Power Amplifiers - Ikom.aau.dk/~hmi/Daniel/TCMT/mm.5.materials/mm.5... · 2007-02-21 · Class BClass B Power amplifier classification • This is an example of a class B power amplifier

Jan Hvolgaard Mikkelsen

Radio Frequency Integrated Systems and Circuits DivisionAalborg University

2007

Power Amplifiers - ITransceiver Components & Measuring Techniques

MM5

Based on material by Saska Lindfors and Tian Tong

Jan Mikkelsen -2-

Agenda

• Introduction to Power Amplifiers

• Classification of Power Amplifiers

• Design example

• System related PA parameters

• Power amplifier non-linearities

Jan Mikkelsen -3-

What is a Power Amplifier (PA)?

• Compared to the other blocks that go into the implementation of a normal wireless transceiver, the power amplifier is a very unique device

• As the name suggest the PA is a special amplifier that has been designed specifically for signal power amplification

• From an implementation point-of-view the PA is also special since it employs large sized active device to achieve the required amplification

• When active the PA always operates in a high power mode which calls for special design considerations compared to other transceiver blocks

• The non-linearity performance of the PA is very important as this more or less dictates the performance of the transmitter

• On top of that the PA induces a high degree of noise and it also results in a heating of the chip

Jan Mikkelsen -4-

Power amplifier integrated in a WLAN systemPower amplifier integrated in a WLAN system


Jan Mikkelsen -5-


Power amplifier integrated in a multiband GSM systemPower amplifier integrated in a multiband GSM system

Jan Mikkelsen -6-

The role of the Power Amplifier

• In both of these system examples the PA serves a number of purposes

Amplification of the signal power and control of the emission level

Drive the antenna load

Provide for isolation and impedance matching

• Compared to a more traditional amplifier the PA is slightly different

It amplifies power rather than voltage or current

The gain per stage is normally 10 dB or less

The size of the active device used in PAs is normally very large, up to a few thousands μm

High power implies high peak drain voltage and/or peak current

Very low output impedance, typically a few Ω, to drive the antenna load which implies that the output matching is tough

The non-linear performance is of significant importance as small-signal operation cannot be assumed.

Efficiency is of course very important in high power amplifiers

Jan Mikkelsen -7-

Power amplifier classification

• Power amplifier are classified in order to distinguish between the different functional principles upon which they are based

• In the good old days the classification was straightforwardly based on the signal conduction angle

• With the more modern switching types of amplifiers this has changed

• Still the mode of operation is used as reference

Linear power amplification (Class A)

Reduces signal angle power amplification (Classes B, AB and C)

Switched mode power amplifiers (Class D, E, F)

Jan Mikkelsen -8-

Class AClass A


• The class A power amplifier is characterized by having a signal conduction angle that is equal to 360o

• This implies that the active device is switched on for the full duration of the signal conduction angle

• From a biasing point of view this implies that the active device has to remain in the linear region always

Bipolar devices must not enter cut-off or saturation

MOS devices must operate in saturation

• This operation provides for verygood linearity performance

• The operation however also impliesrunning at a high bias level

• Very Low Power Added Efficiency(PAE)

Jan Mikkelsen -9-

Class AClass A


• This is an example of a class A power amplifier intended for 1 GHz use

• To accommodate the output signal this implementation operates with a bias current of no less than 825 mA which implies that the amplifier runs with a 2.72 Watt ‘DC’ power consumption

Jan Mikkelsen -10-

Class BClass B


• For class B amplifiers the active device is also operated in the linear region

Bipolar devices must not enter cut-off or saturation when switch on

MOS devices must operate in saturation when they are switch on

• Unlike class A the signal conduction angle is less than 360o for the classB. Here the active device is switch on for 180o (half circle of signal)

• The bias requirement is reducedas the device is shut down for half of the time

• Medium bias level

• Medium distortion level

• Medium PAE

• Since the Class B output stage is only ’on’ for halft the time it alsoprovides for only half the gain of a Class A stage

Jan Mikkelsen -11-

Class BClass B


• This is an example of a class B power amplifier intended for 2 GHz use

• The active device here ‘tracks’ the input signal during the positive half period

• When there is no positive signal swing at the input, no current flows through the active device

• Achieving exactly 180o is extremely difficult

• In practice most amplifiers operate in Class AB or C

Jan Mikkelsen -12-

Class CClass C


• Also for the Class C amplifier the active device is operated in the linear region when it is switched on

• The active device is now switch on for less than a 180o signal conductionangle

• This arrangemen results in a very low (oreven zero) bias current requirement

• This bias operation leads to a very non-linear operation which then of course leadsto a high level of distortion

• On the positive side the resulting PAE is high (depending on the actual signal conduction angle)

Jan Mikkelsen -13-

Class CClass C


• This is an example of a Class C power amplifier intended for 1 GHz use

• Compared to the Class A PA on slide 9 the biasing arrangement has been replaced by M2

