Composite power amplifier

SWEDEN (12) PATENT (13) C2 (id 522 479 (19) SE ₍₅₁₎

<img file="SE522479C2_D0001.tif" />

International class ⁷

H03F 1/02, 3/24 // H03F 1/07

PATENT AND ^ί62)

REGISTRATION! ”!

Patent filed Application widely available The patent application was submitted on expiration date

Tribal application number

2004-02-10

2003-07-17

2002-07-03

2002-07-03 (21) Patent Application Number () 202120-2

Application received as:

(83)

International filing day

Filing date for European patent application Deposit of microorganism Swedish patent application completed international patent application with number □ converted European patent application with number (30)

2002-01-16 SE 0200127-9 (73) (72) (74) (54) (56) (57)

PATENT HOLDER Telefonaktiebolaget LM Ericsson (publ), 126 25 Stockholm SE

INVENTOR Richard Hellberg, Huddinge SE

OMBUD Aros Patent AB

NAME Composite power amplifier

CALLED PUBLICATIONS: - - SUMMARY:

A composite power amplifier comprises a first and a second power amplifier (16, 18) connected to an input signal over an input network and to a load over an output network (21). The output network includes phase shift elements for generating different phase shifts from each power amplifier output to the common load. The input network includes means for powering both power amplifiers to form (1) first output current components with an amplitude which increases linearly with increasing output amplitude below a transition point (TP) and monotonically decreases with increasing output amplitude above said transition point, and (2) second output current which increases linearly with increasing output amplitude both below and above the transition point.

<img file="SE522479C2_D0002.tif" />

ΛΝΝΜΙ.Ι.Ι-.ΚΛ ICKI Ι.ΙΝίΛΚΙΊΊΊ

The numbers in brackets indicate international identification code, INID code. Letters in clamps indicate international document code.

522 479

SUMMARY

A composite power amplifier comprises a first and a second power amplifier (16, 18) connected to an input signal over an input network and to a load over an output network (21). The output network includes phase shift elements for generating different phase shifts from each power amplifier output to the common load. The input network includes means for driving both power amplifiers to form (1) first output current components with an amplitude that increases linearly with increasing output amplitude below a transition point (TP) and monotonically decreases with increasing output amplitude above said transition point, and (2) second output current which increases linearly with increasing output amplitude both below and above the transition point.

522 479

<img file="SE522479C2_D0003.tif" />

TECHNICAL FIELD

The invention generally relates to composite power amplifiers.

BACKGROUND

In many wireless communication systems, the power amplifier (PA) in the transmitter is required to be very linear in addition to being able to amplify many radio channels (frequencies) spread over a fairly wide bandwidth at the same time. It must also do this in an effective way to reduce power consumption and cooling requirements and to increase its service life. High linearity is required because non-linear amplifiers would cause interfering signal energy to leak between channels.

The amplitude distribution of a mixture of sufficiently independent radio frequency (RF) channels, or of a multi-user Code Division Multiple Access (CDMA) signal, tends to resemble a Rayleigh distribution with a large ratio of peak and average power. Since a conventional RF power amplifier generally has an efficiency proportional to its output amplitude, the average efficiency is very low for such signals.

Many methods have been proposed to address the low efficiency of conventional linear power amplifiers. Two of the most promising ones are the Chireix phase-out method [1] and the Doherty method [2]. Since they were published in 1935 and 1936 respectively, theories about implementations and improvements of these two RF amplifier systems have remained separate from each other. The function of the Chireix amplifier has thus been described in terms of constant voltage operation of the input amplifiers, with the phasing out of complex voltage with constant amplitude ("phasors") as a method for

522 479 obtain amplitude modulation. The efficiency enhancing effect of the Doherty amplifier has been described in terms of (dynamic) impedance modulation of the load or, which is equivalent, variable load distribution.

At least three problems with these previously known compound power amplifiers have been identified.

First, the output network of Chireix and Doherty amplifiers must be set very carefully to obtain the proper characteristics at the desired frequency. This is a costly and time-consuming procedure, especially when large quantities of transmitters are to be manufactured.

Secondly, the choice between manufacturing a Chireix or Doherty amplifier must be made very early in the design process, as the respective implementations are very different. Furthermore, it is difficult to change this decision later, as it leads to a lot of restructuring.

