Fullscreen HTML5Med Res (??MB)HTML4

The History of Transistors Part 3: by Ian Batty Left: a Texas Instruments SN7400 quad NAND gate die after its plastic encapsulation was dissolved. Source: https://w.wiki/4mri Below: the 2N2222 50V NPN bipolar junction transistor. Source: https://w. wiki/4pAP Over the last two months, I described the invention of transistor technology and the subsequent innovations and improvements that led to the current transistor technology. In this third and final instalment, we take a more in-depth look at how transistors work, including bipolar junction transistors (BJTs) and both main types of fieldeffect transistors (JFETs and Mosfets). T he previous two articles in this series covered the history of transistor development, from the first-­ generation point-contact transistors to the modern epitaxial planar type. More advanced types exist but are relatively uncommon. Descriptions of devices such as heterojunction and unijunction transistors are available online. Wikipedia is a good starting point; see https://w.wiki/4SJw Those articles described the physical 30 Silicon Chip construction of transistors, their manufacturing processes and the details of how and when they were invented. This one will concentrate on explaining how they behave, starting with some basic semiconductor physics. After that will come information on performance limitations, the origin of the circuit symbol and some typical model numbering schemes. We’ll also cover field-effect transistors (FETs) in some detail, including Australia's electronics magazine junction FETs (JFETs) and metal-oxide semiconductor FETs (Mosfets), used individually and in CMOS (complementary Mosfet) ICs. Let’s start with some fundamental semiconductor theory. Semiconductor physics We’re accustomed to electric current as a flow of electrons. Electrons flow freely in most metals, which is why they are conductive. Fig.43 shows siliconchip.com.au how metal atoms allow electrons from their outer shells to leave the influence of the nucleus and ‘wander about’ in the metal’s crystalline structure. These outer electrons are known as valence electrons. Although electron-deficient atoms become positively-charged ions, the net charge in the metal is zero; the positive and negative charges still balance. The population of free electrons is sometimes known as the ‘electron gas’. Metals have so many free electrons that they are not suitable for the active parts of transistors. We want to influence the conductivity of a transistor’s structure, and metals are already such good conductors that transistor action is impossible only with metals. Fig.43: in conductive metals like copper, the valence electrons are free to roam among the lattice of atoms and hence provide high conductivity. Fig.44: silicon has four valence electrons and forms a very regular crystal with those electrons more-or-less trapped between each pair of adjacent silicon atoms. It therefore has poor natural conductivity, so it is classed as a semiconductor. Semiconductors For simplicity, I’m going to use the atomic model that was standard prior to quantum physics, considering atoms and electrons as distinct objects. Fig.44 shows a crystal of silicon, but the following applies equally well to germanium. With four valence (outer) electrons, pure silicon/germanium crystals form very regular lattices with near-perfect atom-to-atom bonds. These perfect bonds mean that few free electrons can exist. This scarcity of free electrons explains silicon’s poor natural conductivity – it’s a semiconductor. Pure silicon is better known as intrinsic silicon. Its four outermost (valence) electrons class it as a tetravalent element. This tetravalent nature allows silicon atoms to form tight, perfect bonds between each other. Ideally, each set of covalent bonds completely ‘captures’ the electrons in each atom’s outer shell and binds them tightly between their parent atom and its neighbours. The bonding is not totally perfect, however. Some electron motion is possible, which gives silicon a resistance much higher than a true metal such as copper, but less than that of a true insulator such as sulfur. It’s possible to add small amounts of impurity atoms to the crystal to tailor conductivity very exactly. The improved conductivity that comes from this doping by impurities is at the heart of semiconductor technology of all kinds. The effect of doping is to create free charge carriers that are not tightly bound into the silicon-to-­ silicon lattice. siliconchip.com.au Fig.45: when impurities are introduced into the silicon crystal (it’s ‘doped’), in this case, a phosphorus atom, the situation changes. The phosphorus atom has five valence electrons, so one is left free to roam the crystal, giving it a permanent negative charge (making it N-type) and increasing its conductivity. Fig.45 shows the result of heating the silicon to melting point and adding a tiny amount of phosphorus. With five valence electrons, the phosphorous atoms will slot into the crystal structure on solidification, but with only four of each set of five phosphorus electrons taken up into the crystal lattice. Now, each phosphorus atom’s excess electron is free to drift about in the doped silicon crystal. In a true metal, each free electron Australia's electronics magazine leaves behind a positively charged metal ion in the crystal so that the charges balance out over the entire piece. But our doped N-type silicon has a permanent negative charge of electrons. Remember that, although a metal does have free electrons, these are not a permanent surplus. The metal is electrically neutral. Phosphorus, a pentavalent element, is a donor impurity – the “n” in donor reminds us that by phosphorus’ May 2022  31 donation of an electron, the intrinsic semiconductor becomes N-type. Current would flow in this material, pretty much as in a metal. The main difference is that we have exact control over the N-type silicon’s conductivity; heavier phosphorous doping gives more conductivity, light doping gives less conductivity. This helps explain the near-fanatical search for silicon (and germanium) of near-absolute purity. Any ‘foreign’ atoms can have a dramatic and uncontrolled effect on a semiconductor, considering that the doping ratios are so tiny: some as low as one part in 107 (one in 10,000,000). Unwanted contamination must be much less to give reliable doping effects. What if we dope with aluminium, as in Fig.46? It’s a trivalent element, and with only three electrons, there will be a net loss of charge as one silicon atom cannot achieve its preferred ‘take-up’ of four valence electrons. A loss of negative charge must be a supply of positive charge, and this is a positive ‘hole’ – the opposite of an electron. Aluminium, a trivalent element, is an acceptor impurity, with the “p” reminding us that, by aluminium’s acceptance of an electron, the intrinsic