WO2012166060A1

WO2012166060A1 - Method and device for predistortion of nonlinear wideband amplifier

Info

Publication number: WO2012166060A1
Application number: PCT/SI2012/000033
Authority: WO
Inventors: Matjaz BRESKVAR; Rok PIRNAT; Blaz KELBL
Original assignee: Beyond Devices D.O.O.
Priority date: 2011-06-03
Filing date: 2012-06-01
Publication date: 2012-12-06

Abstract

The presented invention discloses a system for nonlinear transfer function compensation of high power amplifiers especially in the case of wideband signal transmission. According to the presented invention the input signal comprising an inphase and quadrature component of an arbitrary modulated digital baseband signal is predistorted in predistortion circuitry and converted form baseband to the radiofrequency band. After digital to analog conversion the analog signal is further processed with shaping filter and amplified by high power amplifier. Predistortion of complex baseband signal is based on a process of splitting the input signal into upper and lower sideband. The individual bands are independently predistorted with complex multiplication with coefficients stored in two independent memories. Selection of the coefficient is based on the instantaneous power of the total input signal. Inphase and quadrature components of the upper and lower sidebands are pairwise summed into inphase and quadrature component of output predistorted signal.

Description

Method and device for predistortion of nonlinear wideband amplifier

The presented invention relates to the method for compensation of amplifier nonlinearity, especially in case of wideband signal transmission, as well as to the device based on that method.

To obtain the linear transfer characteristic of a high power amplifier with high efficiency in the entire range of the input signal is impractical in most cases. This is especially the case when modulation techniques with high peak to average power ratio are employed. As a consequence, nonlinear signal distortion and power spectral dilatation into adjacent frequency bands occupied by other communication channels or systems occur. The latter prevents their operation and infringes regulated power spectral density of the transmitted radio-frequency (RF) signal.

To avoid the aforementioned shortcomings and guarantee high transmitted power with high amplifier efficiency different techniques for compensation of high power amplifier nonlinearity are being used (i.e. nonlinear precorrection, nonlinear predistortion) with baseband digital predistortion being the most commonly employed.

An alternative solution represents frequency selective nonlinear predistortion which is based on the principle of splitting the signal into multiple narrow frequency bands. Each of the bands is independently processed through a circuitry for digital baseband predistortion with complex multiplication and the result combined into a final signal. Since implementing a large number of narrow band signals is technically too complex, a suitable approximation may be achieved by splitting the signal into two frequency bands, an upper frequency band above the carrier frequency and a lower frequency band below the carrier frequency. For the purpose of splitting the signal into two frequency bands the Hilbert transform may be used. However, since the presented solution is based on processing of a baseband modulated signal comprised of an inphase and a quadrature component use of the Hilbert transform is limited. Namely, the Hilbert transform does not transfer DC and low frequency components in its basic form suitable for implementation. This property limits its use to systems which lack a DC component in modulated signal (e.g. IEEE 802.1 l g) since generally modulated signals exhibit the highest power spectral density close to the DC component.

The goal of the invention is to devise a method for nonlinear predistortion of a wideband amplifier to address shortcomings of known solutions as well as to design a device for nonlinear predistortion of a wideband amplifier.

In accordance with the presented invention the inphase component and quadrature component of an arbitrarily modulated wideband signal are transformed into a real-valued signal at intermediate frequency whereby said real signal is split into an upper band and a lower band signal using filtering. Each band is further split into an inphase and a quadrature component which enables the technique for baseband digital predistortion with complex multiplication to be employed. The baseband signal scaled in amplitude and rotated in phase is converted into a radio signal with an arbitrary carrier frequency using a quadrature mixer and a digital to analog converter.

In the embodiment of the presented invention the device for nonlinear predistortion of a wideband amplifier includes a predistortion circuit fed by parallel inphase and quadrature components of an arbitrarily modulated wideband signal. Said predistortion circuit is cascaded with at least one corresponding mixer, followed by an adder, the output of which is further connected to a digital-analog converter. Said converter is cascaded with a signal shaping filter and a high power amplifier.

