Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

Document Sample
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers Powered By Docstoc
					    Current Techniques for Enhancing the Efficiency of
             Ultra Linear Power Amplifiers
                                     By Richard L. Abrahams
                            RF/Microwave Engineering, Senior Staff
                    Harris Corp., Government Communications Systems div.
                                        Palm Bay, FL


Modern, spectrally efficient high data rate wireless communications systems require linear Power
Amplifiers in order to faithfully reproduce their complex modulation waveforms while maintaining
minimum occupied bandwidth. For example, the popular 64QAM Orthogonal Frequency Divis ion
Multiplex (OFDM) modulation used in 54 Mbps 802.11a and g “Wi-Fi” WLANs requires an Error Vector
Magnitude (EVM) no greater than -25 dB (i.e. <5.6%) in order to meet consortium specifications. Practical
tests on WLANs have verified that deterioration of EVM much beyond this point does indeed cause higher
Bit Error Rate (BER) and therefore significantly reduced system throughput.

To further complicate matters, these PA’s must often be capable of driving non-ideal (i.e. high VSWR)
variable antenna load impedances such as commonly found in personal portable equipment such as .
Notebook Computers. This requires even more PA power back-off than in the ideal case and therefore
further reduces the PA’s efficiency.

Many of these wireless appliances are battery operated, placing severe constraints on operating current. PA
efficiency is therefore high on the list of desired attributes.
This paper begins by establishing a performance baseline for the relatively inefficient Class A Power
Amplifier most commonly used today in linear applications and then discusses possible performance
enhancements derived from more sophisticated techniques, such as:
     • Class B Amplifiers
     • Dougherty PA’s
     • Feed Forward Amplifiers
     • Digital Baseband Pre-distortion
     • LINC (Linear Amplification using Nonlinear Components)
     • Envelope Elimination and Restoration (EER)
     • Practical combinations of the above techniques

We live in a world of ever increasing technological complexity. There are recent marketing trends requiring
more wideband PA performance, approaching one octave operation, while maintaining the high DC
conversion efficiency. Unfortunately, several classical approaches to obtaining more efficient PA’s do not
scale well to this requirement. This paper will therefore emphasize designs providing requisite broadband

In each case, tradeoffs of potential power savings, relative complexity, size and cost are presented. Of note:
In spite of the proliferation of PA applications in the modern wireless world, we do not yet have a neat,
simple solution to the problem of achieving broadband linear high power amplification while coupling to a
somewhat variable load as is typically the case in personal portable equipment. So, the research continues
and this paper is therefore a “work in progress”.

The Classical Class A Amplifier

The Class A amplifier is a good baseline for our discussion and is well documented in the literature 1 . It is
characterized by the continuous flow of current in the PA. The bias and Load Line are adjusted so as to bias
the amplifier well into its linear region. It is therefore capable of low distortion. However the DC

                                                Page 1 of 6
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

conversion efficiency is relatively poor and will not practically exceed 15-20% when used with complex
digital modulation such as 64QAM. In the case of 64QAM, the Class A PA must typically be backed off as
much as 5-6 dB from its 1 dB compression point in order to achieve required linearity – perhaps even
further when finite VSWR is factored in. On a positive note, it is relatively low cost, simple to design with
predictable performance and capable of relatively broadband performance.

Class “AB” and “B” PA’s

This class of PA’s, as stand-alone devices, are capable of incrementally improved performance, particularly
when using fairly simple digital modulation, say up to and including the QPSK format. A practical PA
would really be biased for Class AB operation rather than Class B, since the latter implies current only
flowing ½ the time which would not yield adequate linearity. Efficiencies typically run 25-30% and they
are capable of relatively broadband performance. This is however a fairly low cost extension of Class “A”
PA technology. However, it is often desirable, in the case of battery powered equipment, to achieve higher
efficiencies than this PA class can yield. I’ll therefore limit the discussion of this class to the brief overview
presented above

Doherty PA’s

Of historical note, this technology was originally developed back in the mid 1930’s where it was used in
high power AM Broadcast Vacuum Tube PA’s. As a compromise, it did not require the high power audio
generation of a Plate Modulated AM PA design, i.e. 50% of the C         arrier PA DC input, but achieved a
significantly higher efficiency than a Class AB Linear Amplifier producing the same output power. The
recent interest in complex digital modulation has caused a substantial rebirth of this technology.

