Robust method for providing tap leakage in adaptive...

Details Robust method for providing tap leakage in adaptive... Robust method for providing tap leakage in adaptive...

1997-04-30

2001-11-06

Lee, Thomas (Department: 2182)

Electrical computers: arithmetic processing and calculating

Electrical digital calculating computer

Particular function performed

C708S301000

Reexamination Certificate

active

06314441

ABSTRACT:

BACKGROUND OF THE INVENTION
A. Field of the Invention
This application is related to the art of digital signal processing and more particularly to methods for providing tap leakage in adaptive equalizer systems.
B. Background Art
It has become commonplace to transmit a wide variety of information across a transmission medium as a digital signal—i.e., a signal for which both time and amplitude are discrete, whether that information is inherently represented in an analog or digital form. In the case of information which is originally in an analog form, the continuous analog signal is sampled at predetermined intervals to arrive at a sequence of discrete numbers—each being representative of a value of the continuous signal at that sample point—such numbers being expressed in a numbering system (generally binary) compatible with the digital transmission scheme. After such a “digitizing” procedure, there is no difference from the standpoint of the transmission infrastructure between such analog-originated information and information which originates in a digital form.
Signal processing of information signals transmitted over a channel occurs in a wide variety of applications and with many objectives. Typical reasons for signal processing include: estimation of characteristic signal parameters; elimination or reduction of unwanted interference; and transformation of a signal into a form that is in some manner more useful or informative. Such processing of discrete (or digital) information signals is carried out by Digital Signal Processing (“DSP”) techniques. Applications of DSP techniques currently occur in such diverse fields as acoustics, sonar, radar, geophysics, communications and medicine.
Processing elements which operate on a digital signal frequently occur as filters or equalizers, which are typically represented in the form of a tapped delay line, such as illustrated in
FIG. 1
, where the “T” of each element
10
represents the period of the sampling frequency for a signal of interest. A characteristic of such a tapped delay line is that an output is a function of an input signal (including, in some cases, prior values of that input signal) and coefficients corresponding to the taps. Algebraically, that relationship would generally be of the form:
y
n
=C
0
+C
1
x
1
+C
2
x
2
+ . . . +C
n
x
n
where y represents an output signal, x represents an input signal and C
0
, C
1
, . . . C
n
are representative of the coefficients.
A comparatively recent variation in digital signal processing is known as adaptive signal processing which has developed concurrently with rapid advance in processing power for DSP hardware devices. A significant difference between classical signal processing techniques and the methods of adaptive signal processing is that the latter are generally applied for time varying digital systems. For the adaptive signal processing case of adaptive filtering, a filter (or equalizer) is caused to adapt to changes in signal statistics so that the output is as close as possible to some desired signal. Adaptive filtering will often be applied for the recovery of an input signal after transmission of that signal over a noisy channel.
Various adaptation algorithms are well known in the art and need not be discussed herein. However, it should be observed that the general adaptation process for an adaptive filter or equalizer operates on the tap coefficients of such a filter or equalizer by iteratively adjusting such coefficients until a desired objective is achieved—e.g. a signal to noise ratio above a defined threshold. The general adaptation process can be described algebraically as:
C′=C±u
where C′ is the value of coefficient C after an adaptation iteration and u represents an update term added by the adaptation iteration. It should be understood of course that each coefficient in a filter will be updated in this same manner and that the update term u may, and likely will, vary from coefficient to coefficient. In a conventional digital system those coefficients will be expressed as binary numbers.
With adaptive filters and equalizers, it is well known that the coefficients must be reduced by a small quantity (independent of the update term), on a periodic basis—generally coincident with each iteration of the adaptation process—in order to promote stability of the filter or equalizer. This small periodic reduction in the magnitude of a coefficient is known as “leakage”. Without such leakage, some of the coefficients will tend to become too large—primarily due to the effect of truncation of some less significant bits of a coefficient due to use of finite length registers (e.g., 24 bit, 32 bit, etc.) in the computer or DSP hardware used to carry out the processing—which can lead to instability in the system.
In the prior art, leakage is accomplished by providing an adder/subtractor for each coefficient, programmed to subtract a defined small quantity from the coefficient at each cycle of the leakage period. It is, however, common for adaptive filters and equalizers to contain several hundred tap coefficients, and thus a corresponding number of adder/subtractors must be added to the filter system to provide for the required leakage at each coefficient. It is easy to see that this significantly increases the complexity of the filter system.
SUMMARY OF INVENTION
It is accordingly an object of the invention to provide a method for accomplishing leakage for coefficients of an adaptive filter/equalizer which does not rely on the use of an adder/subtractor for each coefficient. To that end, a method is provided for incrementally reducing the magnitude of one or more coefficients in an adaptive filter/equalizer solely through a defined manipulation of certain bits of that coefficient, such bit manipulation being carried out on a periodic basis where such period generally tracks the frequency of adaptation iteration for such a filter/equalizer.


REFERENCES:
patent: 4773034 (1988-09-01), Debus, Jr.
patent: 5175745 (1992-12-01), Kaku et al.
patent: 5260896 (1993-11-01), Iwasaki
patent: 5644597 (1997-07-01), Ueda
patent: 5920599 (1999-07-01), Igarashi
patent: 5937007 (1999-08-01), Raghunath
patent: 5949819 (1999-09-01), Bjarnason et al.
patent: 6021161 (2000-02-01), Yamaguchi et al.
patent: 6163572 (2000-12-01), Velez et al.

Affiliated with

Also associated with

Rating

Robust method for providing tap leakage in adaptive... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Robust method for providing tap leakage in adaptive..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Robust method for providing tap leakage in adaptive... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2616369