SEMICONDUCTORS

Materials and Devices

Solid State Technology

The world of wireless communications really began to take off with the development of semiconductor technology. Another name for semiconductor technology is solid state technology. The reason it is referred to as solid state is that semiconductors are a solid material. (It also helps distinguish it from liquid state technology, in which the electrical components are made out of chocolate milk.) The other type of technology, soon to be extinct, is gaseous technology. In gaseous technology, all the electrical stuff happens in a gas. An example of a gaseous product are the old vacuum tubes found in the television sets of the sixties. Many of those old tubes were actually RF amplifiers in disguise. In those tubes, the RF signal got bigger (amplified) while it floated around in a gas in the tube.

Gas-based electrical products (tubes) are, with only a few exceptions, no longer used in RF communications. It is not that the old tubes can't do what semiconductors can—they can, it is that semiconductors have two really attractive properties which tubes do not: they can be made very small and very cheap (not to mention they don't break when they're dropped).

Silicon and Gallium Arsenide

There are two primary semiconductor materials used to manufacture RF components: silicon (Si) and gallium arsenide (GaAs), also known as "gas." In general, GaAs is used for higher frequency applications. If some RF component utilizes GaAs, chances are that silicon didn't work at the (high) frequency of intended use. When given a choice, an RF engineer will choose a silicon device over a GaAs device because it is less expensive.

There are other, more exotic materials used to make RF components in addition to silicon and GaAs. However, they are all relatively new, not widely used yet, and tend to be a combination of GaAs or Silicon and some other obscure material. So for the time being, just know that there are other materials out there.

Diodes and Transistors

There are only two basic semiconductor building blocks used in the RF world: diodes and transistors. Before you start thinking that things are pretty simple with only two basic building blocks, know that there are a lot of different kinds of diodes and transistors and they are all used for different reasons. Table 5-1 highlights the most popular diodes and transistors used in RF systems.

Table 5.1. RF Diode and Transistor Types

Diode Types	Transistor Types
PIN	MESFET
Schottky	MOSFET
Gunn	Bipolar
Impatt	HEMT & PHEMT
Tunnel	JFET
Varactor	LDMOS

Before you start feeling overwhelmed with all the different types of diodes and transistors, understand that the primary difference between them is in how they are fabricated and from what material they are made. All the diodes pretty much do the same thing, but because they are fabricated differently, they have superior electrical performance in different areas. For instance, Schottky diodes are fabricated to be fast, while PIN diodes are fabricated to handle a lot of power. The same goes for transistors. Most diode types indicated in Table 5-1 can be made of either silicon or GaAs, while the transistors are made of one material or the other, but not both.

Diodes

Diodes are used in many different components in the RF world, but they are primarily used in three components: switches, mixers, and voltage variable attenuators (VVA). (If you want to know what a diode looks like in block diagram form, it is the weird shape on the attenuator shown in Figure 4-6.) If Schottky diodes are fast and PIN diodes can handle substantial amounts of power, what diode is used if the switch needs to be fast? How about if it has to handle a lot of power?

As mentioned in Chapter 3 on sources, most oscillators utilize some sort of material to determine the actual frequency of oscillation. This technique works well for "lower" RF frequencies. For "higher" RF frequencies (greater than 10 GHz), material choice is limited and so oscillators frequently use a diode to determine the frequency of oscillation. Gunn, Tunnel, and Impatt diodes are all used to generate these "higher" RF frequencies in oscillators, especially Impatts, which are used for super high RF frequencies (greater than 100 GHz).

Varactor diodes can be thought of as "varible diodes" and are used in voltage controlled oscillators (VCO). Recall from Chapter 3 that VCOs are oscillators whose output frequency can vary over some range. Well, the reason it can vary is because it utilizes a "varible diodes" (varactor) to determine its frequency. (All of this stuff really does make sense.)

Transistors

Transistors are used extensively in RF—from low noise amplifiers to high power amplifiers, to switches, attenuators, mixers, oscillators, you name it. Transistors are the workhorse of RF. Block diagrams of the two most common transistor types are shown in Figure 5-1, strictly for entertainment purposes.

Here is a great rule to remember: if there is gain, there is at least one transistor. All solid state amplifiers use transistors of one kind or another. To produce a lot of gain, amplifiers require two or more transistors.

