# SYSTEM AND METHOD FOR IM3 REDUCTION AND CANCELLATION IN AMPLIFIERS

Sets of power amplifier branches (which comprise an RF/microwave amplifier stage) are power combined within each stage and each set of power amplifier branches are biased in different classes of operation by bias circuits possessing different impedance characteristics such that the fundamental frequency components present at the output are in-phase with one another and the IMD3 components are anti-phase over a broad range of power levels. The RF input signal is provided by the output of the previous stage of the RF/microwave amplifier. The output of each stage is formed by power combining sets of these power amplifier branches, each of which are separately biased, so the fundamental components are additive resulting in the maximum possible output power and the IM3 components cancel partially or completely. IM3 reduction or cancellation can be achieved over a large range of output powers with the use of a feed forward control loop monitoring the input power and appropriately adjusting the bias currents and impedance characteristics of the bias circuits feeding the individual power amplifier branches in each stage of the RF/microwave amplifier.

## Latest VT SILICON, INC. Patents:

## Description

#### CLAIM OF PRIORITY

This application is a continuation of U.S. application Ser. No. 11/673,422, filed on Feb. 9, 2007, now U.S. Pat. No. 7,573,329, issued on Aug. 11, 2009, which claims priority to U.S. Provisional Application No. 60/771,924, filed on Feb. 9, 2006 and U.S. Provisional Application No. 60/850,146, filed on Oct. 6, 2006, the entire disclosures of which are incorporated by reference herein.

#### FIELD OF THE INVENTION

This invention generally relates to radio frequency/microwave amplifiers for communication devices, and in particular, to a system and method for automatically reducing third order intermodulation product levels to reduce signal distortion and maintain high data rates for communication systems over a wide range of frequencies.

#### BACKGROUND OF THE INVENTION

Amplifiers are used in all cell phones to amplify a modulated signal from a transceiver prior to passing the signal through the front end of the cell phone, which typically consists of passive components and a switch into an antenna where the signal is transmitted to a base station. **1** provides a signal to a transceiver **2**. The transceiver transmits an RF signal to a power amplifier **3**, which outputs an RF signal to a front end module **4** and out into the atmosphere by an antenna **5**. All of these components are housed within a typical cell phone **6**.

Two important factors driving the design of RF/microwave power amplifier **3** are linearity and efficiency. As used herein, linearity refers to the device's ability to amplify without distortion, and efficiency refers to the device's ability to convert DC power to RF/microwave power with as little wasted energy as possible. In conventional power amplifier designs, improvement in one area typically causes degradation in the other. An RF/microwave amplifier has two regions of operation: linear and nonlinear. In the linear region, the input signal envelope is amplified and no distortion is present at the output. For large peak to peak input signal levels, the amplifier enters the non-linear region and the output signal becomes distorted.

Distortion in RF/microwave amplifiers is generally caused by amplitude clipping, phase variations as a function of signal amplitude, and intermodulation products. Amplitude clipping occurs when the peak to peak input signal envelope amplitude extends beyond the linear region of the amplifier. Phase variations with signal amplitude also result when the peak to peak input signal envelope amplitude extends beyond the linear region of the amplifier. Intermodulation distortion (IMD) occurs as a result of nonlinearities in the amplifier transfer function resulting in mixing products being generated at the sum and difference frequencies of the input signals.

The third order intermodulation product (IMD3) is of great interest since this product is very close to the carrier signal on the frequency spectrum. Being located so close to the carrier signal, IMD3 is very difficult to eliminate or even reduce and is often the limiting factor in the linearity of the RF/microwave amplifier. The output third-order intercept point (OIP3), also known as OTOI (Output Third Order Intercept), is defined as the intersection point between the extrapolated 1:1 slope of the fundamental output power and the extrapolated 3:1 slope of the third-order intermodulation products. If the extrapolations are done well within the linear region, the OIP3 (OTOI) becomes a useful specification for predicting the linearity of the power amplifier. Thus, the higher the OIP3 (OTOI) point, the more linear the power amplifier. As mentioned above, reducing the IMD of the power amplifier improves its linearity and thus improves the OIP3 (OTOI).

Referring to **3** comprises one or more stages **7***a*, **7***b *. . . **7***n *connected in series by matching networks **8***a*, **8***b*, **8***c *. . . **8***n *that consist of a circuit topology of zero or more resistors, capacitors, and inductors. Each stage **7***a*, **7***b *. . . **7***n *consists of a number of branches **18***a*, **18***b *. . . **18***n *connected in parallel. Each branch **18** has of one or more unit cells **20** connected in parallel. A unit cell is composed of one or more transistors in a circuit topology of zero or more resistors, capacitors, and inductors.

