SYSTEM AND METHOD FOR DIGITAL MEMORIZED PREDISTORTION FOR WIRELESS COMMUNICATION

Abstract

A power amplifier system includes an input operable to receive an original value that reflects information to be communicated and an address data former operable to generate a digital lookup table key. The power amplifier system also includes a predistortion lookup table coupled to the address data former and a power amplifier having an output and coupled to the predistortion lookup table. The power amplifier system further includes a feedback loop providing a signal associated with the output of the power amplifier to the predistortion lookup table and a switch disposed in the feedback loop and operable to disconnect the predistortion lookup table from the output of the power amplifier.

Claims

1. A power amplifier system comprising: an input operable to receive a signal that reflects information to be communicated; an address data former operable to generate a digital lookup table key; a predistortion lookup table coupled to the address data former; a power amplifier having an output and an input, wherein the input to the power amplifier is coupled to the predistortion lookup table; a feedback signal representing the output of the power amplifier; and a switch coupled to the feedback signal and the predistortion lookup table, wherein the switch is configured to disconnect the feedback signal from the predistortion lookup table.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more fully describe embodiments of the present invention, reference is made to the accompanying drawings. These drawings are not to be considered limitations in the scope of the invention, but are merely illustrative.

[0016] FIGS. 1 and 2 show typical AM-AM and AM-PM behaviors of PA to be predistorted.

[0017] FIG. 3 shows a training schematic diagram of digital adaptive predistorter to linearize power amplifier in handset, which is an embodiment of the present invention.

[0018] FIG. 4 shows schematically a structure of adaptive predistorter in handset when the training finished, which is an embodiment of the present invention.

[0019] FIG. 5 shows schematically the lookup tables' arrangement in the adaptive digital predistortion scheme.

DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS

[0020] The description above and below and the drawings of the present document focus on one or more currently preferred embodiments of the present invention and also describe some exemplary optional features and/or alternative embodiments. The description and drawings are for the purpose of illustration and not limitation. Those of ordinary skill in the art would recognize variations, modifications, and alternatives. Such variations, modifications, and alternatives are also within the scope of the present invention. Section titles are terse and are for convenience only.

[0021] Preferred embodiments of the present invention relate to a novel digital adaptive predistorter to linearize power amplifier (PA) in RF transmitter of mobile stations, including variety of mobile potable equipment, handsets and PDA, for CDMA, TDMA, 20 GSM, GPRS, 3G (UMTS, W-CDMA, CDMA2000, 3GPP and others), WLAN system that transmits the complex modulated signal with aid of quadrature modulator and power amplifier. Because power amplifier in RF transmitter distorts RF output signal, the digital predistorter is used to correct non-linearity of PA by predistortion in opposite sense to PA input. The preferred circuit arrangement in embodiments of the present invention is specially designed for all wireless mobile stations or handsets, and also can be used in base stations or access points and other wireless communication systems such as microwave and satellite communications. Preferred embodiments of the present invention present a practicable DSP-based predistortion algorithm and organization using to handset without increasing manufacture cost but improving greatly handset performance.

[0022] In accordance with the preferred embodiment architecture, a presented digital adaptive predistorter is designed to linearize wireless RF transmitter for all handsets system, such as for CDMA, TDMA, GSM, 3G and WLAN system, and the like. For the application of predistortion technology in handset, the following issues are especially of interest: [0023] 1. Dynamic predistortion control range to track and correct non-linearity of PA in wider range; [0024] 2. Fast convergence speed to swiftly track characteristics of PA; [0025] 3. Requirement for low power consumption to extend handset battery life-time, which needs time-saving software algorithm and less complicity hardware structure; [0026] 4. Being able to use the existing source and chip room in handset for manufacture cost consideration.

[0027] Based on above considerations, a new design of adaptive memorized predistorter 15 for wireless handsets, according to an embodiment of the present invention has the following properties: [0028] 1. Using a stored compensation principle's structure to make the predistorter be able to memory great amount non-linear characteristics of PA for much better predistortion performance in wider dynamic range; [0029] 2. Using a time-delay adaptive structure to improve non-linearity tracking; [0030] 3. Using a simple architecture in implementation to use the existing DSP room in handsets without additional hardware circuit for low power consumption and low manufacture cost. 25

