signal at 10 Mbps based on a signal level of 2.8 V on an impedance of 100 ohms. This PSD is obtained by multiplying the normalized PSD for Manchester encoding with the power factor of P = V²/R = 2.8²/100 = 0.0784.

To meet the standards requirements on transmit signal, a high-order low-pass filter is necessary to make the PSD of above 20 MHz at least 20 dB lower. The transfer function of a typical 10th-order low-pass Butterworth filter is shown in Figure 6.12 together with the filtered PSD of a Manchester encoded 10BaseT transmit signal. In time domain, the transmit signal should fall within the shaded area of the mask defined by the standards as shown in Figure 6.13. After passing through the twisted pair and another receiver low-pass filter (used for out of band noise reduction), the received signal needs to conform to either of two time domain masks (Figure 6.14 and Figure 6.15, respectively) depending on if there is an information bit transition. Based on a transmit signal peak voltage of 2.8 V, the maximum allowed peak signal loss is about 13 dB.

6.2.2 10BaseT Transceiver Structure and Performance Analysis

Standards are usually established based on some prototype products made possible by validated technologies. Technical debates based on working prototypes, in these early days, or computer simulation results, most recently, are normally parts of the standardization process. However, these standards do not usually document the details of techniques currently under debate. Theoretically, one can produce a compliant product by following the details of standard documents using primarily off-the-shelf but a few specific, sometimes patented, technologies. To produce standards conforming signals and decode them after a specific transmission medium for the Ethernet case, lots of implementation details must be verified using computer simulations.

An Ethernet device consists of a transmit part for producing the information-bearing signal and a receive part for recovering the information from the signal after the transmission medium. There are different methods for producing a standards-compliant signal. Computer simulations are normally used to test these different alternatives and to select the optimal choice based on available means of manufacturing. Computer simulations are also useful as a reference model for confirming implementation details at every functional level. There are also different ways of recovering and decoding a signal after a transmission medium. The synchronization of the decoding circuits to that of the transmitter is often an important part of the receiving process. Some compensation measures such as filtering and equalization are also parts of the receiving process. Important choices of the receiver circuits are also validated through computer simulation.

The 10BaseT Manchester encoding can be implemented by using a clock of 10 MHz with two opposite phases and a multiplexer selecting switch. The original clock phase is selected by the multiplexer for a 0 bit, and the opposite phase is selected for a 1 bit. At the receiver, two phase comparators, using opposite phases of the 10-MHz clock as references, can be used to detect the original information bits. The 10-MHz clock at the receiver end is recovered by a phase-locked loop. Corresponding transceiver block diagrams of 10BaseT are illustrated in Figure 6.16.

In Figure 6.16, shift registers are used in the transmitter as well as in the receiver to get in and send out information bits. The clock in the transmitter on the left side of the block diagram drives the shift register, and its two phases are inputs to the multiplexer switch. The information bits coming out of the shift register determine which clock phase to be used as the transmit signal. The phase-locked loop recovers the 10-MHz clock in the receiver. The received signal and the recovered clock as well as its opposite phase are used by a pair of phase detectors. Outputs of phase detectors are compared to determine the correct phase of the received signal. The phase-locked loop should be designed to be invariant to the frequent 180 phase shift of the received signal while recovering a stable 10-MHz clock. Additional important functions are also required in a real implementation of a 10BaseT transceiver. Collision detection and link integrity tests are partially implemented in the transceiver physical layer as well. Additional details of an example implementation of a 10BaseT transceiver can be found in the data sheet of the National Semiconductor's DP83907 chip [2].

The theoretical performance of a 10BaseT transceiver under the worst case environment as defined by the standards can be analyzed by comparing the Signal-to-Noise Ratio at the front of the receiver. Figure 6.17 shows receiver front-end signal and crosstalk PSDs for a frequency band of up to 20 MHz. The R-, L-, and C-based channel and the four-pair crosstalk models are used to generate these PSDs, respectively. The corresponding SNR is shown in Figure 6.18.

Based on these receiver front-end SNRs, the channel capacity for the 10BaseT transmission environment is found using

Equation 6.10

At about 60 Mbps, the channel capacity is about six times the throughput of the 10BaseT transmission protocol.

