The Transistor reference article from the English Wikipedia on 24-Jul-2004
(provided by Fixed Reference: snapshots of Wikipedia from wikipedia.org)

Transistor

Support a children's charity online
Photo with transistor typesEnlarge

Photo with transistor types

The transistor is a solid state semiconductor device used for amplification and switching, and has three terminals. A small current or voltage applied to one terminal controls the current through the other two, hence the term transistor; a voltage- or current-controlled resistor. It is the key component in all modern electronics. In digital circuits, transistors are used as very fast electrical switches, and arrangements of transistors can function as logic gates, RAM-type memory and other devices. In analog circuits, transistors are essentially used as amplifiers.

Transistor was also the common name in the sixties for a transistor radio, a pocket-sized portable radio that utilized transistors (rather than vacuum tubes) as its active electronics. This is still one of the dictionary definitions of transistor.

Table of contents
1 Importance
2 Invention
3 How Does a Transistor Work?
4 Advantages of Transistors over Thermionic Valves
5 Use in audio amplifiers
6 Use in computers
7 How semiconductors work
8 Light-sensitivity
9 History
10 See also

Importance

The transistor is considered by many to be one of the greatest discoveries or inventions in modern history, ranking with banking and the printing press. Key to the importance of the transistor in modern society is its ability to be produced in huge numbers using simple techniques, resulting in vanishingly small prices. Computer "chips" consist of millions of transistors and sell for dollars, with per-transistor costs in the thousandths-of-pennies.

The low cost has meant that the transistor has become an almost universal tool for non-mechanical tasks. Whereas a common device, say a refrigerator, would have used a mechanical device for control, today it is often less expensive to simply use a few million transistors and the appropriate computer program to carry out the same task through "brute force". Today transistors have replaced almost all electromechanical devices, most simple feedback systems, and appear in huge numbers in everything from computers to cars.

Hand-in-hand with low cost has been the increasing move to "digitizing" all information. With transistorized computers offering the ability to quickly find (and sort) digital information, more and more effort was put into making all information digital. Today almost all media in modern society is delivered in digital form, converted and presented by computers. Common "analog" forms of information such as television or newspapers spend the vast majority of their time as digital information, being converted to analog only for a small portion of the time.

Invention

The transistor was invented at Bell Laboratories in December 1947 (first demonstrated on December 23) by John Bardeen, Walter Houser Brattain, and William Bradford Shockley, who were awarded the Nobel Prize in physics in 1956. Ironically, they had set out to manufacture a field-effect transistor (FET) predicted by Julius Edgar Lilienfeld as early as 1925 but eventually discovered current amplification in the point-contact transistor that subsequently evolved to become the bipolar junction transistor (BJT).

How Does a Transistor Work?

A transistor is a three-terminal device. In a BJT, an electrical current is fed into the base (B) and modulates the current flow between the other two terminals known as the emitter (E) and collector (C). In FETs, the three terminals are called gate (G), source (S) and drain (D) respectively, and it is the voltage applied to the gate terminal that modulates the current between source and drain.

Bipolar Junction Transistor (BJT)

Image:iconsBJT.png
The schematic symbols for pnp- and npn-type BJTs.

Conceptually, one can understand a bipolar junction transistor as two diodes placed back to back, connected so they share either their positive or their negative terminals. The forward-biased emitter-base junction allows charge carriers to easily flow out of the emitter. The base is made thin enough so that most of the injected carriers will reach the collector rather than recombining in the base. Since small changes in the base current affect the collector current significantly, the transistor can be used in an electronic amplifier. The ratio of the collector current to the base current, called current gain or β, is on the order of 100 for most types of BJTs. That is, one milliampere of base current usually induces a collector current of about a hundred milliamperes. BJTs prevail in all sorts of amplifiers from audio to radio frequency applications and are also popular as electronic switching devices.

