Pulse or digital communications – Transceivers – Modems
Reexamination Certificate
2000-08-22
2004-02-03
Ghebretinsae, Temesghen (Department: 2631)
Pulse or digital communications
Transceivers
Modems
C375S346000, C370S290000, C379S417000
Reexamination Certificate
active
06687288
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to modem telecommunications in general, and more particularly to NEXT cancellation for modem pools.
BACKGROUND OF THE INVENTION
Near-end cross talk (NEXT) is defined as the cross talk interference between the receiving path and the transmitting path of different transceivers that make use of wiring that share the same cable. The NEXT effect in a cable depends on the number of interfering lines, and increases as the bandwidth that the signals occupy increases. In a modem pool environment where streams of data are distributed to many lines within a single cable, the NEXT that the receivers need to overcome is mainly generated by the transmissions that the modem pool itself generates. Since such a system has access to the transmitted information for a plurality of modems, such information may be used to cancel the interference that leaks into the receivers, thus increasing the noise floor of each receiver.
In classic NEXT cancellation, a transmitter transmitting via one wire or wire grouping (e.g., twisted pair) affects the receiver receiving via another wire or wire grouping. A hybrid circuit separates the received signal from the transmitted interfering signal, but since the hybrid cannot completely separate the transmit path from the receive path, some of the transmitted signal leaks into the receiver and becomes an interfering signal. A canceller then filters out the effect of the interfering signal, resulting in a “cleaned” received signal. For a single modem, this problem may be addressed using classic echo cancellation techniques. In a modem pool environment where several modems transmits via a shared cable, the canceller for each receiver must take into account all the interfering transmitters. Thus, in order to cancel the NFM resulting from several modems in addition to each modem's echo, one approach might to apply echo canceling techniques to each transmitting modem/receiving modem pair. However, such a solution would be relatively complex to implement, since the number of filters required would equal the square of the number of modems in the pool. A system implementing such a solution is described in U.S. Pat. No. 5,970,088.
There are two types of echo cancellation solutions: static and adaptive. In static echo cancellation, the echo canceling filter is adjusted during a learning phase, and the adjusted filter is then applied at steady state. In adaptive echo cancellation, the effect of each interferer is be studied over time using an adaptive filter which continually adjusts the signal to be cancelled out in accordance with changing line conditions in order to achieve optimal cancellation. One well-known adaptation method is the least mean squares (LMS) adaptive filter in which the “cleaned” signal is multiplied by a function of the interfering signal and is integrated over time, resulting in only the “non-cleaned” part of the interfering signal remaining. This remaining signal represents the remaining energy of interference from a single transmitter. Both static and adaptive filters are estimates of the transfer function that a transmitter passed on the way to becoming an interferer.
Many static and adaptive echo canceling algorithms employ Finite Impulse Response (FIR) filters as the echo canceling filter. For a single NEXT canceller employing a FIR filter with N taps in the time domain, computing each output sample of a transmitted signal through the FIR filter requires approximately N multiplications and N additions. In a modem pool environment of M modems where any given receiver may be affected by a number of interfering transmitters, as well as by its own echo, up to M finite impulse response (FIR) filters may be required for each modem in order to cancel the NEXT from all interferers. The maximum number of arithmetic operations required for computing all output samples at a given time would therefore be of the order of N·M·M=N·M
2
.
It is well known that time-domain filtering may be alternatively implemented in the frequency domain for a single modem. In this technique, the coefficients of a FIR filter are transformed to the frequency domain using techniques such as Fast Fourier Transform (FFT), thereby generating frequency domain coefficients. A stream of incoming time samples is divided into blocks. Each block of time samples is then modified, such as by zero padding the block or by concatenating some samples from a previous block to the current block. These modified blocks of samples are transformed to the frequency domain using techniques such as Fast Fourier Transform (FFT), thereby generating blocks of frequency samples. Each block of frequency samples is then multiplied coordinate-wise with the frequency domain coefficients, and the results are transformed back into the time domain using techniques such as inverse FFT to generate a second sequence of blocks of time domain samples. These time domain samples are then processed, such as by considering only certain samples from each block, or by overlapping and adding adjacent blocks, to ultimately arrive at a sequence of samples which represents the same sequence that would have been encountered were the original stream of incoming samples convolved with the FIR filter coefficients. A detailed description of this technique can be found in J. G. Proakis & D. G. Manolakis, “Digital Signal Processing Principles, Algoritbms, and Applications, 2nd Edition,” Macmillan Publishing Co., 1992, pp. 703-709, where two well known variants of block overlapping, “overlap add” and “overlap save,” are presented.
Computing N output samples for a single FIR of N taps would then require the following computations: 2 FFTs of length N (or 2N, if zero padding is applied) for transferring to the frequency domain and back to the time domain, and N complex multiplications for effecting the filter in the frequency domain. The complexity of each FFT operation, as expressed in terms of the number of multiplication and addition operations, is on the order of Nlog(N), and the complexity of effecting the filter in the frequency domain is N. Thus, the total complexity for computing N output samples is on the order of Nlog(N)+N, resulting in a total complexity per output sample of log(N)+1.
The use of frequency domain techniques for NEXT canceling is described in U.S. Pat. No. 5,887,032. However, the system described in U.S. Pat. No. 5,887,032 is a DMT system in which frequency domain techniques are applied on a modem by modem basis. No mention is made of a specific application in a modem pool environment, and the techniques described would be applied in exactly the same manner for one modem as they would be for M modems. Moreover, the NEXT cancellation method described by U.S. Pat. No. 5,887,032 is disadvantageous in several respects which are now described.
In the DMT NEXT cancellation system a cyclic prefix is added to each signal block prior to its transmission. For example, if a signal block has a length of 512 bytes, and the cyclic prefix has a length of 32 bytes, the transmitted signals will comprise blocks having a length of 544 bytes, only 512 bytes of which contain information symbols, with the remaining bytes considered a redundancy to be removed at the receiver. Typically, the processing of the received signal for DMT systems is done by:
1. Converting the received signal from analog to digital (A/D converter).
2. Applying an optimized filter known as a Time Domain Equalizer (TEQ) to find an optimal frame (e.g., of 512 bytes out of the 544 bytes in the preceding example) in which the effects of Inter Symbol Interference (ISI) and Inter Channel Interference (ICI) are minimized.
3. After finding the optimal frame, removing the cyclic prefix, leaving only the signal samples in the frame.
4. Transforming the signal samples to the frequency domain.
5. Applying the NEXT cancellation algorithm, including multiplying the data transmission coefficients (which are given in the frequency domain in DMT systems) by the complex coefficients of an adaptive Filter.
6.
Actelis Networks Inc.
Eitan, Pearl, Latzer & Cohen Zedek LLP
Ghebretinsae Temesghen
LandOfFree
NEXT cancellation for modem pools does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with NEXT cancellation for modem pools, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and NEXT cancellation for modem pools will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3302800