                                         Figure 1 – Doherty PA

With reference to Figure 1, the RF signal is split into two paths, one driving the Main PA and one driving
the Auxiliary (Aux) PA. The Main (Carrier) PA is biased for Class B operation whereas the Aux (Peaking)
PA is biased in Class C (i.e. saturated)2 . When producing peak power output, both amplifiers are saturated.
The theoretical collector efficiency for a Class B amplifier operating under these conditions is equal to p /4
or approximately 78%. The RF currents from the two PA’s sum equally and the load impedance seen by
each PA is given by:
                                           R50 =          = 50Ω
                                                 0.5 * RL
At lower power output levels, the Aux PA is turned off while the Main PA is operating in its linear region.
The load impedance seen by the Main PA, in this case, is given by:

                                                 Page 2 of 6
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

         Z    
R Aux   = 2
         Z     * RL = 102Ω .
          1   
The ?/4 transmission line at the input of the Aux PA ensures in-phase summing of the RF collector currents
of the two PA’s. Since the saturated Main PA output is now four times less than under peak power
conditions, the collector efficiency is twice that of a conventional Class B PA and can theoretically achieve
approximately 78% even though it is backed off in power. We conclude therefore that the collector
efficiency is reasonably high at both peak power output and its transition point.

A major drawback of the Doherty design is its narrowband nature. Unfortunately, the operation of the
Doherty design depends on ?/4 tuned transmission lines. As such, it is inherently a relatively narrow band
amplifier and difficult in practice to broadband with constant efficiency. As a narrowband design, however,
it yields greater than twice the collector efficiency of a traditional Class “A” design.

Feed Forward Amplifiers

The Feed Forward amplifier has long been the mainstay of Base Station Power Amplifier design and is well
documented in the literature 3 . It is a purely analog method of canceling distortion products and consists of
a Main Amplifier, three Couplers, two Phase Shifters, and an Auxiliary error Amplifier. It is capable of
significantly reducing distortion at moderate power output levels and can yield a useful power output
approximately 2 dB higher than the same PA without its added linearizing circuitry.

Its drawbacks are that it is relatively narrowband, costly to produce and bulky in size. Recently, other
digital techniques such as Baseband Pre-Distortion are beginning to supplant this approach due to the low
cost of Digital Signal Processing (DSP). This technique will therefore not be discussed further in this

LINC (Linear Amplification using Nonlinear Components) Amplifiers
Note: This technique also is known as Outphasing Amplification

Historically, this technique also goes back to the mid 1930’s where it was used in high power AM
Broadcast transmitters. Theoretically it is possible, by combining two saturated PA’s with a known RF
phase relationship relative to each other, to generate any RF output voltage from zero up to twice the
individual PA’s output. With reference to Figure 2 and assuming equal PA output voltages, if the two
outputs are combined in phase, twice the individual output voltage would be generated. In the other
extreme, if they are combined 180o out of phase, than zero volts would result. At phase angles in between
these two points, any intermediate output voltage may be generated.

                                                Page 3 of 6
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

                          Figure 2 – PA Vectors for LINC Amplifiers

This technique has had a recent rebirth in popularity due to the availability of an ASIC chip which performs
the task of generating the two phase signal at baseband. Baseband pre-distortion is often implemented to
further enhance the performance of LINC designs. In the 5 GHz WLAN band, authors of a recent paper
describing an application of this ASIC report achievement of a DC Collector efficiency of 65% using
64QAM modulation while maintaining an EVM of -25 dB, an impressive result.4 To verify reproducibility
of the design in a production environment, the authors built a fairly large run of assemblies and achieved
consistent results.

The challenge in a LINC design is to combine the signals from the two PA’s with low loss. A practical
design using a Chireix Power Combining circuit is shown in Figure 3.

