 |
Claims  |
|
|
What is claimed:
1. An equalization apparatus for the equalization of electrical signals codified into symbols and transmitted on a transmission channel comprising:
a first feed forward filter with a received symbol sequence as input, the first feed forward filter having a series of delay means for delaying said received symbol sequence to provide one or more delayed symbol sequences, a means for multiplying
the received symbol sequence and each of the delayed symbol sequences by an associated one of a first plurality of adjustable coefficients, a summation means for adding the multiplied symbol sequences to obtain signal samples;
a first decision means for assigning a decided symbol to each signal sample from the first feed forward filter using a first decision process;
a second feed forward filter with the received symbol sequence as input, the second feed forward filter having a series of delay means for delaying the received symbol sequence to provide one or more delayed symbol sequences, a means for
multiplying the received symbol sequence and each of the delayed symbol sequences by an associated one of a second plurality of adjustable coefficients;
a feedback filter with decided symbols from the first decision means as input, the feedback filter having a series of delay means for delaying the decided symbols to provide one or more delayed symbol sequences, a means for multiplying each of
the delayed symbol sequences by an associated one of a third plurality of adjustable coefficients;
a combiner for combining the multiplied symbol sequences from the second feed forward filter with the multiplied symbol sequences of the feedback filter to provide combined signal samples;
a second decision means having the combined signal samples as input, the second decision means for assigning a decided symbol to each input signal sample using a second decision process, providing the decided symbol as the output of the equalizer
apparatus;
a first coefficient adjustment means for adjusting the coefficients of the first feed forward filter using a first coefficient adaptation process and a first objective function;
a second coefficient adjustment means for adjusting the coefficients of the second feed forward filter using a second coefficient adaptation process and a second objective function; and
a third coefficient adjustment means for adjusting the coefficients of the feedback filter using a third coefficient adaptation process and a third objective function.
2. The equalization apparatus of claim 1 wherein the first objective function uses an error signal determined from the output of the first feed forward filter and the assigned symbols from the first decision means.
3. The equalization apparatus of claim 2 wherein the second objective function uses an error signal determined from the assigned symbols from the second decision means and the combined signal samples.
4. The equalization apparatus of claim 1 wherein the coefficient adaptation processes used by the second coefficient adjustment means and the third coefficient adjustment means are the same.
5. The equalization apparatus of claim 1 wherein the second objective function and the third objective function use an error signal determined from the assigned symbols from the second decision means and the combined signal samples.
6. The equalization apparatus of claim 1 wherein the coefficient adaptation processes used by the first coefficient adjustment means, the second coefficient adjustment means, and the third coefficient adjustment means are the same.
7. The equalization apparatus of claim 3 wherein the first coefficient adjustment means, the second coefficient adjustment means, and the third coefficient adjustment means each use the same coefficient adaptation process.
8. The equalization apparatus of claim 1 wherein the first coefficient adjustment means, the second coefficient adjustment means, and the third coefficient adjustment means each use a least mean square coefficient adaptation process.
9. The equalization apparatus of claim 1 wherein the first coefficient adjustment means, the second coefficient adjustment means, and the third coefficient adjustment means each use a means square error objective function.
10. The equalization apparatus of claim 1 wherein the first coefficient adjustment means, the second coefficient adjustment means, and the third coefficient adjustment means each use a recursive least square error coefficient adaptation process.
11. The equalization apparatus of claim 1 wherein the first decision means and the second decision means use the same decision process.
