Chapter 4: Overview of Simulation Program for Data ...



Performance Evaluation of a Novel Compacting Code System and Digital Subscriber Line System under Gaussian Noise and Intersymbol Interference

Mazen Alkhatib1, M. S. Alam2, and K. S. Tai2

1Department of Information Technology, Colorado Technical University, Malkhatib@faculty.ctuonline.edu

2 Department of Electrical and Computer Engineering, University of South Alabama, malam@southalabama.edu

Abstract

In this paper, we investigated the error level performance of fast digital subscriber line (VDSL2 Plus) under Gaussian noise and intersymbol interference. We also investigated the error level performance of a novel intersymbol interference (ISI) compacting code, and we compared the performance of this new code to DSL systems. Simulation results show that the performance of the proposed novel ISI compacting code system is significantly better than the performance of VDSL2 Plus system for signal compression above 40%.

1. Introduction

Bit error rate plays the most critical role toward effective data transmission in any communication system. Current communication systems require very high speed data transmission rates. This requirement is challenged by the presence of intersymbol interference (ISI) and other sources of noise. To achieve ultrahigh bit transmission rates an efficient pulse shaping and energy usage scheme can be used as reported in [1] and [2]. It is also possible to tackle the ISI challenge by treating the hazards resulting from the effect of ISI as the bit transmission rate is increased. Various studies have been reported in the literature during the last few years [3-8] to calculate the probability of error under Gaussian noise and ISI.

VDSL2 Plus is a standard that was approved in 2006 [10] for fast transmission of digital information at speeds up to 200 Mbps. However, it is difficult to achieve such a high speed in the presence of Gaussian noise and intersymbol interference (ISI). To overcome this problem, recently we designed a novel code [9] that is capable of reducing the effect of ISI. This code depends on searching for code words that are mapped to signals in a manner to minimize the effect of ISI.

In this paper, we developed detailed simulation models to measure the bit error rates (BER) of VDSL2 Plus systems and we compared it to the BER of our coding system in the presence of Gaussian noise and ISI. We verified the effectiveness of the proposed model for different compression ratios and measured the corresponding BERs.

The rest of the paper is organized as follows. Section 2 describes the simulation of the VDSL2 Plus system, Section 3 deals with the simulation for the novel code system. Section 4 shows the results obtained, and section 5 includes the concluding remarks.

2. VDSL2 Plus System Simulation Model

The proposed simulation model for the VDSL2 Plus system consists of 4-bit symbols. Ten such symbols are called a Discrete Multi-Tone (DMT) symbol. The simulated system consists of a transmitter that has Trellis Coded Modulation (TCM) 3-bit generator, and a mapper to a 4-bit output that is connected to an array of 4 parallel queues. The output of these queues is connected to four Binary Phase Shift Keying (BPSK) modulator subsystems.

The receiver consists of BPSK demodulators, Viterbi decoder, and comparators. Figure 1 shows the structure of the simulated transmitter model, and Figure 2 shows the TCM diagram where the Viterbi algorithm was used at the receiver end. As shown in Figure 1, at the transmitter, three bits enter the TCM sub-module which are then mapped into 4 bits (V1V0W1W0). These bits are then queued in four parallel queues, and are BPSK modulated for transmission using four different carrier signals at different frequencies.

Once the signals are transmitted, the simulation model adds Gaussian noise and ISI components to these signals. The Gaussian noise is generated using the Box-Muller technique [7], which uses inverse transformation to turn two uniformly distributed random variable U1 and U2 into two unit normal random variables X and Y in the interval (0, 1). The random variables U1 and U2 are generated in such a way that U1 ≠ 0. The unit normal random variables X and Y can be defined as

[pic] (1)

and

[pic]. (2)

[pic]

Figure 1. Structure of the simulated transmitter model.

At the receiver end, signals are demodulated and then Viterbi decoded. The resulting digital pattern is then compared to the transmitted pattern to determine the bit error rate. Figure 2 shows a Trellis diagram for the Viterbi algorithm that was applied at the receiver end.

[pic]

Figure 2. Trellis diagram where Viterbi algorithm was applied at the receiver end.

The simulation program was designed to provide the capability to compress the generated pulses before transmission. This compression affects the ISI component added to the transmitted signal. Figure 3 shows a flow chart for the main simulation model used for implementing the VDSL2 Plus system.

[pic]

Figure 3. Flow chart of the simulation program model for the VDSL2 Plus system.

3. Novel coding to reduce ISI:

Intersymbol interference occurs due to overlapping of pulses as signals are compressed. As the result, the receiver will detect more than one pulse in a prescribed time period. The probability of receiving an erroneous pulse is denoted as[pic]. In order to reduce[pic], a new coding system is developed in which it can achieve [pic] that is less than the value calculated by the series method developed by Beaulieu in Reference 1. For the sake of simplicity, [pic] is measured in terms of BER in the simulation of the new code for transmission.

