7.1 HomePNA 1.0
| The idea of an existing telephone wiring based home network system was originated from Tut System. Matt Taylor, the founder of Tut System, worked with his engineers to develop a proper signaling method over the home telephone wiring system. They had constructed a patch panel with telephone cables of different lengths to emulate some worst-case topologies. They found that the channel dispersion becomes a major impairment when the pulse-signaling rate approaches 200 kHz. To carry more information bits per signaling symbol while avoiding extensive signal processing, the PPM was used. In a PPM system, the time interval between adjacent pulses is slightly different depending upon the encoded information. The next pulse's position starts when the reflection becomes negligible. A number of next pulse positions are located to represent data bits to be transmitted. Two positions can be used to carry 1 bit, four positions carry 2 bits, eight positions carry 4 bits, and so on. Furthermore, Tut Systems' HomePNA 1.0 proposal has used the Run Length Limited (RLL) codes to increase the coding efficiency by about 10%. 7.1.1 Summary of HomePNA 1.0 SpecificationsThe HomePNA 1.0 signaling method can be analyzed by examining its general frame structure as shown in Figure 7.1. A HomePNA 1.0 transmitter encapsulates the binary information of an Ethernet packet by adding a header to it. The header consists of eight synchronization symbols followed by two data training symbols and a Proprietary Communication (Reserved), or PCOM, period. This header replaces the preamble and Start Frame Delimiter of the Ethernet packet. Symbol 0 is the null synchronization symbol consisting of two pulses. Symbols 1 through 4 carry transceiver Access ID (AID) information. With each AID symbol carrying 2 bits, a total of 8 bits are allocated for AID. This leads to 256 different AID combinations. Symbols 5 and 6 are used for transmitting remote control management commands across the network. Symbol 7 is a silence interval. On the other hand, many more data bits are encoded into an equivalent synchronization symbol interval. Because of the run length encoding, there is no exact timing boundary for every data-bits-carrying symbol. The end symbol is a nonvalid data symbol with the pulse position anywhere beyond those defined for data symbols. Training symbols and the PCOM period (all zeros by default) are defined by operation procedures through the management layer. Figure 7.1. HomePNA 1.0 Frame Structure (From HomePNA specification 1.0. Copyright © 1998 HomePNA. All rights reserved.)
This frame structure is established on a timing base called TIC (Time Interval Clock). A TIC is equal to 7/60 x 10 6 seconds. Each synchronization symbol occupies a fixed time period of 129 TICs. Synchronization symbol zero has one pulse at TIC = 0 and a second pulse at TIC = 126. For AID symbols 1 to 6, two bits are encoded for each symbol. Pulse positions for AID symbols are defined 20 TICs apart. Pulse position 1 of an AID symbol is at TIC 66 representing the bit combination of 00, pulse position 2 is at TIC 86 representing 01, pulse position 3 is at TIC 106 representing 10, and pulse position 4 is at TIC 126 representing 11. Two transmission throughputs, high rate and low rate, are defined for the HomePNA 1.0 by different data symbol pulse starting positions. The same header structure and synchronization symbols are used for both rates. In contrast, pulse positions for data symbols are defined only one TIC apart. There are 32 (0 31) pulse positions defined for the purpose of data symbol encoding. Pulse position 0 is defined at TIC 28 and TIC 44 for high rate and low rate, respectively. The next position is one TIC away (i.e., pulse position 1 is at either TIC 29 or TIC 45). The counting of TICs starts from the time of the previous pulse position. If all 32 pulse positions are used, 5 bits of information can be carried by each symbol with the conventional PPM encoding method. However, only pulse positions 0 to 24 are valid for data encoding. This results in a slightly lower transmission throughput (i.e., 4.64 bits per symbol). In practice, a procedure similar to that of run length coding for data storage is used to encode a fractional number of information bits. A Run Length Code (RLC) is a conceptually simple form of compression. An RLC consists of the process of searching for repeated runs of a single symbol in an input stream and replacing them by a single instance of the symbol and a run count. For the HomePNA 1.0 case, up to three consecutive zeros can be counted by using three different groups of positions for subsequent bit encoding. Right after a group of previously coded bits, if the first bit is 1, then the next three bits are coded using positions 1 through 8. If the first bit is zero and the second bit is one, then the next three bits (third, fourth, and fifth) are encoded using positions 9 through 16. If the first and the second bits are zeros and the third is one, then the subsequent three bits (fourth, fifth, and sixth) are encoded using positions 17 through 24. If the first three bits are all zeros, position 0 is used. The pulse position representations of binary bit sequences are summarized in Table 7.1.