Each power amplifier stage **7** is typically biased with a bias circuit (not shown) that provides the appropriate current or voltage for the branches to operate in a single class of operation. As used herein, a class of operation is determined by the percentage of an input sinusoidal signal during which the unit cells in each branch is on and conducting current. For example, in Class A operation, the branches are all biased such that they are on and conducting current for 360 degrees of the input sinusoid signal. In Class B, the branches are on and conducting for 180 degrees of the input signal. The near-Class B biasing condition is for the case where the branches are on and the conduction angle is close to but above 180 degrees. In Class AB, the branches are on and conducting typically at or around 270 degrees but may vary between 180 degrees and 360 degrees. The limits for each class of operation are not rigorously set and are used herein for purposes of general understanding of the operating condition of an amplifier.

Several methods have been tried to improve the response of power amplifiers. One method is to operate the RF/microwave amplifier at lower power levels to ensure that the device remains in the linear region. A drawback to this method is that when the device is operated at a lower power level it is operating less efficiently than it does at higher power levels.

The present invention seeks to improve the output response of RF/microwave power amplifiers by reducing IM3 levels and improving the OIP3 (OTOI) of the power amplifier. One solution to this problem is to reduce the distortion present in the linear and nonlinear region so that the operation of the amplifier can be extended into the nonlinear region where higher power levels are possible.

#### SUMMARY OF THE INVENTION

The present invention recognizes and addresses the foregoing considerations, and others, of prior art constructions and methods.

These and/or other objects are achieved in one preferred embodiment of a discrete amplifier for canceling at least one distortion component at the output of the discrete amplifier comprising a first stage having at least a first branch and a second branch in parallel with one another, wherein each of the first and second branches is formed of one or more transistor legs having one or more unit cells in parallel with one another and a second stage having at least a third branch and a fourth branch in parallel with one another, wherein each of the third and fourth branches is formed of one or more transistor legs having one or more unit cells in parallel with one another. The discrete amplifier further comprises a first biasing circuit having a first impedance and a first biasing level, the first biasing circuit being operatively connected to the first stage first branch for biasing the first branch into a first mode of operation, a second biasing circuit having a second impedance and a second biasing level, the second biasing circuit being operatively connected to the first stage second branch for biasing the second branch into a second mode of operation, a third biasing circuit having a third impedance and a third biasing level, the third biasing circuit being operatively connected to the second stage third branch for biasing the third branch into a third mode of operation, and a fourth biasing circuit having a fourth impedance and a fourth biasing level, the fourth biasing circuit being operatively connected to the second stage fourth branch for biasing the fourth branch into a fourth mode of operation. At least one of the first mode of operation and the second mode of operation, and the third mode of operation and the fourth mode of operation are chosen so that at least one distortion component is substantially canceled at a respective output of at least one of the first stage and the second stage.

In another embodiment, the first biasing circuit first impedance and first biasing level differs from the second biasing circuit second impedance and second biasing level. In other embodiments, the impedance of the first and second biasing circuits and a biasing level of the first and second branches are different, and the impedance of the third and fourth biasing circuits and a biasing level of the third and fourth branches is different and are chosen so as to maximize the cancellation of the at least one distortion component at the output of at least one of the first stage and the second stage of the discrete amplifier.

In yet other embodiments, the first, the second, the third and the fourth branches are formed on a single integrated circuit chip and the first biasing circuit, the second biasing circuit, the third biasing circuit and the fourth biasing circuit are physically located on the same single integrated circuit chip to minimize electrical parasitics so as to maximize the cancellation of the at least one distortion component. Other embodiments further comprise a sensing circuit that senses the input power to the discrete amplifier and causes at least one of the first, the second, the third and the fourth biasing circuits to adjust the operating mode of its respective branch by adjusting at least one of the bias circuit current and impedance so as to maximize the cancellation of the at least one distortion component.

In some embodiments the first mode, the second mode, the third mode and the fourth mode are chosen from one of Class A, Class B, Class C, and Class AB modes of operation, and the first, the second, the third and the fourth biasing circuits are adjusted to change the impedance of the respective biasing circuits to maximize the cancellation of the at least one distortion component.

In another embodiment, a circuit for canceling at least one distortion component at the output of the circuit comprises a power splitter having a first input port, a first output port, and a second output port, wherein a first signal is input on the power splitter first input port and is split into a second signal at the power splitter first output port and a third signal at the power splitter second output port, and a power combiner having a first input port, a second input port, and a first output port. A first amplifier has an input port coupled to the power splitter first output port, and an output port coupled to the power combiner first input port. A second amplifier is in parallel with the first amplifier, the second amplifier has an input port coupled to the power splitter second output port, and an output port coupled to the power combiner second input port. At least one biasing circuit is operatively connected with one of the first and the second amplifiers for biasing the one of the first and the second amplifiers in a different class of operation than the other of the first and the second amplifiers, and at least one control logic circuit senses the first signal and causes at least one of a size of the one of the first and the second amplifiers to change based on the first signal, and a magnitude of a biasing level of the one of the first and the second amplifiers to change based on the first signal. A fourth signal is output from the power combiner first output port having at least one distortion component reduced.