The preferred embodiment of the invention is set forth below with references to the drawings in which:

fig. 1 illustrates a device with nonlinear predistortion of the wideband amplifier in accordance with the preferred embodiment of the invention

fig. 2 illustrates the predistortion circuit of the device of fig.1 To achieve the compensation of wideband amplifier nonlinearity, the inphase and quadrature component of the arbitrarily modulated signal and their inverse components are fed into a multiplexer which uses sampling frequency to select between input samples. According to the invention the said inverse components are generated with a -1 multiplier arranged in front of said multiplexer. The multiplexer with the -1 multiplier at the input represents a very straightforward realization of a mixer which simplifies the entire procedure. Since a predistorted signal occupies a wider frequency band (compared to the input signal) a sampling frequency of at least four times the highest frequency component of the sampled signal is required to preserve integrity of the predistorted signal in discrete form regardless of the embodiment of the predistortion device.

The resulting signal generated from said multiplexer is then split into two equal parallel signals with the first signal fed into a lowpass filter rejecting the upper frequency band and the second signal fed into a highpass filter rejecting the lower frequency band from said signal. The filtered signals generated from each of the said filters are further fed into a lower and upper transmission circuit whereby, according to the presented invention, it is assumed that both transmission circuits are identical and mutually parallel. According to the invention the characteristic of said filters is chosen in such a way that the sum of their amplitude responses in the frequency domain forms unity.

In accordance with the invention it is further assumed that the characteristic of the lowpass filter passes signals unmodified in frequency range from the DC component to the fs/4 -fc frequency component where the characteristic of said lowpass filter equals unity (in a strictly mathematical sense) and does not allow signals to pass through a stopband of said filter above the fs/4 + fc frequency component. The transition band between frequencies fs/4 - fc and fs/4 + fc takes the form of a raised cosine in the range between 0 and π. It is further assumed that the characteristic of said highpass filter does not allow a signal to pass through a stopband from DC to fs/4 - fc frequency component and passes signals unmodified above fs/4 +fc where the characteristic of said highpass filter equals unity (in a strictly mathematical sense). The transition band between the frequency components fs/4 - fc and fs/4 + fc takes the form of a raised sine in the range between -π/2 and π/2. Here fs denotes sampling frequency and fc denotes the parameter of said filters to adjust the slope of the transition band of said filters. The signal generated from said lowpass filter is split through a demultiplexer in the lower transfer circuit into two signal pairs which consist of a signal and its inverse value generated with -1 multiplier. Each pair of a signal and its inverse value is fed from the demultiplexer into its own interpolation filter which alternately takes samples of a signal and its inverted value every second clock cycle. Consequently, the generated signal from each said interpolation filter represents the inphase and quadrature component of the lower frequency band. Similarly, a filtered signal generated from said interpolation filter of the upper transfer circuit represents the inphase and quadrature component of the upper frequency band. Each pair of said filtered signals generated from said pair of said interpolation filters is fed into a complex multiplier which simultaneously scales the inphase and quadrature component according to a selected signal from a table stored in a memory which is not mandatorily equal for the upper and the lower frequency band. The address of the table in said memory which determines the signal for amplitude scaling and phase rotation of a baseband signal is defined by piecewise logarithmic index according to the expression: i = Ki x (K.2 - ind) + rem, in which:

K], K2 denotes constants whose product equals the size of each said memory block;

ind denotes the value which in turn determines how many times the -instantaneous power a of non-predistorted signal needs to be doubled in order not to exceed a predefined threshold in front of a complex multiplier;

rem denotes the value dependent on the amplified instantaneous power of a non- predistorted signal.