The performance of a typical 10BaseT transmission system under different channel and noise environments can be studied through computer simulation. A Simulink simulation model can visually show the transceiver architecture and hierarchically provides implementation details while presenting timing-accurate simulation results. Figure 6.19 shows a 10BaseT Simulink model block diagram at its highest hierarchical level. This Simulink model consists of a random binary sequence generator as the signal source, a 10BaseT transmitter to convert the information sequence into Manchester encoded line voltage on the twisted pair cable, another pair of binary sequence generator and transmitter as the crosstalk source, a four-pair Category 3 channel model, and a 10BaseT receiver. Four scopes are placed after the information sequence, after the transmitter, after the channel, and after the receiver to collect simulation results.

Figure 6.20 shows the structure of this model 10BaseT transmitter. The Manchester encoding is accomplished with a 10-MHz clock signal and a phase modulator. The original or the opposite phase of a 10-MHz clock is selected depending on the input data sequence through a switch multiplexer. The original phase is selected if the data bit is one, and the opposite phase is selected otherwise. The phase-modulated clock, with zero and positive levels, is then converted into positive, +1, and negative, 1, levels. This signal is passed through an eighth-order Butterworth low-pass filter with a corner frequency of 15 MHz before entering the transmission channel.

Figure 6.21 shows the structure of this model 10BaseT receiver. It consists of a Costas Phase-Locked Loop (PLL) with the received signal as its input and a recovered clock and its 90° phase-delayed version as outputs, a phase comparator, and a sample-and-hold device. The phase comparator compares the input signal and its negated version with the recovered clock and generates corresponding zeros and ones as recovered data bits. The glitches at phase transitions are effectively removed by the sample-and-hold device driven by the 90^o phase-delayed clock.

Figure 6.22 shows the structure of the Costas PLL for the clock recovery. The amplitude-adjusted received signal is multiplied by the sine and cosine phases of a local clock signal generated from a VCO (Voltage-Controlled Oscillator) individually. Voltages dependent only on phase differences are derived after loop filters of second-order Butterworth low-pass type with corner frequencies at 796 kHz. The effect of the received signal polarity is void in the product of these two phase differences. Therefore, this Costas PLL is blind to the frequent 180° phase reversal in the received signal. When the phase of the VCO is lagging behind that of the received signal, a positive voltage is generated and applied to the VCO to raise the frequency slightly, therefore eliminating the phase difference.

The Category 3 channel model consists of a Category 3 twisted pair channel model and a 15-dB/decade crosstalk model. These channel and crosstalk models have a complex frequency s, that is dependent as shown in Equations 6.11 and 6.12, where s = j2pf. The input of the channel model is connected to the transmitter driven by the information sequence, while the input of the crosstalk model is connected to the transmitter driven by the crosstalk sequence. The channel model is constructed based on the circuit channel model of R, L, and C components as defined in the standards. ABCD matrices are established for each or a group of components representing a simple serial or shunt impedance. Each ABCD parameter is a two-element vector representing the first- and zero-order coefficients of the polynomial. For example, the polynomial for an impedance of a resistor R, in serial with an inductance L, is and the corresponding two-element vector is [L R]. ABCD matrices are then obtained for each section of the RLC network through parallel or serial ABCD matrix manipulations. Because of the polynomial representation, multiplications between different ABCD parameters are carried out through convolution. Four sections of the RLC networks are finally combined to form the ABCD matrix of the circuit cable model. The insertion loss of this circuit cable model in the complex frequency s domain is described by

Equation 6.11

The 15-dB/decade crosstalk model with a loss of 26 dB at 10 MHz is approximated with an RLC network using the basic sectional structure of the circuit channel model. Figure 6.23 shows the insertion loss of this RLC network as well as that of the 15-dB/decade model. Loss differences are less than 0.5 dB in the frequency range of 1 30 MHz. The insertion loss of this RLC crosstalk model is described by

Equation 6.12

By comparing the input sequence to the recovered one in the Simulink simulation, some errors can be observed only after amplifying the crosstalk signal by more than four times. This corresponds to a crosstalk noise margin of about 12 dB.