12. The equalization apparatus of claim 11 wherein the first decision means and the second decision means are slicers.
13. An equalization apparatus for the equalization of electrical signals codified into symbols and transmitted on a transmission channel comprising:
a first feed forward filter with a received symbol sequence as input, having a first plurality of adjustable coefficients, the first feed forward filter for filtering the received symbol sequence;
a first decision means for assigning one or more decided symbols to the filtered symbol sequence from the first feed forward filter using a first decision process;
a second feed forward filter with the received symbol sequence as input, having a second plurality of adjustable coefficients, the second feed forward filter for filtering the received symbol sequence;
a feedback filter with decided symbols from the first decision means as input, having a third plurality of adjustable coefficients, the feedback filter for filtering the decided symbols;
a combiner for combining the filtered received symbols from the second feed forward filter with the filtered decided symbols from the feedback filter;
a second decision means having the combined symbols as input, the second decision means for assigning one or more decided symbols to the combined symbols using a second decision process, and providing the decided symbols as the output of the
equalizer apparatus;
a first coefficient adjustment means for adjusting the coefficients of the first feed forward filter using a first coefficient adaptation process and a first objective function;
a second coefficient adjustment means for adjusting the coefficients of the second feed forward filter using a second coefficient adaptation process and a second objective function; and
a third coefficient adjustment means for adjusting the coefficients of the feedback filter using a third coefficient adaptation process and a third objective function.
14. An equalization method for the equalization of electrical signals codified into symbols and transmitted on a transmission channel comprising the steps:
delaying a received symbol sequence with a first series of delay elements;
multiplying the received symbol sequence and the delayed symbol sequences from the first series of delay elements with a first plurality of adaptable coefficients;
adding the multiplied symbols to obtain signal samples;
assigning a decided symbol to each signal sample using a first decision element;
delaying the received symbol sequence with a second series of delay elements;
multiplying the received symbol sequence and the delayed symbol sequences from the second series of delay elements with a second plurality of adaptable coefficients;
delaying the decided symbol sequence from the first decision element with a third series of delay elements;
multiplying the delayed symbol sequence from the third series of delay elements with a third plurality of adaptable coefficients;
combining the symbol sequences multiplied with the third plurality of coefficients with the sequences multiplied by the second plurality of coefficients to obtain a combined signal sample;
assigning one or more decided symbols to each combined signal sample using a second decision element, and
updating each of the plurality of adjustable coefficients.
15. An equalization method for the equalization of electrical signals codified into symbols and transmitted on a transmission channel comprising the steps:
filtering a received symbol sequence with a first feed forward filter having a first plurality of adjustable coefficients;
assigning one or more symbols to the filtered symbol sequence by a first decision element;
filtering the received symbol sequence with a second feed forward filter having a second plurality of adjustable coefficients;
filtering the assigned symbols from the first decision element by a feedback filter having a third plurality adjustable coefficients;
combining the filtered symbol sequence from the second feed forward filter with the filtered symbol sequence from the feedback filter;
assigning one or more symbols to the combined symbol sequence by a second decision element and outputting the assigned symbols from the second decision element as an equalized symbol sequence; and
updating the adjustable coefficients associated with the first feed forward filter, the second feed forward filter and the feedback filter. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
RELATED PATENT APPLICATIONS
U.S. patent application Ser. No. 07/846,651 filed Mar. 5, 1992, entitled "System and Method of Estimating Equalizer Performance in the Presence of Channel Mismatch", IBM Docket No. EN992026 and U.S. patent application Ser. No. 07/866,928
filed Apr. 10, 1992, entitled, "System and Method of Robust Sequence Estimation in the Presence of Channel Mismatch Conditions"IBM Docket No. EN992057, both assigned to the same assignee as the present invention.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a device and method for performing adaptive equalization in a communications system. In particular, the present invention provides for blind or referenced trained adaptive equalization for use in digital
communication systems.
2. Description of the Prior Art
In communications systems multiple reflections lead to a confluence at a receiver of several signals which all stem from the same signal generated at a transmitter but differ in arrival time, carrier phase and amplitude. This can impair the
transmission performance and cause fading or even signal elimination at the receiver. These so-called multipath effects particularly appear in urban environments which are at the same time those areas with the highest demand for communication systems.