Bipolar transistors can be fabricated to match with like devices much better than FETs, making them useful for high precision analog circuit design. This is because an op-amp (and discrete transistor amplifier) input structure called a 'current mirror' can be constructed out of them. Bipolar devices can be used to measure temperature and to compute analog logarithms as well.

Field-Effect Transistor (FET)

Image:iconsMOSFET.png
The schematic symbols for p- and n-channel MOSFETs. The symbols to the right include an extra terminal for the transistor body (allowing for a seldom-used channel bias) whereas in those to the left the body is implicitly connected to the source.

The most common variety of field-effect transistors, the enhancement-mode MOSFET (metal-oxide semiconductor field-effect transistor) consists of a unipolar conduction channel and a metal gate separated from the main conduction channel by a thin layer of (SiO2) glass. This is why an alternative name for the FET is 'unipolar transistor.' When a potential difference (of the proper polarity) is impressed across gate and source, charge carriers are introduced to the channel, making it conductive. The amount of this current can be modulated, or (nearly) completely turned off, by varying the gate potential.

Because the gate is insulated, no DC current flows to or from the gate electrode. This lack of a gate current (as compared to the BJT's base current), and the ability of the MOSFET to act like a switch, allows particularly efficient digital circuits to be created, with very low power consumption at low frequencies. The power consumption increases markedly with frequency, because the capacitive loading of the FET control terminal takes more energy to slew at higher frequencies, in direct proportion to the frequency. Hence, MOSFETs have become the dominant technology used in computing hardware such as microprocessors and memory devices such as RAM. Bipolar transistors are more rugged and hence more useful for low-impedance loads and inductively reactive (e.g. motor) loads.

Power MOSFETs become less conductive with increasing temperature and can therefore be applied in shunt, to increase current capacity, unlike the bipolar transistor, which has a negative temperature coefficient of resistance, and is therefore prone to thermal runaway. The downside of this is that, while the power FET can protect itself from overheating by diminishing the current through it, high temperatures need to be avoided by using a larger heat sink than for an equivalent bipolar device. Macroscopic FET power transistors are actually composed of many little transistors. They are stacked (on-chip) to increase breakdown potential and paralleled to reduce Ron, i.e. allowing for more current, bussing the gates to provide a single control (gate) terminal.

The depletion mode FET is a little different. It uses a back-biased diode for the control terminal, which presents a capacitive load to the driving circuit in normal operation. With the gate tied to the source, a DFET is fully on. Changing the potential of a DFET (pulling an N-channel gate downward, for example) will turn it off, i.e. 'deplete' the channel (drain-source) of charge carriers. MOSFETs, formerly called IGFETs (for Insulated Gate Field-Effect Transistor) can be depletion-mode, enhancement-mode, or mixed-mode, but are almost always enhancement mode in modern commercial practice. This means that, with the source and gate tied together (thus equipotential) the channel will be off (high impedance or non-conducting). The n-channel device (reverse for P-channel), like in the DFET, is turned on by raising the potential of the gate. Typically, the gate on a MOSFET will withstand +-20V, relative to the source terminal. If one were to raise the gate potential of an n-channel device without limiting the current to a few milliamps, one would destroy the gate diode, like any other small diode. Why do we typically think of n-channel devices as the default? In silicon devices, the ones that use majority carriers that are electrons, rather than holes, are slightly faster and can carry more current than their P-type counterparts. The opposite is true in GaAs devices.

The FET is simpler in concept than the bipolar transistor and can be constructed from a wide range of materials.

The most common use of MOSFET transistors today is the CMOS (complementary metallic oxide semiconductor) integrated circuit which is the basis for most digital electronic devices. These use a totem-pole arrangement where one transistor (either the pull-up or the pull-down) is on while the other is off. Hence, there is no DC drain, except during the transition from one state to the other, which is very short. As mentioned, the gates are capacitive, and the charging and discharging of the gates each time a transistor switches states is the primary cause of power drain.