The pulse code is developed based on trinary signaling levels i.e., the pulse code belongs to the set {-1, 0, +1}. The series of bits (bit words) that contain values of {-1, +1} are mapped to series of pulses (pulse series) with certain coefficients of {-1, 0, +1} where some of the pulses are omitted to reduce the effect of intersymbol interference. Consider a word of [pic] bits which may lead to [pic] possible bit words. As a result, in the new coding system, [pic]pulse codes are required to map to [pic]-bits length words. Figure 4 shows the general structure of the novel coding system.

[pic]

Figure 4: General structure of the novel coding system.

The following criteria were used in generating the novel code [9]:

• Number of pulse coefficients within a series per code is[pic].

• Each coefficient within a pulse code series is selected from {-1, 0, +1}.

• Percentage of the zero coefficients (PZ) should be as high as possible.

• Percentage of the negative and positive coefficients (PN and PP) should be as equal as possible for each type.

• All hazard cases should be prevented as much as possible. These pattern are (+1, -1, 0) or (-1, +1, 0) or (+1, -1, +1) or (-1, +1, -1) or (0, +1, -1) or (0, -1, +1) or (+1, 0, +1) or (-1, 0, -1).

• Pulse code series starting with coefficients (0, 0) or (0, +1), (-1, 0) or (-1, +1) are not allowed.

• Pulse code series ending with coefficients (0, -1) or (+1, 0) or (+1, -1) or (-1, 0) or (-1, +1) or (-1, -1) are not allowed.

• Sequence Free Distance (FD) separation should be greater than or equal to minimum free distance MFD. Values of MFD can be 2, 3, and 4. FD is calculated as the absolute differences of pulse coefficients for any two pulses series in a code.

3.1. Novel code structure:

A detailed simulation software was designed and developed to simulate the new coding system. There are two sub-programs in this simulation model. The first sub-program generates the N pulse series or code words which will meet all the criteria as indicated previously. The second sub-program simulates the performance of transmission under the presence of intersymbol interference and Gaussian noise using the novel code words.

3.1.1 Code words generation:

Figure 5 shows the flow chart of the code words generation software model. This program model allows the user to input the number of signals and the number of included code words to be generated. In addition, the program model allows the user to set the number of allowed hazard cases within a code word, the number of minimum zero coefficients, and the minimum free distance in between code words. A value of 4 for the minimum free distance was used, a value of 0 for the minimum zero coefficients within a code word was used, and no hazard cases were allowed in the code words.

[pic]

Figure 5(a): Flow chart of code-generation program (Continued to Fig. 5(b))

[pic]

Figure 5(b): Continuation of flow chart of the code generation program.

Numerous code search trials were run to find a code that yields the best performance under Gaussian noise and ISI. A code of 64 words, where each code word is mapped to 12 Gaussian pulses representing 6 bits each was found to have the best performance.

3.1.2 Transmission modeling:

Gaussian pulses are used to transmit data which is encoded using the optimized code obtained using the procedure described in the previous section. Pulses are compressed at different levels to measure the performance in terms of the level of received erroneous bits. Soft decision decoding using the maximum likely-hood sequence detection principle is used to decode the received pulses into the original data bits. Figure 6 shows the flow chart for the novel code transmission modeling simulation.

[pic]

Figure 6: Flow chart of transmission simulation (Continued next page).

[pic]

Figure 6: Continuation of flow chart of transmission simulation.

4. Results and Discussion

A detailed simulation software was developed for the VDSL system with 10 stages utilizing a convolutional encoder at the transmitter and a trellis decoder (Viterbi decoder) at the receiver. Another detailed simulation software was also developed for the novel coding system utilizing trinary signaling levels with 64 code words with each code word consisting of 12 pulses representing 6 bits. Soft decision decoding using the maximum likelihood sequence detection principle is used to decode the received pulses at the receiver.

Gaussian pulses with various compression ratios ranging from 10% to 90% were used for transmission between the transmitter and the receiver. Gaussian pulses with signal-to-noise ratio of 15 dB were used to be comparable to the results of the research performed in [3]. The resulting bit error rates as a function of the compression ratio for the TCM and the novel code are summarized in Figure 7. From Fig. 7, it is evident that when pulses are compressed above 40%, there is a considerable degradation in the value of BER of the DSL/TCM system. For example, the BER rises sharply from 2.67x10-05 at a pulse compression ratio of 40% to 0.056283 for a compression ratio of 50%. Whereas for the case of the novel code, BER remains less than 1x10-7 until the pulse compression ratio approaches 60%.BER. Thereafter, the BER rises to 2.69x10-05, which corresponds to the best performance obtained under ISI and Gaussian noise. Table 1 shows the BER for DSL/TCM and new ISI compacting code system versus pulse compression percentage.

Table 1: BER for DSL/TCM and new ISI compacting code system versus pulse compression percentage.

|Percentage of pulse compression% |BER(DSL/TCM) |BER Novel Code |

|90 |2.27E-01 |2.56E-05 |

|80 |1.94E-01 |2.18E-05 |

|70 |1.60E-01 |1.79E-05 |

|60 |0.200943 |2.69E-05 |

|50 |0.056283 | ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download