Said instantaneous power of a non-predistorted signal sample is defined by the expression a = (lii + uii)² + (Iqi + uqi)², in which

/ _/ denotes the signal generated from the first filter of the lower transfer circuit which represents the inphase component of a lower frequency band signal,

denotes the signal generated from the first filter of the upper transfer circuit represents the inphase component of an upper frequency band signal,

Iq, denotes the signal generated from the second filter of the lower transfer circuit which represents the quadrature component of a lower frequency band signal, uqi denotes the signal generated from the second filter of the upper transfer circuit which represents the quadrature component of an upper frequency band signal. A pair of scaled signals generated from said multiplier of each transfer circuit is fed into an adder summing the inphase component of the lower frequency band with the inphase component of the upper frequency band and the quadrature component of the lower frequency band with the quadrature component of the upper frequency band. Each sum of said signals now represents said inphase component and said quadrature component of the predistorted signal which is further fed through the mixer of the device of the preferred embodiment of the invention. In a preferred embodiment of the invention the device for predistortion of a nonlinear wideband amplifier includes predistortion circuit 1 which is fed by parallel inphase component 2 and quadrature component 3 of an arbitrarily modulated wideband signal. Said predistortion circuit 1 transforms said components 2, 3 into inphase component 4 and quadrature component 5 of the predistorted signal fed from said circuit 1 through corresponding mixer 6,7 which converts the baseband signal into a radiofrequency signal and further into adder 8. Said components 4, 5 of the predistorted signal are summed in said adder 8 and fed further into a digital-to-analog converter as a single predistorted signal 9. The analog signal generated from said converter 10 is fed into signal shaping filter 11 and high power amplifier 12 transmitting resulting signal 13. In accordance with the invention the inverse amplitude-amplitude and amplitude-phase characteristic of the upper and lower output signal frequency band (relative to the input signal) of said amplifier 12 is stored in memory 14, 15 of said predistortion circuit.

In accordance with the presented invention said predistortion circuit 1 includes multiplexer 16 fed by said inphase component 2 and said quadrature component 3 of an arbitrarily modulated signal and their inverse components 2', 3'. Said inverse components 2', 3' are generated with -1 multiplier 17, 18 of each signal 2,3 arranged in front of said multiplexer 16. Said multiplexer 16 selects between said signals 2, 3, 2' and 3' at the input with sampling frequency. Since a predistorted signal occupies a wider frequency band (with regard to the input signal) the sampling frequency of at least four times the highest frequency component of the sampled signal is required to preserve integrity of the predistorted signal in a discrete form regardless of the embodiment of the predistortion device. Resulting signal 19 generated from said multiplexer 16 is split into two equal parallel signals with the first signal fed into lowpass filter 22 which rejects the upper frequency band from said signal 19 and the second signal fed into highpass filter 23 which rejects the lower frequency band from said signal 19. The filtered signal generated from each filter 22, 23 is fed into lower transfer circuit 20 and upper transfer circuit 21 whereby according to the invention it is assumed that both transmission circuits are identical and mutually parallel. The frequency characteristic of said filters 22, 23 is chosen in such a way that the sum of their amplitude responses forms unity. In a preferred embodiment of the presented invention the characteristic of said lowpass filter 22 passes signals unmodified in frequency band range from the DC component to the fs/4 - fc frequency component where the characteristic of said lowpass filter 22 equals unity (in a strictly mathematical sense) and does not allow signals to pass through a stopband of said filter 22 above fs/4 + fc frequency component. The transition band between frequencies fs/4 - fc and fs/4 + fc takes the form of a raised cosine in the range between 0 and π. Furthermore, the characteristic of said highpass filter 23 does not allow the signal to pass through a stopband from the DC to fs/4 - fc frequency component and passes signals unmodified above fs/4 + fc where the characteristic of said highpass filter 23 equals unity (in a strictly mathematical sense). The transition band between the frequencies fs/4 - fc and fs/4 + fc takes the form of a raised sine in the range between -π/2 and π/2. Here fs denotes sampling frequency and fc denotes the parameter of said filter 22, 23 to adjust the slope of the transition band of said filter.