The relative motion of the receiver with respect to the transmitter and/or the transmitter with respect to the receiver can cause a doppler effect which can cause fading which also impairs transmission performance. These effects are particular
troublesome with mobile communication systems. Mobile channels are generally characterized as fading multipath channels with time dispersion (multipath spreads). The mobility of such systems creates transmission channel characteristics that are
constantly changing as the geometries, transmission path, interference and transmission medium change.
The high bit or data rates of modern digital mobile radio systems cause a significant part of the typical multipath effects to appear as inter-symbol interference (ISI). Because of the non-ideality of the frequency response of the transmission
channel, each transmitted symbol interferes with the others, generating ISI. To remove the inter-symbol interference, the systems are usually equipped with equalizers (see Lucky R. W., "Automatic equalization for digital communication", Bell System
Technical. Journal, 1965, 44, pp. 574-588). Equalizers are widely used in communications systems and can employ either dedicated hardware or a programmable digital signal processors (DSP) or DSPs. There are two primary types of equalizers: linear and
non-linear. Both types of equalizers can be classified as either reference trained or blind. Both types of equalizers typically utilize an adaptive filter. The adaptive filter, often referred to as transversal filter or moving average filter, is made
with a chain of delay elements, at the output of each of which is placed a variable gain amplifier (tap gain). The variable tap gains are usually referred to as adjustable coefficients. The outputs of the variable tap gain amps are then added to
provide a signal sample which gives an indication of the transmitted symbol. This signal sample is then sent to a decision element or symbol detector to obtain a decided symbol. Assuming no errors, the decided symbol should be equal to the symbol fed
into the transmission channel by the transmitter.
By appropriate selection of the delay elements and the coefficients, equalizers can reduce the inter-symbol interference according to a given criterion. Some types of equalizers, referred to as adaptive, provide for automatic coefficient
adjustment. In these equalizers, starting from arbitrary initial coefficients often quite far from the optimum, the coefficients can be modified iteratively until an optimal configuration is reached. To minimize inter-symbol interference many adaptive
equalization systems adopt the criterion of minimizing the mean square error (MSE) defined from the signal samples at the adaptive filter output before the decision element and the corresponding transmitted signals using estimated gradient methods. For
a given transmission channel, the mean square error is a quadratic function of the tap gains for referenced trained adaptive filters. The mean square error is minimized by estimating its gradient with respect to the filter coefficients. The filter
coefficients are modified in the direction opposite to the estimated gradient.
More particularly, starting from arbitrary tap gain values, differences are found between the transmitted reference symbols and the signal samples at the equalizer output. Using these differences, in combination with the signals present at the
equalizer input, the tap gains are modified to obtain the minimum mean square error. It can be shown that a tap gain configuration which minimizes the mean square error exists and is unique (see Gersho A., "Adaptive equalization of highly dispersive
channels for data transmission", Bell System Technical Journal, 1969, 48, pp. 55-70). When the optimum configuration has been reached the outputs of the receiver decision element, i.e. the self-decided symbols, are correct with very high probability
and can be used instead of the reference symbols to obtain the present value of the error to be used in the adaptation process. Many other coefficient adjustment schemes have been suggested. The basic assumption for the adaptive equalizer is therefore
that the current output samples for the adaptive equalizer can be compared with the corresponding transmitted symbols, which have to be known a priori.
However, if the channel characteristics change during transmission, as is particularly the case with mobile systems, the self-decided symbols may become incorrect and the equalizer is unable to reconfigure the tap gains to the new optimum values. In this case, to obtain reliable self-decided symbols at the receiver output, the above described start-up procedure ( i.e., the transmitted reference sequence and adjustment of the coefficient) must be repeated with considerable loss of time. To remedy
this serious drawback, blind equalization techniques have been proposed. Blind equalization techniques are capable of converging in a configuration of limited distortion without the necessity of using a predetermined reference symbol sequence (see Y.