                                         Figure 3 – LINC PA

The load on the Power Amplifiers is only purely resistive if the two signals are being combined in phase. In
the other extreme case of combining the signals completely out of phase, the loads are purely reactive. The
effect of the reactive loads can be somewhat compensated by the addition of the two shunt susceptances,
+jB and –jB. In particular, the load may be made resistive at one particular power setting by satisfying the

                                         2 * RL VL               VL 2
                                    B=       2
                                                         1− (         )
                                          Z o VL PEP            VLPEP

For convenience, we may define the normalized susceptance:

                                                   B * Zo
                                              B' =
                                                   2 * RL

It may be shown that a value of B’=0.2 will provide good efficiency over the top 6 dB of operating power
range and is therefore an optimum value for best performance. If the shunt network is not employed, i.e.
B’=0, efficiency approaches that of a traditional Class B amplifier

                                               Page 4 of 6
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

Due to the use of tuned transmission lines, this design is limited to only a moderate operating bandwidth.
However, using classical filter theory, i is possible, to somewhat extend the operating bandwidth by
replacing the single ¼ wave transmission lines with cascaded sections of lines. Three sections is a practical
limit. It is also possible to build a lower frequency version of the Outphasing Amplifier using lumped
elements such as PI sections.

Envelope Elimination and Restoration PA’s

This technique was patented in 1952 by L    eonard Kahn, primarily for HF SSB applications, however it is
particularly well suited to amplification of digital modulation signals . The EER PA relies on the principle
that any complex time varying waveform may be broken down into a phase component and an amplitude
component. Kahn, in his original implementation, used a hard limiter to strip the amp litude variations from
the signal, leaving only the phase component. Furthermore, he used a Peak Detector circuit to derive the
amplitude component. With reference to Figure 4, in a contemporary design transmitting digital modulation
and using low cost DSP components, we can generate these signals right at baseband. A Mixer, then
translates the Phase component to the final operating frequency to drive the PA.

               Figure 4 – A Contemporary Implementation of the EER PA

One key limitation in the efficiency of the original Kahn implementation was the high level modulator used
to derive the variable voltage feeding the PA (note the similarity to the Plate Modulator or Collector
Modulator used in an AM transmitter). Kahn had to implement this function using relatively low efficiency
Class AB amplification (but also, please keep in mind that the circuit was developed in the early 1950’s as
noted above!). Today, we implement this function using digital Pulse Width Modulation (PWM) techniques
which can achieve efficiencies greater than 90%.

On a positive note, the EER PA is capable of good wideband performance. It is also relatively tolerant of
moderate changes in PA load impedance. There are, however some practical problems which typically have
to be overcome in order to provide true ultralinear operation:

    1.   At low PA collector voltages, the input impedance of the PA may vary significantly due to the
         well known Miller effect. This can cause undesirable AM-PM conversion in the PA which can

                                               Page 5 of 6
Current Techniques for Enhancing the Efficiency of Ultra Linear Power Amplifiers

         degrade the EVM . Fortunately this phenomenon is fairly predictable and can be compensated for
         in the DSP processing.
    2.   It is important that the time delay in the PWM modulator be low compared to the chip rate, since
         time coincidence of the Phase and Amplitude signals is necessary for low distortion. Alternately,
         time delay equalization may be easily implemented in DSP.

Classical Digital Pre-Distortion techniques may also be used to dynamically correct these two problems. A
portion of the output signal is sampled, translated to baseband and compared with the desired signal. A
DSP correction matrix may then be generated to cancel the above effects.


This paper has presented some contemporary methods of yielding high efficiency Power Amplifiers
suitable for amplification of digitally modulated signals. The “jury is not in” and there is still significant
research needed to produce an ideal high efficiency ultralinear PA-to-Antenna interface. Contemporary
systems demand low distortion, wideband operation and the ability to preserve these characteristics under
conditions of varying PA load impedance – a daunting challenge.

  “Reference Data for Radio Engineers” 7th ed., Howard W. Sams, 1988, p 17-2
  Grebennikov, Andrei “RF and Microwave Power Amplifier Design”, McGraw-Hill, 2005, pp 372-380
  Grebennikov, Andrei, op.cit. pp 389-392
  Grundlingh, Johan, Parker, Kevin and Rabjohn, Gordon, “A High Efficiency Chireix Out-phasing Power Amplifier
for 5GHz WLAN Applications”, Proceedings of the 2004 IEEE MTT-S Conference, Ft. Worth, TX

                                                 Page 6 of 6

Shared By: