WO2000079746A1

WO2000079746A1 - Companion nyquist filter and linear equalizer within a data transmission system

Info

Publication number: WO2000079746A1
Application number: PCT/US2000/017117
Authority: WO
Inventors: Edwin Twitchell
Original assignee: Harris Corp
Current assignee: Harris Corp
Priority date: 1999-06-22
Filing date: 2000-06-22
Publication date: 2000-12-28
Anticipated expiration: 2001-12-22
Also published as: AU5631000A

Abstract

A transmission system (110) has a Nyquist filter (114) that operates upon a signal to shape signal spectrum. A first processed version of the signal is output from the Nyquist filter (114). A linear equalizer (112) of a system (110) operates upon the signal to pre-distort the signal in a manner that compensated for subsequent linear distortion. A second processed version of the signal is output from the linear equalizer (112). The linear equalizer (112) is located in parallel with the Nyquist filter (114). A combiner (118) combines the first and second processed versions of the signal.

Description

COMPANION NYQUIST FILTER AND LINEAR EQUALIZER WTTHLN A DATA TRANSMISSION SYSTEM

The present invention relates to broadcast transmission systems and is particularly directed to compensation of distortion within a digital transmission system, such as a digital television ("DTV") transmission system.

A broadcast transmission s stem, such as DTV broadcast system, includes an amplifying device that increases the power of an electrical information signal such that an antenna is excited to emit a broadcast system at a desired strength. The amplifying device is referred to as a power amplifier. In order to optimize the quality of the broadcast signal, the electrical signal is conditioned prior to amplification. The signal conditioning includes bandpass filtering of the electrical signal to limit the frequency band of the electrical signal that is input to the power amplifier.

Several issues arise during operation of such a transmission system. One issue is that the components of the transmission system, including the power amplifier and the signal conditioning devices, distort the electrical information signal away from intended values. Specifically, the power amplifier imposes non-linear distortion upon the signal. Also, some of the signal conditioning devices (e.g., a band-limiting filter) impose linear distortions upon the information signal.

As a result of such distortions within the transmission system, instantaneous amplitude variations (AM/ AM) and instantaneous phase variations (AM/PM) occur. In addition, frequency dependent amplitude and phase variations occur. It is to be appreciated that within a phase-amplitude modulated system, amplitude and phase integrity of the system must be preserved for optimum system performance.

Another issue that presents itself is that the power amplifier may impose a frequency spectrum spread on the signal during amplification. The spreading may include smearing of the frequency and generation of unwanted frequency components. The frequency spread results in a broadcast signal of dimensioned quality. Additional signal conditioning, prirnarily in the form of bandpass filtering, after amplification will improve the quality of the broadcast signal. However, each additional signal-conditioning component (e.g., a bandpass filter) causes additional distortions to the signal. An increase in the number of distortion causing components within the system is associated with an increase in the distortions that must be corrected.

One technique to correct for the linear and non-linear distortions is to pre-distort the information signal. In a system such as a DTV system, the baseband signal is a digital signal and pre-distortion can be accomplished in the digital domain. However, a typical known microprocessor may have difficulty in providing the distortion correction and correction adaptation needed for high data-rate signals conveyed within such systems. Thus, there are continuing needs for improved techniques to provide distortion correction and correction adaptation for such a system, and to provide such correction and adaptation via microprocessor components that have an acceptable performance/ cost consideration. The present invention includes a *1*

Advantageously, a transmission system that has a Nyquist filter for operating upon a signal to shape signal spectrum and for outputting a first processed version of the signal. The system also includes a linear equalizer for operating upon the signal to pre-distort the signal in a manner that compensates for subsequent linear distortion and for outputting a second processed version of the signal. The linear equalizer is located in parallel with the Nyquist filter. A combiner of the system combines the first and second processed versions. The present invention will now be described, by way of example, with reference to the accompanying drawings in which:

Fig. 1 is a block diagram of a system that utilizes one technique of sequential pre- distortion;

Fig. 2 is a block diagram of a system that has pre-distortion in accordance with the present invention; ,