[0031] The components of embodiment architecture shown in FIG. 3 are described as follows: [0032] 00: Coder to generate the required modulation data under wireless system specifications [0033] 10 I-Q: Address data formers to obtain the required address data [0034] 11 I-Q: Base band symbol wave shaping filters [0035] 12 I-Q: N-bit vector multipliers to result the predistorted signals for I-Q channel respectively [0036] 13 I-Q: Digit to analog converters [0037] 14 I-Q: Analog reconstruction filters [0038] 15: Quadrature modulator [0039] 16: Nonlinear power amplifier [0040] 17: Antenna [0041] 20 I-Q: Address registers [0042] 21 I-Q: Memory tables to store linearity of PA and time-delay characteristics of channels [0043] 22 I-Q: Square circuits to generate the envelope of reference signals v.sub.d [0044] 23 I-Q: Envelope comparator to generate the error signals e.sub.p [0045] 24 I-Q: Switch T/C: T is in the training status when system is trained and C is in the calling status when system finishes training [0046] 25 I-Q: Square circuits of feedback channels to generate the envelope of feedback signals v.sub.f [0047] 26 I-Q: Analog to digit converters [0048] 27 I-Q: Analog receiving filters [0049] 28: Demodulator [0050] 29 Local oscillator to generate a high frequency carrier signal [0051] 30 I-Q: Address registers of lookup tables [0052] 31 I-Q: Predistortion lookup tables to store nonlinear characteristic of PA [0053] 32 I-Q: Adders of lookup tables [0054] 33 I-Q: Step size factors of adaptive algorithm [0055] 34 I-Q: Switch On/Off: On is when system is trained, and Off is when the training finished

Power Amplifier (PA) Model

[0056] Prior to further describing the predistorter according to embodiments the present patent, the characteristic of PA is discussed initially.

[0057] Power amplifier is the final stage of handset RF transmitter, and normally consumes the most electrical power. For the high efficient linear modulations such as QPSK or QAM, a linear high power amplifier must be used to maintain optimum spectral efficiency and low out-of-band emission. Such a linear amplifier, however, does not usually have a good DC to RF power conversion efficiency. For instance, traditional class-A power amplifier has a maximum theoretical power conversion efficiency of 50%. This is a major drawback for mobile application, especially for the portable battery operated equipment such as handsets, where battery life is of major importance. The power conversion efficiency can be obtained by using nonlinear power amplifier, such as class AB, C or D power amplifier. However, these amplifiers distort the input-output signal and cause spectral broadening and high out-of-band power emission of the output signal.

[0058] The PA in mobile communication system should typically be operated close to saturation or even saturated so as to maximize power efficiency. The saturation has serious repercussions on the signal to be amplified, and exhibits nonlinear characteristics such as amplitude and phase distortion that lead to an undesirable inter-modulation interference in the neighboring frequency band. The amplitude and phase characteristics used in this patent are shown in FIGS. 1 and 2, respectively.

Adaptive Predistorter

[0059] FIG. 3 shows the organization of handset digital predistorter from base band through RF part. The predistorter consists of address data formers 10I-10Q, linearity and time-delay lookup tables 21I-21Q, predistortion lookup tables 31I-31Q, vector multipliers 12I-12Q, error comparators 23I-23Q, and two types of switches 24I-24Q and 34I-34Q. All signals in the architecture are denoted by v(t) with the corresponding subscripts to express their location in system.

[0060] In accordance with a preferred implementation of the presented invention, the coded 1/Q data symbols to be transmitted are fed into both digital base band waveshaping filters 11I-11Q and address data former 10I-10Q, respectively.

[0061] The address data formers 10I-10Q are designed to generate the required binary signal format. The data formers 10I-10Q receive signal from coder 00 first, and then transform the received signal to the sign symbol with form either 0 or 1. The binary data are used as the address of both predistortion lookup tables 31I-31Q and linearity and time-delay lookup tables 21I-21Q.

[0062] The linearity and time-delay tables 31I-31Q are used to store the linearity characteristics of PA 16 and time-delay parameters of channels. The signal time-delay of channel is caused when I- and Q-signal from the output of shaping filters 11I-11Q pass through the circuit components, such as multiplier 12I-12Q, DAC 131-13Q, reconstruction filters 141-14Q, QM 15, PA 16, DQM 28, receiver filters 271-27Q, ADCs 261-26Q, squire circuits 251-25Q and switchers 24I-24Q to the comparers 23I-23Q. The 20 stored signals in the tables 21I-21Q are first obtained by adaptive iteration, and then are used as linear reference model to compare feedback signal that contains nonlinear distortion component. The lookup tables 21I-21Q are updated by adaptive algorithm in training period. When the training period is over, the required linear and time-delay information are stored in the tables for predistortion tables' training. As the result, there is no necessary to build a special time-delay circuit for channel time-delay estimate of handset.