We can calculate the average number of bits each data symbol carries. For half of the time, the first bit is 1 and the next three bits are encoded into a symbol for a total of 4 bits. For a quarter of the time, the first two 01 bits and the next three bits are encoded for a total of 5 bits. For one eighth of the time, the first three 001 bits and the next three bits are encoded for a total 6 bits. For the remaining one eighth of the time, the first three 000 bits themselves are encoded for a total of 3 bits. Therefore, the average number of data bits a data symbol can carry is Equation 7.1
We can also calculate the average length of a data symbol. The average pulse position for 4 bits encoding (1 followed by 3 bits) is 4.5 TICs away from pulse position zero. The average pulse position for 5 bits encoding (01 followed by 3 bits), is 12.5 TICs away. The average pulse position for 6 bits encoding (001 followed by 3 bits) is 20.5 TICs away. The average symbol length in number of TIC is therefore Equation 7.2
The average symbol duration in time is then Equation 7.3
The transmission throughput is Equation 7.4
To carry 4.375 bits using the conventional PPM, 24.375 = 20.75 pulse positions are required for an average longer symbol length of 28 + 20.75/2 = 38.375 TICs. On the other hand, the average symbol length of 35.9375 TICs can only carry log2[(35.9375 28) x 2] = 3.989 bits by the conventional PPM code. The coding efficiency improvement provided by the RLL code over the conventional PPM code is (4.375 3.989)/3.989 = 0.0967 = 9.67%. Similarly, for low rates we have the following. The average symbol length in number of TICs is Equation 7.5
The average symbol duration is Equation 7.6
The transmission throughput is Equation 7.7
Assuming every node on the HomePNA 1.0 network can generally hear each other, the collision detection is only performed during AID and silent intervals (AID symbols 0 through 7). During a collision, a transmitter reads back an AID value that does not match its own and recognizes the event as a collision, alerting other stations with a JAM signal. A JAM pattern consists of 1 pulse every 32 TICs and continues until at least the end of the AID intervals. When a transmitter receives pulses in a position earlier than the position it transmitted, it recognizes it as a pulse transmitted by another transceiver and signals a collision. Guaranteed collision detection is possible only as long as the spacing between successive possible pulse positions in an AID symbol (20 TICs or 2.3 µs) is greater than the round trip delay between the colliding nodes. At approximately 1.5-ns propagation delay per foot, the maximum distance between two HomePNA 1.0 transceivers must therefore not be greater than 500 ft for collision detection purposes. A HomePNA 1.0 pulse can be generated by passing four cycles of a 7.5-MHz square wave through a 10th-order Butterworth filter with a passband between 5.5 and 9.5 MHz as shown in Figure 7.2. The peak voltage level is defined as 1.2 and 0.6 V for high and low power versions, respectively, with a tolerance of about 15%. Figure 7.2. Symbol Waveform
Figure 7.3 shows the PSD of the HomePNA 1.0 low-power version. The high-power version's PSD is 6 dB higher between 5.5 and 9.5 MHz as a result of the doubled voltage level. Figure 7.4 shows a HomePNA 1.0 test wiring configuration as a typical in-house wiring channel model. Figure 7.5 shows the frequency response of this channel model over a frequency range of 20 MHz. At frequencies above 5 MHz, the attenuation is between 20 and 30 dB for a configuration with total wiring length less than 700 ft. These attenuations are caused by branching and reflection losses. A deep frequency notch caused by the coincidence of many reflections has a heavy loss of more than 60 dB at around 4 MHz for this particular wiring configuration. Figure 7.3. PSD of HomePNA 1.0
Figure 7.4. A Test Wiring Configuration (From HomePNA specification 1.1. Copyright © 1998 HomePNA. All rights reserved.)
