The key to power amplifier design: performance of output matching circuit

JasonYoo

The key to power amplifier design: performance of output matching circuit [Copy link]

Power losses reduce efficiency and power output capability.

Soeren Laursen
RF Micro Devices

For any power amplifier (PA) design, the performance of the output matching circuit is critical. However, during the design process, there is a problem that is often overlooked, that is, the power loss of the output matching circuit. These power losses appear in the capacitors, inductors, and other energy-consuming components of the matching network. Power losses will reduce the operating efficiency and power output capability of the PA.

Because the output matching circuit is not a 50Ω component, the dissipation loss is very different from the sensor gain. The specific circuit of the output matching is different, and the loss is also different. For the designer, even if he does not have the room to choose different technologies, there are still many design compromises between bandwidth and dissipation loss.
Matching networks are used to achieve impedance changes, just like power is transmitted from one system or subsystem to another system or subsystem, and RF designers put a lot of effort into this. For power amplifiers, the impedance controls the amount of power delivered to the output, its gain, and the noise it generates. Therefore, the design of the power amplifier matching network is the key to achieve optimal performance.

There are different definitions of loss, but what we are concerned about here is the loss of RF power dissipated as heat in the matching network. This lost power is not used for any purpose. Depending on the function of the matching circuit, the acceptable range of loss is also different. For power amplifiers, output matching loss has always been a concern because it involves a lot of power. Low efficiency not only shortens talk time, but also causes great problems in heat dissipation and reliability.
For example, a GSM power amplifier operates at 3.5V with an efficiency of 55% and can output 34dBm. At maximum output power, the power amplifier current is 1.3A. The matching loss is on the order of 0.5dB to 1dB, which is related to the specific circuit of the output match. Without dissipative loss, the efficiency of the power amplifier is 62% to 69%. Although losses cannot be completely avoided, this example tells us that losses are the primary problem in the power amplifier matching network.
Dissipative loss

Now let's look at a network and study the dissipative losses in a matching network (Figure 1a). The power supply delivers power to a passive load through a passive matching network. There are no other constraints between the source and load impedance. Considering the matching network and the load together, the source outputs a fixed amount of power Pdel to this network (Figure 1b). Part of the output power is dissipated in the matching network as heat. The rest is delivered to the load. Pdel is the total power delivered to the matching network and load (Figure 1c), and PL is the part of the power delivered to the load.

Knowing these two quantities, we can know how much power is actually transferred from the power source to the load as useful power, and the ratio is equal to PL/Pdel.
This is a correct measurement of the dissipation loss of the power amplifier output matching, because it only considers the actual transmitted power and the dissipated power. The reflected power is not calculated.
From this, it can be seen that this ratio is equal to the power gain GP of the matching network when it is working. The complete expression of the power gain when working is:

Here, is the load reflection coefficient, is the S parameter of the matching network, and 740)this.width=740" border=undefined>. Note that in this expression, no information about the source impedance is included.
The loss is the inverse of the gain. Therefore, the dissipation loss can be defined as:
Ldiss = 1/GP.
For the power amplifier, the load we design for it is generally 50Ω. Usually, the system impedance we use to measure the S parameters is also 50Ω. If the system impedance and load are both 50Ω, then it is 0, so the above expression can be simplified to:

To calculate the dissipative loss of a matching network, we only need to know the transmission and reflection parameters, which can be easily obtained from the S-parameter calculation process because network analyzers usually display S-parameter values in a linear manner. When evaluating input and interstage dissipative losses, the load impedance is not 50Ω, but the above rules still apply.
Because reflection and dissipative losses are easily confused, RF engineers sometimes use the wrong method to calculate dissipative losses. The worst method is to use unprocessed S21 for calculation. A typical matching network at 1GHz (Figure 2) has a load impedance of 4+j0Ω for the power amplifier. The matching network is simulated using lossless components, so there is no power dissipation in the matching network. However, S21 is -6dB because there is a huge mismatch between the 50Ω source impedance and the 4Ω load. As a lossless network, the simulated dissipative loss is 0dB except for some digital noise.
In the circuit simulation, we may be able to use S21 to calculate the correct dissipative loss. This process involves using the conjugate impedance of the complex simulated load line as the source impedance. Since the dissipative losses are independent of the source impedance, this is the correct method, but it is not convenient to use.
Another common method is to use the maximum gain calculation in the circuit simulator. Since this measurement is made using ADS, it is convenient to use. However, it may give the wrong answer. In a simple circuit with only a 50Ω series resistor, obviously the load is also 50Ω, and the dissipative loss of the 50Ω series resistor is 3dB because the transmitted power is evenly divided between the series resistor and the load (Table 1). In this case, the simulator can choose a load impedance of 1GΩ. When the 50Ω resistor is connected in series with the 1GΩ load, the voltage drop across it is very low, and the power dissipated is very small. The
correct calculation method should be to use the operating power gain. It may be possible to get the same result using other methods, but there is no guarantee that the result will be obtained. It is very simple to find the operating power gain when the load is 50Ω, and there is no reason not to use it.

