Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.

Millimeter-wave frequencies offer great capacity for a wide variety of short-distance communications links—for systems ranging from wireless local area networks (WLANs) and collision-avoidance radars to satellite-communications (satcom) systems and radiometry systems. But to use these bands, low-noise amplifiers (LNAs) are needed for receiver front ends, and these amplifiers must provide suitable gain along with minimal noise.

Fortunately, a high-gain LNA was developed using a commercial 0.15-μm pseudomorphic high electron-mobility transistor (pHEMT) process and a balanced topology for increased instability with the high gain. The LNA achieves more than 34-dB gain from 30 to 36 GHz, with gain flatness of better than 0.8 dB and noise figure of less than 2.6 dB. And it all comes in a chip measuring just 4.3 × 1.9 mm2.

Many high-frequency LNAs have been designed and based on gallium-arsenide (GaAs) semiconductor technology for its noteworthy high gain and low noise at millimeter-wave frequencies.1-8 A balanced configuration offers numerous benefits in performance at these frequencies, with a typical balanced amplifier configuration consisting of two identical unit amplifiers and two quadrature hybrid couplers (Fig. 1). Input signals are split into quadrature signals by a coupler and fed to the two unit amplifiers. A coupler at the output combines the outputs of the two unit amplifiers. 

Design A Ka-Band High-Gain LNA, Fig. 1

At the input of the balanced amplifier, reflected signals from the two unit amplifiers are 180° out of phase and cancel each other out. For the same reason, at the output of the balanced amplifier, reflected signals will also be canceled, resulting in good input and output return-loss behavior for the balanced amplifier. The balanced amplifier also provides better stability compared to a single-ended amplifier.9-11 In a balanced amplifier, each unit amplifier is terminated with a fixed load, which is close to 50 Ω. So the actual loads at the input and output of the balanced amplifier only minimally affect the loads of the two unit amplifiers. Instead, the single-ended amplifier is affected by actual loads.

The amplifier may be unstable in some frequency ranges, especially if the actual load is a filter. A filter behaves as a 50-Ω load only in the passband. In the stopband, the filter will behave as a purely reactive load to the amplifier. The purely reactive load can cause the single-ended amplifier to become unstable. Furthermore, the failure of one unit amplifier will only cause a gain drop of 6 dB instead of catastrophic failure for the balanced amplifier.

To demonstrate the use of this balanced amplifier design approach, an LNA was designed based on a 0.15-μm GaAs pHEMT process. By employing more amplifier stages, the LNA exhibits a higher gain than the previous balanced LNAs.1,5 The input and output return losses of the LNA are improved by using two optimized Lange couplers at the RF input and output ports, respectively.

With application of a self-bias technique, the LNA can be biased from a single power supply. The LNA exhibits high gain of more than 34 dB with gain flatness within 0.8 dB, and a noise figure of less than 2.6 dB from 30 to 36 GHz. The LNA is stable at any frequency within its operating range.

Design A Ka-Band High-Gain LNA, Fig. 2Active devices for the LNA are grown on 6-in. GaAs substrates using a commercial 0.15-μm pHEMT process. The process provides GaInAs/AlGaAs pHEMT devices, using electron-beam lithography to define 0.15-μm T-gates. The low-noise pHEMTs consist of single-sided doped structures with a single recess. Connections are made by means of air bridge or slot viahole connections.

The process achieves devices with peak transconductance (gm) of 550 mS/mm at a gate-drain voltage of -0.1 VDC, a gate-drain breakdown voltage of +9 VDC, and a threshold voltage of -0.45 VDC. The process delivers devices with transition frequency, ft, of 95 GHz, and maximum frequency of oscillation, fmax, of 160 GHz.

Low input and output return loss can be achieved by means of Lange couplers at the LNA’s input and output ports. Each of the LNA’s unit amplifier’s incorporates five common-source field-effect transistors (FETs), combining to form the balanced monolithic Ka-band amplifier shown in Fig. 2. The active devices are all 2 × 40 μm. The input matching networks of the first two amplifier stages were designed for optimum noise match for low noise figure, while the remaining amplifier stages were conjugate-matched for increased gain. The output matching networks were conjugate-matched for high gain.

Since the input and output return losses depend on the Lange couplers, the two unit amplifiers were optimized for gain, gain flatness, and noise figure. To bias the LNA from a single power supply rail, a self-bias circuit, consisting of a bypass capacitor (C2) and a resistor (R2), was used. For the best noise figure, bias voltages were selected to ensure that the amplifier operates in its low-noise region.

The drain-source voltage (Vds) and the gate-source voltage (Vgs) were selected at +2 VDC and -0.19 VDC, respectively, with the aid of computer software simulations. The source current was selected at 12.7 mA, so that resistors (R2 were chosen with value of 15 Ω to set a gate-source voltage (Vgs) of -0.19 VDC.

Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.