Pulse or digital communications – Equalizers – Automatic
Reexamination Certificate
1999-05-19
2003-03-18
Tse, Young T. (Department: 2634)
Pulse or digital communications
Equalizers
Automatic
C375S222000, C375S348000, C370S292000
Reexamination Certificate
active
06535552
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to fast training of equalizers in DMT systems, and more particularly to multi-tone (multi-carrier) systems and the fast training of pre-equalizers for high-speed communications over severely distorting and/or severely long channel lines/loops.
BACKGROUND OF THE INVENTION
In the telecommunications industry, information is transmitted over imperfect communication lines such as copper wire, TV cables, fiber optics, twisted pair, and the like. Such transmissions are also made in imperfect conditions and environments. Generally, any and all communication channels, even wireless systems that transmit signals through air, have undesirable parasitic characteristics (e.g., interference, line resistance, line capacitance, signal reflection, etc.) and external influences (e.g., crosstalk from other communication sources). These parasitics and influences result in dispersion of a transmitted signal and the smearing of adjacent data values over one another in the time domain. Such a dispersion phenomenon, referred to as intersymbol interference (ISI), is illustrated in FIG.
1
.
FIG. 1
depicts a communication line or communication channel
10
. Inside the box representing the channel
10
is depicted a time domain impulse response
12
for a typical channel
10
. Specifically, if the channel
10
were subject to a transmitted impulse input &dgr;(t) of very short time duration (e.g., lasting for a time period of only one sample interval at a given sampling frequency ƒ
s
) the distorted response
12
would result at the receiving end of the channel
10
. Due to the undesired parasitics of the channel
10
and external interference, response
12
is a signal that is spread over &ngr; sampling intervals where &ngr; is greater than the one sample interval in which the transmit signal &dgr;(t) was initially contained. In short, response
12
indicates that the energy produced at the end of the channel
10
in response to a short-time duration impulse at time
0
is a smeared and distorted response that spans v sampling intervals. Such dispersion of a signal is common in all communications channels under various conditions. Therefore,
FIG. 1
illustrates that if even a single bit of data (a single binary one value) is communicated through the channel
10
from a transmit side to a receive side (using some modulation scheme), that the receive side will receive a widely time dispersed and distorted “image” of that one transmitted binary one value.
FIG. 1
further illustrates the impact the response
12
has on communicated data when transmitting asymmetric digital subscriber line (ADSL) data symbols over the channel
10
. In an ADSL system, data is sent in discrete packets containing many frequency-coded digital bits. Each of the frequency-coded packets is transmitted by the transceiver for a time duration of about 250 microseconds whereby each packet is modulated using either 32 carriers (upstream direction, i.e., from remote to central office (CO) side) or 256 carriers (downstream direction, i.e., from central office to remote side). The packets are transmitted serially in time, one after the other, in order to communicate larger blocks of related data, voice, video, sound, or other information between users. Each packet, also referred to as a symbol, is physically sent through the channel using a digital-to-analog converter at the transmit side and retrieved using an analog-to-digital converter at the receive end. Due to the impulse response
12
of
FIG. 1
, each 250 microsecond transmitted symbols
14
a
and
16
a
is distorted and/or smeared at the receiving end over a longer time period of time than the desired 250 microseconds duration.
FIG. 1
illustrates the smeared receive symbols
14
b
and
16
b
that are provided at,the receive end of the channel
10
.
In most cases, adjacent smeared symbols
14
b
and
16
b
will overlap each other in time thereby causing an intersymbol-interference (ISI) region
18
as shown in FIG.
1
. The ISI region is a period of time where data from symbol
14
b
is distorting data from symbol
16
b
and vice versa. One solution to the ISI problem is to throw away all ISI-distorted data that lies, within the ISI region
18
. Another solution is to spread the symbols
14
a
and
16
a
farther apart from each other in time by using a larger dormant time period between symbols at the transmit end. By transmitting fewer symbols close to each other, it is possible to eliminate or reduce the size of the ISI region
18
at the receive end. Both of these “solutions” greatly reduce the available data transmission rate and may increase the bit-error rate (BER) of the receive signal. Neither, reduced data rate nor increased BER is desired by the industry.
Furthermore, some channel lines
10
are so adversely subject to parasitics that a single 250 microsecond ADSL symbol may be smeared in time to result in a receive symbol that spans a time duration of more than one symbol. When viewed another way, any one single ADSL symbol at the receive end may be experiencing interference from several other ADSL symbols. To add to the problem of data recovery and integrity, an ADSL channel that transmits 1 unit of initial power at a transmit end will easily attenuate that power to 10
−6
units of power or less by the time the signal reaches the receive end of the channel
10
. The combination of ISI and signal attenuation makes ADSL data transmission and recovery complex.
One common way to reduce ISI is to place a hard-wired time domain equalizer (TEQ)
20
serially in-line with the channel
10
. Such a TEQ-based system is shown in FIG.
2
. The main objective of the time domain equalizer (TEQ)
20
of
FIG. 2
is to reduce the channel dispersion and the resulting intersymbol interference (ISI) by shortening the channel impulse response in time (see response
12
of
FIG. 1
) before discarding data samples. In other words, the channel
10
is initially analyzed to find its impulse response, and the TEQ
20
is then hard-wired to fixed filter coefficients that create a near inverse of this impulse response. With this fixed length channel and fixed TEQ, the combination (convolution) of the channel impulse response and the response of the TEQ
20
nulls out a substantial portion of the ISI whereby only the actual symbol data survives intact from the transmit side to the receive side.
Generally, reduction of ISI is achieved in the prior art by cascading the channel
10
(having an impulse response h with &ngr; samples) with a finite impulse response (FIR) filter w of length N
w
, which is know as the TEQ filter
20
, where N
w
, is a finite positive integer representing a number of data samples. Herein, underlined variable names are used to represent the signals or filter taps in a vector notation. The FIR filter response w is hard-wired, in response to data received by proper signal analysis of the channel
10
, such that the combination of h and w has a Target Impulse Response (TIR)
22
shown in FIG.
2
. Note that the response
22
is shorter in time than the unfiltered response
12
of
FIG. 1
(i.e., the samples &ngr; over which the response
12
extends is greater than the number of samples N
b
over which the response
22
extends). The desired TIR
20
is an FIR filter b of length N
b
, where N
b
of
FIG. 2
is much smaller than the length &ngr; of the channel impulse response h of
FIG. 1
(where h is a design constraint due to parasitics and external interference). By adding the fixed TEQ
20
in
FIG. 2
, the adverse ISI of
FIG. 2
is contained within a window of N
b
samples where N
b
<<&ngr; (see &ngr; in
FIG. 1
) so that less ISI occurs and reduced discarding of data is required in order to effectively transmit the original data
14
a
and
16
a
to the receive end.
In discrete multi-tone (DMT) modulation, the ISI will generally occur over the cyclic prefix (CP) portion of the data (which is non-data overhead in the ADSL symbol/packet). Since the CP is non-data overhead in the system, the
Motorola Inc.
Tse Young T.
LandOfFree
Fast training of equalizers in discrete multi-tone (DMT)... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Fast training of equalizers in discrete multi-tone (DMT)..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Fast training of equalizers in discrete multi-tone (DMT)... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3003142