semiconductor becomes P-type. If an electron escapes an adjacent atom, it may wander in and fill the hole, but that will leave another hole behind. Thus, the P-type silicon has a permanent net positive charge and is also conductive to an extent determined by the doping concentration. Do holes really exist? Are holes really only a flow of electrons in the opposite direction? This was a critical step in the understanding of semiconductor physics. The way that holes flow is different enough from that of electrons that we are justified in describing hole flow as a distinct kind of current flow. One critical difference is diffusion/flow speed. Holes move more slowly than electrons, and this accounts for NPN transistors having better high-frequency characteristics than PNPs. Electron flow in the N-type emitter and collector of an NPN transistor (the bulk of the entire transistor) is faster than hole flow in the P-type emitter and collector of a PNP transistor. Hole flow actually already exists in some metals, it is just much less common than electron flow. It seems no sooner had we discounted ‘current from positive to negative’ by the discovery of electrons than we needed to call it back from obscurity. Be aware, though, that this is not the conventional current flow model, which – coming so many decades before the identification of hole flow as a real-but-uncommon phenomenon – did not include current carriers. It’s now clear why semiconductor action was initially so hard to describe and understand. Valve theory can be handled pretty well with classical Newtonian physics and the conventional ‘tiny solar system’ model of the atom with electrons orbiting the central nucleus. But semiconductor theory is impossible without delving into the weird world of quantum physics. It’s that complexity which bedevilled Welker, Mataré, Bardeen, Brattain, Shockley and all of the other physicists, chemists and engineers who brought us the transistor. Majority & minority carriers The description so far has shown the intended result of doping: a surplus of electrons in N-type, a surplus of holes in P-type. These are the ‘majority carriers’. In reality, thermal agitation of the crystal lattice (occurring at all temperatures above absolute zero, -273.15°C) will liberate some charges of the opposite polarity to those created by doping; N-type semiconductors will exhibit a small numbers of holes while P-type will exhibit a small numbers of electrons. These are ‘minority carriers’. We might expect minority carriers to be obliterated by the overwhelming number of majority carriers. But in practice, new minority carriers are continually being generated by thermal agitation. Because they are thermally generated, they increase with temperature. Germanium is especially productive in this regard and this is why leakage currents (which are caused by minority carriers) are so troublesome in germanium semiconductors. The combination of high leakage currents and an inability to operate over about 75°C contributed to silicon’s supplanting of germanium in semiconductor devices. This is also the basis of thermal runaway, where leakage at high temperatures causes increased current flow, which causes increased heating and possibly, eventual self-destruction. The semiconductor diode Fig.46: in contrast to phosphorus, aluminium has three valence electrons, so when a silicon crystal is doped with aluminium, it obtains a permanent positive charge (P-type). This results in a ‘hole’ (lack of electron) that can also roam the crystal lattice, albeit with lower mobility than an electron. 32 Silicon Chip The following figures show holes and electrons travelling in straight lines – this simplicity makes the drawings easier to understand. Be aware that, in reality, their paths are random and wandering. Let’s take two pieces of doped semiconductor: N-type and P-type. If we join them, as in Fig.47, we find the junction region ‘populated’ with both holes (P-type) and electrons (N-type). We now have a two-element device – a diode. Holes in the P material and electrons in the N material are mutually attracted and will flow to the junction. On crossing the junction, holes will meet the excess electrons in the N material and will recombine with them. Likewise, electrons crossing the junction will meet excess holes in the P material and recombine with them. Australia's electronics magazine siliconchip.com.au This means that a small region on each side of the junction will contain only the crystal lattice, with the formerly-­ polarised atoms (whether originally P or N) neutralised by the inflow of opposite-polarity charge carriers – see Fig.48. In practice, the ‘rush’ of charge happens progressively in a diffused device, as the top layer diffuses into the bulk of the substrate. The depletion zone will assume a small potential dependent on the type of semiconductor. As this potential prevents charge carriers from crossing it, it appears that any applied voltage must exceed this depletion zone’s effective potential before current can flow. Fig.49 shows that making the P-type more negative and the N-type more positive will cause holes to move to the negative end and electrons to move to the positive end. This is reverse bias for the diode; the depletion zone widens, and no current flows. Minority carriers will cross the depletion zone, and these constitute the diode’s leakage current. As noted above, minority carriers increase with temperature, and occur in much higher numbers in germanium than silicon. If the reverse bias is excessive, minority carriers can reach such high numbers and travel so quickly that they collide with the crystal lattice and ‘knock off’ extra charge carriers. This is the avalanche effect, and it can cause reverse current to skyrocket, destroying the diode through overheating. Alternatively, with a limited current applied as in the case of a zener diode, it is the intended operating mechanism. At least, this is the case for zeners above about 5.1V; they conduct in avalanche mode, whereas below 5.1V, a different conduction mechanism (tunnelling) is used. Fig.50 shows that applying the opposite polarity to the diode (negative to the N-type, positive to the P-type) creates a forward bias. Electrons move away from the negative terminal and towards the depletion zone. Likewise, holes move towards the depletion zone. As the forward bias increases, the depletion zone narrows and is eventually overcome. Current flows through the diode, with a small voltage drop in the depletion zone. Electrons and holes meet and recombine at the junction, and this recombination allows current to flow continuously. Fig.50 appears to show two ‘channels’ in the diode: one for electrons and the other for holes. In reality, it’s a mess. Holes and electrons move like clouds – chaotic when you look closely, but with an overall, predictable direction. For germanium, current