The C in CMOS stands for 'complementary.' The pull-up is a P-channel device (using holes for the mobile carrier of charge) and the pull-down is N-channel (electron carriers). This allows busing of the control terminals, but limits the speed of the circuit to that of the slower P device (in silicon devices). The bipolar solutions to push-pull include 'cascode' using a current source for the load. Circuits that utilize both unipolar and bipolar transistors are called Bi-Fet. A recent development is called 'vertical P.' Formerly, BiFet chip users had to settle for relatively poor (horizontal) P-type FET devices. This is no longer the case and allows for quieter and faster analog circuits.

A clever variant of the FET is the dual-gate device. This allows for two opportunities to turn the device off, as opposed to the dual-base (bipolar) transistor which presents two opportunities to turn the device on.

FETs can switch signals of either polarity, if their amplitude is significantly less than the gate swing, as the devices (especially the parasitic diode-free DFET) are basically symmetrical. This means that FETs are the most suitable type for analog multiplexing. With this concept, one can construct a solid-state mixing board, for example.

The power MOSFET has a 'parasitic diode' (back-biased) normally shunting the conduction channel that has half the current capacity of the conduction channel. Sometimes this is useful in driving dual-coil magnetic circuits (for spike protection), but in other cases it causes problems.

The high impedance of the FET gate makes it rather vulnerable to electrostatic damage, though this is not usually a problem after the device has been installed.

A more recent device for power control is the insulated-gate bipolar transistor, or IGBT. This has a control structure akin to a MOSFET coupled with a bipolar-like main conduction channel. These have become quite popular.

(For more on FET's, and MOSFETs in particular, see Field effect transistor)

Advantages of Transistors over Thermionic Valves

Before the transistor, the thermionic valve, or vacuum tube, was the main component of an amplifier. The key advantages that have allowed transistors to replace their valve predecessors in almost all applications are

Use in audio amplifiers

Some audio amplifiers still use valves, their enthusiasts claiming that their sound is superior.

In particular, some argue that the larger numbers of electrons in a valve behave with greater statistical accuracy, although this ignores the facts that tubes generally have a high-impedance control terminal, and that discrete-transistor (as opposed to, say, op-amp) circuits can also be designed to use large currents. (1 milliamp of current carries about 6.24 million billions of electrons per second.)

Others detect a distinctive "warmth" to the tone. The "warmth" is actually distortion caused by the valves, but some audiophiles find a certain amount of "fuzziness" pleasing. This is 'soft-saturation' and occurs when the valves are overdriven, causing poorly designed tube amplifiers to sound better than poorly designed transistor amplifiers.

Tubes are also preferred in guitar amplifiers which are designed to be overdriven, because they have a different non-linear transfer characteristic than transistors, and create a different, more pleasing spectrum of harmonic distortion or "fuzz". Digital signal processing (DSP) can be used to achieve similar effects in the digital domain.

It is possible to mix transistors and valves in the same circuit.

Simple transistor amplifiers use emitter degeneration to achieve negative feedback, which gives a relatively predictable gain compared to the gain of the transistor itself, which varies widely.

Use in computers

The "second generation" of computers through the late 1950s and 1960s featured boards filled with individual transistors, and magnetic cores. Subsequently, transistors, other components (capacitors, but not high-value inductors or transformers), and the necessary wiring were integrated into a single, mass-manufactured component: the integrated circuit. In modern digital electronics, single transistors are very rare, though they remain common in power and analog applications. Recently, inroads have been made in the integration of analog circuits, also, with the advent of 'programmable analog' circuits. DSP is a technique that can (among other things) be used with A/D and D/A converters to simulate analog circuits. Linear integrated circuits got a bad reputation early on because of the difficulty of creating (high-quaity) PNP transistors, but are much better now.