Sato, "A method of self-recovering equalization for multi-level amplitude-modulation systems", IEEE Transaction on Communication, Vol. COM-23, N. 6, pp. 679-682, June 1975; D. N. Godard, "Self-recovering equalization and carrier tracking in
two-dimensional data communication systems", IEEE Transaction on Communication, Vol. COM-28, N. 11, pp. 1867-1875, November 1980; A. Benveniste and M. Goursat, "Blind equalizers", IEEE Transaction on Communication, Vol. COM-32, N. 8, pp. 871-883,
August 1984).
To minimize inter-symbol interference these blind techniques typically use new non-convex cost functions different than the mean square error used for the self-learning equalizer. Under weak conditions, these cost functions characterize the
inter-symbol interference sufficiently well while their stochastic minimization can be performed by using locally generated control signals with no knowledge of the transmitted data. However, these methods of adaptive blind equalization are not fully
satisfactory because they do not converge smoothly, and particularly because under steady state operating conditions they maintain a very high residual variance of the error signal. In other words, they do not reach the point of minimal inter-symbol
interference but oscillate continually around the minimum. This leads to operation with unacceptable results.
Blind equalization techniques are attractive not only because they provide for uninterrupted data transmission (because there is no need to send a training sequence when incorrect decisions are made or the transmission channel characteristics
change) but, also because they are quite easy to implement in practice. Most of the existing blind equalization techniques can be categorized as decision-directed-type techniques which use a nonlinear estimator at the output of the equalizer to generate
a decision-directed estimated error. This error is then utilized to adjust the coefficients in a feed forward filter. Thus, the decision directed type equalizer uses a feed forward filter to compensate for the non-ideal channel. However, a feed
forward filter is not very effective in equalizing channels containing spectral nulls. In an attempt to compensate for the channel distortion, the equalizer places a large gain in the vicinity of the spectral null and as a consequence significantly
enhances the additive noise present in the received signal. Consequently, decision directed blind equalizers are not effective for equalizing channels containing spectral nulls. Spectral nulls in the transmission channel are encountered in practice
wherever there is multipath propagation. Mobil radio channels, as discussed above, are generally characterized as fading multipath channels with time dispersion (multipath spreads). The ability of the equalizer to compensate for spectral nulls is
particularly important where multi-path propagation is present.
Additionally, a decision directed equalizer does not efficiently compensate for postcursor ISI. Postcursor ISI is the effect of previously detected symbols on the present symbol. Because detected symbols are not used as feedback, the effect of
ISI from previously detected symbols is not effectively removed from the present estimate. The decision directed equalization with a feedforward filter attempts to invert the transmission channel without directly using previously detected symbols.
Decision feedback equalization techniques use feedback to provide for better compensation for spectral nulls and attempt to eliminate postcursor ISI. Decision feedback equalization permits the removal of ISI by using decision feedback to cancel
from the present symbol the interference from symbols which have already been detected. The basic idea of decision feedback equalization is that if the value of symbols already detected are known then the ISI contributed by these symbols in the present
symbol can be determined and canceled exactly by subtracting the previously detected symbol values with appropriate weighing. A typical decision feedback equalizer combines the output of a feed forward filter and feedback filter, and provides the
combined outputs to a decision element. The output of the decision element is then utilized by the feedback filter. The output of a feedback filter can be thought of as representing the postcursor ISI imposed by previously detected symbols on the
present symbol.
The adjustment of the feed forward filter and feedback filter are typically based on the current value of the filter coefficients and an objective function. The objective function typically uses an error signal which can be defined as the
difference between the symbol sequence input to the decision element and the output symbol sequence of the decision element. Because the error signal is based upon the output sequence and the output sequence is used as input to the feedback filter,
decision feedback equalizers are susceptible to decision error propagation. Decision error propagation can cause the equalizer to "blow up", diverge or oscillate.