Figure 7.5. Channel Frequency Response
Because only one pair of the in-house wiring is used and one transceiver is allowed to transmit at one time, no NEXT or FEXT noises are experienced at the front end of a receiver. The strength of the received signal is relative only to the background noise with a PSD of 140 dBm/Hz. Matching the telephone wall jacket, the RJ11 plug is used to connect a HomePNA 1.0 transceiver to the in-house wiring. Following the telephony convention, the wire pair connected to the middle two pins (pins 3 and 4) is used for transmission as specified by the HomePNA 1.0 standards. 7.1.2 Transceiver Structure and Performance EstimationA HomePNA transceiver consists of a bandpass filter, a transmitter, and a receiver, as shown in Figure 7.6. Because the transmission is not full duplex, the same passive bandpass filter is used by both the transmitter and the receiver. Within the transmitter, there are an encoder function and a preamble function. Information bits are first transformed into different RLL symbols, and then a predefined preamble is attached to every sequence of symbols. The synchronization function of the receiver uses the preamble of a signal sequence. To make the detection of the preamble and subsequent symbols reliable, a special detection function is defined in the receiver. The detection function can identify the average noise level, the peak signal level, and the proper data detection threshold. The synchronization circuit in the receiver identifies the exact pulse position of each AID symbol. Once synchronized, the reference timing is used for subsequent data symbol position identification. The received sequence of symbols is then decoded. Figure 7.6. HomePNA 1.0 Transceiver Structure
The purpose of the bandpass filter is to shape the transmit signal according to the defined PSD, as shown in Figure 7.3. These frequency characteristics of the bandpass filter will help to minimize band noise in the receive path. The in-band PSD level is 62 dBm/Hz. This PSD level is designed to achieve satisfying transmission performance over average condition in-house wiring while minimizing interference to other transmission systems such as those in the Amateur Radio band. For in-house wiring with higher attenuation, this PSD level can be raised to 56 dBm/Hz. Because the symbol is a modulated passband signal, an envelop detection circuit is used for all threshold identification and data detection purposes. The performance of a HomePNA 1.0 transmission system can be analyzed by examining the Signal-to-Noise ratio at the receiver end. The received signal level is that of the transmit signal subtracted from the attenuation level of the channel. The nominal PSD of a HomePNA 1.0 signal is at 62 dBm/Hz. The noise floor of in-house wiring is assumed to be at around 140 dBm/Hz for an average home. On the other hand, the duration of the symbol in comparison to the average symbol rate and the rise time of a symbol in comparison to the TIC interval could also affect the performance. If the duration of the symbol is longer than the average symbol interval, the tail of the previous symbol might touch the beginning of the current symbol, reducing the effective SNR. If the rise time of the symbol is slower than that of a TIC interval, the beginning of a symbol might cover multiple symbol timing positions and, therefore, also reduce the effective SNR. Figure 7.7 shows the received signal in time domain along with the transmitted signal scaled down by 10 times in magnitude. In comparison, the received signal is attenuated by about 20 dB. The pulse duration is about 3 µs compared with about 2 µs for the transmitted signal. The extra duration is caused mainly by reflection. To examine the intersymbol interference, we compare the symbol duration with the average symbol period. The average symbol period is about 6 µs. The reflection has died out before the next symbol arrives. To examine the accuracy of symbol position detection, we look at the pulse rise time in comparison with the TIC interval. The rise time is about 0.6 µs. Because the TICs are 7/60 µs apart, rising peaks are well separated. Figure 7.7. Received Signal