How semiconductors work

Operationally, semiconductors and vacuum tubes are similar. Think about a container (borosilicate glass or platinum, say) filled with distilled water. Drop in a pair of resistance probes and one would find that this arrangement does not conduct electricity, because there is no charge carrier, as long as the sense potential is kept below the electrolysis point--i.e. breakdown. Add a pinch of (any) salt, and conduction begins, because mobile carriers (ions) have been added. Increasing the salt concentration helps with conduction, but not very much. A dry lump of salt is non-conductive, because the charge carriers are immobile.

In a valve, on the other hand, the charge carriers (electrons) are generated by thermionic emission from a cathode heated by a resistance wire.

A pure silicon lattice is an insulator, but when boron atoms are added, in a concentration small enough not to disturb the regularity of the lattice, they donate free electrons and allow conduction. This is because we have introduced a mobile carrier of charge. The lattice seems like a vacuum to the charge carrier. Alternatively, we can introduce an arsenic impurity, which has only three electrons to silicon's four, and we introduce another kind of (virtual) carrier of charge called a 'hole.' This conducts also, yielding p-type material. Valves, in contrast, do not contain positive charge carriers.

Introduce a slightly higher doping concentration, and we have an even more (slightly) conductive material, so long as the lattice remains intact. The part of the bipolar transistor designated as the collector has a higher doping concentration (the ratio directly determines gain) than the emitter. Back-biasing of a diode, or turning off a transistor, is analogous to somehow removing the salt from the water, or freezing it. Note that in all cases the charge carriers repel one another, evening out their distribution in their respective matrices, in the absence of an outside electrostatic force. The grid in a tube deflects the charge carriers rather than extinguishing them. This structure is metallic, rather than semi-metallic as in an transistor, so it can withstand higher ambient temperatures, and dissipate local sources of heat (counter emf from say, droping an inductive load) better, because of its lower specific heat.

Do not think you can make a bipolar transistor out of two diodes. You need to have them on the same crystal, otherwise you don't have epitaxy.

Light-sensitivity

Bipolar transistors can be turned on with light as well as electricity. Devices designed for this purpose are called phototransistors, but these are often standard transistors in a transparent package.

History

All transistors rely on the ability of certain materials, known as semiconductors, to change their electrical resistance under the control of an electric field. In bipolar transistors, the semiconductor is formed into structures called p-n junctions that allow electricity to flow in only one direction through them – that is, they are a conductor when voltage is applied in one direction, and an insulator when it is applied in the other direction.

1900s

Semiconductors had been used in the electronics field for some time before the invention of the transistor. Around the turn of the 20th century they were quite common as detectors in radios, used in a device called a "cat's whisker". These detectors were somewhat troublesome, however, requiring the operator to move a small tungsten filament (the whisker) around the surface of a crystal until it suddenly started working. Then, over a period of a few hours or days, the crystal would slowly stop working and the process would have to be repeated. At the time their operation was completely mysterious. After the introduction of the more reliable and amplified vacuum tube based radios, the cat's whisker systems quickly disappeared. The "cat's whisker" is an example of a special type of diode still popular today, called a Schottky diode.

World War II

In WWII, radar research quickly pushed the frequencies of the radio receivers inside them into the area where traditional tube based radio receivers no longer worked well. On a whim, Russell Ohl of Bell Laboratories decided to try a cat's whisker. After hunting one down at a used radio store in Manhattan, he found that it worked much better than tube-based systems.

Ohl investigated why the cat's whisker functioned so well. He spent most of 1939 trying to grow more pure versions of the crystals. He soon found that with higher quality crystals the "oddness" went away, but so did their ability to operate as a radio detector. One day he found one of his purest crystals nevertheless worked well, and interestingly, it had a clearly visible crack near the middle. However as he moved about the room trying to test it, the detector would mysteriously work, and then stop again. After some study he found that the behaviour was controlled by the light in the room – more light, more conductance. He invited several other people to see this crystal, and Brattain immediately realized there was some sort of junction at the crack.