flow begins at around 0.1V for junction construction or around 0.4V for alloy-diffused construction. For silicon, it’s around 0.6V for common types. Germanium’s low forward voltage drop was its only real advantage over silicon. Silicon devices such as the schottky diode (using a metal-­semiconductor junction) have lower forward drops of about 0.3-0.4V. This is about half that of a P-N junction diode because the depletion zone is about half as wide; the metal side of the diode has no depletion zone. Schottky diodes withstand lower reverse bias voltages though (for a similar reason) and also have higher leakage currents. The maximum forward current is principally limited by heating in the diode junction due to Ohm’s Law losses. A silicon diode passing a current of 1A will drop as much as one volt, thus converting about 1W of the electrical energy to infrared emissions and heat. The diode must be siliconchip.com.au Fig.47: when N-type and P-type doped silicon crystals meet, the roaming electrons and holes are attracted to each other and ‘cancel out’. Fig.48: the cancellation noted in Fig.47 results in a “depletion zone” forming at the junction of the two zones, where there are neither free-roaming electrons nor holes, thus blocking the flow of current between the zones. Fig.49: by applying a reverse-biased voltage across this PN junction, the depletion zone widens, so current will still not flow. However, that would change if the bias voltage was increased to the point of avalanche breakdown, at which point a high current would suddenly start to flow. Fig.50: on the other hand, if a forward-biased voltage is applied to the PN junction, the depletion zone shrinks, and if the bias voltage is high enough, it is eliminated and the roaming electrons and holes can once again meet. The result is that current will flow, with a slight voltage loss as it crosses the junction (the diode’s forward voltage). Australia's electronics magazine May 2022  33 capable of dissipating this without melting its junction. To handle higher currents, the junction and/or package have to be increased in size, a heatsink needs to be attached or a schottky type (with a lower forward voltage and thus dissipation) used – or a combination of all three. Now for amplification Fig.51: the basic structure of an N-channel JFET. The negatively doped (N-type) channel is connected to the drain and source electrodes on either side via ohmic contacts. The P-type gate(s) form diode junctions with the channel. In operation, a negative voltage is applied to the gates relative to the drain/source, so these junctions are reverse-biased and virtually no current flows. Fig.52: if the negative bias on the JFET gate is high enough, the depletion zone extends all the way through the channel, ‘pinching off’ the current flow between drain and source. Fig.53: with a less negative JFET gate bias, the depletion zone still narrows the conducting channel, decreasing its conductivity, but current can now flow between the source and drain. Fig.54: even with zero gate bias, a depletion zone still exists. This narrows the channel, so a JFET typically does not allow a high current to flow. This property is taken advantage of in ‘current regulator diodes’ (a component you don’t often see these days). 34 Silicon Chip Next, let’s look at the field-effect transistor (FET), the device patented in 1925 by Julius Lilienfeld, and the device that William Shockley and his team tried and ultimately failed to develop. Lilienfeld’s 1925 patent provided a starting point for William Shockley’s efforts at Bell Labs in the 1940s. After much frustration and with only very weak demonstrations of any effect, Shockley’s team (led by Bardeen and Brattain) abandoned the field-effect approach and successfully embarked on point-contact and then junction transistor research. Looking back, it appears that Shockley’s efforts were frustrated by the imperfect nature of his feedstock. Without germanium of near-perfect purity, and without a crystal surface of near-perfect regularity and alignment, his intended electrostatic influence could not penetrate the chaotic and tangled surface of what would be the conducting channel. Ironically, Shockley could well have succeeded had he listened to Gordon Teal’s insistence on using feedstock of the highest possible purity and regularity. Shockley’s field-effect efforts were frustrated by the poorly-understood concept of surface states, the understanding of which eventually led to the successful construction of FETs. Remarkably, this device’s operation is very similar to a triode valve, as had been Shockley’s aim. The FET has a single conducting path between its source (‘cathode’) and its drain (‘anode’), and it presents a very high input impedance at its gate (‘grid’). Two major FET technologies exist. The junction FET (JFET) uses a diode structure for its gate. During regular operation, the diode is reverse-biased, so it allows minuscule current to flow, in the low nanoamps (1/1000 of a microamp) and presents impedances easily exceeding 1000MW. This contrasts with vacuum tubes, where grid currents due to emission and gas effects are commonly in the low microamps range, to give input impedances well under 100MW. JFETs are suitable as low-noise amplifiers, gain control devices and radio-frequency amplifiers into the hundreds of megahertz. Working models were presented in 1953 by George F. Dacey and Ian M. Ross (see http://en.wikipedia. org/wiki/JFET). Actual operation is simplicity itself. Let’s say we use an N-type channel, as in Fig.51. Electrons flow into the channel via the source connection. This is a simple ohmic connection, not a diode junction, so the electron flow continues as electrons; ideally, there are no holes to recombine or carry current in an N-channel FET’s conducting channel. This differs both from the junction transistor and from the vacuum triode. Junction transistors and triodes both create a space charge, either within the base (transistor) or surrounding the hot cathode (vacuum triode). The junction FET needs neither forward bias (transistor action) nor a heated cathode (triode action) to permit conduction. Australia's electronics magazine siliconchip.com.au Fig.55: you can see here how insensitive the JFET’s is to changes in drain-source voltage above a few volts; the channel current remains more-or-less constant, determined mainly by the gate bias. A current-regulator diode is just a JFET with its gate permanently connected to the source, so it always has 0V bias. You can see from this plot how they provide a semi-constant current. Current flows through the channel towards the drain. Again, this is a simple ohmic connection. The device so far appears to be simply a resistor, its initial resistance controlled by the amount of doping in the semiconductor channel. With the N-type channel, a P-type gate is added to the side of the