Output matching circuit

The specific circuits used to match the output will result in different losses. At the low end of the microwave spectrum, transmission lines take up too much space, so a lumped-component approach is used. In a typical output-matching circuit for a PA module, a large DC-blocking capacitor is used to prevent DC current from flowing from the PA power supply to the load (Figure 2). A two-section low-pass matching network consisting of surface-mount capacitors and printed inductors and surface-mount inductors transforms the nominal 50Ω load impedance into an appropriate load line. The load line is set based on the specified PA output power and the available supply voltage. The load line for a cell phone amplifier varies from 1Ω to 5Ω.
Standard or high-Q capacitors can be used. Another approach that is becoming increasingly popular is to use integrated capacitors. High-quality metal-high-k-metal storage capacitors are available in many process technologies, including GaAs and CMOS. One supplier offers a complete GSM PA module that uses no surface-mount components; all matching networks use leadframe traces and integrated capacitors. In addition to reducing size, using integrated capacitors has cost advantages, which can be achieved through better production lines, reduced assembly complexity, saved logistics work, and shorter delivery times.

Minimize losses

Even if designers cannot choose different technologies, they still have a lot of room to make design trade-offs between bandwidth and dissipative losses. One way to understand the loss mechanisms of an output match is to simulate the match with lossless components and then introduce the loss mechanisms one component at a time (Table 2).
The quality factor of a capacitor is inversely proportional to its capacitance. To minimize the dissipative losses of the output match, the value of Cl must be as small as possible in the output match. The trade-off is between bandwidth and dissipative losses.
Dissipative losses are critical to the efficiency of a power amplifier. The value of dissipative losses is equal to the inverse of the operating power gain of the matching network and is independent of any characteristics of the source impedance. The formula for calculating dissipative losses when the load impedance is 50Ω is very simple and easy to apply in the design.
There are other ways to measure the loss of an output match, but these measurement methods sometimes give erroneous results. Different capacitor technologies in the output match circuit will have different losses. Integrated capacitors are very suitable for low-loss output matches. Even once the capacitor technology has been selected, there is still a lot of room for design trade-offs between bandwidth and dissipation losses.

About the Author
Soeren Laursen is a senior design engineer at RF Micro Devices. He holds a Ph.D. from Aalborg University in Denmark. The company is headquartered in Greensboro, New York, USA. Tel: (1) 336-664-1233. Website: www.rfmd.com .

Table 1 Dissipation loss
simulation results of 50Ω series resistor -3.5 dB
Maximum gain 0.0 dB
Gp -3.0 dB

Table 2 Mechanical loss of output matching
Lossy components Dissipation loss at 1 GHz
L1 0.17 dB
C1 0.66 dB
L2 0.15 dB
C2 0.11 dB
Cout 0.03 dB
Total 1.11 dB

Figure 1. The network constructed to calculate the dissipation loss of the matching network (a). Considering the matching network and the load together, the power supply outputs a certain amount of power to this composite network (b). When the power supply outputs Pdel to the composite network of the matching network and the load, PL is the part of the power transmitted to the load (c).

Figure 2 A typical matching network at 1GHz presents a load impedance of 4+ j0Ω to the power amplifier. The matching network is simulated using lossless components, so no power dissipation occurs in the matching network.

chuchuda

Downloaded and saved, thank you