Third, neither the Chireix nor Doherty amplifiers are suitable for use with a plurality of relatively densely spaced bands (such as

1800/1900/2100 MHz).

In light of the problems just described, there is a need for an amplifier that works without excessive setting of the output network. There is also a need for an amplifier platform, where the choice between Chireix and Doherty25 implementation can be made late in the design process to match the advantages of each amplifier system to what the application requires. In addition, a need for an amplifier that works on a plurality of bands without re-engineering or setting changes can be identified.

522 479

<img file="SE522479C2_D0004.tif" />

SUMMARY

An object of the invention is a more flexible composite power amplifier, which can be used in both Doherty and Chireix type modes without excessive restructuring or setting change.

Another object is an improved composite amplifier of the Chireix type.

These objects are achieved in accordance with the appended claims.

In short, the invention provides a flexible, robust, possibly multi-band, effective composite amplifier by the application of several new techniques. The strategy is to rectify the input and output network structures and control the signal formation of Doherty and Chireix amplifiers to make the drive signals as equal as possible. In addition to the pure Chireix and Doherty modes, new efficient operating modes that lie between and outside these pure modes can also be used. The resulting continuum of high efficiency mode can be obtained by using an output network and a uniform way of operating the input network.

The modified output network can also be used in a new Chireix-type composite amplifier, thereby eliminating the need for the compensating reactants used in the prior art.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention, and further objects and advantages thereof, are best understood by reference to the following description taken in conjunction with the accompanying figures, in which:

Fig. 1 is a block diagram of a typical Chireix amplifier of the prior art;

522 479

Fig. 2 is a block diagram of a typical Doherty amplifier of the prior art;

Fig. 3 is a block diagram illustrating a Chireix output network that has been modified in accordance with the invention;

Fig. 4 is a block diagram illustrating the output network of Fig. 3 used as a Doherty output network;

Fig. 5 is a diagram illustrating the relationship between the amplitudes of the linear / non-linear components of the output current and the output voltage of a conventional Chireix amplifier;

Fig. 6 is a diagram illustrating the relationship between the linear / non-linear components of the output current and the output voltage of a Chireix amplifier modified in accordance with the invention;

Fig. 7 is a block diagram of an exemplary embodiment of one

Chireix amplifiers in accordance with the invention;

Fig. 8 is a model of the modified output network of the Chireix amplifier;

Fig. 9 is a block diagram of another exemplary embodiment of a Chireix amplifier in accordance with the invention;

Fig. 10 is a diagram illustrating the relationship between the amplitudes of the output amplifier output current and the output voltage of a conventional Doherty amplifier;

Fig. 11 is a diagram illustrating the relationship between the amplitudes of the linear / nonlinear components of the output current and the output voltage of a Doherty amplifier modified in accordance with the invention;

Fig. 12 is a block diagram of an exemplary embodiment of a composite amplifier in accordance with the invention;

Fig. 13 is a diagram illustrating the efficiency of a composite

0 amplifiers in accordance with the invention for various modes of operation; and

Fig. 14 is a diagram illustrating the efficiency of an asymmetric composite amplifier in accordance with the invention for various modes of operation.

522 479

DETAILED DESCRIPTION

In the following description, the same reference numerals for the same or similar elements are used throughout the drawings in the drawings.

Furthermore, the output networks of both Chireix and LINC amplifiers are referred to as output networks or Chireix-type combiners, although they are not identical.

Fig. 1 is a block diagram of a typical Chireix amplifier of the prior art. The term outphasing, which is the key method in Chireix and LINC amplifiers, generally stands for the method of obtaining amplitude modulation by combining two phase modulated signals of constant amplitude created in a signal component separator 10. After upconversion and amplification in RF chains 12, 14 (mixer, filter, amplifier) ​​and power amplifier (PA) 16, 18, the phased out signals are combined to create an amplified linear signal in a Chireixt output network 20. The phases of these phased out signals of constant amplitude are selected so that the result of their vector sum gives the desired amplitude. The output network 20 includes two quartz wavelength lines λ / 4 and two compensating reactances + jX and -jX, which are used to extend the high efficiency range to include lower output power levels. In [3, 4], the efficiency of Chireix systems is analyzed. [5, 6, 7] describe methods to overcome non-linearity due to imbalances in gain and phase. The Chireix method has also been used in radio transmitters under the brand name Ampliphase [6, 7).