channel. A negative voltage will act as a reverse bias on the P-N diode, so current flow between gate and channel is virtually zero, as shown in Fig.52. The bias penetrates the full depth of the channel and forces current flow to stop. In a valve, we would call it cut-off. In the JFET, this is pinch-off. Fig.53 shows the JFET with a reduced negative bias while Fig.54 shows it with zero bias. There is some depletion zone effect even at zero bias, since the right-hand end of the channel becomes more positive with respect to the zero voltage bias at the gate is closer to the positive drain connection voltage. This makes the gate progressively negative compared to the channel. JFET operation is similar to valve action: with zero bias, about 10mA flows. As the negative bias increases, current falls until the point where the bias voltage causes current flow to cease. If we think of the JFET’s channel as a resistor, it’s having its cross-sectional area reduced. This increases its resistance. For the valve, the effect is like a resistor of constant cross-sectional area but of poorer conductivity (higher ‘natural resistance’) with increasing bias. Remember that the junction transistor has its current carriers diffusing slowly and randomly across the base region. In contrast, the FET’s channel experiences a significant voltage difference (similar to the anode-cathode field in a vacuum tube) that does accelerate current carriers in their path from source to drain. Because the JFET’s gate is not within the channel’s current flow, we don’t get anything similar to the Edison effect we see in valves, where the grid is naturally weakly negative. With the moderate negative bias shown in Fig.54, the depletion zones widen, restricting current flow and the –3V bias reduces the drain current to 2.5mA, ¼ of maximum. siliconchip.com.au So the JFET shows a non-linear transfer characteristic: 50% of cut-off bias allows only 25% of zero-bias current. The curves flatten off to an almost constant current after a few volts are applied across the device, as shown in Fig.55. At lower voltages, the gate voltage to drain current relationship is not linear. The JFET’s curves, in valve terms, are most similar to those of a remote-cutoff pentode. The JFET has no semiconductor junctions in its conduction path, so there is no ‘noisy’ recombination of holes and electrons. Lacking a heated cathode, electron flow does not suffer thermal agitation, so internal noise is low. The JFET is a naturally low-noise device, with noise figures less than 1dB for many types. Unlike valves (but like bipolar transistors), FETs are made in both polarities: a P-channel FET would give exactly the same characteristics as those above, but would be pinched off by a positive gate voltage relative to the source. The JFET’s gate-channel junction overcomes the surface-­ state problem that frustrated Shockley: its reverse-biased diode readily accepts a control voltage and widens its barrier region in response. Mosfets The metal-oxide semiconductor (silicon) FET (Mosfet), also known as the insulated gate FET (IGFET), uses a thin insulating layer between the gate connection and the bulk of the device. Fig.56 shows a simplified version. These FETs offer impedances in the millions of megohms with gate leakage currents below 1nA. As well as high-impedance, radio-frequency and low-noise applications, Mosfet technology is used in high-power switching and linear devices such as for RF and audio power amplifiers, and DC applications such as power controllers in electric cars and switch-mode power supplies. The greatest usage of Mosfets is found in the millions of active sites in microprocessors, where it is known as CMOS (complementary metal-oxide semiconductor) due to the use of both N-channel and P-channel devices. Again, Shockley’s surface-state problems are averted. The semiconductor-insulator interface is a continuation of the highly-regular, highly purified silicon lattice. It’s just that the channel is doped (and is therefore conductive), while the oxide layer is not (and is thus a very good insulator). The bias voltage field is propagated across the oxide layer by the ordinary process of dielectric strain, and is Fig.56: a Mosfet is similar to a JFET, but instead of using a reverse-biased PN junction to isolate the gate from the channel, it uses an extremely thin layer of semiconductor oxide; typically silicon dioxide, SiO2, basically glass – an excellent insulator. The gate’s electric field typically enhances electron/hole flow in the channel when applied; it is pinched off otherwise. These are thus known as ‘enhancement mode’ devices. Australia's electronics magazine May 2022  35 Fig.57: a dual-gate Mosfet is pretty much what you’d expect, like a regular Mosfet but with two separate gate terminals. They are useful as mixers or variable-gain amplifiers. Fig.58: a simplified model of a bipolar junction transistor (BJT) operating as a common-emitter. Note how the emitter current (Ie) is the sum of the collector current (Ic) and the base current (Ib). Here beta or hfe = 50 (50mA ÷ 1mA). Not exactly a tetrode: the dual-gate Mosfet Fig.59: we’ve removed the collector from consideration so we can examine what is happening in the base. Holes from the emitter enter the base region, but the base’s light doping means that few of them recombine with base electrons, leaving a surplus “space charge” of holes in the base. It’s this space charge that will become collector current. 36 Silicon Chip thus able to directly influence charge carriers in the channel; the interfering jumble of irregular surface states that bedevilled Shockley is absent. FETs offer transconductances in the 1000~10,000μS (microsiemens) range, roughly the same as valve tetrodes and pentodes. Despite this, FETs are rarely used in the main parts of audio or RF amplifiers, where bipolar junction transistors (BJTs) are most common. Since BJTs offer transconductances some ten times that of FETs, FETs need very high load resistances to give comparable gains. But you will find FETs of all kinds as low-noise “front ends” and in amplifiers, especially op amps. Various sub-types exist, and it’s possible to build depletion-­mode Mosfets that require bias to reduce current to give operational usefulness (like the vacuum triode), or enhancement-mode types that must have bias applied to conduct at all (just like bipolar transistors!). William Shockley’s foundation patent described the familiar ‘triode’ transistor. But he also described a multilayer device (mentioned in the first article of this series) intended for use as a mixer. So, why not a multi-gate Mosfet? The dual-gate Mosfet looks like a tetrode – one source/drain pair and one channel with two independent gates (see Fig.57). The extra gate, however, does not act as does the screen grid in a valve tetrode. It gives little if any increase in gain, and little if any