In other embodiments, a second biasing circuit is operatively connected to the other of the first and the second amplifiers for biasing the other of the first and the second amplifiers into a class of operation different than the one of the first and the second amplifiers so that the at least one distortion component is reduced in the fourth signal. In some embodiments, the at least one distortion component is a third harmonic of the first signal. In yet other embodiments the modes of operation are chosen from one of Class A, Class B, Class C, and Class AB modes of operation. In other embodiments the power splitter and the power combiner are quadrature hybrids.

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate one or more embodiments of the invention and, together with the description, serve to explain the principles of the invention.

#### BRIEF DESCRIPTION OF THE DRAWINGS

A full and enabling disclosure of the present invention, including the best mode thereof directed to one of ordinary skill in the art, is set forth in the specification, which makes reference to the appended figures, in which:

Repeat use of reference characters in the present specification and drawings is intended to represent same or analogous features or elements of the invention.

#### DETAILED DESCRIPTION OF THE INVENTION

Reference will now be made in detail to presently preferred embodiments of the invention, one or more examples of which are illustrated in the accompanying drawings. Each example is provided by way of explanation of the invention, not limitation of the invention. In fact, it will be apparent to those skilled in the art that modifications and variations can be made in the present invention without departing from the scope and spirit thereof. For instance, features illustrated or described as part of one embodiment may be used on another embodiment to yield a still further embodiment. Thus, it is intended that the present invention covers such modifications and variations as come within the scope of the appended claims and their equivalents.

The present invention seeks to reduce or significantly cancel the IM3 levels in power amplifiers at one or both of the component level and the sub-component level of the amplifier. In a first embodiment shown in **110**, an amplifier stage consisting of two amplifiers **114** and **116** in parallel and biased in different classes of operation, and a post-amplifier quadrature hybrid **112**. The input, V_{in}, consists of a sum of sinusoids at two different frequencies expressed as:

*V*_{in}*=A *sin(*w*_{1}*t*)+*B *sin(*w*_{2}*t*)

The input signal V_{in }is passed to quadrature hybrid **110** positioned before the amplifier stage. One example of a suitable quadrature hybrid is the SHY550 90 degree hybrid coupler manufactured by Mid-Atlantic RF Systems, Inc. The input is divided into two signals with the same frequency components, but with a fraction of the power of the input and where one, V_{I}, is shifted by radians and the other signal, V_{Q}, is shifted by −π/2 radians resulting in a relative phase difference of 90 degrees:

These signals are then passed to amplifiers **114** and **116**, respectively. One example of suitable amplifiers are Model EIC5359-8 10-Volt internally matched power FET manufactured by Excelics Semiconductor, Inc. The signal amplification performed by each amplifier can be modeled respectively using the following approximate equations:

*V*_{A1}*=K*_{1}**V*_{1}*+K*_{2}**V*_{I}^{2}*+K*_{3}**V*_{I}^{3 }

*V*_{A2}*=K*_{1}*′*V*_{Q}*+K*_{2}*′*V*_{Q}^{2}*+K*_{3}*′*V*_{Q}^{3 }

where K_{i }represents the coefficients of the third-order voltage transfer function of amplifier **114**, and K_{i}′ represents the coefficients of the third-order voltage transfer function of amplifier **116**. One important feature of this amplifier design is that the K_{3 }term in amplifier **114** and the K_{3}′ term in amplifier **116** are equal and opposite. In addition to this condition, the phase of the fundamental frequency components are substantially equal resulting in maximum power combining of the fundamental components.

The following equations examine one set of IMD3 terms. The other IMD3 terms are similar in form and follow the same mathematical operations.

The outputs of the amplifiers are then passed to second hybrid **112**, which is the same type as the first hybrid and is used to combine the RF signals present at the output of the two amplifiers into one RF output signal. One suitable example of a hybrid is the Model SHY550 90 degree hybrid coupler from Midatlantic RF Systems, Inc. Hybrid **112** has two inputs (ports **1** and **4**) and two outputs (ports **2** and **3**). The “top” output port, port **2**, consists of the addition of a signal from port **1** with a fraction of the power and shifted by radians and a signal from port **4** with a fraction of the power and shifted by /2 radians. The “bottom” output port, port **3**, consists of the addition of a signal from port **4** with half the power and shifted by radians and a signal from port **1** with half the power and shifted by /2 radians. The bottom output port is taken to be the output of the device and, under ideal conditions, the IMD3 products will be substantially removed from the frequency spectrum and the fundamentals still remain. The following equations show what happens to the IMD3 terms present at the “bottom” output port, VO2.

