Achieving low noise figures in an amplifier requires starting with the right transistor and then finding the optimum input and output matching networks while also meeting other performance needs.
Low-noise amplifiers (LNAs) are usually the most critical component in setting an RF/microwave receiver's noise figure and sensitivity. Over the years, designers have taken advantage of emerging device technologies to achieve the lowest possible noise figures for a given operating frequency. But not all LNAs are created equal, and it takes more than just a low-noise transistor to make a high-performance LNA. It also requires careful impedance matching and circuit optimization, often with tradeoffs to be made on whether to sacrifice some of the transistor's signal gain in favor of a lower noise figure.
Designing an LNA starts with a choice of device technology. Silicon bipolar transistors were once the favorite device for microwave LNAs, but gallium-arsenide (GaAs) metal-epitaxial- semiconductor field-effect-transistor (MESFET) devices have all but replaced them at higher RF and microwave frequencies. In recent years, the development of smaller-dimensioned silicon CMOS devices has enabled the design of silicon CMOS LNAs for RF/microwave wireless applications requiring low power consumption. The evolution of pseudomorphic high-electron-mobility transistors (pHEMTs) and heterojunction bipolar transistors (HBTs) on different substrate materials, such as GaAs, indium phosphide (InP), and silicon germanium (SiGe), have given LNA designers an unprecedented choice of starting points in low-noise devices.
In a receiver, an LNA usually follows a front-end filter placed between the receive antenna and the amplifier. The LNA must be able to boost the amplitude of low-level signals without adding unwanted noise. Yet, it must also be able to handle larger signals without adding distortion. Depending upon the receiver architecture, the output of the LNA may feed a mixer or the first in a series of mixer stages, or may feed a high-speed analog-to-digital converter (ADC) in a software-definedradio (SDR) configuration. In all cases, the LNA must provide an adequate dynamic range, should be unconditionally stable (lack of oscillation) over all operating frequencies, and maintain a low noise figure over the frequency band of interest.
Because of this, the design of an LNA is never as simple as aiming for the lowest noise figure from a given active device over a desired bandwidth. The amplifier must also provide good stability under a wide range of source and load impedance conditions. It may also have requirements for small-signal gain, output power at 1-dB compression, and third-order intercept point. These requirements are often in conflict, since stability often comes by sacrificing some gain, while good noise-figure performance requires low current operation, in contrast to the higher-current levels needed for good third-order-intercept performance. At best, an LNA design is a compromise in achieving satisfactory performance levels in as many of these areas as possible.
LNA design can be approached by prioritizing the performance requirements of the amplifier. For example, classical noise matching (CNM) techniques are used when low noise figure is the most important requirement for the amplifier. In such a case, the active device is presented with an optimum noise impedance (which may not be optimum for the highest gain), usually implemented in the form of an impedance matching network between the signal source, such as the filter or antenna in a receiver, and the input port of the amplifier. Similarly, simultaneous noise and input matching (SNIM) techniques typically employ series feedback to obtain the required matching parameters for very low noise figure from a given active device. When the available power supply for an LNA may be limited, such as in a battery-powered or portable application, a power-constrained noiseoptimization (PCNO) design technique may be applied for achieving low noise figure with limited bias. In many cases, the available values of discrete matching components, such as resistor, inductors, and capacitors, or the values possible through microstrip circuit fabrication, may limit the matching networks for an LNA to suboptimal designs.
The source load presented to the LNA can further complicate its design, since the impedance of the source is often not ideal. Typically, it is a filter or an antenna, with impedance that can change over time, temperature, and other conditions. In addition, the interactions of the impedances of a filter and the LNA, for example, can render the LNA unstable, especially a filter that presents dramatic shifts in impedance in its rejection bandwidth.
But properly matching the input of an LNA's device(s) to the source is not enough for low noise figure. To achieve low noise figure, the input of the output load connected to the LNA must also be impedance matched to the conjugate noise impedance of the active device(s). This minimizes noise from the transistor that is reflected from the load and re-amplified in the transistor. Achieving this condition often means accepting less gain from the active device(s), which can also help LNA stability by avoiding oscillation conditions for the amplifier. Transistor characterization data typically includes the measured impedance of the load at which the minimum device noise figure occurs, Gopt.
Any transistor considered for an LNA should provide adequate characterization information on its data sheet or on the supplier's web site. For example, its scattering parameters (S-parameters) should have been measured by the manufacturer with a calibrated microwave vector network analyzer (VNA) and available at different operating voltages and current levels for a frequency range of interest. Noise parameters should also be available.
Transistor models (linear models for an LNA) that can be used in industrystandard design software tools, such as the Advanced Design System (ADS) from Agilent Technologies and Microwave Office from AWR can help simplify the design process by allowing a designer to simulate an amplifier circuit's behavior under different bias conditions. The software also helps in the optimization process by allowing an engineer to experiment, for example, with different values of source resistance, to examine the effects on noise figure, gain, and linearity, without having to rebuild a prototype circuit each time. Of course, components used as parts of models in computer-aided-engineering (CAE) programs tend to be ideal versions of the real things, lacking the parasitic circuit element of surface-mount packages and their pin connectors, for example. These differences typically result in deviations between measured and simulated results in LNA designs
To determine the effectiveness of an LNA design, a means of measuring noise figure is needed. A number of different approaches are used to measure noise figure, with the Y-factor method being among the most popular. The method is based on the use of a noise source with precisely known output noise power. The noise source can be turned on and off, and is used as the signal source to the LNA to be tested. The output power of the LNA is measured with the noise source turned on and then with it switched off, and the ratio of the two noise powers, called the Y-factor, is determined. The noise source is often used with a spectrum analyzer, a power meter, or noise-figure meter. The absolute power level accuracy of the measuring instrument is not critical since the technique involves the measurement of a ratio. For more information on noise-figure measurements, a 31-page application note (No. 57-1) from Agilent Technologies, "Fundamentals of RF and Microwave Noise Figure Measurements," provides a solid background on different measurement approaches.
What types of noise figures are currently available in standard (rather than cryogenic) commercial LNAs? Most LNAs serve bandwidths targeted at specific communications bands, such as the SKY67100-396LF LNA from Skyworks, which is designed for GSM and CDMA cellular applications. It is specified for use from 1700 to 2000 MHz, with a noise figure of 0.55 dB at 1750 MHz and 0.61 dB at 1950 MHz. The gain, which is a function of the bias supply, is typically 17 dB or more. The company also offers the model SKY67101-396LF LNA for applications from 700 to 1000 MHz. It boasts a noise figure of 0.49 dB at 900 MHz with 17.9-dB gain.
The diminutive HMC902 and HMC903 LNA amplifiers from Hittite Microwave are GaAs pHEMT monolithic-microwaveintegrated- circuit (MMIC) chips covering 5 to 10 GHz and 6 to 18 GHz, respectively, and ideal for use in military systems and microwave radios (Fig. 1). They measure less than 1.5 x 1.5 mm but provide as much as 20 dB gain with noise figures as low as 1.6 dB. For even more broadband coverage, CTT recently introduced its AMX/0118- 30xx series of LNAs with more than four octaves of bandwidth from 1 to 18 GHz. Based on GaAs pHEMT technology, amplifiers in the series (Fig. 2) exhibit noise figure of typically 3 dB with gain levels ranging from 11 to 28 dB.