reduction of output-input feedback capacitance in most circuits. The second gate’s effective transconductance is about that of the first gate. The dual-gate Mosfet can have gain control voltages applied to its second gate, and the device is often used in the famous ‘cascode’ circuit at VHF and in high-voltage wideband video amplifiers. This gives high gain with virtually no troublesome feedback, especially the Miller Effect that limits gain at higher frequencies in conventional single-stage amplifiers. The dual-gate device is also close to being an ideal mixer. The remainder of this article details operation of the ‘transistor’ as we usually think of it – the bipolar junction transistor or BJT. The BJT behaves unlike any thermionic device that came before, and is also completely unlike its later solid-state ‘cousin’, the field-effect transistor already described. The transistor Let’s consider the most common real-world BJT circuit, the common-emitter amplifier. Fig.58 shows a BJT with bias applied. It’s a PNP device (P-type emitter and collector, N-type base) like the BC107 (silicon) or OC71 (germanium). Notice that the emitter current (51mA) is the base current (1mA) plus the collector current (50mA). This gives a base-to-collector current gain of 50mA ÷ 1mA = 50. Considering the OC71, the transistor has a typical input resistance at low frequencies of around 500~5000W. Let’s say it’s 1kW. Its output resistance is much higher, but let’s say 10kW for simplicity, and let’s use quite a small input signal of just 1μA AC. A quick back-of-the envelope calculation shows this: 1μA into 1kW ohms is 1nW (10-9W). This is the signal’s input power to the transistor. Australia's electronics magazine siliconchip.com.au A current gain of 50 means the collector signal current is 1μA × 50 = 50μA. Now, 50μA in the output resistance of 10kW gives us 25mW (2.5 × 10-5W). This is the potential output power delivered to the next amplifying stage. The power gain works out to 2500 times or around +44dB. This is around the theoretical maximum for the venerable OC71, but given that the common BC107 has a current gain around 250 with an output impedance of up to 50kW, you can see that a modern transistor’s maximum power gain is quite impressive. Power gain derives from two main factors: the current gain, and the fact that the transistor’s output resistance is considerably higher than its input resistance. These combine to give high power gains. How is this possible? With sufficient negative bias on the base, electrons are attracted out (down) from the P-type emitter, liberating holes that flow upwards to cross the bulk of the emitter and enter the base-emitter junction. Arriving in the N-type base, the holes meet the resident majority electrons. This sounds like diode action, and it is. But it’s a pretty poor diode, because the base is very lightly doped, and has very few electrons compared to the flood of holes entering. Fig.59 shows the movement of holes in the emitter and their interaction with electrons within the base, but with no collector voltage. Electrons leaving the emitter connection to flow to the battery’s positive pole liberate holes in the emitter region. Electrons enter at the base from the battery’s negative pole to recombine with holes in the base region, thus forming the base current. The few electrons that do meet holes and recombine with them become the base current in the base lead. Since there are many more holes than the base electrons can recombine with, the base electrons form a positive ‘cloud’ similar to the space charge that forms around a thermionic valve’s filament, with the important difference that the valve’s space charge only ever consists of electrons. Notice, though, that the base is at pretty well the same potential throughout; there is no powerful electric field to either attract or repel the cloud of holes in the base. The holes naturally repel each other and diffuse throughout the base. This diffusion is augmented by more and more holes flooding in to the base. Some holes diffuse all the way across to the base-­ collector junction, and more particularly, to the base-­ collector depletion zone, changing the effective base width, as shown in Fig.60. With the base at about 0.3V and the collector at 10V, there is a powerful electric field across the extremely thin depletion zone – it’s probably a micrometre or less in width. As soon as holes diffuse into the depletion zone, they rapidly cross the collector’s P-type material. Reaching the collector connection, they recombine with “incoming” electrons to become collector current. Or, in point form: 1. Holes cross the emitter-base junction and enter the base according to the amount of bias applied. 2. With enough bias, holes enter the base region and combine with the resident electrons to form the base current. Since the base doping is light, there are not many electrons available to do this. siliconchip.com.au Fig.60: this plot illustrates how the effective base width is reduced at higher collector voltages, providing shorter transit times for electrons and holes. 3. Holes in the base overwhelm the few electrons, so a space charge of holes floods the base. 4. The holes, by mutual repulsion, diffuse to fill the base region. 5. Some holes diffuse all the way to the base-collector junction’s depletion zone. 6. Once holes diffuse into the depletion zone, they encounter a powerful electric field and become collector current. 7. Arriving at the collector terminal (connection), holes recombine with entering electrons which form the external collector current. The base current may be one-fiftieth, or as little as one-thousandth, of the emitter current. The collector current is almost the same as emitter current (it’s the emitter current minus the much smaller base current). Therefore, this device has high current gain. Compared to valves, BJTs have very high mutual transconductance (gm). This is the ratio of change in collector (or anode) current to the change in base (or grid) voltage that caused it, and is measured in microsiemens (or micromhos for us “oldies” – mho is ohm backwards, and this is the inverse of resistance). The iconic 6AC7 set a benchmark gm of 9000μS in valve technology (you may know this as 9mA/V). A grid voltage swing of 1V would cause the anode current to change by 9mA. The humble germanium OC70 has a gm of around 30,000μS or 30mS; a base voltage swing of only 100mV gives a collector current swing of 3mA. A silicon BC109 transistor has a gm of about 90mS or 90mA/V. Australia's electronics magazine May 2022  37 Table 1 Common emitter Common base Common collector (emitterfollower) Voltage gain High, 30~1000 High, 30~1000 Low, 0.95~0.999 Current gain High, 30~1000 Low, 0.95~0.999 High, 30~1000 Power gain Up to 1,000,000x Up to 1000x Up to 1000x Input impedance Medium, 500W~5kW Low, 10~50W High, 5kW~1MW Output High, 30kW+ High, 30kW+ impedance Feedback impedance Signal inversion Low Greatest effect Least effect Not usually considered Yes No No Properties of different transistor circuit configurations Fig.61: a bipolar transistor’s collector-emitter current flow mostly depends on the base-emitter current flow and not the collector-emitter voltage. This is a valuable property as it means they provide substantially constant collector current regardless of collector voltage. In this sense, they operate similarly to a pentode valve, not a triode. However, bipolar transistors are not commonly characterised for transconductance (although FETs often are). The most useful single parameter for a BJT is base-to-­collector current gain, written as β (beta), hfe (h parameter, Forward, common Emitter) or h21 (h parameter, output current to input current). Beta values range from around 30 (OC70) to 900 (BC109) in small-­signal transistors, and from about 150 down to only about 12 in power transistors; for example, a 2N3055 has a typical hfe of 120 at 0.3A Ic and 12 at 10A The 2SD2153 high-gain transistor has a specified hfe at 500mA of between 560 and 2700. Plotting collector current against base current (for differing collector voltages) gives the curves in Fig.61. Notice that, like the field-effect transistor, the bipolar transistor has a ‘pentode characteristic’: at any collector voltage above a few volts, collector current is pretty much independent of collector voltage. In other words, the bipolar transistor has a high output resistance. However, unlike the FET’s non-linear voltage-current characteristic, the BJT’s base current to collector current characteristic is quite linear. This means that the base-to-collector current gain (β, hfe) is pretty much the same over a range of collector currents. Outside that range, though, hfe varies considerably. It usually falls off as the collector current approaches the transistor’s maximum, and can sometimes drop off a little at very low currents, although some transistors maintain their mid-current hfe down to basically leakage current levels. Many other transistor performance parameters exist. Some of the most useful are maximum collector-emitter voltage, collector current and power (dissipation), Vce (the collector-emitter saturation voltage), the transition frequency (Ft), input resistance and capacitance, output Fig.62: three different ways to use a PNP transistor as an amplifier. Each has its advantages and disadvantages. Commonbase has the best high-frequency performance but a low input impedance and low current gain. Common-emitter has the highest power gain but suffers from feedback capacitance. Common-collector (emitter-follower) provides a high current gain but low voltage gain. 38 Silicon Chip Australia's electronics magazine siliconchip.com.au resistance and capacitance & feedback resistance and capacitance. Some of these depend on the circuit configuration. Transistor circuit configurations The first circuit generally used was common base, shown in Fig.62(a). The signal is coupled to the emitter, the output comes from the collector and the base is held at a constant bias voltage. This was, and still is, used to give maximum gain at the upper end of a transistor’s frequency capabilities, as with thermionic triodes in grounded grid configuration. Notice that the example circuit uses transformer coupling to match the transistor’s low emitter and collector impedances. Shown in Fig.62(b), common-emitter gives the highest power gain with moderately high input impedance. As with thermionic triodes in common cathode, it’s the most commonly-used configuration. Common collector, Fig.62(c) is like common anode/ cathode follower for thermionic triodes. This gives very high input impedance and very low output impedance, useful for driving low-impedance loads. Table 1 summarises the performance of small-signal transistors in these three configurations. The upside-down world Fig.62 is drawn using PNP transistors. NPN is the structure of preference for very high and ultra-high frequencies, and for high powers; electrons travel more quickly than holes throughout the transistor. The configurations for an NPN transistor would be identical but with the voltages inverted. PNP-NPN combinations (complementary designs) are common in transformerless power output stages. Mosfets and JFETs are also available in complementary designs, and are used similarly to complementary BJTs. Like BJTs, N-channel Mosfets are closer to ideal than P-channel. Using bipolar transistors for gain control Many valve radios used automatic gain control (AGC) circuitry to control RF amplifier/converter/IF amplifier stage gains, allowing full gain when needed to amplify weak signals, but reducing it to prevent overload with strong input signals. This was reverse AGC: a stronger signal would push the valve to a lower anode current, so the gain was reduced. Junction transistors had similar characteristics, as shown in Fig.63, with stage gains dropping to zero at low currents (in the μA range). Thus, designers applied reverse AGC to bipolar transistors as had been done with valves, reducing device current to reduce gain. However, modern planar transistors have much flatter hfe vs Ic curves. For example, the BF115 has an hfe of around 150 at 1mA, dropping to only about 50 at 10μA. That is not enough for any useful gain control, and some newer transistors such as the BC807 have an essentially flat hfe curve from about 20mA down into the microamps range. All transistors show some drop in gain at high collector currents, so it’s possible to reduce stage gain by pushing the collector current above the usual operating point. This is forward gain control, where a stronger signal increases the device current to reduce stage gain. Like the availability of PNP-NPN complements, it’s another fundamental difference between valves and planar transistors. So for planar transistors, gain control is usually implemented using forward gain control. In other words, the DC bias is increased until the hfe drops. Lower-current transistors or specially-designed transistors exaggerate this effect, so are very useful for gain-control applications. For example, the BF167 is specified for forward AGC. It has a transducer gain of some +28dB at low collector currents, dropping to around -32dB at a high collector current. That means that, used in a radio, they can give an AGC control range of 60dB in one stage (see Fig.64). The more traditional (reversed) method struggles to better 30dB gain control per stage. Factors limiting performance With valves, the ‘flight time’ between cathode and anode (transit time) sets an absolute limit to operating frequency. Extremes of triode valve technology, with cathode-anode spacings in the sub-millimetre range, reach their limits at about 5GHz. We would like the transistor to be perfect: a simple input resistance (rb) and an output