|
 |
|
 |
  |
|
      |
||
|
|
         
 |
|
IMD performances of harmonically tuned microwave power amplifiers This paper by P.Colantonio*, F.Giannini*, G.Leuzzi** and E.Limiti* won the prize for Best Poster Paper at GAAS 2000. Using a simplified active device model the effects of different amplifier harmonic input and output terminations on the intermodulation distortion are investigated and experimentally validated for four hybrid amplifier designs. *Department of Electronic Engineering, University of Roma "Tor Vergata",
Via di Tor Vergata 110, 00133 Roma -
Italy.
The linearity performance in terms of intermodulation (IM) is one of the crucial requirements of microwave power amplifiers (PAs), especially for handy-phone or mobile communication systems. The most troublesome effect of the non-linear behaviour of the device is represented by the third order IM product (IMD3), since it is often difficult to filter out or it is even an in-band signal. Thus, many PA design solutions have been developed to reduce the IMD3 to acceptable levels. The proposed approaches are usually based on linearisation techniques, often requiring complex and expensive circuitry, like the feedforward or predistortion strategies, or ones that worsen the amplifier performances, like the feedback or back-off approaches. If a simpler low IM PA design technique is required, the solution can be in the active device carrier to intermodulation (C/I) load pull characterisation [1]: the active device is measured and characterised for different load values and the resulting amplifier design is a trade-off between output power and IMD3 performances. The main target of published works on the topic is to understand the role of the PA, harmonic loads on the IM behaviour and their effects on the appearance of sweet-spots (i.e. null values) of IMD3. This paper gives a brief discussion on the mechanisms that generate IMD3. It then presents and comments upon the IMD3 performance of four hybrid amplifiers that were designed for different operating classes and harmonic manipulation strategies, that is, with different harmonic input and output terminations. Non linear device IMD3 model
In order to obtain high efficiency PAs, it is well known that harmonic manipulation design approaches are helpful and in the past the authors have suggested some strategies [2, 3]. Nevertheless, it is a common idea that the use of harmonic terminations could be detrimental for IMD3 performance. This idea is not completely correct, since the role of the harmonic terminations and their effects depend on the harmonic number and the topological position (input or output of the stage). To understand the IMD3 generating mechanism and to predict the active device intermodulation performances, modelling techniques have been devised [4] or analytic approaches have been developed, based on a Volterra Series analysis, in which the active device is characterised at small signal regimes while its large signal behaviour is approximated using the describing function approach [5]. In particular, the active device non-linear behaviour could be simplified through a simple three-terms power series expansion for the drain current (see Equation 1) in which v in is the gate-source voltage (in this approach the dependence of i d on v ds has been neglected). To infer some useful information about the effect of input signal harmonics on IMD3, an input signal composed by two tones and their second harmonics can be assumed (see Equation 2) obtaining the IMD3 product given by Equation 3, while the fundamental output is shown in Equation 4. Thus, remembering that usually g m2 and g m3 are opposite in sign, the effect of a second harmonic component at the input port is evidenced by the term g 2 , 2 A 1 A 2 that can reduce the overall IMD3 value. Obviously, this is a coarse approximation, but the underlying idea that a proper second harmonic input component can reduce the IMD3 performance is stressed also by Aitchison [6], who has experimentally demonstrated that a second harmonic feedback improves the amplifier linearity. The simple results in Equation 3 can be applied to a more realistic device model, in order to compare different harmonic tuning strategies. As an example, if Tuned Load and Class G approaches are considered, the small-signal IMD3 performances are shown in Figure 1, showing a remarkable improvement in linearity.
In the following, measurements of four PAs designed with different harmonic tuning approaches will be shown, giving a practical confirmation of this property and demonstrating that the harmonic manipulation design approach can be helpfully used both to increase power performances and improve IMD3. Experimental results Four amplifiers were realised implementing different tuning approaches, ranging from the classical Tuned Load (TL) strategy, in which the output voltage harmonic components are shorted, to Class F [2], Class G [3] and Class FG [7] strategies, where respectively 3rd, 2nd and 2nd and 3rd harmonics are properly terminated. The active device used is a medium power GaAs MESFET by Alenia Marconi Systems, and the amplifiers have been designed to operate at 5GHz, with a 5V drain bias and a quiescent drain current of 30% I dss . The PA's matching networks have been realised on an alumina substrate 25mm thick, with different criteria. For TL and Class F approaches, the input matching networks were synthesised to fulfil fundamental frequency conjugate matching and to minimise the input distortion, i.e. zeroing the harmonic components of driving signal represented by the voltage across the gate-source capacitance C gs . Such a function, as represented in Table 1, has been accomplished by means of an external input termination that, at the internal terminals, actually synthesises a short-circuit starting from the reported values. For Class G and Class FG strategies, the input networks were designed, in addition to fulfiling the same matching conditions, to manage the output current harmonic components (2nd and 3rd) to obtain the proper phase relationships, allowing purely resistive output loads [7].
The synthesised external loads are reported in Table 1, where their baseband values are also shown, These are practically independent from harmonic load selected and are correlated to the DC bias. The IMD3 measurements have been performed using a spectrum analyser, driving the amplifiers with two equal amplitude input signal at 5 and 5.05GHz respectively. The scheme of the measurement set-up is depicted in Figure 2. The measured performances of the four realised PAs are in Figure 3, shown demonstrating the improvement due to the harmonic manipulation approach, both in terms of output power and power added efficiency, as expected when a harmonic tuning strategy is applied. Finally, Figure 4 depicts the measured IMD3 performances.
Some interesting observations can be made from the results shown; it can be noted from Figure 4 that with a Class F approach, the use of a third harmonic actually degrades the IMD3 behaviour, even if its effect is not so relevant, as predicted with Volterra Series analysis. On the other hand, the use of a second harmonic component, as Class G and FG PAs decreases the IMD3 distortion, as observed in the previous section. In particular, the Class G amplifier exhibits the lower IMD3, and its different behaviour with respect to the Class FG amplifier can be attributed to two reasons. Firstly, from Equation 3, the second harmonic component 2 changes the IMD3 sign, reducing the overall amplitude (in the case of the Class G the amplitude obtained is less than for Class FG, due to the lower second harmonic contribution). Secondly, the presence of an input third harmonic component as is in the Class FG amplifier, which is necessary to obtain the output in-phase third harmonic voltage component, not taken into account in Equation 1, actually increases the overall IMD3. Finally, from the measured results, the sweet-spot position (the null in IMD3) is basically unchanged by the different harmonic manipulation strategies, demonstrating its major dependence on the choice of the bias point only . Conclusions The intermodulation generating mechanisms has been discussed and the role of the amplifier harmonic termination, both at the input and at the output, has been clarified by a simplified model. Measured performances of four realised PAs, implementing different harmonic terminating solutions, have confirmed the assertion made in this paper. Moreover, the presented results show the improvement due to the harmonic manipulation approach, both in terms of output power and power added efficiency, as would be expected when a harmonic tuning strategy is applied. References
|
|
||||||||||||||
|
||||||||||||||||||