For the one set of IMD3 terms that was derived above, we obtain the following:

As can be seen above, when K_{3}=−K_{3}′ this set of IMD3 terms will cancel.

Similarly all of the other IMD3 terms will cancel as well which leads to an improvement in OIP3. In particular, the reduction in IMD3 levels will move the intersection point between the extrapolated 1:1 slope of the fundamental output power and the extrapolated 3:1 slope of the third-order intermodulation products to a higher power resulting in a higher OIP3 value. Looking at the first order terms present in the amplifiers,

it can be seen that the output of quadrature hybrid **112** due to these terms is as follows:

This term reduces to:

If the magnitude and sign of K_{1 }are substantially equal to those of K_{1}′, then the fundamental term will be present at the output of quadrature hybrid **112**. Having the magnitude and sign of K_{1 }substantially equal to those of K_{1}′ also ensures that the phases of the fundamental components for both amplifiers are equal.

_{3 }on the bias level of the amplifier. For example, in the class AB mode of operation an amplifier has a K_{3 }equal to −α. In order for the second amplifier in the parallel configuration to achieve a K_{3}′ equal to α, the bias level for this amplifier must be changed. As the bias level is dropped, the K_{3}′ term crosses the zero axis and becomes positive and reaches α as the amplifier approaches the class B mode of operation. By choosing bias points from this graph where K_{3 }is substantially equal to −K_{3}′, the IMD3 components present at the output will be greatly reduced if not cancelled completely.

In yet another embodiment, the phase of K_{3 }can be changed by operating the amplifier in either the gain compression mode or the gain expansion mode. In the gain compression mode, the gain of the amplifier decreases with increasing input power whereas in the gain expansion mode, the amplifier gain increases with increasing input power. The output power at the fundamental frequency can be modeled using the following terms: K_{1}*A+K_{3}*A^{3}+K_{5}*A^{5 }where A is the amplitude of the input voltage signal and the K terms represent the Taylor series expansion coefficients that describe the amplifier's output voltage to input voltage transfer characteristics. The even order K terms do not contribute to the IMD3 components and hence have not been included in the equations. For gain expansion, the K_{3 }and K_{5 }terms are positive provided K_{1 }is positive and for gain compression, the K_{3 }and K_{5 }terms are negative. The third order IMD3 components for a two tone excitation also contain the K_{3 }and K_{5 }terms. This implies that the IMD3 components are in phase during gain expansion and 180 degrees out of phase with respect to the fundamental for gain compression. In RF amplifiers, this can be achieved by adjusting the biasing of the amplifier by varying the bias levels and/or using a biasing circuit block with specific input impedance values.

To ensure that the magnitudes of the K_{3 }terms are substantially equal in amplifiers **114** and **116**, the correct size must be chosen for each amplifier. The size of the amplifier is directly related to how much power the amplifier can produce. The amplifier itself is made of many single transistors connected in parallel. Each individual transistor has a standard size depending on the fabrication process. Each transistor also contributes a portion of the total output power when the transistors are connected in parallel. As the number of transistors connected in parallel increases, so does the total output power provided by the amplifier. Therefore, size, in this case, refers to the number of transistors multiplied by the amount of active area for each transistor.

When designing an amplifier for use in the present invention it is important to know how much output power is needed so that the appropriate number of parallel transistors can be used. This total number of transistors then sets the size of the amplifier. However, two amplifiers of the same size biased at different points will generate different IMD3 levels. For example, an amplifier of a certain size biased in the class AB mode of operation will generate significantly less IMD3 terms than an amplifier of the same size biased in the near-class B mode of operation at the same input power level. Thus the near-class B mode amplifier must be sized so that its IMD3 magnitudes are substantially equal to the IMD3 magnitudes of the class AB mode amplifier at the same input power levels. One method for determining the size is by testing a number of near-class B mode amplifiers of differing sizes for the appropriate input power levels to determine the appropriate IMD3 magnitudes. That is, as the IMD3 magnitudes of the two parallel amplifiers diverge, the improvement in OIP3 seen at the output of the output quadrature hybrid **112** will not be as great had the IMD3 levels been substantially equal in magnitude and opposite in phase. Moreover, the IMD3 magnitudes of amplifiers **114** and **116** may be substantially equal at one power level, but unequal at another.

As shown in **110** to control the size and biasing level of amplifier **116**. That is, the RF input signal V_{in }is sampled and fed into a control logic block **118** that generates a digital word or correction signal by comparing input power level V_{in }with levels stored in a look-up table, which are associated with the appropriate size and bias level for near-class B amplifier **116**. The correction signal controls the size of near-class B amplifier **116** by controlling the number of parallel transistors that are active in the near-class B amplifier by turning on and off certain portions of the near-class B amplifier.