The problem of decision error propagation can be explained as follows, if the decision element incorrectly decides (or detects or assigns) a symbol this incorrect symbol is provided to the feedback filter as input. It should be noted that this
incorrect decision will then be utilized by the feedback filter to compensate for postcursor ISI for a number of present symbols (the exact number will depend upon the number of delay elements in the feedback filter). The incorrect determination by the
decision element not only impacts the symbols provided to and propagated in the feedback filter but, also impacts the error signal which is utilized by the feedback filter to adjust its coefficients. The incorrect adjustment of the coefficients along
with the incorrect symbols used by the feedback filter causes an incorrect cancellation to be made from the present symbol (i.e., the postcursor ISI from previously detected symbols is incorrectly determined). The sequence provided to the decision
element is thus incorrect and the decision element is more likely to make another incorrect decision. This cycle repeats. In severe cases decision error propagation can cause the equalizer to diverge rather than converge. Thus, the decision error
propagates through the equalizer resulting in the equalizer not minimizing the ISI.
One proposed solution to reduce the effects of decision error propagation is to provide a reliability criterion for the self-decided symbols that prevents updating of the adjustable coefficients when the reliability criterion is low. (See U.S.
Pat. No. 4,847,797 entitled Adaptive Blind Equalization Method and Device, to Picchi et. al.) Thus a binary consent function prevents the equalizer from updating the adaptive coefficients. This technique requires the additional complexity of a consent
or inhibit function. It also prevents the equalizer from tracking changes in the transmission channel when the reliability criterion is low. The propagation error still exists but, the binary consent function allows adaptation to proceed when the
propagation error is small and stops adaptation when the propagation error is large. Thus, this technique only stops adaptation and not the propagation error.
SUMMARY OF THE INVENTION
The object of the invention is to overcome the above mentioned drawbacks.
It is an object of the invention to equalize received signals without the need to provide training sequences.
It is an object of the invention to provide for equalization for fading multipath channels with time dispersion.
It is still a further object to provide equalization for channels with spectral nulls without significantly enhancing the noise.
It is an object of the invention to minimize the effect of inter-symbol interference.
It is yet another object to reduce or eliminate the effects of decision error propagation.
It is an object of the invention to provide for equalization for fading multipath channels with time and frequency dispersion.
Accordingly, the present invention provides a device and method for the equalization of received signals or received symbol sequences. An equalization device for the equalization of electrical signals codified into symbols and transmitted on a
transmission channel has a first feed forward filter with a received symbol sequence as input, the first feed forward filter having a series of delay means for delaying said received symbol sequence to provide one or more delayed symbol sequences, a
means for multiplying each of the symbol sequences by an adjustable coefficient associated with the symbol sequence, a summation means for adding the multiplied symbol sequences to obtain signal samples; a first decision means for assigning a decided
symbol to each signal sample from said first feed forward filter using a first decision process; a second feed forward filter with the received symbol sequence as input, the second feed forward filter having a series of delay means for delaying said
received symbol sequence to provide one or more delayed symbol sequences, a means for multiplying each of the symbol sequences by an adjustable coefficient associated with the symbol sequence; a feedback filter with decided symbols from the first
decision means as input, the feedback filter having a series of delay means for delaying said input symbols to provide one or more delayed symbol sequences, a means for multiplying each of the delayed symbol sequences by an adjustable coefficient
associated with the delayed symbol sequence; a combinet for combining the multiplied symbol sequences from the second feed forward filter with the multiplied symbol sequences of the feedback filter to provide combined signal samples; a second decision
means having the combined signal samples as input, the second decision means for assigning a decided symbol to each input signal sample using a second decision process, providing said decided symbol as the output of the equalizer apparatus; a first
coefficient adjustment means for adjusting the coefficients of the first feed forward filter using a first coefficient adaptation process and a first objective function; a second coefficient adjustment means for adjusting the coefficients of the second
feed forward filter using a second coefficient adaptation process and a second objective function; and a third coefficient adjustment means for adjusting the coefficients of the feedback filter using a third coefficient adaptation process and a third
objective function.