• A simple resistor from the gate of M1 to ground would work as well

• Only requirement is that a high resistance is presented to the coupling capacitor

Jan Mikkelsen -14-

Class ABClass AB


• As a compromize between classes A and B there is the class AB power amplifier

• In a class B amplifier the device draws no current when it is not active

• For the class AB each device is biased at a low bias current always

• The active device is operated in linear region when on

• The signal conduction angle is between 180 and 360o

• Medium bias level (quiescentbias between cut-off point and class A point)

• Very good linearity

• Medium PAE

Jan Mikkelsen -15-

Class ABClass AB


• This is an example of a class AB power amplifier

Jan Mikkelsen -16-


• In summary, the schematics for Class A, B, AB, and C are very similar and it is more or less only in the biasing that we see a difference

• Reducing the conduction angle is an efficient method to enhance the efficiency of an amplifier

• The increase in efficiency comes at the cost of an increase in harmonic levels and hence a higher degree of distortion

Jan Mikkelsen -17-


• As a generic compromise the Class AB seems to be the best choice

• For RF applications running a constant-envelope signal scheme the linearity of the PA is less important and other more efficient, and non-linear, amplifiers can be used

• For a carrier based output signal a tuned load can be used to filter out the harmonic components to “reconstruct” the original signal shape

• The constant envelope applications the very non-linear but also very efficient PA classes are of great interest

Jan Mikkelsen -18-

Class DClass D


• In the case of a Class D PA the active device is operated in switch mode

• Two or more transistors are used to generate a square-wave like drainvoltage waveform

• A series tuned filter attenuates all but the fundamental

• Working in on/off mode (low or zero bias) leads to very good PAE

• Due to finite switching speed and parasitic effects the output will not bea square wave

Jan Mikkelsen -19-

Class EClass E


• A Class DE PA makes use of a single active device that is being operatedin switch mode

• A high-order reactive output network provides for enough freedom to shape the switch voltage to have zero value and zero slope at switch turn-on

• At switch turn-on the the transition is still less than ideal

Jan Mikkelsen -20-

Class FClass F


• The Class F PA also makes use of a single active device operated in switch mode

• It extents the Class E reactive load approach through the use of an output tank that is tuned to resonate at the carrier frequency and is assumed to have a high enough Q to act as a short/open circuit for all frequencies outside the wanted signal band

• As a result the only current that ideally flows into the load is thefundamental component

• Again the device runs at a very low or even zero bias

• As a result the Class F PA provides for a very goodPAE performance

Jan Mikkelsen -21-

Class FClass F


• As a conceptual solution the Class F approach is very efficient and alsovery elegant

• A quarter-wave transmission line makes sure that the drain sees a short at all even-order harmonicsand an open at all odd-orderharmonics

• In a practical implementation thetransmission line has to be avoidedand it is replaced by a lumpedcomponent approximation

• Often the 3ωο stage is sufficient

Jan Mikkelsen -22-

The first stageThe first stage

The second stageThe second stage

Output matchOutput match

Input matchInput match Internal matchInternal match

Exampel: Class E PA for UMTS Exampel: Class E PA for UMTS

Power amplifier design example

Bondwire

inductance

Bondwire

inductance

LoadLoad

Jan Mikkelsen -23-

Exampel: Class E PA for UMTS Exampel: Class E PA for UMTS

Power amplifier design example

• These performance measures represent a very practical Class E power amplifier

• The topology is a 2 stage Class E

• The center frequency is 1.9 GHz with a 200 MHz bandwidth

• The PA is implemented in a UMC 0.25um CMOS process

• Most components are on-chip and only the output matching is implemented off-chip

Jan Mikkelsen -24-

System related PA parameters

PA parameters and typical valuesPA parameters and typical values

• Center frequency and bandwidth (typically a few Hundreds MHz)

• Power gain (10 dB – 30 dB)

• Output power level (20 dBm – 30 dBm)

• Operating conditions (power supply: typically 3.8 - 5.8 V while for CMOS IC implementation typically down to 2.2 V or 1.8 V)

• Matching parameter (input/output impedance, S11, S22, ...)

• PAE (30%-60%)

• K (stability factor >1)

• Non-linear parameters (IIP3, 1-dB compression point)

Jan Mikkelsen -25-


Gain, output power and efficiencyGain, output power and efficiency

Jan Mikkelsen -26-


Non-linearitiesNon-linearities

Jan Mikkelsen -27-

dc

inrf

PPP

PAE−

=


Power Added Efficiency (PAE) Power Added Efficiency (PAE)

• To describe a PA’s ability to transfering DC power to RF power thepower added efficiency has been defined

• Often, when considering the efficiency of an amplifier only the drainefficiency is considered.