More specifically, if control logic **118** determines that certain portions of the near-class B amplifier should be turned off, then some of the transistors in the near-class B amplifier are shut off by reducing their bias level to zero. If, on the other hand, control logic **118** determines that certain portions of the near-class B amplifier should be turned on, then some transistors that had their bias level at zero will be raised to the near-class B level. Bias control logic **118** is located in the bias circuitry for near-class B amplifier **116**. As the near-class B amplifier decreases in size, so does its IMD3 components. Likewise, as the near-class B amplifier increases in size, its resultant IMD3 components increase as well. As a result of the size adjustment capability, it possible to see OIP3 improvement across a wide range of input power levels.

**120**, and the information is used to decide what changes should be made to the near-class B amplifier at step **124**. At step **126** and **128**, a decision will be made whether to increase or decrease the size of the amplifier and to increase or decrease the bias of the amplifier, respectively, to enhance IMD3 cancellation. Once a decision is made at steps **126** and **128**, the appropriate changes are made at step **130** to the near-class B amplifier to make the magnitudes of the IMD3 components of the two amplifiers substantially equal, the phases of the IMD3 components 180 degrees out of phase with respect to one another, and the phases of the fundamentals equal.

One example of the effect of biasing in the proposed topology will next be described by the experimental results obtained using commercial off the shelf components at 5.7 GHz.

In the second case, one amplifier was biased in the class AB mode of operation as above while the other was biased in the near-class B mode of operation. With each amplifier in a different mode of operation, the IMD3 components are 180 degrees out of phase with respect to one another, but their magnitudes are substantially equal over a short input power range. This experiment is denoted by Class A&B in

**118** shown in

In yet another embodiment of the present invention, IM3 levels can be significantly reduced or canceled at the sub-component level. More particularly, the IM3 levels at each stage **7**_{a}, **7**_{b }. . . **7**_{n }of power amplifier **3** may be significantly reduced or canceled by biasing two sets of branches in a power amplifier stage in different conduction angles of operation as described by class of operation (i.e. Class A, Class B, Class AB) with two separates bias circuits. Each biasing circuit would have distinct output impedances, with the impedance at baseband frequencies being of particular interest. Baseband frequencies are defined as the range of modulation frequencies for the input signal of interest, which typically ranges from the kHz to tens of MHz. Biasing parallel branches allows the phase difference between the IM3 components of each set of branches to approach 180 degrees. Additionally, the difference in the output impedance at baseband frequencies for the bias circuits allows for additional phase shift for the IM3 components, enabling a 180 degree difference between the IM3 components to be achieved. The number of branches in each set is adjusted such that the magnitudes of the IM3 components are substantially equal. That is, because the amplification of the sets of branches must be substantially equal to maximize performance, the number of branches in each set is adjusted to compensate for amplification differences.

The conduction angle of operation (i.e. Class A, Class AB, Class B, near-Class B) of the power amplifier branches is adjusted by varying the dc operating current in the branches with the bias circuit. Typically, control voltages and/or control currents are varied in the bias circuit to change the dc operating current and hence the conduction angle of operation. The impedance of the bias circuit can be modified, in one implementation, by using a resistor in series with the bias circuit. The value of the resistor can be varied to obtain different impedance values.

Referring to **7***a *are shown. The first stage consists of the following three components: an on-chip power divider **10**, two sets of branches **12** and **14** belonging to the first stage of the amplifier with separate bias circuits B**1** and B**2** biasing each set of branches, followed by an on-chip power combiner **16**. Each set of branches **12** and **14** comprises one or more branches **18***a*, **18***b *. . . **18***n *of cells **11**. Cells **11** can consist of one or more transistors and other passive circuit elements. A second stage of power amplifier **7***a *is connected in series with the first stage by a matching network and consists of the following three components: an on-chip power divider **10***a*, two sets of branches **12***a *and **14***a *belonging to the second stage of the amplifier with separate bias circuits B**3** and B**4** biasing each set of branches, followed by an on-chip power combiner **16***a*. Each set of branches **12***a *and **14***a *comprises one or more branches **18***a*, **18***b *. . . **18***n *of cells **11**. Cells **11** can consist of one or more transistors and other passive circuit elements. The input, V**1**, consists of a sum of sinusoids at two different frequencies:

*V*_{1}*=A *sin(*w*_{1}*t*)+*B *sin(*w*_{2}*t*)

The input signal V_{1 }is passed to on-chip power divider **10**. One example of a suitable on-chip power divider is a microstrip line with a characteristic impedance of Z_{o }split into two quarter wave length lines with a characteristic impedance of Z_{o}*sqrt(2). The input is divided into two signals with the same frequency components, but with a fraction of the power of the input and where one, V_{2}, and the other signal, V_{3}, have no phase difference.

