Low-noise amplifiers (LNAs) are invaluable for increasing the sensitivity and range of a microwave receiver. For applications at 4.9 to 6.0 GHz, which include IEEE 802.11a, HiperLAN2, and HiSWANa wireless-local-area-network (WLAN) receivers, a two-stage design based on enhancement-mode, pseudomorphic high-electron-mobility-transistor (E-pHEMT) device technology delivers 22 dB gain at 5.5 GHz, with a low noise figure of 1.4 dB and +11.5-dBm output power at 1-dB compression. The amplifier provides an output third-order intercept point of +28 dBm.
With its broad frequency range, this LNA design can be applied to WLAN receivers in North America, Europe, and Japan. These bands include 5.15 to 5.35 GHz and 5.725 to 5.825 GHz in North America (802.11a), 5.15 to 5.35 GHz ad 5.470 to 5.725 GHz in Europe (HiperLAN), ad 5.15 to 5.25 GHz in Japan (HiSWANa). The amplifier is designed around the model ATF-551M4 low-noise E-pHEMT transistor from Agilent Technologies for both stages (a data sheet can be downloaded from http://literature.agilent.com/litweb/pdf/5988-9006EN.pdf). The transistor, which features a 400-µm gate width for low noise and high intercept point from 2 to 10 GHz, is housed in a leadless surface-mount plastic package that measures 1.4 3 1.2 3 0.7 mm.
Besides having a very low typical noise figure (0.5 dB), the ATF-551M4 is specified at 2 GHz and 2.7-V bias to provide a +24.1-dBm output third-order intercept point (OIP3) at 10 mA drain current. The advantage of an E-pHEMT versus a depletion-mode pHEMT is that biasing the device is simplified by the fact that the E-pHEMT requires a positive voltage on the gate for normal biasing as opposed to a negative voltage for a depletion-mode pHEMT. Biasing an E-pHEMT requires a simple voltage divider from the drain to supply a small positive voltage to the gate for nominal drain current.
To meet the goals for noise figure and gain, drain source current (Ids) was chosen to be 15 mA. According to the device data sheet, this value provides good IP3 combined with a very low minimum noise figure. The data sheet also indicates that a 2.7-V drain-to-source voltage (Vds) gives a slightly higher gain and easily allows the use of a regulated 3.3-V supply.
Using the EEsof Advanced Design System (ADS) software from Agilent Technologies, the amplifier circuit can be simulated in both linear and nonlinear modes of operation. For the linear analysis, the transistors can be modeled with a two-port S-parameter file in the Touchstone file format: file ATF551M4.s2p can be downloaded from the Agilent Wireless Design Center website (http://www.agilent.com/
view/rf). In addition to information regarding gain, noise figure, and input/output return loss, the simulation provides important insight into circuit stability. The computer simulation simplifies the calculation of the Rollett stability factor (K) and eases the creation of stability circles.
ADS's nominal optimization (also known as performance optimization) tool was used to select the values of supporting components for optimum performance. This tool can be used to modify a set of parameter values to satisfy predetermined performance goals by comparing computed and desired responses and modifying design parameter values to bring the computed response closer to target performance. Nominal optimization is available in the ADS simulator for analog/RF systems simulation using any analysis type, such as AC, DC, S-parameter, harmonic-balance, circuit-envelope, and transient simulations. Goals were set for gain, noise figure and return loss over the 4.9-to-6.0-GHz band, out-of-band gain, as well as for unconditional stability from 0.1 to 18 GHz. An example of nominal optimization, optex1_prj, is available in Chapter 2 of the ADS help library under tuning, optimization, and statistical design.
Accurate equivalent-circuit models for the resistors, inductors, and capacitors are required for the optimization tool to work at 6 GHz. These models must include package parasitic inductance, resistance, and capacitance, which allows the component values to be varied over a small range using the optimization tool and accurately correlate to measured data. Examples of passive component models and ADS optimization tool terminology are shown in Fig. 1. It should be noted that each manufacturer's passive elements exhibit slightly different parasitic properties.
A demonstration board (Fig. 2) was developed primarily for 5-to-6-GHz applications. The printed-circuit board (PCB) is a three-layer configuration for rigidity. The top layer is the signal layer, which is 0.010-in.-thick FR4 with a dielectric constant of 4.2. The second and third layers are included for rigidity. The board utilizes small EIA 0402 form-factor surface-mount components. The use of microstrip lines in place of the 0402 inductors would reduce circuit losses but would produce a larger layout. The actual 6 3 15 mm area required for the circuitry is outlined in blue.
Figure 3 shows a schematic diagram of the two-stage amplifier. The amplifier uses a bandpass network for input matching and a highpass network for output matching. Interstage matching is provided by a highpass network.