current generator with Fig.64: planar transistors don’t suffer from such a large hfe drop at low currents as junction transistors (if at all). So forward gain control is used, reducing the stage gain by increasing the collector current. Fig.63: the dB gain of a junction transistor as a function of its DC collector current. Like valves, junction transistors can have their gain reduced by dropping the bias current below the optimum value. siliconchip.com.au Australia's electronics magazine May 2022  39 Transistor Family Tree Metal Semiconductor All-Semiconductor Point Contact Junction Diffused Micro-Alloy Grown Junction Drift-Field Base Micro-Alloy Diffused Alloyed Junction Alloy-Diffused This family tree serves to demystify the history of semiconductors and how they developed as one fabrication method superseded another. a current (β × ib) directly proportional to the input current. The output should appear as a current generator shunted by the output resistance (rc). The output resistance is very high because the transistor draws a nearly constant current regardless of the collector voltage. Similarly, its signal output current is nearly independent of the load resistance. Combining these elements, Fig.65 is pretty much the same as a simplified model of the tetrode/pentode valve. Could transistors, with their micrometre-wide base regions, also suffer from transit time effects? Yes. There are two principal effects even at relatively low frequencies. Firstly, unlike valves, transistors have no powerful accelerating field to sweep charge carriers across the entire device. Once charge carriers enter the base region and form a space charge, they only move by diffusion, a slow process. Very narrow base regions help to reduce diffusion times, yet these remain finite, rather like the transit time in a valve. Secondly, and more frustratingly, reducing diffusion times by using a very thin base gives a fairly high spreading resistance from the contact side across to the other extreme of the base. This comes from two factors: the base is very thin, and it has quite low doping compared to the emitter and collector regions. Any thin conductor will have high resistance, and a poor conductor (the result of low base doping) compounds the problem. This is rbb, the base spreading resistance. This is not simply the base lead resistance, it’s in the base itself, so it’s impossible to eliminate rbb – it can only be minimised. Unlike the valve’s grid, where one can expect any voltage Fig.65: a very straightforward model of the bipolar transistor – this is how we’d like an ideal bipolar transistor to behave, but in reality, they are not this simple. 40 Silicon Chip Multi-Diffusion Mesa Planar Epitaxial Epitaxial Mesa Planar change on the connecting terminal to appear almost instantaneously at every point across the entire grid, there can be a significant time lag across the expanse of a transistor’s base at radio frequencies. Complicating this, the considerable base-emitter capacitance must be charged and discharged by the base voltage. The base spreading resistance limits the maximum charge/ discharge rate of the base as a whole, and thus contributes to limiting high-frequency performance. We can now create a more realistic common-emitter transistor model, shown in Fig.66. The input is partly composed of the internal base-emitter resistance (rb’e), the result of bias voltage and base current. But there is also the base spreading resistance (rbb), and the base-emitter capacitance (ce). This last feature seems odd. The base-emitter is forward-­ biased and should surely appear as a resistance. Why do we appear to have a capacitance? This is due to complex hole (or electron) generation in the base and emitter areas and the hole-electron recombinations. These effects can be described mathematically, and the maths reduces to a non-resistive, reactive component: capacitance. This can be over 400pF, as the data for the OC44 germanium transistor shows. The output circuit is more like our expectations: the current generator (β × ib) is shunted by the transistor’s high output resistance (rc) and its output capacitance (cc). Finally, we must expect some collector-base feedback. This is essentially capacitive, but transit time effects change it according to frequency. For an AF118, the phase Fig.66: a more comprehensive transistor model, including the parasitic resistances and capacitances that limit their performance. Australia's electronics magazine siliconchip.com.au 99mm 193mm Fig.67: ‘pallet’ amplifiers like this are used for high-power RF transmitters. This board uses two Ampleon BLF989 900W RF Mosfets to provide 1000W of peak analog power for digital TV broadcasts at 470-705MHz (the Mosfets are rated to 860MHz). The voltage gain is 19dB and the cost is ~US$1200. Source: https://broadcastconcepts.com/180wuhf-digital-400w-analog-tv-pallet-amplifier.html angle (in common-emitter configuration) ranges from the expected 270° at 455kHz to around 210° at 100MHz. The complex nature is represented by rb’c and cb’c. We might have expected the tiny dimensions of transistor construction to free us from the tyranny of high-­ frequency limits. Alas, not so. Recent developments, however, have yielded transistors with impressively high frequency limits, in the hundreds of gigahertz; frequencies simply impossible with triode valves. High power and frequency? Returning to the field-effect transistor, its conducting channel does provide an accelerating field for charge carriers. This ‘valve-like’ characteristic allows FETs to be the device-of-choice at ultra-high and microwave frequencies. Gallium arsenide (GaAs) FETs can easily give noise figures of 0.2dB at 432MHz, a barely measurable contribution to the theoretical minimum noise figure. Power FETs are used in ‘pallet’ amplifiers of powers up to around one kilowatt (see Fig.67)! Want more power? Just put some in parallel. High-power solid-state transmitters at HF, VHF & UHF do this; a 20kW transmitter might use twenty individual 1kW pallets, paralleled and combined to deliver the final output. Transistor devices have replaced virtually all valve designs. A few niches, such as megawatt, gigahertz radars still use the powerful magnetron. But its microwave fellows, such as low-power klystrons and travelling-wave tubes, are obsolete in new equipment. That circuit symbol The semiconductor diode symbol had been in existence for some time when the transistor was invented, so it made sense to adopt it. Since the emitter admits current to the transistor, it was denoted as an arrow that indicated the direction of current flow. Engineers accepted conventional current (positive to negative) as the direction of current flow, so the emitter arrow obeys this convention. The shape of the circuit symbol represents the physical construction of a point-contact transistor – see Fig.68(a). Point-contact devices had