2015年6月23日 星期二

Viewpoint: GaN power amps start with the transistor

The first step in designing a microwave frequency power amplifier monolithic microwave integrated circuit on a given GaN foundry process is to select the transistor size and bias, writes Liam Devlin.

24jun15GaNFigure 1The high breakdown voltage and operating temperature capability of high electron mobility transistors (HEMTs) fabricated on a gallium nitride (GaN) semiconductor process makes the technology well suited to building power amplifiers (PAs).

A number of commercial GaN foundry processes are now available and the creation of custom microwave GaN PA monolithic microwave integrated circuits (MMICs) is a practical option for many applications.

 

Bandgap size matters

At 3.4eV GaN is a wide bandgap material. This compares to 1.4eV for gallium arsenide (GaAs) which is the other main high-frequency semiconductor process technology. Materials with a wider bandgap can withstand a higher electric field, so transistors made with them can have a higher breakdown voltage, which leads to a higher RF output power capability.

Most of the commercially available GaN foundry processes used for microwave frequency amplifiers are built on semi-insulating silicon carbide (SiC) substrates. The devices are fabricated in a thin layer of GaN on top of the much thicker SiC substrate (Figure 1).

GaN transistors often include a source-coupled field-plate (‘Source FP’ in Figure 1). Its main purpose is to reduce the fields at the surface at high drain‑gate voltages. The presence of the field plate increases the breakdown voltage of the transistor. The downside of having a field plate is that it also increases the gate‑source capacitance, which degrades the maximum operating frequency compared to a device of the same gate length without one.

Another effect of source field plates is that they shield the gate from the drain. This reduces the feedback between the drain and the gate, increasing the RF gain. GaN HEMTs tend to have a higher maximum gain than their GaAs counterparts.

The transistors on GaN IC processes can be operated at significantly higher bias voltages than corresponding GaAs devices – typically 20V to 40V compared to 4V to 12V – so have a higher power density. This is the available output power per milimetre of gate width.

Another effect of the higher bias voltage is that for a given output power a GaN transistor will have a significantly higher output and input impedance than a GaAs PHEMT transistor, so it can be matched more easily, and matching bandwidths can be wider, while matching network losses are normally lower.

Better thermal conductivity

The thermal conductivity of the SiC substrate is good – around 10 times that of GaAs at room temperature. The higher power density of GaN transistors necessitates better thermal conductivity than GaAs, and even with an SiC substrate GaN PA transistors normally run at a higher channel temperature than their GaAs counterparts. However, GaN transistors are able to operate at higher junction temperatures than GaAs for a given reliability.

The RF output power that a transistor can produce increases with the physical size (total gate periphery). However, as the number of transistor fingers increases and/or the unit width of each finger increases, the available gain at microwave frequencies decreases as a result of distributed effects and increased parasitics. Microwave frequency PA MMICs therefore tend to make use of multiple power combined transistors to simultaneously achieve high output power and acceptable gain.

Start with transistor size and bias

The basic approach to designing a microwave frequency GaN PA MMIC on a given foundry process is to start by selecting the preferred transistor size and bias. Devices are normally operated in Class AB, and a quiescent current of 50mA to 100mA per millimeter of gate width is typical. The size of the device needs to be chosen to provide the required output power while achieving adequate gain. The power-added efficiency and linearity performance should also be assessed.

For power-combined reactively matched designs, a single stage design is normally implemented in the first instance. The transistor is first rendered unconditionally stable across the operating band by the addition of appropriate input losses. A load-pull bench is then used to optimise the matching structures. Multiple parallel copies of the single stage can then be combined to generate higher output power levels.

Two-, four- and eight-transistor designs are most common, as they lend themselves to a more practical layout. It is also common to absorb part of the matching network as an integral part of the combiner. Biasing components and networks to ensure broadband stability can now be added. It is normal to start with the output stage and then add the driver stages.

GaN transistors have a soft compression characteristic, and therefore the driver stages must be of sufficient size to avoid a requirement to drive the overall amplifier far into compression to achieve the expected output power.

Liam Devlin is CEO of Plextek RFI



from News http://ift.tt/1HauENM
via Yuichun

沒有留言:

張貼留言