An advantage of the Chireix amplifier is its ability to change the efficiency curve to fit different peak / medium power ratios by changing the size (X) of the reactants. The peak output value is distributed equally among the amplifiers despite this adjustment, which means amplifiers of equal size (capacity) can be used.

Fig. 2 is a block diagram of a typical Doherty amplifier of the prior art. The Doherty amplifier uses a linear and a non-linear

522 479 power amplifiers. According to the published theory, the main power amplifier 16 is operated as a linear class B amplifier and the auxiliary output current auxiliary power 18 (by class C operation or other technique represented by block 22) "modulates" the impedance experienced by the main amplifier through the impedance inverting quartz wave 2. , 10] in the output network 20. Since the non-linear output current from the auxiliary amplifier is zero value under a certain transient voltage (output voltage), the auxiliary amplifier does not contribute to the power loss below this voltage. An alternative output structure with built-in impedance matching is described in [11,

12].

For a standard Doherty amplifier, the transition point is at half the maximum output voltage. With this transition point, the efficiency curve is best suited for moderate peak / average power ratios and the peak power is equally distributed between the two input amplifiers. The transition point in

The Doherty amplifier can be changed by changing the impedance of the quartz wavelength transmission line (or an equivalent circuit). The efficiency curve can then be adjusted for higher peak / average power ratios and the peak power output value is evenly distributed between the amplifiers. Amplifiers of various sizes

0 is thus needed for optimal utilization of the available peak power.

Although the principles of the Chireix and Doherty amplifiers are very different, it will be shown below that it is in fact possible to combine these principles into a very effective composite power amplifier by appropriately modifying the input and output networks.

A first step towards this new composite amplifier is to change the structure of the Chireix amplifier output network to get rid of the reactants + jX and -jX. This can be done by using abbreviated and extended ones

0 variants of the quartz wavelength lines used in the prior art. The impedance Z1 of these is the optimal load impedance Ropt for a class B amplifier and the load impedance Rload should be half of this value. The shortened and extended (with equal Δ)

522 The 479 transfer lines function as quartz wavelength conduits with appropriate reactances. The new output network structure is illustrated as an output network 21 in Fig. 3.

If the operating frequency is changed, this output network structure can also be used to produce Doherty amplifiers, as illustrated in Fig. 4. The criteria that must be met are:

1st The length difference between the two transmission lines should be 1/4 of

Doherty amplifier wavelength λο.

2nd The total length of the transmission line around the output network should be (2n + l) * λϋ / 4, where n is a positive integer.

For example, if the electrical lengths of the wires are λο / 6 and λο / 3 at

The Chireix operating frequency is a Doherty operating mode at 1.5 times the Chireix frequency, where the electrical lengths of the transmission lines are instead λϋ / 4 and λϋ / 2 and where λο = 1.5λυ. This will be realized (with reference to Figures 3 and 4) by solving the equations:

_{2A =}lt _lc

6

2A = -

This gives λε = 1.5λϋ. The total length of the transmission line around the Doherty network is (the total length is the same in both cases):

A? j Oh? _ A? _

3 2 4 which means that the second criterion is also met.

522 479

At 2 times the first Chireix operating frequency there is another Chireix mode. Other relationships between the lengths of the wires give rise to other (sometimes useful) patterns of courage. The different modes can of course also be placed at the same frequency by changing the physical lengths of the transmission lines. Thus, to obtain the Doherty mode, the lengths of the transmission lines in the modified Chireix network are multiplied by a factor of 1.5.

A second step towards the new composite amplifier in accordance with the invention is to change the structure of the input network. Before describing this new structure, it is appropriate to first describe structures of typical Chireix and Doherty input networks according to the prior art.

The Chireix amplifier drive signal from the signal component separator 10 of Fig. 1 includes a linear plus or minus a non-linear component (plus for one power amplifier and minus for the other power amplifier). These components are illustrated in Fig. 5. By separating each phase-modulated signal of constant amplitude from a standard component signal separator 10 into a linear part and a modified non-linear part, changing the amplitude and phase of these components individually according to a set of rules and recombining the parts into a signal with new properties, it is possible to obtain a Chireix type amplifier with significantly lower drive power consumption than the standard Chireix amplifier.