become obsolete by the 1960s, so attempts were made to refashion the symbol to more siliconchip.com.au Fig.68: the standard PNP and NPN transistor symbols, shown at the top, are based on the physical configuration of point-contact transistors. The symbols at the bottom, designed to look more like a junction transistor, never really caught on even though they would probably make circuit drawings neater. closely represent the junction transistor, as shown in Fig.68(b). Wireless World and our own Radio, TV and Hobbies carried the charge, but the rest of the publishing world did not adopt their more rational and descriptive form. Interestingly, current mesa and planar transistor technologies have reverted to a physical structure more similar to point-contact technology. Type numbering As with valves, US manufacturers took a haphazard approach to numbering. The Joint Electron Devices Engineering Council (JEDEC) simply numbered junctions: 1N for diodes, 2N for triode transistors, 3N for the now obsolete junction tetrodes and current dual-gate Mosfets, and 4N for optocouplers. JEDEC’s 2N series were issued in order of application, with no indication of function. The 2N1066 is a germanium PNP RF type rated at 240mW, 80V and 120MHz in a four-wire TO-33 case. The 2N1067 is an NPN silicon power transistor rated at 5W, 60V and 1.5MHz in a threelead TO-8 package. Like JEDEC, the Japanese Industrial Standards Committee’s JIS numbers were simply allocated in order of registration with no indication as to application or voltage/power rating. Frequency ratings and polarity can be deduced to some extent by the prefix (see Table 2). For example, the 2SA120 is a high-frequency PNP, akin to a higher-power OC170, while the 2SD43 is a low-power NPN audio type. Australian transistors, either licence-manufactured or local types, took a bit from everywhere. We have a mess. The saying goes, “the great thing about standards is that there are so many to choose from”. Table 2 – JIS transistor code categories 2SA high-frequency PNP BJTs 2SB audio-frequency PNP BJTs 2SC high-frequency NPN BJTs 2SD audio-frequency NPN BJTs 2SJ P-channel FETs (both JFETs and Mosfets) 2SK N-channel FETs (both JFETs and Mosfets) Australia's electronics magazine May 2022  41 • AWA’s licensing from RCA produced many 2N types, plus their own AS series. • Ducon licensed from Compagnie Générale de Télégraphie Sans Fil (CSF), producing SFD diodes and SFT transistors. • Electronic Industries Ltd (EIL) owned Radio Corporation Pty Ltd, makers of Astor brand radios and TVs, and Eclipse Radio Pty Ltd, makers of Peter Pan and Monarch radios. They made semiconductors under their Anodeon brand: 2N series and their own AT and AX series. • Devices from, or licensed from, General Electric in the UK use the GET prefix. • Fairchild Australia produced 2N series devices and their own, unique, SE, AX and AY series. • Early Philips/Mullard devices followed their European parents, adopting O (for ‘no heated cathode’), using OA for diodes and OC for transistors. Like the JEDEC series, device numbers were allocated on demand, running to at least OC977 and with very little indication of device type. The OC45 is a low-performing version of the OC44 PNP germanium converter, but the OC16 is a 10W germanium power transistor. Between the OC44/45 and OC70/71 junction transistors we find the (then) obsolete OC50/51 point-contact types. The OC206 is PNP silicon with a cutoff frequency of 850kHz. • Standard Telephones and Cables released their own TS series. • As with valves, the European Electronic Component Manufacturers Association (EECA) Pro Electron system took an organised approach and provided semiconductor type and intended application via the type number. Notable Australian adopters, Philips and Mullard, deserve praise for adopting Pro Electron which aids in decoding those metal and plastic devices that populate transistor radios. The first letter shows the type of semiconductor: A for germanium, B for silicon, C for gallium arsenide (GaAs). The second letter shows device type (see Table 3), followed either by a three-digit code (such as AF118, BC107 etc), or a third letter (X, Y or Z) and a two-digit code for professional devices, such as AFY40, BUX84 and BCZ10. Pro Electron also includes diodes, with the second letter: A = signal diode, B = varicap diode, X = varactor/ step recovery diode, Y = power diode and Z = zener/ reference diode. Table 3 – Pro Electron transistor prefixes AC Germanium small-signal AF transistor AD Germanium AF power transistor AF Germanium small-signal RF transistor AL Germanium RF power transistor AS Germanium switching transistor AU Germanium power switching transistor BC Silicon small-signal transistor (‘general purpose’) BD Silicon power transistor BF Silicon RF (high-frequency) BJT or FET BS Silicon switching transistor (BJT or Mosfet) BL Silicon high-frequency, high-power (for transmitters) BU Silicon high-voltage (eg, for CRT horizontal deflection circuits) CF GaAs small-signal microwave transistor (MESFET) CL GaAs microwave power transistor (FET) This series is extracted from Chapters 1 to 4 of How Your Transistor Radio Works by Ian Batty. The remaining nine chapters cover transistor receivers – from biasing and power supplies, through converters, RF/IF amplifiers and demodulation, audio amplifiers, to detailed analysis of actual circuits, including AM/FM radios. How Your Transistor Radio Works contains 102 pages of valuable information in the one volume – you won’t find a better combination of basic theory and practical circuit description anywhere. It’s available through the HRSA’s Valve Bank at the very reasonable price of $20.00 (plus postage). Visit https://hrsa.org.au/training-manuals/ to order this and other fine HRSA books. Joining the HRSA gives you access to our Valve Bank, and you’ll get our quarterly magazine, Radio Waves with 60 pages packed full with everything from Marconi radios and restorations of Australian classics to helpful contacts around Australia. And while you’re there, consider Ian’s previous How Your Radio Works, which covers similar topics in the Valve Universe. At only $12.00 (plus postage), it’s a must have for any restorer of valve radios from TRF sets to modern SC superhets. Raspberry Pi Pico BackPack With the Raspberry Pi Pico at its core, and fitted with a 3.5inch touchscreen. It's easy-to-build and can be programmed in BASIC, C or MicroPython. There's also room to fit a real-time clock IC, making it a good general-purpose computer. This kit comes with everything needed to build a Pico BackPack module, including components for the optional microSD card, IR receiver and stereo audio output. $80 + Postage ∎ Complete Kit (SC6075) siliconchip.com.au/Shop/20/6075 The circuit and assembly instructions were published in the March 2022 issue: siliconchip.au/Article/15236 Australia's electronics magazine siliconchip.com.au