Fig. 3 is a simplified block diagram of the system shown in Fig. 2; Fig. 4 is a block diagram that illustrates a mathematical representation of the system shown in Fig. 3;

Fig. 5 is a block diagram of a system in accordance with a second embodiment of the present invention;

Fig. 6 is a block diagram of a system in accordance with a third embodiment of the present invention; and

Fig. 7 is a block diagram of a system in accordance with a fourth embodiment of the present invention. A representation of a system 10 is shown in function block format in Fig.1 as a plurality of components that define a path of a data stream 12 (the data stream flows clockwise in Fig. 1, starting in the upper left corner). An information data signal proceeds along the data stream 12. Preferably, the information signal has a relatively high data rate. The high data rate is related to the environment in which the system 10 functions. Specifically, the system 10 is preferably a high definition ("HD") digital television ("DTV") system. Preferably, the DTV system broadcasts signals in the radio range of frequencies. In one embodiment, the broadcast signal is in the ultrahigh frequency range (300-3,000 MHz), and is preferably in the range of 470- 860 MHz. A DTV system includes an 8NSB exciter and a transmitter. The components shown in

Fig. 1 are located within the 8VSB exciter and the transmitter. Specifically, the system 10 includes a power amplifier 20 that amplifies the information signal to a power level that is suitable for broadcast transmission of a RF signal. In one example, the amplified power level is 50 kilowatts. The power amplifier 20 may be comprised of an array of amplifying devices. If a plurahty of amplifying devices is present within the power amplifier 20, a combiner device is present to combine amplifier device outputs.

Turning now to the components located upstream of the power amplifier 20, many of these upstream components operate in digital format and at certain predetermined data rates. In distinction, the power amplifier 20 amplifies an analog signal at a desired frequency to convey the relatively high rate of data. Thus, a series of components is located upstream of the power amplifier 20 to condition and convert the information signal to provide the desired input to the power amplifier. Specifically, (at the bottom of Fig. 1) a digital form of the information signal is provided to a digital-to-analog (D/A) converter 22, which outputs the information signal in analog form. A series 24 of filter(s) and up-converter(s) is located downstream of the D/A converter. The filter and converter series 24 contains at least one bandpass filter. The output of the filter and converter series 24 is provided to the power amplifier 20.

A post-amplification filter 26 is located downstream of the power amplifier 20. Herein, the post-amplification filter 26 is referred to as a high power filter 26. The high power filter 26 is a band-lirniting filter. Focusing now upon a theoretical "ideal" design, all of the components (e.g., the power amplifier and the filter components) would be ideal. Specifically, the power amplifier transfer curve would be linear. Thus, an information signal that has a given pre-amplification power level will be amplified to a predetermined power level by the power amplifier, based solely upon a linear relationship that dictates the amount of amplification. Also, the filters of such an ideal system would not impose any frequency dependent distortions.

The actual power amplifier 20 is, however, not ideal. The power transfer curve of the power amplifier 20 is not linear. A non-linear distortion is imposed by the power amplifier 20 upon the information signal during amplification of the information signal. Specifically, the non-linear distortion is directed to changes in instantaneous amplitude and phase variations. Accordingly, a correction is desired upon the information signal to compensate for distortion caused by the power amplifier 20.

The filter and converter series 24 and the high power filter 26 impose linear, frequency- dependent deformations to the information signal. For example, the distortion imposed by the high power filter 26 is directed to group delay and amplitude response (i.e., amplitude variation versus frequency). Thus, for each distortion that occurs within the system 10, an amount of correction or equalization is to be imposed upon the information signal to compensate.

Turning again to a theoretical ideal system, any action (i.e., amplification or filtering) imposed upon the information signal would be time-invariant. Specifically, in the ideal s stem, the actions imposed upon the information signal would not change over time. Thus, for a given input stimulus, the ideal system always produces the same amount, independent of the time at which the stimulus occurs.

However, in actuality, the system 10 is time-variant. Specifically, for a given input stimulus, the outputs of the components change over time. One reason for time-variance is thermal effects that cause variations in the amount of signal deformation caused by the filter and converter series 24, the power amplifier 20, and the high power filter 26. Thus, it is desirable to compensate for all of the signal distortion (i.e., the sequence of linear, non-linear, and linear), and to adapt to change in the distortion.

