Efficient implementation of error approximation in blind...

Details Efficient implementation of error approximation in blind... Efficient implementation of error approximation in blind...

1999-11-01

2002-04-09

Chin, Stephen (Department: 2734)

Pulse or digital communications

Equalizers

Automatic

C375S350000, C708S300000, C708S323000

Reexamination Certificate

active

06370191

ABSTRACT:

CROSS-REFERENCE TO RELATED APPLICATIONS
Not applicable.
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
Not applicable.
BACKGROUND OF THE INVENTION
This invention is in the field of data communication, and is more particularly directed to equalization techniques in modem communications.
Data communication among computers and terminals is now commonplace. Much of this communication traffic is carried out by way of modulated signals transmitted and received over existing communications facilities, such as the Public Switched Telephone Network (PSTN) and existing cable television (CATV) networks. The modulation and demodulation of these signals are generally carried out by way of modulator/demodulators (“modems”), residing as system functions in many personal computers, workstations, palm-size computers, and other devices that perform data communications over these communications facilities. As is fundamental in the art, a modem is a device or system function that converts signals to be communicated over the telephone network from a digital bitstream generated by its host computer into an analog signal within the voice band, and vice versa, permitting data communications to be readily carried out over the same telephone network as other communications, such as voice or television programming.
The frequency response of the path, or channel, over which a given modem communication session is carried out is far from ideal, causing the transmitted signal is significantly distorted by its transmission over the channel. Such distortion can give rise to effects such as intersymbol interference (“ISI”), which prevents the receiving modem from accurately distinguishing adjacent received data symbols from one another. Because the channel frequency response varies widely from location to location, it is not feasible to define signal processing characteristics such as data rate, sub-channel frequencies, and signal filtering a priori; rather, the particular frequency characteristics of a communications channel are specifically measured and compensated within each modem communications session.
One common approach for deriving the appropriate channel compensation utilizes the transmission of a known “training sequence” of data from one location to a receiving modem (the string sequence being “known” by the receiving modem). The receiving modem demodulates and decodes the training sequence as received from the transmitting modem, determines the nature of any error between the received training sequence and the known sequence, and sets various filter parameters accordingly, thus compensating for frequency-dependent distortion that is presented by the communications channel. These parameters may also be communicated to the transmitting modem, if appropriate, for use in its own compensation of return traffic (or for pre-compensation of the transmitted signal, if desired).
However, in certain situations, a known training sequence is not available for use by the modem in establishing communication. One such situation is the use of the cable television (CATV) network as the medium of communication, particularly in the provision of Internet access to homes. The attractive features of the CATV network include the relatively large installed base of homes having cable television service, and also the inherently high data rates that may be carried by the coaxial cable with which cable television programs are delivered, especially when compared with twisted-pair copper wiring commonly used in telephone service. Data communication along the CATV network typically consists of multiple user modems connected to a headend, which is a cable company office from which the CATV programming is forwarded to the cable subscribers. The transmission medium in a CATV network is typically coaxial cable, or a combination of coaxial cable and fiber optic links, giving rise to the so-called Hybrid Fiber Coax (HFC) infrastructure. Modems at the user locations can also establish dial-up connections with the headend in the CATV network, for example to access the Internet, in which case signals are bidirectionally communicated, rather than simply unidirectionally communicated as in the distribution of television programming. However, according to conventional protocols for the establishment of communications sessions using cable modems, no training sequences are communicated between the headend and the user modems.
Even if training sequences are not to be used, the appropriate filter parameters must still be determined in order for data communications to be carried out at a reasonable data rate. Absent the training sequences, the equalization parameters must be determined, and updated, from the transmitted data itself. One conventional type of equalizer that is useful for this operation is referred to as the decision feedback equalizer, or “DFE”. In a typical DFE, the received digital signal is processed by an equalizer, such as may be implemented by a finite impulse response (FIR) filter, and forwarded to a decision block which presents a “decision” about the value of the filtered symbol. The output of the decision is applied to an update function that implements an adaptation algorithm for updating the equalizer coefficients (in the FIR realization, these parameter will be the “taps” along the digital delay line) in response to a comparison between the decision output and the signal applied to the equalizer. Over time, the equalizer coefficients converge upon the appropriate channel compensation. However, because decision feedback equalization depends upon reasonable accuracy of the decision process, the equalizer coefficients may never converge if the intersymbol interference is significant. To address this issue, a process referred to in the art as “blind equalization” can be performed by a receiving modem to coarsely define the appropriate filter parameters for use in later transmissions. Blind equalization does not utilize any assumptions about the actual data transmitted, but rather relies upon statistical properties of the transmitted data in order to set the equalization coefficients. Specifically, the update, or adaptation, performed in blind equalization compares the input and output of the equalizer FIR to one another, and updates the FIR by minimizing a cost function based upon the desired statistics.
Typically, blind equalization is used as an initial process in order to remove ISI and other distortion to such an extent that other equalization techniques, such as DFE, may then be applied. An example of such a combination equalization technique is schematically illustrated for a conventional cable modem in FIG.
1
. As illustrated therein, signal s(t) is applied to communications channel
2
, and communicated to a modem which receives the signal r(t). Signal r(t) thus corresponds to the transmitted signal, as distorted by channel
2
; in the time domain, r(t)=h(t) * s(t), where h(t) is the response of channel
2
and where * is the convolution operator. Typically, the analog signal r(t) is sampled upon receipt by an analog-to-digital converter (ADC) (not shown), which generates a digital sample stream r
k
. Blind equalizer function
3
receives the sample stream r
k
at its adaptive equalizer function
4
. Adaptive equalizer function
4
applies a digital filter, such as an FIR, to the received sample stream r
k
. The equalization coefficients applied by equalizer function
4
are defined according to an initial default state that is updated by update function
6
. The output of adaptive equalizer function
4
is a signal y
k
, which is applied to update function
6
along with the received sample stream r
k
. Decision block
7
also receives the output of adaptive equalizer
4
, determines the symbol value therefrom, and forwards detected symbols ŝ
k
to the remainder of the modem.
Decision feedback equalizer (DFE) function
8
is in parallel with blind equalizer function
2
. In the conventional manner, DFE function
8
includes feed-forward equalizer
9
that applies a compensating fil

Affiliated with

Also associated with

Rating

Efficient implementation of error approximation in blind... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Efficient implementation of error approximation in blind..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Efficient implementation of error approximation in blind... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2858879