Fig. 6 is a diagram similar to Fig. 5 illustrating the modification according to the invention. The basic idea of ​​the modification is to drive the composite amplifier linearly below a certain transition point TP In a regular Chireix amplifier, the amplitude of the drive signal is constant and the phase difference between the amplifiers is used to create amplitude modulation at the output. Below the transition point, this leads to excessive power consumption as the voltage and current of each power amplifier will become more and more out of phase. The output power decreases but the transistor RF current (which can be converted to direct current) does not decrease. At some point (defined below) it is

522 479 thus better reducing the driving force and keeping the phase difference constant instead of continuing with phase-out.

To generate the drive signals, the transition point at which the amplifier 5 should go from phase-out to linear operation must first be calculated. This point can be in one

Chireix amplifier power charts (or in a charger efficiency chart) are easily identified as the point where a straight line through the origin tangent to the power curve. It is also possible to calculate this point analytically.

It is also possible to use a linear approximation of the circle segment without any significant loss of efficiency, as indicated by the dashed line in Fig. 6. Below the transition point, the phase difference between the drive signals is kept constant and the amplitude is reduced linearly to zero to maximize efficiency.

Fig. 7 is a block diagram of an exemplary embodiment of a Chireix amplifier in accordance with the invention. The linear drive signal components are created directly from the input signal of amplifiers / phase shifters 26 and 32. Similarly, non-linear drive signal components are generated by a non-linear

0 element 38 and amplifier / phase shifter 28 and 30. The amplitude dependence is linear up to the transition point TP and follows a linear approximation of the circle segment after the transition point. For example, unit 38 may be implemented as a combination of a lookup table and subsequent D / A converters where a digital input amplitude is converted into appropriate drive signals (the input amplitude is assumed to be proportional to the amplitude of the composite amplifier output voltage). However, analogous implementations are also possible. The nonlinear signal from the amplifier / phase shifter 28 is added to the linear signal component from the amplifier / phase shifter 26 in an adder 34, while the nonlinear

0 the signal from the amplifier / phase shifter 30 is subtracted from the linear signal component from the amplifier / phase shifter 32 in an adder 36. For example, in an analog embodiment, the adder 34, 36 can be realized as hybrids. In a digital embodiment, they are digital adder. As indicated

522 479

<img file="SE522479C2_D0005.tif" />

With an antenna, the composite amplifier may be part of a transmitter, for example a transmitter in a radio terminal, such as a base station or a mobile station in a cellular system for mobile radio communication.

It will be appreciated that there is a common phase component in each of the drive signal components. Because of this, only three of the phases need to be adjusted.

One method of adjusting the amplitudes and phases of the linear parts (which do not require an ideal balanced output network) is to adjust them so that maximum output power and maximum efficiency are obtained at the maximum input excitation level (with otherwise linear behavior). This can be done in a model of the amplifier, to obtain adjustment factors, or in the amplifier itself. The result of this process is an amplifier performance and efficiency similar to those of a class B amplifier using the same transistors.

In this embodiment, the amplitudes and phases of the non-linear signal components are adjusted to minimize the mean of current consumption. This condition usually permits at least one of the transistors to have constant and maximum output voltage. Since this procedure is preferably performed when linearity of the output signal is guaranteed, the amplified non-linear signals should take each other out at the output. As will be shown below, such cancellation is possible. However, in order to explain how this is done, it is necessary to introduce a model over the output network.

Fig. 8 illustrates such an output network model. In this model, the active parts of the output of the amplifier transistors are modeled as linear controlled current generators. The final output conductances of the transistors are designated zpi and Zp2, respectively. Of particular interest is a quantity of how RF currents from the output nodes of the amplifiers 16 and 18 are converted to voltages at the output node of the composite amplifier. This quantity is represented

522 479 of the output impedances

ol ^zo \ = fl = 0 vo h

*2=0

If all components are assumed to be reasonably linear, overlay can be used to analyze this model.