[0063] The predistorter lookup tables 30I-30Q are used to store the predistortion signal to linearize power amplifier. The required predistorted signals in the tables are obtained by adaptive training and by comparing the outputs of tables 21I-21Q and feedback signals. The outputs v.sub.r(k) of lookup table 301-30Q are fed to multipliers 12I-12Q, and multiply with signals v.sub.m(k) from shaping filters 11I-11Q to result a predistorted signal v.sub.p(k) that is an inverse non-linearity of PA to predistort the input of PA.

[0064] The switch ON/OFF controllers 34I-34Q are set to ON status when predistortion lookup tables 31I-31Q are trained by adaptive algorithm. After the training procedure completed, the switches are set to Off status and the predistortion lookup tables 34I-34Q are no longer updated adaptively.

[0065] The switch T/C controllers 24I-24Q are used to choose handset circuit status. When the lookup tables are trained, the switches are set to Training status so that two types of lookup tables 21I-21Q and 31I-31Q are updated by adaptive algorithm. When the training processing is completed, the controllers are switched to the calling status and all lookup tables in handset stop being updated.

[0066] Handset is trained by two phases. First one is to obtain and store the linearity of PA and time-delay characteristics of channels to the linearity and time-delay tables. Second one is to generate the required predistortion signals and to store them into predistortion lookup tables. The training time for two phases will take about 0.5-1 second. Preferably, the training time for two phases is less than about 4 seconds. When all adaptive training finished, the handset predistortion can be implemented by the organization shown as FIG. 4. This is a very simple structure and may use the existing DSP room in handset for programming without adding any extra hardware circuit, and therefore is a cost-saving and high efficiency approach.

Predistortion Lookup Tables

[0067] The predistortion lookup tables for I- and Q-channel store the non-linearity of PA in inverse to AM-AM and AM-PM distortion to correct nonlinear distortion of RF transmitter. The predistortion lookup tables can be configured in a number of ways. However, a drawback of lookup table structure in traditional predistorter is memory less system, slow tracking speed and complicity structure, and therefore is unable to be used in handsets.

[0068] The new memorized predistorter structure presented in this patent introduces memory function into predistortion scheme by addressing operation. When the training procedure completed, the predistortion lookup tables just need to read out the signal stored in table entries according to the corresponding address for distortion correction. Thus, the predistortion of handset is completed by a simple and fast approach without great amount calculating for each correcting sampler so as to reduce of power consumption of handset.

[0069] The lookup tables of predistorter are based on a stored compensation principle that maps a set of input data into a digital output, and are updated adaptively by a time delay algorithm. The output signal of lookup table is actually related to the previous N transmitted data, and therefore has a memory function when compensate the nonlinearity of PA.

Linearity and Time-Delay Tables

[0070] The linearity and time-delay tables for I- and Q-channel are used to store the linearity of PA and time-delay characteristics of channel by adaptive training processing. In order to obtain linearity of PA, two larger sizes of lookup tables should be designed for I- and Q-channel, respectively, to obtain the linear reference signal with time-delay. The tables' outputs are compared with the feedback signals to result in the weighted error signals with nonlinear distortion for predistortion tables updating by a adaptive algorithm. After the training procedure finishes, the linearity and time-delay tables accomplish their mission.

Address of Lookup Table

[0071] The address of lookup table in predistorter is formed by the following approach. At first, the signal complex envelope of PA output can be expressed as

v.sub.a=v.sub.QG(|v.sub.Q|.sup.2)=H(v.sub.p)G(|H(v.sub.p)|.sup.2)  (1)

where v.sub.Q is the output of quadrature modulator, v.sub.p is the predistorted signal, H is an assumed transform function from DAC to quadrature modulator, and G is a level-dependent complex gain of power amplifier. We see from predistorter architecture of FIG. 3 that the complex gain predistorter is described by the following complex gain equation

vp=v.sub.mF(V.sub.m)=v.sub.mv.sub.r  (2)

where F is the mapping function of lookup table, which maps a N-dimensional vector V.sub.m to real output. In fact, the N-dimensional vector represents a set of N-bit address of lookup table, expressed by

V.sub.m=(sg.sub.1,sg.sub.2, . . . ,sg.sub.N).sup.T  (3)

where each binary symbol sg.sub.i in above vector can be obtained by the following expression

[00001]$\begin{array}{cc}s{g}_i==\{\begin{array}{ccc}1& v_m=1,& 1\le i\le N\\ 0& v_m=-1,& 1\le i\le N\end{array}& \left(4\right)\end{array}$

which transforms the data from coder to the binary sign with the form of 0 or 1.