In order to provide for compensation, the system includes three corrector or equalizer (i.e., compensating) components 28-32. The corrector/ equalizer components 28-32 are located upstream of the distortion-causing components. Specifically, all of the corrector/ equalizer components 28-32 are upstream of the filter and converter series 24, the power amplifier 20, and the high power filter 26. The corrector/ equalizer components 28-32 also happen to be upstream of the D/A converter 22 and an interpolation component 34. Thus, the correction/ equalization is done in digital format. Further, the location of corrector/ equalizer components 28-32 upstream of the interpolation component causes the correction/ equalization to occur at baseband or at a relatively low IF compared to the amplification and filtering that occurs within the components 20-26 (recalling that the interpolation component 34 is upstream of the power amplifier and the filters). It is to be appreciated by the person of ordinary skill in the art that the pre-distortion of the information signal is such that once distortion subsequently occurs, the signal has desired values.

Turning to the specifics of the corrector/ equalizer components 28-32, an adaptive linear equalizer 28 imposes a pre-distortion onto the information signal to compensate for the linear distortion caused by the high power filter 26. Preferably, the linear equalizer 28 includes at least one finite impulse response ("FIR") digital filter that has suitable structure for pre-equalizing the information signal. It is be appreciated that other filter types could be employed.

An adaptive non-linear corrector 30 imposes a pre-distortion onto the signal to compensate for the non-linear distortion caused by the power amplifier 20. The non-linear corrector 30 may have any suitable structure for pre-correcting the signal. An adaptive linear equalizer 32 imposes a pre-distortion onto the information signal to compensate for the pre- amplification linear distortion that is primarily caused by the filter and converter series 24.

In the example illustrated in Fig. 1, the linear equalizer 28 and the linear equalizer 32 happen to be real format filters, and the non-Linear corrector 30 is a complex format filter. Accordingly, a complex-to-real converter 36 is located upstream of the linear equalizer 28, a real- to-complex converter 38 is located between the linear equalizer 28 and the non-linear corrector 30, and a complex-to-real converter 40 is located between the non-linear corrector 30 and the linear equalizer 32.

Turning to the signal input provided to the portion of the system 10 shown in Fig.1, the information signal is provided to a Nyquist filter 42 that constrains the spectral energy of the signal to be contained within the Nyquist bandwidth. Intersymbol interference will be reduced to the extent that the system response does not deviate from the "ideal" Nyquist filter shape. Ideally, the information signal is provided at 3 bits per symbol (8 level VSB), and the 3 dB bandwidth is equal to the symbol rate of 5.38 MHz. A roll-off factor of α=0.1152 allows for some excess bandwidth for a total bandwidth of 6 MHz.

In the illustrated example, the Nyquist filter 42 has 127 filter taps. An output of the Nyquist filter 42 is a 32 bit complex signal that is provided to the complex-to-real converter 36. The output of the complex-to-real converter 36 has 16 bits. Further, at this point, the information signal can be considered to be perfect. As stated above, the amount of correction/ equalization imposed by the linear equalizer 28, the non-linear corrector 30, and the linear equalizer 32 can be adapted (i.e., updated). A controller 41 determines the amount of change of the correction/ equalization for each of the linear equalizer 28, a non-linear corrector 30, and the linear equalizer 32 (e.g., the filter coefficients are changed). In order to make determinations regarding correction/ equalization adaptation, the information signal is sampled prior to each correction/ equalization component 28-32. The signal sample taken prior to the linear equalizer 28 is held within a Dl memory 44. The signal sample taken prior to the non-linear corrector 30 is held within a D2 memory 46. The signal sample taken prior to the linear equalizer 32 is held within a D3 memory 48. In turn, the memories 44-48 are connected to the controller 41 to provide the signal sample values to the controller.

Determinations of whether a correction/ equalization requires adaptation (i.e., change) require comparisons between the information signal prior to the correction/ equalization and the information signal after respective distortion occurs. Thus, samples of the information signal are taken for each distortion. Specifically, the information signal is coupled off 50 just prior to the power amplifier 20, such that the linear distortion of the filter and converter series 24 is discernible. The information signal is coupled-of f 52 just after the power amplifier 20, such that the non-linear distortion of the power amplifier is discernible. The information signal is coupled-off 56 just after the high power filter 26, such that the linear distortion of the high power filter is discernible. A sampler 58 selectively samples at one of the three available sample locations. The output of the sampler 58 is passed, via a low-pass filter 60 and an A/D converter 62, to a Y memory 64. The Y memory 64 is connected to the controller 41. Adaptation is performed as desired.

Turning now to the position of the linear equalizer 28 along the path, it can be seen that because the output of the N quist filter 42 is 32 bit in a complex format, which is then converted to a 16 bit real format, the linear equalizer must have sufficient capacity to handle such a signal. It can be further appreciated that capacity within an adaptive filter translates directly to requirements for processing ability, and in turn costs. It would be beneficial to be able to achieve the same results that are provided by the system 10 shown in Fig. 1, but utilizing components that have lesser capacity, and thus lesser cost.

Advantageously a modification to the system 10 is shown in Fig. 1 such that a lower capacity and lower cost linear equalizer can be employed. An example of a system 110 is shown in Fig.2. The system 110 of Fig.2 has many of the same components as the system 10 of Fig. 1. Components that are identical are identified by identical reference numbers. The primary difference in the system 110 of Fig. 2 from the system 10 of Fig. 1 is that a linear equalizer 112 is located in parallel with a Nyquist filter 114. Specifically, the input to the Nyquist filter 114 is also provided as an input to the linear equalizer 112. The input to the linear equalizer 112 includes a delay 116 such that processing within the Nyquist filter 114 and the linear equalizer 112 remain in synchronous. The delay amount is dependent upon the number of filter taps within the Nyquist filter 114 and the number of filter taps within the linear equalizer 112. Specifically, if the Nyquist filter 114 has N filter taps and the linear equalizer 112 has M filter taps then the delay D is shown by the equation: M - N

D

In the illustrated example, the Nyquist filter 114 has 127 filter taps and the linear equalizer 112 has 63 filter taps thus:

127 -63

Further, the outputs provided by the Nyquist filter 114 and the linear equalizer

112 are summed by summation device 118. The Nyquist filter and the linear equalizer are complex filters. Accordingly, because the linear equalizer 112 is operating upon a 3-bit signal, instead of a 16-bit signal as in the system 10 of Fig. 1, less processing power is required within the linear equalizer 112 of the system 110 of Fig. 2. In order for adaptation to occur for the linear equalizer 112, a comparison between signal values (i.e., initial/ desired and result) must be performed. The sampler 58 that derives a sample output 56 from the high power filter 26 still provides Y memory values for use within the comparison. However, the proper sample location to provide samples for the Dl memory must be after the Nyquist filter 114, but not include pre- distortion imposed by the linear equalizer 112. Thus, proper sample point is downstream of the Nyquist filter 114 and upstream of the summation device 118.

Fig. 3 illustrates a simplified block diagram of the system 110 of Fig. 2, and further identifies the various signal values within the system. The simplified version is useful to illustrate the mathematical relationships showing operation of the present invention. Specifically, the error is identified by the equation: e=d-y If the system were to be perfect, then e=0 and d-y=0. However, in actuality, during operation: y = (d + w ) * h W = X * f d = x * g y = x * (g + /)* h knowing y=d

8 = (.? + / ) * * g + f = g * h ^{' 1}

This requires that the summation of the Nyquist filter and the equalizer taps must be equal to the convolution of the Nyquist filter and the inverse system response. This is exactly the desired response of the parallel configuration. When there is no system impairment (h=δ[n]) then g+f=g. This forces the taps f_k=0 for all k as expected. Thus, this illustrates that the parallel-located linear equalizer 112 can properly pre-distort the signal.

The benefit of performing equalization as shown in Fig. 3 is that the filter coefficients operate on small data numbers (in this case 3 bits). Since the Nyquist filter is fixed, it can be implemented in ROM (read only memory) or hard- wired to elirrtinate expensive multiples. The equalizer need only perform 3xN multiplies where N is the coefficient resolution in the bits.