Further research cleared up the remaining mystery. The crystal had cracked because either side contained very slightly different amounts of the impurities Ohl could not remove – about 0.2%. One side of the crystal had impurities that added extra electrons (the carriers of electrical current) and made it a conductor. The other had impurities that wanted to bind to these electrons, making it an insulator. When the two were placed side by side the electrons could be pushed out of the side with extra electrons (soon to be known as the emitter) and replaced by new ones being provided (say from a battery) where they would flow into the insulating portion and be collected by the filament (the collector). However, when the voltage was reversed the electrons being pushed into the collector would quickly fill up the "holes", and conduction would stop almost instantly. This junction of the two crystals (or parts of one crystal) created a solid-state diode, and the concept soon became known as semiconduction. 'Anode' and 'Cathode' are the terms used to denote the two terminals of a diode. The mechanism of action when the diode is off has to do with the separation of charge carriers around the junction. This is called a 'depletion region.'

Development

Armed with the knowledge of how these new diodes worked, a crash effort started to learn how to build them on demand. Teams at Purdue University, Bell Labs, MIT, and the University of Chicago all joined forces to build better crystals. Within a year germanium production had been perfected to the point where military-grade diodes were being used in most radar sets.

The key to the development of the transistor was the further understanding of the process of the electron mobility in a semiconductor. It was realized that if there was some way to control the flow of the electrons from the emitter to the collector, one could build an amplifier. For instance, if you placed contacts on either side of a single type of crystal the current would not flow through it. However if a third contact could then "inject" electrons or holes into the material, the current would flow.

Actually doing this appeared to be very difficult. If the crystal were of any reasonable size, the amount of electrons (or holes) supplied would have to be very large – making it less than useful as an amplifier because it would require a large current to start with. That said, the whole idea of the crystal diode was that the crystal itself could provide the electrons over a very small distance. The key appeared to be to place the input and output contacts very close together on the surface of the crystal.

Brattain started working on building such a device, and tantalizing hints of amplification continued to appear as the team worked on the problem. One day the system would work and the next it wouldn't. In one instance a non-working system started working when placed in water. The two eventually developed a new branch of quantum mechanics known as surface physics to account for the behaviour.

Essentially, the electrons in any one piece of the crystal would migrate about due to nearby charges. Electrons in the emitters, or the "holes" in the collectors, would cluster at the surface of the crystal where they could find their opposite charge "floating around" in the air (or water). Yet they could be pushed away from the surface from any other location with the application of a small amount of charge. So instead of needing a large supply of electrons, a very small number in the right place would do the trick.

Their understanding solved the problem of needing a very small control area to some degree. Instead of needing two separate semiconductors connected by a common, but tiny, region, a single larger surface would serve. The emitter and collector would both be placed very close together on one side, with the control lead on the other. When current was applied to the control lead, the electrons or holes would be pushed out, right across the entire block of semiconductor, and collect on the far surface. As long as the emitter and collector were very close together, this should allow enough electrons or holes between them to allow conduction to start.

First transistor

They made many attempts to build such a system with various tools, but generally failed. Setups where the contacts were close enough were invariably as fragile as the original cat's whisker detectors had been, and would work briefly, if at all.

Eventually they had a practical breakthrough. A piece of gold foil was glued to the edge of a plastic wedge, and then the foil was sliced with a razor at the tip of the triangle. The result was two very closely spaced contacts of gold. When the plastic was pushed down onto the surface of a crystal and voltage applied to the other side (on the base of the crystal), current started to flow from one contact to the other as the base voltage pushed the electrons away from the base towards the other side near the contacts. The point-contact transistor had been invented, a primitive variation of the BJT.

While the device was constructed a week earlier, Brattain's notes describe the first demonstration to higher-ups at Bell Labs on the afternoon of December 23, 1947, often given as the birthdate of the transistor. The PNP point-contact germanium transistor operated as a speech amplifier with a power gain of 18 in that trial.