[0072] Based on mobile system architecture, the two predistortion tables are used for I and Q-channel, respectively, to map any possible combinations of input binary symbol to the lookup table output, written as

v.sub.rI=F.sub.I(V.sub.mI) v.sub.rQ=F.sub.Q(V.sub.mQ)

where v.sub.rI=Re(v.sub.r), v.sub.rQ=Im(v.sub.r) and V.sub.mI=Re(V.sub.m), V.sub.mQ=Im(V.sub.m).  (5)

Adaptive Methodology for Predistortion Lookup Table

[0073] The adaptive memorized lookup table consists of address register and memory table. The address of lookup tables are formed in a shift register by first taking binary symbol by means of Eq. (3) and (4) and then sequentially feeding the binary symbol into the shift register. The address determines the corresponding entry of lookup table that stores information related to previous N transmitted data to correct non-linearity of PA.

[0074] Assume that the address of lookup table is formed by N-bits shift register, then the lookup table contains M=2.sup.N entries. Clearly, each output of lookup table is a function of last N transmitted data. Thus, there exist M possible estimates for PA nonlinearity at any compensation instant, and only one of them is selected as the output of lookup table by means of the address formed by N transmitted signal. The output signal is used to compensate and correct nonlinear distortion of PA.

[0075] The lookup table structure, based on the memory compensation principle, is involved only very simple logical operation and less complexity hardware structure, therefore, is better suitable to handset and higher bit rate wide band wireless communication system. We denote the address vector of table at the kth time as

V.sub.m(k)=(sg.sub.1(k),sg.sub.2(k), . . . ,sg.sub.N(k)).sup.T  (6)

All possible input vectors from k to previous k−M+1 time are expressed by the following set

A={V.sub.m(k),V.sub.m(k−1), . . . , V.sub.m(k−+1)}  (7)

which records M possible estimates of PA nonlinearity, denoted by

R(k)=(v.sub.r(k),v.sub.r(k−1), . . . ,v.sub.r(k−M+1)).sup.T  (8)

At the same time, only one of the estimates is read out from the lookup table

v.sub.r(k)=r.sub.q(k)(k) q(k)=V.sub.m(k)∈A∈{0,1}.sup.N  (9)

where the subscript q(k)=V.sub.m(k) denotes the address determined by N input binary symbol characters at kth time. Thus, the mapping function F of lookup table can be written by

F(V.sub.m(k))=r.sub.q(k)(k) V.sub.m(k)∈{0,1}.sup.N,r∈R  (10)

Note that F is unknown and hardly expressed mathematically before adaptive procedure beginning. However, F may be determined adaptively by updating lookup table entries under the adaptive algorithm to realize all possible mapping that corresponds to the relations with {0,1}.sup.N.fwdarw.R.

[0076] The entries of lookup table can be updated by the following iteration

[00002]$\begin{array}{cc}{r}_i\left(k+1\right)=\{\begin{array}{cc}{r}_i\left(k\right)& 1\le i\le M,i\ne q\left(k\right)\\ r_i\left(k\right)+\mu e_p\left(k\right)& i=q\left(k\right)\end{array}& \left(11\right)\end{array}$

where e.sub.p(k) is the error signal, and μ is the step size ranged from 0<μ<1 to control the convergence rate and steady-state of algorithm.

[0077] The error signal e.sub.p(k) contains both AM-AM and AM-PM components of PA so that the memorized predistorter could track adaptively variation of amplitude and phase characteristics of PA. The envelope error using to update lookup table entries is given by

e.sub.p(k)=v.sub.d.sup.2(k)−v.sub.f.sup.2(k)  (12)

where v.sub.j.sup.2(k) and v.sub.d.sup.2(k) express the envelopes of feedback signal v.sub.f(k) and reference signal v.sub.d(k) (12) respectively.