These signals are then passed to branch sets **12** and **14**, respectively. The signal received by each amplifier is amplified according to the amplifier's transfer function which may be approximately modeled using the following respective equations:

*V*_{4}*=K*_{1A}*×V*_{6}*+K*_{2A}*×V*_{6}^{2}*+K*_{3A}*×V*_{6}^{3 }

*V*_{5}*=K*_{1B}*×V*_{7}*+K*_{2B}*×V*_{7}^{2}*+K*_{3B}*×V*_{7}^{3 }

where K_{iA }represents the coefficients of the third-order voltage transfer function of branch set **12**, and K_{iB }represents the coefficients of the third-order voltage transfer function of branch set **14**. One important feature of this amplifier stage design is that the K_{3A }term in branch set **12** and the K_{3B }term in branch set **14** are substantially equal and opposite. In addition to this condition, the phase of the fundamental frequency components are substantially equal resulting in maximum power combining of the fundamental components. The following equations examine one set of IMD3 terms. The other IMD3 terms are similar in form and follow the same mathematical operations.

The outputs of branch sets **12** and **14** are passed to second on-chip power combiner **16** that may or may not be the same type as first on-chip power divider **10** and used to combine the RF signals present at the output of the two branch sets into one RF output signal. As can be seen above, when K_{3A}=−K_{3B }this set of IMD3 terms will cancel. Similarly all of the other IMD3 terms will cancel as well leading to an improvement in OIP3. To specify, the reduction in IMD3 levels will move the intersection point between the extrapolated 1:1 slope of the fundamental output power and the extrapolated 3:1 slope of the third-order intermodulation products to a higher power resulting in a higher OIP3 value. Looking at the first order terms present in amplifier branches **12** and **14**,

It can be seen that the output of on-chip power combiner **16** due to these terms is as follows:

If K_{1A }is substantially equal to K_{1B}, then the fundamental term will be present at the output of on-chip power combiner **16**. Having K_{1A }substantially equal K_{1B }also ensures that the phases of the fundamental components for both sets of branches are equal.

To ensure that the magnitudes of the IMD3 components are substantially equal in parallel sets of branches **12** and **14**, the correct size must be chosen for each set of branches. The size of each set of branches is directly related to how much power it can produce. Two branches with the same number of unit cells biased at different points (i.e. different conduction angles of operation for each branch) will generate different IMD3 levels. For example, a branch with a certain number of unit cells **20** biased in the class AB mode of operation will generate significantly less IMD3 terms then a branch of the same size biased in the near-class B mode of operation at the same input power level. Thus, the near-class B mode amplifier must be sized so that its IMD3 magnitudes are substantially equal to the IMD3 magnitudes of the class AB mode amplifier at the same input power levels. One method for determining the size is by testing a number of near-class B mode branch sets to determine the appropriate IMD3 magnitudes. That is, as the IMD3 magnitudes of the two parallel set of branches diverge, then the improvement in OIP3 seen at the output of the output power combiner **16** will not be as great had the IMD3 levels been substantially equal in magnitude and opposite in phase. Moreover, the IMD3 magnitudes for branch sets **12** and **14** may be substantially equal at one power level, but unequal at another.

In addition to the size of the individual branch sets, the number of branch sets in a particular amplifier stage may also change based on the application of the amplifier. Thus, if a stage in the power amplifier has three branch sets, then on chip power dividers **10** and **16** must be designed to split input V_{1 }three ways and also to combine the output of three branch sets into a single output signal V_{6}. Thus, it should be understood that the number of branch sets in any one amplifier stage **7** may vary depending on the application of power amplifier **3**.

In yet another embodiment, the phase of K_{3 }can be changed by operating the amplifier in either the gain compression mode or the gain expansion mode. In the gain compression mode, the gain of the branch or set of branches decreases with increasing input power whereas in the gain expansion mode, the branch gain increases with increasing input power. The output power at the fundamental frequency can be modeled using the following terms: K_{1}*A+K_{3}*A_{3}+K_{5}*A^{5 }where A is the amplitude of the input voltage signal and the K terms represent the Taylor series expansion coefficients that describe the branch's output voltage to input voltage transfer characteristics. The even order K terms do not contribute to the IMD3 components and hence have not been included in the equations.

For gain expansion, the K_{3 }and K_{5 }terms are positive provided K_{1 }is positive, and for gain compression, the K_{3 }and K_{5 }terms are negative. The third order IMD components for a two tone excitation also contain the K_{3 }and K_{5 }terms. This implies that the IMD3 components are in phase with respect to the fundamental during gain expansion and anti-phase with respect to the fundamental for gain compression. In RF amplifiers, this can be achieved by adjusting the biasing of one or more branches by varying the bias levels and using biasing circuits with different impedances.