The input network represents a compromise between best noise figure and reasonable input return loss, and consists of series capacitor C1, shunt inductor L1, and shunt capacitor C12. By using the demonstration board layout shown, the mounting pads before L1 will have to be bridged with copper foil: the mounting pads are included to allow for a lowpass impedance matching network topology.
The output highpass network consists of series capacitor C3 and shunt inductor L4. The bandpass networks provide additional low frequency gain roll off which serves several purposes. First, enhanced rejection of lower-frequency signals may decrease the susceptibility of the LNA to stronger low-frequency emitters, which could drive the LNA to compression and adversely affect the in-band performance. Second, the roll off of the low-frequency gain also improves the stability of the LNA, since a low-frequency gain peak is generally associated with a decreased value of K.
Since L1 is also used for the insertion of gate voltage, the inductance must be adequately bypassed at the normal operating frequency by C4. Resistor R1 and capacitor C5 provide a good low-frequency resistive termination, which enhances low-frequency stability. If the value of C4 is too large, the series-resonant frequency of L1 and C4 can often produce a low-frequency gain resonance, which may prove difficult to stabilize with R1 and C5. With the help of the optimization function in ADS, values for L1, C4, and R1 can be selected for best low-frequency stability. Capacitor C5 is chosen as 10 nF to improve the output third-order intercept; its value is kept constant during the optimization process.
The output highpass network consists of series capacitor C3 and shunt inductor L4. Since L4 is used to apply drain voltage to Q2, it is bypassed with capacitor C10. Resistor R8 and capacitor C11 provide a low-frequency resistive termination.
Since the primary goal of the input network is to provide a low noise figure coincident with good S11 and the primary goal of the output network is to provide the desired power output with good S22, the interstage network can be called upon to help flatten gain over the desired bandwidth, to decrease gain at lower frequencies, and to help with overall stability. With inductors required to apply drain voltage to Q1 and gate voltage to Q2, and a capacitor required for DC isolation between the two stages, this forms the basis for a highpass network. The ADS software can be used to optimize this network for the various parameters.
The importance of properly grounding the transistors' source terminals cannot be overemphasized. Although the shortest length between the device and signal ground plane will generally provide the highest gain at lower frequencies, some controlled amount of source inductance can be used to lower gain, increase stability, and improve S11 and S22 with minimal effect on noise figure. Accurately modeling the microstripline dimensions between each source terminal and the plated throughhole, and the plated throughhole dimensions between the microstrip and the signal ground plane, will allow a designer to use ADS to determine the optimum amount of source inductance. Since source inductance generally makes the transistors regenerative at higher frequencies and degenerative at lower frequencies, a plot of K from 100 MHz to 18 GHz will reveal an optimum source inductance to be used in the circuit.
Once the RF matching has been established, the next step is to establish DC biasing. Figure 3 shows a passive biasing example. In this example, the voltage drop across resistors R4 and R8 sets the drain current (Id). The values of these two resistors can be calculated using Eq. 1:
Vdd = the power supply voltage, 3.3 V;
Vds = the device drain to source voltage, 2.7 V;
Vg = the device gate to source voltage, 0.515 V;
Ids = the device drain to source current, 15 mA;
Ibb = the DC stability (around 10 times the typical gate current, 0.1 mA).
A voltage divider network using resistors R2 and R3 establishes the typical gate bias voltage (Vg):
The complete passive bias example may be found on page 10 of the product data sheet (http://literature.agilent.com/litweb/pdf/5988-9006EN.pdf). Note that there are differences between calculated values and the actual values due to the use of preferred component values as shown in the table.
The fabricated amplifier was tested at a power supply voltage Vdd of 3.3 V, which provides each device with a bias point of Vds = 2.7 V @ Id = 15 mA. Figure 4 shows the measured and simulated noise figures. The noise figure is nominally 1.4 dB at 5.8 GHz. The loss of the input microstripline has been measured at 0.15 dB, making the noise figure of the device plus the loss of the matching network around 1.25 dB. The output power at 1-dB gain compression (P−1 dB) was measured as +11.5 dBm. The output third-order intercept point (OIP3), was measured at +28 dBm.
The measured and simulated gain of the amplifier is nominally 22 dB at 5.8 GHz. The swept gain plot in Fig. 5 shows moderate gain roll off at lower frequencies. Figures 6 and 7 show measured and simulated input and output return losses, respectively. The input return loss measures greater than 15 dB and output return loss greater than 9.5 dB across the 4.9-to-6.0-GHz band.
Special thanks are due to Irene Armenta of Agilent Technologies for all her hard work in turning around the printed-circuit-board (PCB) layout during a factory holiday.
- Al Ward, "Agilent ATF-54143 Low Noise Enhancement Mode Pseudomorphic HEMT in a Surface Mount Plastic Package," 2001 . Available from: http://literature.agilent.com/litweb/pdf/5988-9006EN.pdf.