Adaptive Updating of Linearity and Time-Delay Lookup Table

[0078] The adaptive training for linearity and time-delay table is based on the traditional MSE updating algorithm. The entries of lookup table are updated by the following iteration

[00003]$\begin{array}{cc}{r}_i\left(k+1\right)=\{\begin{array}{cc}{r}_i\left(k\right)& 1\le i\le M,i\ne q\left(k\right)\\ r_i\left(k\right)+\mu e_i\left(k\right)& i=q\left(k\right)\end{array}& \left(13\right)\end{array}$

where the error signal e, (k) is given by

e.sub.i(k)=v.sub.d(k)−v.sub.f(k)  (14)

where v.sub.f(k) and v.sub.d(k) express the feedback signal with linear characteristic and reference signal that is the output of table, respectively.

(Other) Observations

[0079] Embodiments of the present invention may be adapted for use for all wireless systems regardless the modulation types (such as QAM, QPSK, OFDM and others) and PA models used in wireless systems including variety of mobile stations, handsets, base stations and access points such as, for example: [0080] Current wireless system: CDMA, TDMA, GSM, GPRS and their extension systems; [0081] Next generation broadband wireless system: CDMA2000, UMTS, WCDMA, 3GPP, WLAN (802.11 a, b specifications) and their extension systems; [0082] PDA and potable mobile PC for WLAN (802.11 a, b) system and their extension systems.

[0083] In an embodiment, an arrangement of lookup table address introduces a stored compensation function into predistorter. As the result, the predistorter is of memory function when it estimates and corrects nonlinear distortion of RF transmitter, which is extremely effective to provide a dynamic predistortion correction of AM-AM and AM-PM distortion in wider range.

[0084] In an embodiment, an arrangement the stored compensation function depends on the address data in lookup table. Therefore, each predistortion output of lookup table is not only related to the current input data but also related to last N transmitted data. Actually, the output of predistorter is a function of last N transmitted symbol.

[0085] In an embodiment, the predistortion correction for AM-AM and AM-PM distortion can be implemented by a measure of vector multiplication between the outputs of lookup tables and the output shaping filters. Also, other approach such as vector adding can be used to implement the predistortion correction. Because all entries of lookup table are adaptively updated by a weighted envelope error that contains both AM-AM and AM-PM characteristics of PA, the outputs of lookup tables contain the inverse AM-AM and AM-PM distortion components.

[0086] In an embodiment, the predistorter uses two types of lookup tables for different roles in handset predistortion. One is the predistortion table to provide predistortion signal for the non-linearity correction of PA. Another is the linearity and time-delay table to store the linearity of PA and time-delay characteristics of channel, which will be used as reference signal for the training of predistortion table.

[0087] In an embodiment, the linearity of PA and time-delay characteristics of channel are obtained by adaptive training. After the algorithm converges, the memorized information in the table is the reference signal required by predistortion table. Because the reference signal also contains the time-delay characteristics of channel of handset, there is no necessary to build a special time-delay circuit for estimating the time-delay of channel for the power consumption saving in handset.

[0088] In an embodiment, the linearity and time-delay table is trained by adaptive algorithm when PA of handset is set in the linear region so that the linear characteristics of PA could be pick up and stored into the entries of table. After algorithm converges, the required linear and time-delay characteristics of handset are stored in the table, and PA is set back to its non-linear operating region working as Class C or D amplifier.

[0089] In an embodiment, the linearity and time-delay table will be trained first to obtain the linear and time-delay characteristics of handset. Then predistortion table is trained by the reference signal from linearity and time-delay table to obtain the required predistortion signal and be stored in predistortion table for nonlinear distortion correction.

[0090] In an embodiment, the training for both linearity table and the predistortion table ill take around 0.5 to 1 second. Afterward, the linearity and time-delay table accomplished its mission and becomes static table without any output and being updated. Also the predistortion table is no longer updated by adaptive algorithm, and only responds to the input data in address register to output the corresponding predistortion signal. Thus, in an embodiment, the lookup tables in the predistorter will not be longer be updated after the training. For example, they will not be updated for at least one telephone call. For example, they will not be updated for at least one day of use. An actual handset predistortion structure is very simple and may use the existing DSP chip in handset for all predistortion signal processing.

[0091] Throughout the description and drawings, example embodiments are given with reference to specific configurations. It will be appreciated by those of ordinary skill in the art that the present invention can be embodied in other specific forms. Those of ordinary skill in the art would be able to practice such other embodiments without undue experimentation. The scope of the present invention, for the purpose of the present patent document, is not limited merely to the specific example embodiments of the foregoing description, but rather is indicated by the appended claims. All changes that come within the meaning and range of equivalents within the claims are intended to be considered as being embraced within the spirit and scope of the claims.