Consequently, since the output network (including combiner 21 and input lines to this combiner) may be unbalanced, 201 may differ from

Z02. Since the non-linear signal parts should take each other out at the output, the linearity condition + 02 2 = 2 (1), where q 1 and q 2 are the adjusted non-linear components, is required. This is possible because the non-linear parts are identical except with respect to the sign (thus it is possible to replace the elements 28, 30 with a single amplifier / phase shifter 28). An easy way to meet this condition is by having a compensating filter in one of the nonlinear branches, to

0 example Zoi * Zo2^_1 in the lower branch ("*" stands for convolution in the time domain and multiplication in the frequency domain). By introducing broadband, frequency dependent transimpedances into the linearity condition, linearity is obtained in the output for a wide bandwidth. Thus, the non-linear portion is canceled in the output signal for all frequencies within any given bandwidth.

Another way of fulfilling the condition (1) is to insert the filter Z02 into the upper branch and the filter zoi into the lower branch. In this way, both branches are subjected to the composite filter zoi * Z02.

Fig. 10 is a diagram illustrating the relationship between output current and output voltage for an ideal conventional Doherty amplifier. Huvudeffekt30

522 The 479 amplifier 16 is operated with a linear signal, while the auxiliary power amplifier 18 is operated with a non-linear signal which is zero-value up to the transition point TP and then varies linearly. Thus, in a conventional Doherty amplifier, one amplifier is operated with a linear signal and the other amplifier with a non-linear signal. This is very different from a conventional one

Chireix amplifiers, where both amplifiers are powered by signals containing both linear and non-linear components.

Fig. 11 is a diagram illustrating the relationship between the components of the output current and the output voltage of a Doherty amplifier which has been modified in accordance with the invention. The idea behind this modification is to power both amplifiers with a linear and a non-linear drive signal component, just as in the Chireix case described above. This provides (ideally) the same output currents as in Fig. 10 from each power amplifier.

With this modification, it is possible to operate the input network of a Doherty amplifier in the same way as the modified Chireix amplifier described above (but at a different frequency). The same structure of the input network as in Fig. 9 can thus also be used for the modified Doherty driver.

Consequently, it is possible to use a uniform control strategy for both Chireix and Doherty type amplifiers. The central result of the elaborated theory for linear broadband operation of Doherty and Chireix amplifiers is that the non-linear portions of the input to the amplifiers in the output stage should and

5 can be canceled in the composite amplifier's output node.

When the triangular (or partially rounded or partially monotonically decreasing) non-linear "base" function is used, the linear function is obtained up to the maximum amplitude of the output by applying only the linear

0 base function. When only the linear base function is present, everyone is

RF voltages in the system are less than or equal to their end values ​​because they are linearly dependent on this function. The magnitude and gain of the linear basis function's signal paths can then be adjusted to achieve a

522 479 intermediate targets, ie maximum peak power at the output, without adjusting the signal path for the non-linear (triangular) function. This means that if the output network is properly designed, adjustments are usually made for the sum of the individual input amplifiers' peak effects. The amplifier (without the effect of the triangular base function) then functions as an optimal class B amplifier. For the uniform Chireix-Doherty output network according to the invention, this is possible at all frequencies.

The signal paths for the non-linear base function can be adjusted arbitrarily without being marked in the output signal as long as condition (1) is met. An additional requirement is, of course, that non-linear effects, such as saturation of the amplifiers, play no significant role. The purpose here is to minimize

DC power consumption (ie power consumption) by setting the correct size and phase for the non-linear base function's signal paths under these conditions. This maximizes efficiency. One result is that the RF voltages generally change at both output amplifiers but as long as the voltages are within the maximum allowable this will have no major consequences. Another parameter that can be adjusted is the transition point TP where the output voltage of the triangular base function has its maximum. This is useful, for example, if the optimum transition voltage is not known in advance. In this adjustment, an adjustment of the shape of the non-linear base function (to a rounded shape) can be incorporated if necessary.

5 The control strategy just described has the advantages that it can be used for all composite amplifiers, that the peak output value can be optimized separately, and that the efficiency maximization does not affect the output signal. The resulting structure with the adjustments is shown in Fig. 12. The letter G denotes gain, P stands for phase and TP means transition point.

The phase of the linear portion of one of the amplifiers (as mentioned above) can be omitted from the adjustments. These adjustments refer only to a frequency or operating mode. If broadband function is desired, the additional methods of [13] can be used.

