Mona M. Hella 《RF CMOS Popwer Amplifiers: Theory, Design and Implementation》

Implementing power amplifiers in CMOS technology is considered a major step towards the realization of a complete transceiver on-chip. Modern transceivers require means for adjusting the transmitted power over a finite range to further reduce power consumption and improve channel capacity. A low performance, short-range wireless standard such as the Bluetooth radio requires a high level of integration, and low cost that can onle be achieved using CMOS technology, together with means of controlling the output power up to 20dBm. In addition, power amplifiers are tyupically backed-off relative to their peak power and PAE points in order to meet the Linearity requirement of the system. The degree of back-off varies depending on the modulation scheme employed -- 0dB fro Gaussian-filtered minimum shift keying (GMSK) (GSM and DECT), 7dB for Pi/4DQPSK (IS-54 and PHS), 10dB for QPSK (IS-95) and 12dB for 16QAM are typical. Thus, adding efficient techniques for adjusting the output power level is considered a challenging issue to integrated power amplifiers.

Linear power amplifiers can have their power adjusted by the variation of biasing or dynamic variation of the load seen by the output stage (Doherty amplifier). The output power can be also be controlled through the variation in the input signal amplitude; this can be realized by having a variable gain amplifier (VGA) as a preceding stage. However a large dynamic range of output power requires a linear wide dynamic range variable gain amplifier which is usually power consuming and hard to achieve. Also this configuration suffers from a large reduction in efficiency at lower power transmission because the standing bias current at the output stage does not scale with the output power. This technique requires a very linear power amplifier for any kind of signal shaping at the input.

In non-linear power amplifiers, the input to the amplifier provides only timing information. Thus, the output power cannot be controlled through the variation in input signal amplitude as is done in linear or weakly nonlinear amplifiers. Instead, output power control can be realized effectively through a variable supply, implemented for example by a dc-dc converter. The losses in the DC-DC converter might casuse the effieciency to drop to reach that of a linear power amplifier's case. A new methodology based on switching a combination of power amplifiers in parallel is presented in some articales and represents an extension to the idea of Doherty amplifier to non-linear power amplifiers.

Richard: 关于PA的功率控制没有什么经验。我想具体做法可以采用两种办法：调制，反馈控制。调制包括对PA的供电电源控制，和对PA的输入信号功率进行控制，殊途同归达到控制输出功率的目的。反馈控制，其实他的实现也要借助于调制--检测输出功率水平，反馈到调制电路进行控制。以前做过一个输出功率检测电路，就是为了做输出功率控制--对检测线性度要求较高，譬如输出功率每增加3dB，检测电路输出电平增加0.1V。很多PA被设计成Lamination的结构，其中集成了几个die，如GaAs HBT的PA die，GaAs HEMT的RF Switch die，CMOS的Control Circuit die，针对不同的功能采用不同的工艺，然后封装在一起。现在RFMD，Skyworks，Anadigics，TriQuint都有了BiFET的工艺，那么，在不久的将来，也许Lamination即将消失，取而代之的是全集成的解决方案了。另外，这里再列一下PA线性化技术：

1. Feedforward;

2. Feedback;

3. Predistortion;

4. Envelope Elimination and restoration (EER);

5. Linearization using nonlinear components (LINC);

6. Bias adaption;

7. Doherty amplifiers.