According to the invention, said transmission circuit 20 includes demultiplexer 23, cascaded with a pair of parallel interpolation filters 25, 26. The signal generated from lowpass filter 22 is split by said demultiplexer 24 into two signals and through -1 multiplier also their inverse values. According to the presented invention said filters 25, 26 alternately take samples of a signal and its inverted value every second clock cycle. Consequently, signal 29 generated from said first filter 25 represents the inphase component of the lower frequency band and signal 30 generated from said second filter 26 represents the quadrature component of the lower frequency band. Similarly, signal 31 generated from said first filter 25 of the upper transfer circuit -21 represents the inphase component of the upper frequency band and signal 32 generated from said second filter 26 of the upper transfer circuit 21 represents the quadrature component of the upper frequency sideband. Signal pairs 29, 30; 31, 32 generated from each filter pair 25, 26 are fed into complex multiplier 33 which scales the amplitude of the inphase and quadrature component according to the selected signal from the table stored in memory 14, 1 . The address of said memory 14, 15 which determines the signal for amplitude scaling and phase rotation of a baseband signal is defined by piecewise logarithmic index according to the expression: i = Ki x (¾ - ind) + rem, in which:

Ki, K.2 denotes constants whose product equals the size of each said memory block 14,

15;

ind denotes the value which in turn determines how many times does the instantaneous power of the non-predistorted signal need to be doubled in order not to exceed a predefined threshold in front of a complex multiplier;

rem denotes the value dependent on the amplified instantaneous power of non- predistorted signal a.

Said instantaneous power of a non-predistorted signal sample is defined by the expression a = (lii + uii)² + (Iqi + uqi)², represented by circuit 35 in fig. 2 and in which:

/ ; denotes signal 29 generated from the first filter 25 of lower transfer circuit 20 which represents inphase component 36 of the lower frequency band signal, uii denotes signal 31 generated from the first filter 25 of upper transfer circuit 21 which represents inphase component 38 of the upper frequency band signal, Iqi denotes signal 30 generated from the second filter 26 of lower transfer circuit 20 which represents quadrature component 37 of the lower frequency band signal, uqi denotes signal 32 generated from the second filter 26 of upper transfer circuit 21 which represents quadrature component 39 of the upper frequency band signal.

The pairs of scaled signals 36, 37; 38, 39 generated from said multiplier 33 of each transfer circuit 20, 21 are further fed into adder 40, 41, where inphase component 36 of the lower frequency band is added to inphase component 38 of the upper frequency band and quadrature component 37 of the lower frequency band is added to quadrature component 39 of the upper frequency band. Each sum of each said signal pair 36, 38; 37, 39 represents said inphase component 4 and said quadrature component 5 of the predistorted signal fed through corresponding mixer 6,7 of the preferred embodiment of the invention.

The disclosed solution allows for independent dual band adjustment of frequency selective compensation of a high power amplifier's nonlinear characteristic. Its processing does not require feedback from the output and in case of identical table values in both said memory blocks performs equal to digital baseband predistortion with complex multiplication. In this case the inverse characteristic of a high power amplifier can be determined with the same methods as for digital baseband predistortion with complex multiplication. If all values in both tables equal unity the signal traverses the device unmodified.

Claims

Patent claims

1. The method for predistortion of a nonlinear wideband amplifier is characterized in that it includes the following steps:

a) predistortion of an input signal in a predistortion circuit whereby said input signal comprises an inphase component and a quadrature component of an arbitrarily modulated wideband digital signal;

b) conversion of said predistorted signal components from baseband to radiofrequency band;

c) addition of said components of a predistorted signal and forwarding of a single predistorted signal into a digital-to-analog converter;

d) conversion of said digital signal into an analog signal through digital-to-analog converter;

e) shaping of an analog signal with a filter; and

f) amplification of said analog signal with a high power amplifier with the resulting signal being transmitted at its output.

2. The method of claim 1 characterized in that the predistortion process consist of : a) multiplexing the inphase component of the digital input signal and the quadrature component of the digital input signal and their inverse signals into a real-valued signal with the carrier frequency equal to a quarter of the sampling frequency;

b) division of said real signal into a lower band signal and an upper band signal with filtering through lowpass and highpass filters;

c) feeding said lower frequency band signal and upper frequency band signal into corresponding demultiplexer;

d) interpolation of missing samples of said signal from each demultiplexer with lowpass interpolation filters;

e) phase rotation and amplitude scaling of each sample of said lower frequency band comprising an inphase and quadrature component with a signal from the lower band table;

f) phase rotation and amplitude scaling of each sample of said upper frequency band comprising an inphase and quadrature component with a signal from the upper band table;

g) addition of the predistorted upper frequency band inphase component and the predistorted lower frequency band inphase component into a predistorted signal inphase component;

h) addition of the predistorted upper frequency band quadrature component and the predistorted lower frequency band quadrature component into a predistorted signal quadrature component;

3. The method of claim 1 characterized in that the inverse amplitude-amplitude and amplitude-phase characteristics of the upper and lower frequency band of said high power amplifier output signal (with regard to the input signal) are stored in memory blocks of predistortion circuit.