An equalization method for the equalization of electrical signals codified into symbols and transmitted on a transmission channel having the steps of filtering a received symbol sequence with a first feed forward filter having adjustable
coefficients; assigning one or more symbols to said filtered symbol sequence by a first decision element; filtering said received symbol sequence with a second feed forward filter having adjustable coefficients; filtering said assigned symbols from said
first decision element by a feedback filter having adjustable coefficients; combining said filtered symbol sequence from said second feed forward filter with the filtered symbol sequence from said feedback filter; assigning one or more symbols to said
combined symbol sequence by a second decision element and outputting said assigned symbols from said second decision element as the equalized symbol sequence; and updating the adjustable coefficients associated with the First feed forward filter, the
Second feed forward filter and the feedback filter.
The present invention is an adaptive equalization method and device which provides for the equalization of symbol sequences sent through fading multipath channels with time dispersion. By using a trainer system to supply an estimate of the
received symbol sequence to a trainee system equalization of the received symbol sequences is accomplished without the need for training sequences and with the ability of compensating for spectral nulls without substantial increasing the noise in the
system. The trainer system is configured as a decision directed equalizer with a feed forward filter having the received symbol sequence as input and connected to a decision element that outputs decided symbols. The trainee system is similar to
decision feed back equalization in that it has a feed forward filter, a feedback filter and a decision element but, differs in that the input to the feed back filter is provided by the trainer system. The feed forward filter of the trainee system takes
received symbol sequence as input. The output of the feed forward filter and the feedback filter of the trainee system is provided to the decision element which outputs the equalized symbols sequences.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing and other objects, aspects and advantages of the invention will be better understood from the following detailed description with reference to the drawings, in which:
FIG. 1(a) depicts a typical feed forward transversal filter.
FIG. 1(b) depicts a typical feed back transversal filter.
FIG. 2 depicts decision direct equalization.
FIG. 3 depicts decision feedback equalization
FIG. 4 shows one embodiment of the present invention.
FIG. 5 shows one embodiment of the present invention highlighting the error signals.
FIG. 6 shows one embodiment of the present invention highlighting the trainee and
trainer system elements.
FIG. 7 shows a discrete-time equivalent communication model.
FIG. 8(a) shows the scatter diagram of the distorted received signal before equalization.
FIG. 8(b) depicts the scatter diagram of the equalized signal using the feed-forward only
blind equalization.
FIG. 8(c) shows the scatter diagram of the equalized signal using the present invention.
FIG. 8(d) shows the learning curves in terms of the mean square error (MSE).
FIG. 9 shows one embodiment of the present invention using Viterbi decoding .
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
I. OVERVIEW
Inherent in every communication system are channels which link a transmitter and a receiver. These channels include telephone lines used in voice and modem applications, coaxial cables, fiber cables, under water channels used in acoustic
applications, read/write channels used in magnetic storage devices, and atmospheric or space channels used in radar, satellite, radio, and other wireless communication systems. Although their physical media and propagation characteristics vary greatly,
these channels are an important consideration in any communication system. A communication system consists of a transmitter for sending symbols into a channel and a receiver for receiving the transmitted symbols from the channel. This can be modeled as
shown in FIG. 7. The equalization method and device of the present invention can be considered as part of the receiver and offers advantages to all communications systems.
FIG. 4 provides an overview of the present invention. The present invention as depicted in FIG. 4 can be thought of as two systems: a trainer system and a trainee system. As is shown in FIG. 4 both the trainer and trainee system receive their
input signals from the same source. The input signal is electrically coded symbols from antennae or the front end of a receiver. The input signals can be thought of as electrical signals which are codified into symbols. As is shown, the trainer system
consists of a decision directed equalizer having a feed forward filter and a decision element. The output signal of the trainer system is provided as input to the feedback filter of the trainee system. The trainee system looks similar to a decision
feedback equalizer having a feed forward filter and feedback filter with the important exception that the input signals for the feedback filter are provided from the output of the trainer system instead of the output of the decision element of the
trainee system. The output of the decision element in the trainee system, which is the output of the present invention (i.e., the equalized symbol sequence), is not directly utilized by the trainee system. The output of the decision element of the
trainee system may be utilized in determining one or more of the objective functions which are used to adapt the coefficients of the adaptive filters. This feature of the present invention eliminates decision error propagation.
