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

RF PA oscillation problems can be broadly categorized into two kinds; Bias oscillations, and RF oscillations. Bias oscillations occur at low frequencies, in the magahertz to VHF range, and are caused by inappropriate and unintentional terminations at those frequencies by the bias insertion circuitry, such as the addition of a large-value decoupling capacitors. The oscillations have little to do with the detail of the RF matching circuitry, where the RF blocking and decoupling capacitors become open circuit terminations at lower frequencies. RF oscillations, on the other hand, typically occur either in band or commonly out of band but still quite close to the desired bandwidth from the low frequency side. These latter kinds of oscillations are too common in single ended multi-stage designs, and their elimination will require modifications to the RF matching network topology and element values. Both kinds of instability can be analyzed effectively using the k-factor analysis. Although k-factor analysis assumes a linear two-port device, it is usually a satisfactory assuption to assume that RF oscillations in power amplifiers will more likely occur when the amplifier is backed off into its linear region, where the k-factor analysis is valid. In the case of deep class AB or B operation, it is necesary to increase the quiescent current to perform the stability analysis with a representative amount of gain. A simple way to test stability of the PA, is to run the entire circuit on a linear simulator, sweeping the frequency all the way down to dc.

Higher frequency instability will show up in a k-factor analysis of individual stages. Any single-ended design must show a k-factor greater than unity over the widest frequency sweep, extending from the low-frequency bias circuit range all the way up to the frequency at which the gain rolls off to lower than unity. Designing or midifying a circuit to obtain such a response for the k-factor typically will involve some sarifices in the in-band RF performance through the use of resistive elements, which will affect the efficiency of the PA.

Richard: 振荡的问题，在射频/微波功率放大器设计中司空见惯。往往，设计好的振荡器不振荡，设计好的PA却是很好的振荡器。这个问题，几乎每个射频工程师都曾遇到过，也感到非常头疼。最大的问题是，对于振荡的机理和解决方案，一直没有非常明确的认识。湖哥、江哥和泰哥他们，当然还有我，都被这个问题折磨过。湖哥曾经深层研究过振荡的问题，可惜当时我太懵懂，他的成果没有继承过来，遗憾万分！这里作者将振荡分为两种：Bias Osc. 和RF Osc. ，这也是业界普遍的认识。两者的表现主要是振荡频率的高低上，作者说得很明白，但是对于如何解决却没有说明。结合经验，我认为，低频振荡（可以说就是Bias Osc.）更好解决一些，办法当然focus on偏置电路，如在FET功放中，栅极串联电阻（就是作者说的through use of resistive elements），偏置线加磁环等等，实验证明非常有效。对于RF Osc.，我认为主要是射频接地不良导致寄生正反馈，其次还有匹配不良导致反射严重也会引起RF Osc.。解决办法，当然就是显而易见了。另外，在高频的功放模块设计中，腔体效应和空间耦合也会导致振荡产生，具体机理我完全无知。可以借助HFSS仿真模拟，解决办法则相对直接：结构设计上的挡板、独立Room设计，调试过程中吸波材料的应用等等。上次和RFMD的牛人聊天的时候说到振荡问题，我说了以上这些我的见解，可惜没有能得到指点，遗憾。

Instability occurs when some of the output energy is fed back to the input port in the proper phase so as to mack negative resistance appear at the output or input of the amplifier. Coupling from output to input occurs through capacitances within the active device and through external elements. Because the reactance of the feedback capacitance decreases with increasing frequency, the likelihood of unwanted oscillations is higher in RF amplifiers.

Class E: The usual k-factor stability definition does not actually have a meaning in the case of Class E amplifiers since the transistors in this case are acting as swtiches, while the definition of the k-factor is based on small signal analysis. The practical way to test stability in this case is to perform transient analysis using a step input located at various nodes and check that the output settles down to a fixed level within a short period of time.

____________________________________________________