• The problem here is that the drain efficiency definition can lead to verymisleading results. For instance, a very high efficiency can be assignedto an amplifier that has no power gain .. in other words a useless PA

• PAE just replaces RF output power with the difference between input and output power

• PAE < drain efficiency !

Jan Mikkelsen -28-

PA non-linearities

• As for any other active component the PA is also characterized by somelevel on non-linear performance

• Generally, the non-linearities of power amplifiers are related to thefollowing phenomena

Output power is not partial to input power (generation of IM)

Deviation from linear phase respons

Adjancent channel inteference

• To some extent we use the same power series expansion to describethe non-linearities of PA’s

• But in many cases this proves to beinsufficient as we need to considerboth AM-AM distortion and AM-PM distortion

Jan Mikkelsen -29-

PA non-linearities

Origins of PA non-linearitiesOrigins of PA non-linearities

• The primary reason for the non-linear performance relates to the non-linearity of the transistor (GaAs MESFET, HEMT, HBT, Si MOS) operatedin large signal mode

• For the MOSFET the typical issues are:

Narrow channel effect (channel modulation, hot carrier overshoot)

Parasitic capacitance modulation by signal

Vt modulation by signal

Gm modulation by Signal.

• It is very difficult to characterize a transistor when it is expected to oprate under large signal conditions as is the case for the PA transistors

• Further, some classes of PA’s operate in non-linear mode whereby thecharacterization of transistor non-linearities becomes even more important

Jan Mikkelsen -30-

)cos()cos( 21 ωω vvvi +=

)()()()( 55

331 tvatvatvatv iiio ++=

)cos(425

49

2,15

53

31_ ω⎭⎬⎫

⎩⎨⎧ ++= vavavav fundo

PA non-linearities

Non-linearity .. AM-AM and AM-PM Non-linearity .. AM-AM and AM-PM

• As usual the general approach to non-linear analysis is to make use of a two-tone analysis

• Like that we can describe the input signal as follows:

• Assuming a power series expansion with fundamental, third-order and fifth-order components is adequate the transfer function is gived as:

• The resulting output at the fundamental frequency is the given by thefollowing relation:

Jan Mikkelsen -31-

IM3:IM3: )2cos(825

43

1,22,15

53

33_ ωω −⎭⎬⎫

⎩⎨⎧ += vavav imo

PA non-linearities

Non-linearity .. AM-AM and AM-PM Non-linearity .. AM-AM and AM-PM

• Collecting terms also leads to equations for the resulting inter-modulation components

IM5:IM5:

IM7:IM7:

……

……

Jan Mikkelsen -32-

2,

3outIMout

inIIP

PPPP

−+=

PA non-linearities

Definition of AM-AM and AM-PM Definition of AM-AM and AM-PM

• The AM-AM performance of a device is describe by how much the actualoutput power deviates from an ideally linear relation

• Just as you have gotten accoustomed to in the case of small-signalamplifiers the AM-AM relation for PA’s can be described in terms of IIP3, IM3 or 1-dB compression point

Jan Mikkelsen -33-

PA non-linearities

Definition of AM-AM and AM-PM Definition of AM-AM and AM-PM

• The AM-PM performance is normally determined by how much the actualphase deviates from an ideally linear phase respons

• The key is here that the phase of theoutput signal changes as a function ofthe input (and hence output) power level

Jan Mikkelsen -34-

∫

∫=

ch

cha

dffS

dffSACPR

)(

)(_

PA non-linearities

Definition of ACPR and EVM Definition of ACPR and EVM

• The AM-AM and AM-PM performance will manifest it self through an increased interference power leve in adjacent frequency channels

• This effect is also known as ”Spectral regrowth”

• The Adjacent Channel Power Ratio (ACPR) is a measure of howsignificant the spectral regrowth contribution is from a specific block

• ACPR is a measure of how the non-linearities are hoing to affectadjacent channels

Jan Mikkelsen -35-

PA non-linearities

Definition of ACPR and EVMDefinition of ACPR and EVM

• The effect of AM-AM and AM-PM distortion can also be representedthrough the Error Vecto Magnitude (EVM) measure

• EVM describes how the non-linear behavior of the device affect themudulation error

• To illustrate the levels of the typical EVM performance/requirement takethe following example using an EDGE transmitter

RMS EVM: The resulting RMS value of the EVM measured over 200 bursts ofeach 146 usefull symbols (normal Burst). The maximum acceptable EVM levelis here 0.9% under normal conditions [GSM 05.05 Section 4.6.2.1]

Peak EVM: The average of the peak EVM as measured over 200 single bursts. The maximum value is here 30% [GSM 05.05 Section 4.6.2.3]

Jan Mikkelsen -36-

PA non-linearities

Illustration of error vectorIllustration of error vector

Power Amplifiers - Ikom.aau.dk/~hmi/Daniel/TCMT/mm.5.materials/mm.5... · 2007-02-21 · Class BClass B Power amplifier classification • This is an example of a class B power amplifier

Documents