The range of power levels where improvement is observed may vary as this technique is used in different semiconductor IC processes, and the incorporation of a control loop may assure the improvement over a broad power range. A common problem is that the K_{3 }term, as related to the mathematical derivations, may vary as a function of RF input power. Thus, a control loop may be used to compensate for the changes in K_{3 }by adjusting the bias level (i.e. conduction angle) and the impedance of the bias circuit of both power amplifier branch sets. This dynamic compensation for the changes in K_{3 }as a function of RF power allows the IMD3 magnitudes and phases to be adjusted for optimal IM3 cancellation/reduction over a wide range of input powers.

Referring to **22** is shown connected from input V_{1 }to the biasing circuit (not shown) for branch sets **12** and **14** of amplifier stage **7***a*. A correction signal **24** is generated by monitoring the RF input signal V_{1 }to control the biasing level and impedance of the bias circuit (not shown) of branch sets **12** and **14** as a function of power. That is, the RF input signal V_{1 }is sampled and fed into control logic block **22**, which generates the appropriate bias levels for the applied RF input power. These bias levels may be generated by any number of means, such as (but not limited to) comparing the input power level V_{1 }with levels stored in a look-up table or by using a circuit block to directly generate the respective bias levels for the specific input power level. Correction signal **24** controls the bias levels and bias circuit impedance of branch sets **12** and **14** by varying the control voltages for the bias circuit (not shown). More specifically, if control logic **22** determines that branch set **14** should be biased more towards a Class AB mode, then the control voltages of the bias circuits will be adjusted accordingly. Bias control logic **22**, in one implementation, can be located in the bias circuitry for branch set **14** and also in branch set **12**. Also, the impedance of the bias circuit can be adjusted with switches that turn on or off segments of a parallel resistor array (not shown). As a result of the bias/impedance adjustment capability of the bias circuit (not shown), OIP3 improvement across a wide range of power levels can be obtained.

Referring to **22** is shown. Input power level V_{1 }is first sensed at step **26**, and the information is used to determine the changes that should be made to branch sets **12** and **14** by comparing the sensed signal to, for example, (1) values stored in a look-up table, in step **28** or (2) converting the sensed signal to the respective bias/impedance levels for the bias circuits. The possible changes to the bias circuits are: (1) change the impedance of the bias circuit, step **34** or (2) increase/decrease the bias of the branches, step **32**. Once a decision is made in control logic block **22**, the appropriate changes are made to branch sets **12** and **14**, at step **36**, to make the magnitudes of the IMD3 components of branch sets **12** and **14** substantially equal, the phases of the IMD3 components 180 degrees out of phase with respect to one another, and the phases of the fundamentals nearly equal.

The above described amplifier improvements for IM3 reduction results in very wide band solutions and can be implemented completely on-chip. The techniques are based on a unique power amplifier topology combined with a novel biasing technique at the amplifier branch level. The techniques are very flexible and can be applied to any IC process (i.e. Si CMOS, SiGe BiCMOS, GaAs HBT, etc. . . . )

It should be appreciated by those skilled in the art that various modifications and variations can be made in the present invention without departing from the scope and spirit of the invention. It is intended that the present invention cover such modifications and variations as come within the scope and spirit of the appended claims and their equivalents.

## Claims

1. A discrete amplifier for canceling at least one distortion component at the output of the discrete amplifier comprising at least a first stage and a second stage:

- a. said first stage having at least a first branch and a second branch in parallel with one another, wherein said first stage first branch operates in a first mode of operation, and said first stage second branch operates in a second mode of operation;

- b. said second stage having at least a third branch and a fourth branch in parallel with one another, wherein said second stage third branch operates in a third mode of operation, and said second stage fourth branch operates in a fourth mode of operation;

- wherein one of said first, said second, said third and said forth modes of operation differs from another one of said first, said second, said third and said forth modes of operation so that at least one distortion component is substantially canceled at an output of said second stage.

2. The discrete amplifier of claim 1, further comprising:

- a. a first biasing circuit operatively coupled to said first branch for operating said first branch in said first mode;

- b. a second biasing circuit operatively coupled to said second branch for operating said second branch in said second mode;

- c. a third biasing circuit operatively coupled to said third branch for operating said third branch in said third mode; and

- d. a fourth biasing circuit operatively coupled to said fourth branch for operating said fourth branch in said fourth mode.

3. The discrete amplifier of claim 2, wherein one of a first impedance and a first biasing level of said first biasing circuit differs from one of a second impedance and a second biasing level of said second biasing circuit.

4. The discrete amplifier of claim 2, wherein and chosen so as to maximize the cancellation of said at least one distortion component at said output of said second stage of said discrete amplifier.