Each of the elements of the present invention is described in section II below. A description of a decision feedback equalizer is also provided because it provides a basis for understanding the present invention and its advantageous features. A
detailed description of one embodiment of the invention is provided in section III. One example of the expected performance of the present invention is provided in section V. A discussion of the advantages of the present invention is provided in section
VI.
II. ELEMENTS
A. ADAPTIVE FILTERS
The adaptive filter is typically a finite duration impulse response filter with adjustable coefficients. Adjustments of the adjustable coefficients is usually performed adaptively during the transmission of information by using an objective
function and a coefficient adjustment process. This objective function is usually minimized or optimized by the coefficient adjustment process. The coefficient adjustment process or adaptation algorithm adjusts the adjustable coefficients of the
adaptive filter to effect the objective function. In many systems, the objective function is an error signal and the coefficient adjustment process attempts to minimize the error signal. An error signal may use the difference between the signal input
to the decision element and the signal output by the decision element. Several coefficient adaptation processes/adaptation algorithms are identified below.
FIG. 1 (a) shows one embodiment for a feed forward filter (FFF) where the input is a received symbol sequence which is sent through a series of delay elements. The received input sequence and each of the delayed input sequences are provided to
their own variable gain amplifiers (tap gain). The variable tap gains are usually referred to as adjustable coefficients. The received input sequence and each of the delayed input sequences are multiplied by their respective adjustable coefficients.
The outputs of the variable tap gain amps are then added to provide a signal sample which gives an indication of the transmitted symbol. The signal sample output of the FFF can be thought of as the estimate of the transmitted symbol with a certain
delay. The signal sample output of the FFF is referred to as the present symbol or current symbol. The addition may be carried out by a summer, combiner or adder. This signal sample can then be sent to a decision element or symbol detector to obtain a
decided symbol. As will be described, the present invention utilizes two FFFs one in the trainer system and one in the trainee system.
FIG. 1(b) shows one embodiment for a feedback filter (FBF) where the input is a symbol sequence which is sent through a series of delay elements. The delayed sequences are each provided with their own variable gain amplifiers (tap gain). The
variable tap gains are usually referred to as adjustable coefficients. Each of the delayed input sequences are multiplied by their respective adjustable coefficients. The output of a FBF can be thought of as representing the estimated postcursor ISI
due to previously assigned symbols. By canceling the ISI from the signal sample provided by the FFF the respective ISI is removed. The outputs of the variable tap gain amps of the FBF are then added with the output from a FFF to provide a signal sample
which gives an indication of the transmitted symbol with the ISI cancelled. The feedback filter may use the same summer or combinet or adder as the FFF or a separate summer or combiner or adder. The signal sample with the ISI removed can then be sent
to a decision element or symbol detector to obtain a assigned or decided symbol. The present invention utilizes a single FBF in the trainee system.
FFF and FBF are very similar in design and structure. The main difference between the two types of adaptive filters is where they are placed in the equalizer and the fact that the feed back filter typically delays all the signal sequences while
the feed forward filter provides a coefficient for the input symbol sequence without any delay. The filters can have any number of tap gains and delay elements. The exact number of delay elements and associated tap gain amplifiers to use is a design
decision dependent on one or more of the following factors: the modulation scheme, expected number of multipath signals, expected strength of multipath signals, the time dispersion, the frequency dispersion, ambient noise and data rate.
B. DECISION DIRECT EQUALIZATION (DDE)
FIG. 2 shows an overview of the decision direct equalization (linear equalization) using an adaptive filter. The basic idea of the DDE consists of minimizing or optimizing an objective function based upon the decision element | | |
