Sunghwan Ong

A decision feedback recurrent neural equalizer as an infinite impulse response filter

An adaptive decision feedback recurrent neural equalizer (DFRNE), which models a kind of an IIR structure, is proposed. Its performance is compared with the traditional linear and nonlinear equalizers with FIR structures for various communication channels. The small size and high performance of the DFRNE makes it suitable for high-speed channel equalization.

Performance Of Neural Equalizers On Partial Erasure Model

The increase in the capacity of the digital magnetic recording systems inevitably causes severe intersymbol interference (ISI) and nonlinear distortions in the magnetic channel. In this paper, to cope with severe ISI and nonlinear distortions a neural decision feedback equalizer (NDFE) is applied to the digital magnetic recording channel-partial erasure model. In the performance comparison between the NDFE and the conventional decision feedback equalizer (DFE) via simulations, it has been found that as nonlinear distortions increase the NDFE has more SNR (Signal to Noise Ratio) advantage over the conventional DFE.

A quadratic sigmoid neural equalizer for nonlinear digital magnetic recording channels

A new neural equalizer is proposed in order to compensate for intersymbol interference and to mitigate nonlinear distortions in digital magnetic recording systems. The proposed equalizer uses the quadratic sigmoid function as the activation function. The performance of the proposed equalizer is compared to those of a decision-feedback equalizer (DFE) and a neural decision feedback equalizer (NDFE) in terms of bit-error rate in nonlinear digital magnetic recording channels. Simulation results demonstrate that the proposed equalizer outperforms both DFE and NDFE.

A complex version of a decision feedback recurrent neural equalizer as an infinite impulse response filter

A decision feedback recurrent neural equalizer (DFRNE), a nonlinear infinite impulse response (IIR) filter, has already been proposed successfully for the channel equalization problem (Ong et al. 1997). In this paper, we introduce a complex version of a DFRNE. Its small size and high performance makes it attractive for channel equalization. We evaluate its performance through computer simulations for various complex channels and derive that its convergence is more stable and robust than that of other equalizers. It is also shown that the complex DFRNE outperforms other equalization schemes when the channel distortion is profound.

Partial response maximum-likelihood system and crosstalk cancellation method for high-density optical recording

Methods for partial response maximum-likelihood (PRML) de- tection and crosstalk cancellation are discussed. In accordance with the demand for increased recording density, a 20 Gbyte digital versatile disk (DVD) ROM channel is considered. The channel is modeled to be close to a real optical channel using DIFFRACT™. After comparing the spec- tral characteristics of various PR polynomials, P(D)511D1D 2 1D 3 is proposed as a target PR. The performance of the system is illustrated under the condition that the readout signal is degraded by crosstalk, radial tilt, and nonlinear distortions in optics. Finally, we present the mul- tispot detection technique for crosstalk cancellation. The simulation re- sults show that if no radial tilt is present, crosstalk is fully cancelled.

Equalization techniques using neural networks for digital versatile disk-read-only memory systems

Several equalizers are applied to digital versatile disk-read- only memory (DVD-ROM) systems. As a result of an imperfect writing process, nonlinear distortion may be generated in the replay signal. Neu- ral equalizers, which have strong nonlinear mapping capabilities, are applied to the system to compensate for the nonlinear distortion. We perform computer simulations for two systems: a DVD-ROM system with and without run-length-limited (RLL) modulation code. Nonlinear distor- tion such as domain bloom is the dominant interference of the channel in a DVD-ROM system with RLL code because this code reduces the linear intersymbol interference (ISI). Thus the experimental results with RLL code show that the signal-to-noise ratio (SNR) gain of equalizers using neural networks over conventional equalizers increases in proportion to the nonlinearity of the DVD-ROM channel. When RLL is not used, how- ever, linear and nonlinear ISI is the dominant interference of a DVD- ROM channel. Therefore equalizers with a decision-feedback section in a DVD-ROM system with RLL code outperform equalizers without a decision-feedback section in eliminating ISI. Consequently among the equalizers in our computer simulations, the neural decision-feedback equalizer (NDFE) showed the best bit error rate (BER) performance.

A decision feedback recurrent neural equalizer for digital communication

In this paper, we introduce an adaptive decision feedback recurrent neural equalizer (DFRNE) whose small size and high performance makes it suitable for high-speed channel equalization. By evaluating its performance through computer simulations for various channels, the DFRNE has comparable performance with traditional equalizers when the channel interferences are mild; it outperforms them when the channels transfer function has spectral null or when severe nonlinear distortion is present. In addition, the DFRNE, being essentially an IIR filter, is shown to outperform multilayer perceptron equalizers in linear and nonlinear channel equalization cases.

Performance enhancement of 1/spl times/EV/DV MIMO systems in frequency selective fading channels

In this paper, we evaluate the performance of 1/spl times/ evolution data and voice (1/spl times/EV/DV) systems with multiple antennas at both the transmitter and the receiver. The multicode interference (MCI) in 1/spl times/EV/DV systems is hard to compensate in multipath fading channels. Especially, the effect of MCI is more severe when using multiple antennas. We can improve the receiver performance by using multipath interference cancellation (MPIC) scheme. Simulation results show that 1/spl times/EV/DV systems with the multiple antennas and the MPIC scheme can achieve higher data rate than single antenna system in the frequency selective fading channel.

An Error-controlled Detection Method For PR-IV Magnetic Recording Systems

The major advantage of PR-IV channel modeling on high-density magnetic recording systems is that the maximum-likelihood sequence detector (MLSD) with optimal performance can be used. However, the MLSD has a complex structure and a high cost. Another detection method is the three-level detector (TLD). Its structure is simpler than the MLSDs, but its performance is extremely poor. Sometimes it is necessary to use a detection method that has a simpler structure than the MLSD, yet performs competitively. In this paper, a new detection scheme called an error-controlled detection (ECD), is proposed for PR-IV systems. Its hardware complexity is greatly reduced and the performance is competitive with the MLSD.

Blind direct DFE estimation using multistep prediction

We propose a blind decision feedback equalizer (DFE) that is characterized by the fact that it does not require channel estimation. Because the output of the optimized multistep prediction error filter (PEF) can be represented as a product of the channel partial impulse response and the transmitted sequence, a backward multistep PEF can be used as the blind DFE feedforward filter. The corresponding feedback filter is obtained from the symbol-rate partial channel impulse response. The proposed algorithm has several advantages over existing blind channel estimation techniques, including stable performance without the necessity of exact channel order estimation.