522 479

Although the input network of Fig. 12 is provided before the RF chains 12, 14, it will be appreciated that the same principles can also be applied anywhere in or after these chains.

The uniform control strategy using the triangular and linear base functions works for all composite amplifiers. Amplifiers with the uniform output network are composite amplifiers at all frequencies with the possibility of at least class B function. It will be readily appreciated that the output network acts as a phase-in combiner as long as the signals from the two amplifiers are delayed by amounts that offset the difference in electrical length between the wires leading to the load. This means that it is possible to obtain linear Class B performance from the network at all frequencies. At certain (singular) frequencies, Class B is the best possible mode of operation, but at all other frequencies the efficiency can be increased at least to a certain extent compared to Class B function by using the methods proposed above.

A first class B mode, in addition to the trivial at frequency zero, is found (for the exemplary transmission line network having an electrical length of λο / 3 and Xc / 6) at three times the frequency of the first Chireix mode.

This means that it is possible to increase the efficiency at all frequencies between these two modes. In practice, the potential efficiency increase is not very large at frequencies close to the class B frequencies. However, for the exemplary network, an efficiency comparable to their efficiency may be

5 the first pure Chireix mode (at the first Chireix frequency and one octave higher) and the first Doherty mode (at 1.5 times the first Chireix frequency) are obtained all the way from about 0.7 to 2.3 times the first Chireix frequency. frequency. This is more than a 3-to-1 bandwidth. Efficiency curves for the Chireix mode, Doherty mode and mode at 0.7 and 1.2 times, respectively

0 The Chireix mode frequencies using the same triangular base function are shown in Fig. 13. For comparison, the efficiency of a Class B amplifier has also been indicated by a dashed line.

522 479

Examples of a similar network with other characteristics are power amplifiers with asymmetric capacity. We keep the previous sample lengths on the transmission lines, ie a 1: 2 ratio. Since the first amplifier has 56% of the second amplifier's capacity (output current capacity at equal supply voltage), the load, for an optimal combination, must be 64% of the second amplifier's optimal load. Each of the transmission lines has an impedance equal to the optimum load for each amplifier. The exemplary network is a Doherty network at the frequency where the transmission line from the first amplifier to the load is one-quarter wavelength and the second transmission line is half a wavelength. The optimal transition point for this Doherty mode is 0.36 times the maximum output voltage and ideally provides an average efficiency of 60% for the Rayleigh amplitude distribution with peak average value 10 dB for which it is optimized. What is then obtained at the "Chireix15 frequencies" of two-thirds and one-third of the Doherty frequency is the quasi-Chireix mod with optimal transition points and the corresponding efficiency curves comparable to a Doherty amplifier with equal amplifiers. An interesting point of view is that although these modes are at the Chireix frequencies, it is not possible to operate the amplifier like a traditional Chireix amplifier. This is because the amplifier capabilities (and the network branches) are asymmetric, so the traditional Chireix criterion for balancing the signal paths is not met and the amplifier would thus be non-linear. The efficiency curves (using triangular base functions with different transition points) for the Doherty25 mode, an intermediate mode at 0.8 times the Doherty frequency, and one of the quasi-Chireix modes are shown in Figure 14.

The description has so far dealt with somewhat idealized starting networks. In practice, the harmonics of the Class B amplifier's collector voltage (or

0 D-electrode voltage) is terminated in special ways for the transistors to be fully used. For example, the first harmonic must be terminated by shorting the collector to RF ground at the frequency of the first harmonic. The network that takes care of this usually has limited bandwidth,

522 479 both at the useful (fundamental) and the final (first harmonic) frequencies. This generally means that a continuous band with high efficiency mode within bandwidth limits determined by the requirements for harmonic finishes is obtained. However, to say that the networks "themselves" can be used in a 3-to-1 bandwidth with the methods described is the same as to say that the lengths of the transmission lines in the network can be anywhere within a 1-to-3 range when used at a single frequency as long as the harmonic ending is the right one for that frequency. The tolerance requirements can thus be very mild.

In order to achieve a broad usable bandwidth with high efficiency mode, it is presently preferred with clean ("distributed") transmission line implementations of the output network. Alternatively, output networks for different modes, or modbands, can be implemented with lumped elements or combinations of composite and distributed documents.

