CDMA的PA为什么要用线性放大器，而GSM可以用非线性的？

和调制方式有关，GSM调制方式是GMSK，幅度不包含信息，CDMA手机Tx的调制方式是OQPSK（基站TX调制方式QPSK），幅度不可以失真。楼下的补充？_^o^_

QPSK不是只有相位包含信息吗？
OQPSK和QPSK只是Q之路相差一个Tb，在幅度上有差别吗？
楼上的兄弟能否解释的再详细一点。

原因如下：
GSM调制方式是GMSK，信息寄载在载波频率上,(要注意到GMSK是属于FSK调制的一种),非线性放大器一般工作在C类,为了提高PA效率,用整个GMSK载波余弦信号去激励功放,而PA的输出是半个载波;通过LC谐振电路,可以取出激励PA的同频率的信号（该信号功率已经被放大），信息不会丢失，因为通过PA前后，载波频率不变。
CDMA手机的调制方式是QPSK，属于PSK调制一种，信息寄载在载波相位上，非线性放大器一般工作在C类,为了提高PA效率,用整个PSK载波余弦信号去激励功放,而PA的输出是半个载波;虽然也可以通过LC谐振电路,取出激励PA的同频率的信号，但是PA输出信号的相位信息不能保证输入PA的前后一致，导致信息会被丢失，因为通过PA前后，载波的相位不能保证和PA前信号一样。
而线性放大，输入是怎么样，输出仍旧是怎么样信号，不存在失真（理论上），从而可以保证相位信息不失真。

还是英文说得清楚一些，copy过来共勉。input data 00,01,10,11QPSK:00-->01-->10-->11OQPSK:00-->[01-->00-->10]-->11If the modulation data changes, QPSK modulation makes one symbol change each period. If the
data is the same, then QPSK and OQPSK do not change state. However, if the data changes,
then OQPSK will make at least one symbol change and possibly two changes in a single period.
In the above example, the data input to the modulator is: 00, 10, 01, 11. The upper I/Q diagram
shows the symbol changes made by QPSK (00-->01-->10-->11). The I/Q modulator for QPSK moves from state to
state, for a total of three transitions (with one transition going through the origin,data from 01 to 10). For the same
pattern, OQPSK makes four transition00-->[01-->00-->10]-->11, and it avoids the origin. The change from 00 to 10 results
in a single transition since the value of Q did not change. However, the change from 10 to 01
makes two transitions because the value of Q does change. First the modulator returns to the 00
state. Then one-half period later, the modulator changes to the final 01 state. This action, due to
the delay placed into the Q branch of the modulator, eliminates the transition through the origin.
The change from the 01 to 11 state for OQPSK is a single transition since the value of Q again
remained constant.The modulation is Filtered QPSK in the base station, and Filtered Offset QPSK in the mobile
station. Note that the I/Q diagram for the base station signal is for only a single channel (such as
the pilot channel). In normal operation, many channels are summed together and transmitted on
top of each other by the base station. O-QPSK is used in the mobile stations because it avoids
going through the origin and makes the design of the output amplifier easier. For the base station,
since many channels are summed together, using O-QPSK would not always avoid the origin.
This is due to the random nature of adding many signals together.由于OQPSK调制方式避免了调制信号经过IQ图的原点，也就是避免了调制信号的幅度出现零点，极大的降低了调制信号的动态范围，降低了Tx链路的设计难度。对于基站，过零点是不可避免的，所以用的是QPSK方式。

GMSK is one sort of FM, a nonlinear modulation, has constant envelope. So with nonlinear PA possible.OQPSK = PM + AM, a linear modulation, has nonconstant envelope due to filtering by cosine rolloff filter, so with linear PA.

We distinguish constant amplitude modulation , such as M-PSK, from those inherently involving amplitude modulation, like QAM. Note that the use of a constant amplitude constellation does not imply a sonstant amplitude signal, since Nyquist filtering tends to introduce envelope variations.From "Modulation and Coding for Wireless Communication" by Alister Burr. Page 56Published by 电子工业出版社。

我的理解是：对QPSK，由于有Nyquist滤波器，从一个星座点向另一个星座点迁移时轨迹不是沿单位圆的。对于GMSK，虽然有Gaussian滤波器，但迁移的轨迹还是在单位圆上，只是迁移的速度不时地变以下。

申明：网友回复良莠不齐，仅供参考。