Lower-complexity peak-to-average reduction using...

Details Lower-complexity peak-to-average reduction using... Lower-complexity peak-to-average reduction using...

1998-09-28

2001-05-29

Pham, Chi (Department: 2631)

Pulse or digital communications

Systems using alternating or pulsating current

Plural channels for transmission of a single pulse train

C708S404000

Reexamination Certificate

active

06240141

ABSTRACT:

FIELD OF THE INVENTION
This invention relates to telephone systems, and more particularly to peak-to-average ratio (PAR) reduction for Digital-Subscriber Lines (DSL).
BACKGROUND OF THE INVENTION
High-bandwidth technologies that use the existing copper-cable telephone lines such as Digital-Subscriber Lines (DSL) are now becoming available. One type of DSL can provide bandwidth up to 8 Mbps downstream, or up to 2 Mbps symmetric. DSL approaches the bandwidth of T
1
lines, about 1.5 Mbps. Several variations of DSL technology are being explored, such as HDSL, IDSL, SDSL, ADSL and VDSL. ADSL (asymmetric DSL) is particularly attractive for consumer Internet applications where most of the data traffic is downloaded to the customer. Upstream bandwidth for uploading data can be reduced to increase downstream bandwidth since most Internet traffic is downstream traffic. See U.S. Pat. Nos. 5,461,616, 5,534,912, and 5,410,343 for descriptions of ADSL technology.
DSL Modem—
FIG. 1
FIG. 1
shows a DSL modem. Data from a personal computer or other equipment at the customer premises is sent to transmitter
10
. Transmitter
10
arranges the data into frame packets and symbols using techniques such as trellis encoding or quadrature-amplitude modulation (QAM). The symbols are converted from digital to analog signals by digital-to-analog converter (DAC)
12
. The analog signals then pass through hybrid
14
and the telephone line to the central office (CO). The signal from the central office is separated from the locally-transmitted signal by hybrid
14
and converted from analog to digital signals by analog-to-digital converter (ADC)
16
. Receiver
18
detects the alignment of frames and decodes the symbols. The decoded data is sent to the local data equipment such as a computer. For greater detail see “DSL Simulation Techniques and Standards Development for Digital Subscriber Line Systems”, by Walter Chen (Macnillian, 1998).
Multi-Tone Signaling—
FIG. 2
Rather than use single carrier, multiple carriers can be used. Separate frequencies or “tones” can each carry a portion of the data. The amplitude and phase of each carrier can be modulated to carry the data. This technique is known as discrete multi-tone (DMT).
FIG. 2A
shows a frequency spectrum of a discrete-multi-tone DSL signal. The transmitted signal is composed of several separate carriers that are sufficiently separated in frequency so that the data from separate carriers can be extracted. The carriers are separated into frequency bins
0
,
1
,
2
,
3
, . . . N−1. Each frequency bin contains one carrier wave. Thus a total of N carriers provide N simultaneous data channels.
The N carriers in N frequency bins are combined by the transmitter into a single output signal that contains information from all N carrier waves. Since the output signal is real, the N input carriers have complex-conjugate symmetry. The (N−i)-th carrier is the complex conjugate of the i-th carrier, for i=1,2, . . . N/2−1. Therefore, only about half of the N carriers are independent carriers that carry information.
FIG. 2B
shows a multi-carrier output signal to a DSL telephone line.
FIG. 2B
is a time-domain signal derived from the frequency-domain representation of FIG.
2
A. An inverse Fourier transform is used to convert the N frequency bins of data to a time domain signal for transmission. Typically an inverse fast-Fourier transform (IFFT) is used to accelerate calculations to real-time speed.
Clipping Problem
The average power of the transmitted signal is related to the root-mean-square (RMS) average of the signal. This RMS voltage can be increased to reduce the effects of line interferences, but the specification puts a limit to the average transmit power. Some of the samples of the time-domain signal are below the RMS voltage while other samples are above the RMS voltage. For the same average power, multi-carrier transmission can have larger peak voltages than do single-carrier signals. The increased peak voltage of DMT systems is a problem that can limit its effectiveness.
Peak voltages cause problems when they exceed the dynamic range of the digital-analog converters and the amplifiers such as the line driver. More expensive and power-consuming DACs and line drivers with a wider dynamic range may be needed.
FIG. 2B
shows the maximum voltage of the DAC as the dashed line marked “DAC PEAK THRESHOLD”. The DAC is able to accurately convert signals below this threshold, but signal peaks above this threshold cannot be accurately converted. The DAC or the line driver simply clips the signal to its maximum (peak threshold) voltage, which is less than the peak signal voltage. Peak
20
is then clipped or reduced in voltage.
Clipping is undesirable since some of the signal information is distorted. The ratio of the peak voltage to the RMS voltage is known as the peak-to-average ratio (PAR). The PAR depends on several factors, such as the line code used for symbol encoding, the number of carriers, and the number of signal levels used to encode the symbols. Filtering also affects the PAR, as does the channel (telephone line) and echo path for the received signal.
PAR reduction techniques are being developed for DSL systems.
FIG. 3
is a flowchart of a PAR reduction technique that uses subset-sign inversion. Such a PAR reduction technique has been proposed by Lucent Technologies and Centillium Technology for the International Telecommunication Union (ITU), Telecommunication Standardization Sector, study group
15
, in temporary document NF-074R2.
The signals for N carrier waves are input as X(N). An inverse fast-Fourier transform (IFFT) is performed on the X(N) frequency bins to produce a time-domain output x(N), step
41
. If none of the x(N) signals exceed the peak-voltage threshold, step
42
, then the time-domain output x(N) is transmitted since no clipping occurs.
When step
42
detects that one or more of the x(N) voltages exceeds the peak threshold, then the signal is modified to remove the peak. The X(N) inputs are divided into several is subsets that may or may not overlap. In step
45
, one of these subsets is chosen at random or in a predefined sequence. The chosen subset has the sign changed for all of its frequency bins.
The modified inputs X(N) are then re-input to step
41
and the IFFT re-executed. Often the sign inversion reduced the largest peak voltage so that step
42
does not find any peaks exceeding the peak threshold. Otherwise, another subset is chosen in step
45
and its inputs are sign-inverted, and the IFFT again re-executed in step
41
. Several iterations may be needed before all peaks are within the threshold. Once all possible subsets have been inverted without successfully reducing the peak voltage below the peak threshold, then the x(N) outputs are transmitted with clipped peaks. But with PAR reduction, the probability of clipping is greatly reduced.
The X(N) input is divided into several subsets, and some of the data bits are used to indicate which subsets have been inverted. The receiver reads these subset-sign-inversion bits and inverts the indicated subsets after FFT operation.
While such a subset-sign-inversion technique is effective, the added complexity for PAR reduction is relatively high since in each iteration we have to repeat the entire IFFT operation, which is complex and lengthy, requiring many multiplications and additions. Often, several iterations may be needed. The entire IFFT is re-executed as each way of subset inversion is tried. The subset is chosen at random rather than being chosen as a function of the location (time coordinate) and the value of the peak over the threshold.
What is desired is an efficient and effective peak-to-average reduction technique for multi-tone DSL. It is desired to reduce the probability of clipping of peaks with low computational complexity. It is desired to avoid expensive and power-consuming digital-to-analog converters and line drivers with wider dynamic ranges. A PAR-reduction technique that uses the inherent properties of the IFFT to reduce calculati

Affiliated with

Also associated with

Rating

Lower-complexity peak-to-average reduction using... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Lower-complexity peak-to-average reduction using..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Lower-complexity peak-to-average reduction using... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2550311