The described control scheme can advantageously be used for all traditional Chireix and Doherty networks, as well as for all faulty or otherwise malfunctioning variants of these. Thus, most of the hardware can

0 is designed regardless of which output network it is intended to be used with. This means that a common flexible "amplifier platform" can be built, based on which different properties are obtained through small changes in the output network. It is easy to change between different Doherty and Chireix output networks and switches between output networks optimized for different peak / medium power ratios can be made without the need to modify the rest of the hardware.

The ability of the proposed method to handle the output network of any composite amplifier and under these conditions create a

0 amplifiers with near-optimal efficiency make the amplifier robust. This means that any misalignment of the output network will have small consequences and last minute changes will not be particularly risky. Also, the proposed networks themselves are very robust against misalignment, as a very wide one

522 479 bands with high efficiency function can be obtained. If variants of an amplifier for different systems were still needed, they could be built for a very small extra cost. The amplifier can easily be made to work in a number of bands within the limits set by the condition for harmonic termination.

New (not known until now) operating modes with high efficiency and networks for providing these modes have been identified. Of particular interest is the ability to operate in a quasi-Chireix mode with different capacities of the input power amplifiers. High-efficiency modes located continuously between Doherty and Chireix modes, or between Chireix modes, are also useful, especially as a complement to the Chireix and Doherty modes.

The efficiency improvement is separate from the linear amplifier adjustments and also "invisible" for the output. The number of parameters to adjust is also quite low. This means that the control scheme is a good candidate for dynamic adaptive adjustment.

In summary, the proposed output networks enable the production of efficient, robust amplifiers that do not need much trimming and provide multi-band function and new useful high-efficiency modes. The uniform control scheme with rectified base functions gives a high tolerance to fault setting and enables common amplifier platforms, output networks that can be easily changed and adaptive adjustment. These two parts of the invention can be used separately with high profits. However, the combination of the two increases their respective abilities. Although it is preferable to use the described modified output network in a composite amplifier which also has the described modified input network, this modified output network itself also brings about an improvement in pure Chireix type amplifiers, as it eliminates the need for the reactors used in Chireix. amplifiers according to the prior art.

522 479

Those skilled in the art will recognize that various modifications and changes to the invention may be made without departing from the scope thereof, which is defined by the appended claims.

REFERENCES [1] H. Chireix, High Power Outphasing Modulation, Proc. IRE, vol. 23, no. 2, p. 1370-1392, nov. In 1935.

[2] WH Doherty, A New High Efficiency Power Amplifier for Modulated Waves, Proc. IRE, vol. 24, no. 9, p. 1163-1182, Sept. 1936th

[3] FH Raab, Efficiency of Outphasing RF Power Amplifier Systems, IEEE Trans. Communications, vol. COM-33, no. 10, p. 1094-1099, Oct. , 1985.

[4] B. Stengel and WR Eisenstadt, LINC Power Amplifier Combiner Method Efficiency Optimization, IEEE Trans. Vehicular Technology, vol. 49, no. 1, p. 229-234, Jan. In 2000.

[5] X. Zhang and LE Larson, Gain and Phase Error-Free LINC Transmitter, IEEE Trans. Vehicular Technology, vol. 49, no. 5, p. 19861994, sept. In 2000.

[6] AS Wright, SJ Bennett, US Patent US 6,054,896.

[7] RE Stengel, SA Olson, US Patent US 5,901,346.

[8] Ampliphase AM Transmission System, ABU Technical Review, no. 33, p. July 10-18, 1974.

[9] I. Ullah, Output Circuit of an Ampliphase Broadcast Transmitter, ABU Technical Review, no. 63, p. July 17-24, 1979.

522 479 [10] FH Raab, Efficiency of Doherty RF Power Amplifier Systems, IEEE Trans. Broadcasting, vol. BC-33, no. 3, p. 77-83, Sept. , 1987.

[11] DM Upton et al. A New Circuit Topology to Realize High Efficiency,

High Linearity, and High Power Microwave Amplifiers, IEEE Proc. RAWCON '98, p. 317-320.

[12] JJ Schuss et al., U.S. Patent US 5,568,086, Oct. 1996.

[13] International patent application WO 02/05421 A1, Telefonaktiebolaget LM Ericsson.

522 479