Menu Bar

Main Window

Trash Theme Home News Downloads Tutorials WebLinks Login to Electromaniacs

  Menu Bar
Main Menu
Live Chat
News Feeds
How To
Online tools
Electronics lessons
Download Zone
Contact Us
Forum Wrapper
Books store
About us
Members: 3887
News: 247
Web Links: 21
Visitors: 2975050
Your IP
You are connecting to this site from:
Home arrow Blog arrow Improved InGaP/GaAs HBT technology facilitates high linearity PAs
Improved InGaP/GaAs HBT technology facilitates high linearity PAs PDF Print E-mail
Written by Bogdan   
Article Index
Improved InGaP/GaAs HBT technology facilitates high linearity PAs
Page 2

Gallium arsenide (GaAs) heterojunction bipolar transistor (HBT) has become the dominant technology for handset power amplifier (PA) applications due to its linearity and efficiency. Consequently, GaAs HBT with the emitter made of indium gallium phosphide (InGaP) is widely used in handset PAs, and is gaining acceptance in infra-structure amplifier designs. To address base station applications with power levels above multi-Watt, operation voltage higher than the common 3 V to 7 V now used must be adopted. The base station PA application commonly requires 24 V to 28 V operation, and occasionally even above 30 V[1,2]. An InGaP/GaAs HBT technology was developed in our labs for 28 V power amplification[3-5], which is compatible with standard MMIC technology.

To achieve high breakdown voltage in InGaP/GaAs HBT, a thick collector is required: 3 m thickness is commonly used. The thick collector presents challenges to semiconductor processing in lithography and step coverage of metal interconnection. Power transistors demand large total device size. Multiple fingers are arranged into a single building block. Such building blocks can deliver 2 W RF power in the 1 GHz to 3 GHz frequency band. Multiple building blocks can be arrayed into a single-power HBT for higher power levels. The fT and fmax of the basic HBT finger are 6.4 GHz and 25 GHz respectively.

In a typical lineup of power amplifiers, the power stage is often a class B circuit for the best efficiency, and the driver stage a class AB for a trade-off of linearity and efficiency, and the pre-driver stage may be a class A amplifier. Working with InGaP/GaAs HBT, the goal is to operate the driver and pre-driver stages in near-class B operation for better efficiency, while achieving a superior linearity at the back-off power level.

Class B operation with high linearity can be achieved by a low-frequency low-source impedance, or a dynamic bias circuit[7]. The ruggedness is improved by increasing the ballasting to withstand high output mismatch of 10:1 VSWR with typical input power at 1 dB gain compression and overdrive condition of typical 8 dB gain compression into 50 Ω load at 30 V.

Accelerated lifetime test at 315 C junction temperature and 28 V bias was repeated. More than 3000 hours lifetime test on HBTs was achieved; the Gummel plot before and after the 3000 hours lifetime test shows no increase of the leakage current. The high linearity power performance in class B condition in the back-off power level, the ruggedness under mismatch and overdrive condition, and the long lifetime of the InGaP/GaAs HBT technology makes it a contender for the 28 V power amplifier application

High-voltage HBT fabrication

InGaP/GaAs HBT for 28 V operation is identical to common 5 V counterpart in its epitaxial layers except the collector thickness. Increasing collector thickness to 3 mm enables the base-collector breakdown voltage to go more than 60 V. Likewise, monolithic microwave integrated circuit (MMIC) for 28 V HBT is achieved with two interconnection metal layers, MIM capacitor, thin-film resistor, spiral inductor, and through substrate via holes. The substrate is thinned to 4 mils.

As a result, BVcbo of 70 V is achieved with BVceo over 30 V. With 28 V as the bias voltage, the collector voltage in RF operation will swing much above 28 V and exceed the BVceo. Such condition does not present any concern to the device operation or its reliability[8].

Power HBT design

Power HBT, like other semiconductor technology, is made of multiple small devices strung in parallel. Thermal resistance design is the first task for any power device. Sufficient spreading of the active HBT fingers across the IC die is done in the conventional MMIC approach. Bipolar transistor is known to require ballasting since Vbe has a negative temperature coefficient.

The individual HBT finger size is balanced between the RF performance and thermal resistance. Multiple HBT fingers are linked into the basic building block. The building block in the present design delivers around 2 W RF power at 2 GHz. Each building block has a MMIC prematch circuit.

RF performance

The bias circuit for the 28 V power HBT is implemented on the same chip through the current mirror approach. Excellent thermal stability is achieved: less than 9% change in the quiescent current is achieved over -40 C to +85 C.

Linearity improvement: The driver stage for the power amplifier chain is often biased toward class A in order to provide the needed linearity. This approach sacrifices the operation efficiency. It was found that a low-frequency low-source impedance matching improves the linearity in near-class B operation[4,6]. At the dc side of the choke, a 6 F shunt capacitor is used to provide a low impedance at the modulation frequency and 5 dB improvement in the IM3 is observed. The improvement comes from the elimination of the low-frequency component (ω12) at the input, which if existing will mix with ω1 and ω2 to generate the third-order distortion[6].

Further improvement on the linearity in class AB or class B operations were achieved via a dynamic bias circuit. The major non-linearity in the bipolar transistor is found to come from the exponential I-V relationship. Following the previously reported analysis[11], HBTs are found to have similar behavior as LDMOS in the third-order derivative of the function Ic(Vin) as shown in Figure 2.

The curve of Ic-Vin along the load line follows the exponential relationship with the ballasting resistor effect. The quiescent bias point is at point I. For conventional class B operation, the dc average voltage will remain at point I regardless of the RF swing. With the dynamic bias circuit, the time average bias point is lifted up to conditions II then III as the input power level is increased. The RF voltage avoids swinging into the peaking portion of the gm (Vin) curve as the bias point is lifted by the dynamic bias circuit. Thus, improving the linearity in near-class B operation. Figure 2 shows the analysis of the third-order intermodulation distortion (IMD3) contributed by the transconductance non-linearity in the bipolar transistor.

< Prev   Next >
Sponsored Links

  Privacy Policy