4. The method of any of the previous claims characterized in that the address of each of the said tables depends on the sum of the squared sum of the upper and the lower frequency band inphase component and the squared sum of the upper and the lower frequency band quadrature component and is defined according to the expression /^' = Ki χ (¾ - ind) + rem.

5. The method of any of the previous claims characterized in that simultaneous upper and lower frequency band samples can be predistorted with different coefficients.

6. The method of any of the previous claims characterized in that the signal traverses the circuit completely unmodified if all coefficients in said tables of said memory blocks equal unity.

7. The method of any of the previous claims characterized in that the digital input baseband signal comprising inphase and quadrature component contains all frequency components from the DC component to the highest input signal frequency component.

8. The method of any of the previous claims characterized in that the output predistorted baseband signal comprised of the inphase and quadrature components contains all frequency components from the DC component to the highest predistorted signal frequency component.

9. The method of any of the previous claims characterized in that the inphase and quadrature input signal components are independent of the modulation technique.

10. The device for predistortion of a nonlinear wideband amplifier characterized in that it includes a predistortion circuit (1) fed in parallel by an inphase component (2) and a quadrature component (3) of an arbitrarily modulated wideband signal, whereby said predistortion circuit (1) is cascaded with a corresponding mixer (6, 7) followed by an adder (8) with its output fed further into a digital-to-analog converter (10) whereby the latter is coupled to a signal shaping filter (11) which is coupled with a high power amplifier (12).

1 1. The device of claim 10 characterized in that said predistortion circuit (1) comprises:

a) an input multiplexer (16) fed by a simultaneous inphase (2) and quadrature (3) component of an arbitrarily modulated wideband digital input signal and their inverse values (2', 3');

b) a highpass filter (23) fed by an output signal (19) from said multiplexer (16); c) a lowpass filter (23) fed by an output signal (19) from said multiplexer (16); d) a transfer circuit (20, 21) comprises a demultiplexer (24) fed by filtered signals from said filters (22, 23) and from which the two signals and through -1 multipliers (27, 28) also their inverted values are generated whereby said demultiplexer (24) is coupled to a pair of interpolation filters (25, 26); the complex multiplier (33) fed by pairs of signals (29,30; 31,32) from each pair of filters (25, 26) which scales the amplitude of the inphase and quadrature component based on a selected signal from the table stored in memory (14, 15); the circuit (35) for instantaneous power calculation of the entire signal with the output supplying the address of said memory (14,15) and an adder (40, 41) which adds the lower frequency band inphase component (36) to the upper frequency band inphase component (38) and the lower frequency band quadrature component (37) to the upper frequency band quadrature component (39) generated from each said multiplier (33).

12. The device of claim 10 and 11 characterized in that said highpass filter (23) characteristic does not allow any signals to pass through the frequency band from the DC component to the highest stopband frequency of the filter and passes signals unmodified above the lowest frequency for which the characteristic equals unity (in a strictly mathematical sense) whereby the transition band between said frequencies takes the form of a raised sine between -π/2 and nil.

13. The device of claims 10 and 1 1 characterized in that said lowpass filter (22) characteristic passes signals unmodified through the frequency band from the DC component to the highest frequency for which the characteristic equals unity (in a strictly mathematical sense) and does not allow signals to pass above the lowest stopband frequency of the filter whereby the transition band between said frequencies takes the form of a raised cosine between 0 and π.

14. The device of any of the claims from 10 to 13 characterized in that said interpolation filters (25, 26) alternately take samples of a signal and its inverse value every second clock cycle.

15. The device of any of the claims from 10 to 14 characterized in that the sampling frequency is at least four times higher than the highest frequency of the inphase component (2) and the quadrature component (3) of a digital input signal.