An alternative implementation method would be to combine the equalizer and N quist filter together into a single adaptive equalizer. To ensure that the Nyquist filter shape was always applied in the equalizer, the desired signal would need to be applied to a Nyquist filter in the controller before being subtracted from the return signal y[n] as shown in Fig. 4. When the error signal e[n] is minimized (near zero), then y [n ] = x [n ] * g [n ] which is the desired system response. The advantage here is that only one in-line filter is required. The Nyquist filtering is done off-line by the controller. Implementation choices may actually favor the parallel structure because it allows for a fixed Nyquist filter. The system in Fig. 4 requires an adaptive filter of length sufficient to preserve the specified Nyquist shape. Equalizer truncation to a length (assuming FIR filter) less than the minirnal Nyquist filter length would not be an option.

In both cases, each filter could be implemented as a real filter if the appropriate complex-to-real conversions were performed prior to filtering. Figs. 5-7 provide various examples of such a concept. Specifically, Figs. 5-7 illustrate various modifications of the system 110 of Fig. 2 that are in accordance with the present invention. The modified systems of Figs. 5-7 are designated by suffixes A-C, respectively. Shown components are designated by suffixes A-C to signify that the structure is different than its counterpart of the system 110 of Fig. 2.

In the modification shown in Fig. 5, a real-to-complex converter 38 A is located upstream of both the Nyquist filter 114A and the linear equalizer 112A. This allows the

Nyquist filter 114A and the linear equalizer 112A to both be of the type to receive complex inputs. The outputs of the Nyquist filter 114A and the linear equalizer 112A remain in a complex format. Accordingly, the summation device 118A is a complex format summation device and the output of the summation device is provided directly to the complex format non-linear corrector.

In the modification shown in Fig. 6, the output of the linear equalizer 112 is provided to a real-to-complex converter 38B. Here also, the summation device 118B is a complex format component.

In the modification shown in Fig. 7, a real-to-complex converter 122 is located upstream of the Nyquist filter 114C and is also located upstream of the linear equalizer 112.

The Nyquist filter 114C is such that it operates on a complex input and provides a real output. A complex-to-real converter 124 is located upstream of the linear equalizer 112.

Thus, the linear equalizer 112 operates in the real format.

A transmission system (110) has a Nyquist filter (114) that operates upon a signal to shape signal spectrum. A first processed version of the signal is output from the Nyquist filter (114). A linear equalizer (112) of a system (110) operates upon the signal to pre-distort the signal in a manner that compensated or subsequent linear distortion. A second processed version of the signal is output from the linear equalizer (112). The linear equalizer

(112) is located in parallel with the Nyquist filter (114). A combiner (118) combines the first and second processed versions of the signal.

Claims

CLAIMS:

1. A transmission system comprising a Nyquist filter for operating upon a signal to shape signal spectrum and for outputting a first processed version of the signal, characterized in that a linear equalizer for operating upon the signal to pre-distort the signal in a manner that compensated for subsequent linear distortion and for outputting a second processed version of the signal, said linear equalizer being located in parallel with the Nyquist filter, and a combiner for combining the first and second processed versions.

2. A system as claimed in claim 1, characterized in that said system is a digital television system and includes a power amplifier and a post-amplification filter, said post- amplification filter causing the linear distortion that is compensated by said linear equalizer.

3. A system as claimed in claim 2, characterized in that said linear equalizer is adaptable to change the pre-distort imposed upon the signal, said system mcluding means for determining adaptation for the linear equalizer.

4. A system as claimed in claim 3, characterized in that said means for determining adaptation includes means for sampling the first processed version of the signal output from said Nyquist filter and prior to combination of the first and second processed versions of the signal by said combiner.

5. A system as claimed in daim 1, characterized in that said Nyquist filter has a first number of filter taps, said linear equalizer has a second, different number of filter taps, said system including means for delaying ,input of the signal to said linear equalizer, the delay being related to the first and second numbers of filter taps.