- a. one of an impedance and a biasing level of said first biasing circuit differs from one of an impedance and a biasing level of said second biasing circuit, and

- b. one of an impedance and a biasing level of said third biasing circuit differs from one of an impedance and a biasing level of said fourth biasing circuit,

5. The discrete amplifier of claim 1, wherein each of said first, said second, said third and said fourth branches are formed from a plurality of transistor legs.

6. The discrete amplifier of claim 5, wherein each of said plurality of transistor legs in said first, said second, said third and said fourth branches are formed from a plurality of unit cells coupled in parallel with one another.

7. The discrete amplifier of claim 6, further comprising a sensing circuit that senses an input power to said discrete amplifier and causes a size of at least one of said first branch, said second branch, said third branch and said fourth branch to either increase or decrease depending on said sensed input power.

8. The discrete amplifier of claim 7, wherein said size of said at least one of said first branch, said second branch, said third branch and said fourth branch is decreased by turning off at least one of said plurality of transistor legs in said one of said first branch, said second branch, said third branch and said fourth branch.

9. The discrete amplifier of claim 7, wherein said size of said at least one of said first branch, said second branch, said third branch and said fourth branch is decreased by turning off a portion of at least one transistor leg in said at least one of said first branch, said second branch, said third branch and said fourth branch.

10. The discrete amplifier of claim 1, wherein said at least one distortion component is a inter-modulation distortion component.

11. A discrete amplifier for canceling at least one distortion component at the output of the discrete amplifier comprising at least a first stage and a second stage:

- a. a first stage having at least a first branch and a second branch in parallel with one another, wherein each of said first and second branches is formed of at least one transistor leg having one or more unit cells in parallel with one another, said first stage first branch operates in a first mode of operation, and said first stage second branch operates in a second mode of operation;

- b. a second stage having at least a third branch and a fourth branch in parallel with one another, wherein each of said third and fourth branches is formed of at least one transistor leg having one or more unit cells in parallel with one another, said second stage third branch operates in a third mode of operation, and said second stage fourth branch operates in a fourth mode of operation;

- c. a sensing circuit that senses an input power to said discrete amplifier and causes a size of at least one of said first branch, said second branch, said third branch and said fourth branch to either increase or decrease depending on said sensed input power;

- wherein one of said first, said second, said third and said forth modes of operation differs from another one of said first, said second, said third and said forth modes of operation so that at least one distortion component is substantially canceled at an output of said second stage.

12. The discrete amplifier of claim 11, wherein said size of said at least one of said first branch, said second branch, said third branch and said fourth branch is decreased by turning off at least one of said at least one of said transistor legs in said first branch, said second branch, said third branch and said fourth branch.

13. The discrete amplifier of claim 11, wherein said size of said at least one of said first branch, said second branch, said third branch and said fourth branch is decreased by turning off a portion of at least one of said at least one transistor leg in said first branch, said second branch, said third branch and said fourth branch.

14. The discrete amplifier of claim 11, wherein said at least one distortion component is a inter-modulation distortion component.

15. The discrete amplifier of claim 11, further comprising

- a. a first biasing circuit operatively coupled to said first branch for operating said first branch in said first mode;

- b. a second biasing circuit operatively coupled to said second branch for operating said second branch in said second mode;

- c. a third biasing circuit operatively coupled to said third branch for operating said third branch in said third mode; and

- d. a fourth biasing circuit operatively coupled to said fourth branch for operating said fourth branch in said fourth mode.

16. The discrete amplifier of claim 15, wherein one of a first impedance and a first biasing level of said first biasing circuit differs from one of a second impedance and a second biasing level of said second biasing circuit.

17. The discrete amplifier of claim 15, wherein and chosen so as to maximize the cancellation of said at least one distortion component at said output of said second stage of said discrete amplifier.

- a. one of an impedance and a biasing level of said first biasing circuit differs from one of an impedance and a biasing level of said second biasing circuit, and

- b. one of an impedance and a biasing level of said third biasing circuit differs from one of an impedance and a biasing level of said fourth biasing circuit,

## Patent History

**Publication number**: 20100033242

**Type:**Application

**Filed**: Aug 10, 2009

**Publication Date**: Feb 11, 2010

**Applicant**: VT SILICON, INC. (Atlanta, GA)

**Inventors**: Vikram Bidare Krishnamurthy (Smyrna, GA), Tanveer Kaur Khanijoun (Smyrna, GA), Kyle Mark Hershberger (Marietta, GA), Jeremy Thomas Reed (Atlanta, GA), Paul Erik Pace (Acworth, GA)

**Application Number**: 12/538,428

## Classifications

**Current U.S. Class**:

**330/124.0R;**Hum Or Noise Or Distortion Bucking Introduced Into Signal Channel (330/149); For Plural Stage Amplifier (330/200)

**International Classification**: H03F 3/68 (20060101);