Problem

Shockley was upset about the device being credited to Brattain and Bardeen, who he felt had built it "behind his back" to take the glory. Matters became worse when Bell Labs lawyers found that some of Shockley's own writings on the transistor were close enough to those of an earlier patent that they thought it best that his name be left off the patent application.

Shockley was incensed, and decided to demonstrate who was the brains of the operation. Only a few months later he invented an entirely new type of transistor one day while sitting in his hotel room waiting to give a speech. This new form, the layer transistor, was considerably more robust than the fragile point-contact system, and would go on to be used for the vast majority of all transistors into the 1960s.

With the fragility problems solved, a remaining problem was purity. Making germanium of the required purity was proving to be a serious problem, and limited the number of transistors that actually worked from a given batch of material. One then-small company, Texas Instruments, decided that the solution to this problem was to use silicon rather than germanium, which should be easier to work with. They were right, and germanium disappeared from almost all transistors within only two years of silicon being introduced in the early 1950s.

Now everything was in place, and within a few years, transistor-based products, most notably radios, were appearing on the market. A major improvement in yield came when a chemist advised the fabs to use distilled water rather than tap water: calcium ions were the cause of the problem. 'Zone melting,' a technique using a moving band of molten material through the crystal, makes this whole endeavour possible.

Origin of Name

John R. Pierce coined the name "transistor" in 1949. It was originally thought that the transistor could be usefully considered to be the electronic dual of a vacuum tube. The property equivalent to the transconductance of a tube would have been "transresistance" and the device would then have been a "transresistor," or "transistor" for short. In practice the transistor was not close enough to being a vacuum-tube dual for the concept to have any quantitative usefulness, and the concept of "transresistance" lives on only in the name "transistor."

Early Consumer and Hobbyist Applications

The transistor radio was not the first "mainstream" application of transistors. Even by the 1940s, ordinary consumer radios were rather sophisticated pieces of electronics, using several tubes, and based on Armstrong's brilliantly ingenious superheterodyne technology. To meet consumer expectations, it was necessary for a transistor radio to use similar circuitry. It was not easy in the early days to get transistors to operate reliably as amplifiers and oscillators in the RF range—even the 540-1700 "kilocycle" AM broadcast band. Miniaturized versions of many necessary components such as IF transformers and multiganged tuning capacitors were not available.

The first major consumer application of transistors was the hearing aid, which required only audio amplification, and represented a market where miniaturization was important and low price was not a requirement. Raytheon, which had developed miniaturized and ruggedized vacuum tubes for the military, introduced the first transistorized hearing aids.

Raytheon also introduced the first transistor, the CK722, that was widely available through ordinary commercial channels. Electronics hobbyists of the fifties have a warm place in their heart for the CK722, essentially the only transistor available for almost a decade, and innumerable homebrew projects were designed around it. The CK722's available to hobbyists were, essentially, those that had failed QC for more demanding applications. Germanium based, with low gain and high emitter-to-collector leakage, and high variation in characteristics from unit to unit, designing practical circuits with these components was quite a challenge.

Nowadays, the most visibly popular transistor, at least in the USA, is the 2N3904, a small NPN component that can be found in hundreds, if not thousands, of types of electronic devices. The 2N3906 (its PNP counterpart) and the 2N2222 (NPN) are also very popular. All cost only a few cents in quantity, but even single units can be had quite cheaply, making them popular with hobbyists as well.

Miniaturization

The first CMOS transistor circuit was introduced by RCA in 1963.

Another level of miniaturization later became possible with the invention of the integrated circuit, which included many transistors on one chip of silicon, and led to a new generation of devices such as pocket calculators and digital watches.

NASA was the buyer, paying $4000 in the money of the day for a quad-two-input NAND function. Now this circuit runs for about two cents. It is very costly to put things in space aboard liquid-fueled rockets. Reliability was an even bigger motivation in the space program. Less connections was the advantage here.

See also

External link and references