The lowest frequency (less than 1 GHz) transistor used in RF is the MOSFET, which stands for Metal Oxide Semiconductor Field Effect Transistor. Your intuition should tell you that MOSFETs are made from silicon and they are primarily used in high power amplifiers (HPA).

Above 1 GHz, RF engineers choose between bipolar transistors and MESFETs, which stands for Metal Semiconductor Field Effect Transistor. Bipolar transistors are always made from silicon and MESFETs are always made from GaAs. Which transistor type is better? Well, if cost is an issue, then (silicon) bipolar transistors are the better choice because they are cheaper. If, on the other hand, the transistor needs to operate at a particularly high frequency, then the (GaAs) MESFET is the better choice. As another consideration, at frequencies where either transistor type can be used, bipolar transistors are manufactured to produce more RF power than MESFETs, while MESFETs provide a lower noise figure (NF) than bipolars. In summary, bipolars are less expensive and produce high power, while GaAs FETs cost more but work better at higher frequencies while delivering lower noise figures.

There is a new type of transistor on the block which is known as LDMOS (pronounced l dee' mäs), which stands for Laterally Diffused Metal Oxide Semiconductor. (I couldn't make this stuff up.) Some bright young RF engineer discovered that if a MOSFET is fabricated a little differently, it can be made to work at frequencies above 1 GHz. Where do you suppose these LDMOS transistors get used? If you said high power amplifiers above 1 GHz, give yourself a prize.

HEMT Transistors

There is new family of transistors, specifically designed for super high frequency applications, known as HEMT and PHEMT, which stands for High Electron Mobility Transistor and Pseudomorphic HEMT, respectively. (How many people do you suppose you would have to ask before you could find one who knows what "pseudomorphic" means?) Picture the electrons in an ordinary MESFET as cars traveling on a regular highway (speed limit: 55 MPH). HEMT transistors are nothing more than MESFETs in which the electrons have been given their own high speed autobahn. If the electrons can travel faster, then the transistor can work at higher frequencies. HEMT transistors are particularly suited to high frequency, low noise applications.

Something to note: When I say certain transistors "work" at higher frequencies and others do not, what I really mean is that certain transistors work well at higher frequencies and others perform poorly, even though technically they "work" at higher frequencies. Poor performance manifests itself as degraded electrical parameters like power, gain, and noise figure.

Integrated Circuits (MMIC)

Some clever engineers figured out awhile back that if an amplifier (or other device) needs, say, three transistors, two diodes, and a bunch of other electrical components, why not just put all of these goodies onto a single piece of semiconductor (silicon or GaAs). There are many advantages to doing this, including lower cost and smaller size. When more than one electrical device (transistor, diode, etc.) is combined onto a single piece of semiconductor, it is called an integrated circuit (IC).

Of course, RF engineers couldn't let their devices be called just integrated circuits for fear that people might confuse them with other integrated circuits, like the Pentium microprocessor in a PC. RF engineers complicate matters by calling their ICs Monolithic Microwave Integrated Circuits or MMICs (pronounced mim' iks). (Whatever.)

Most of the components covered to this point can be made as MMICs. A MMIC is not a particular device, it is a manufacturing technology used to realize particular components. For instance, a MMIC which uses a bunch of transistors to make an amplifier is called a MMIC amplifier, while a MMIC which uses a bunch of transistors to make a switch is called a MMIC switch. (You get the idea.)

MMIC Performance

If a component can be made as a MMIC, which is smaller and cheaper (per unit) than other approaches, why aren't all components made as MMICs? There are two drawbacks to MMICs. First, since they involve the semiconductor manufacturing process, they are very expensive if only a few will be needed. Therefore, MMIC technology is only used when the volume requirement is sufficient enough to justify the initial investment to develop the MMIC. Second, quite often MMICs have worse performance (on key parameters) than the same device made out of individual components. Where performance is at a premium, like low noise in a low noise amplifier or high power in a high power amplifier, the devices are made out of individual transistors (or diodes). When performance is not as critical, and cost is, the devices used are MMICs.

To quickly review, there are two main semiconductor materials used in the RF world: silicon and gallium arsenide. Silicon and gallium arsenide are used to make two basic building blocks: diodes and transistors. RF devices can be manufactured two ways: by combining individual diodes and transistors or as integrated circuits (MMICs).