Since the effect of intersymbol interference is minimal as a result of the long separation between adjacent symbols and the rise time of each symbol is shorter than the TIC interval, the transmission performance is very much decided based on the receiver front-end SNR. For a channel loss of 30 dB, the receiver front-end SNR is Equation 7.8
The receiver front-end SNR reduces to 28 dB for a channel loss of 50 dB. The channel capacity of the in-house wiring environment against the background noise for the HomePNA 1.0 frequency band of between 5.5 and 9.5 MHz is calculated for an SNR of 48 dB as Equation 7.9
The channel capacity for a SNR of 28 dB is then Equation 7.10
For a threshold detection based transmission system, a communication error occurs when the magnitude of the noise exceeds the detection threshold. This condition can be described by Equation 7.11
where a is the detection threshold and s2 is the noise power. A multiplication factor of 2 is used to account for both positive and negative noise elements. If we choose a to be the RMS value of the received signal (i.e., a2 to be the received signal power), then we have Equation 7.12
By replacing the variable x with x/s in Equation 7.11 we have Equation 7.13
For SNR = 48 dB, we have SNR Equation 7.14
For SNR = 28 dB, we have SNR
If for some reason, such as a higher receiver front-end electronic noise level combined with a higher channel attenuation, the SNR dropps to 15 dB, and we then have SNR
This analysis shows that a HomePNA 1.0 transmission system can operate reliably under a relatively low SNR unless effects of extensive reflections come into play in forms of intersymbol interference or a retarded rising time. This analysis is based on background white noise. Impulse noise with a magnitude close to that of the received signal can also cause transmission errors. The original in-house wiring based transmission proposal from Tut Systems for the HomePNA 1.0 is a baseband system. Each symbol was a continuous pulse (instead of a sequence of positive and negative pulses with a period of about 0.1333 µs). Figure 7.8 shows the received baseband pulse compared with the transmitted symbol formed by sending a rectangular pulse through a low-pass filter. The transmitted and received symbols have the same scale. Over the same test wiring configuration, the baseband symbol is only slightly attenuated. The peak attenuation is only about 2 dB compared with the passband attenuation of about 20 dB. The move of the spectrum to a passband with a center frequency of 7.5 MHz was necessary for spectrum compatibility with Asymmetrical Digital Subscriber Line (ADSL). Because the attenuation is much lower, the baseband method should have a better performance in terms of either a higher transmission throughput or a longer transmission distance. Figure 7.8. Baseband Received Waveform
7.1.3 HomePNA and MAC InterfaceThe HomePNA 1.0 allows Ethernet packets to be transported over existing in-house telephone wiring, with no modifications, using the standard Ethernet CSMA/CD-based Media Access Control procedures as specified in the IEEE 802.3 standard. A HomePNA 1.0 physical layer to 802.3 MAC interface is defined as shown in Figure 7.9. Figure 7.9. HomePNA 1.0 and MAC Interface (From HomePNA specification 1.1. Copyright © 1998 HomePNA. All rights reserved.)
Much like the MII, a transmit path and a receive path are defined separately with transmit clock, TxClk, and receive clock, RxClk, both from the physical layer. In contrast, only a single bit, instead of 4 bits, is defined for both transmit and receive data paths and the Receive Data Valid, RX_DV, pin is omitted. There are also a Carrier Sense pin, CarSns, and a Collision Detection pin, Coll, defined for the HomePNA 1.0 to MAC interface. There are no management pins defined over this interface. A HomePNA 1.0 transceiver can be managed by either remote control word management commands embedded in the AID header over the wire network or management messages from attached host hardware. |
63095, 
, and the probability of error is
, and the probability of error becomes
, and a probability of error of
EAN: 2147483647
Pages: 97