|
|
| Elliott Sound Products | Amplifier Basics - How Amps Work (Part 2) |
 Main Index
Articles Index
|
Since it was invented, the transistor (from 'transfer resistor') has come a long way. Early transistors were made from germanium, which was 'doped' with other materials to give the desired properties required for a semiconductor. In the beginnings of the transistor era, nearly all devices were PNP (Positive Negative Positive), and it was very difficult to make the opposite (NPN) polarity. The NPN transistors that were available at that time were low power and did not work as well as their PNP counterparts.
When silicon was first used, the opposite was the case, and for quite some time the only really high power devices available were all of silicon NPN construction. More recently, it has become possible to make NPN and PNP transistors that are almost identical in performance. Germanium is rarely used any more, although some examples are still available.
All transistors have three 'elements':
A transistor can be represented as two diodes, with a junction in the middle. This is shown for both polarities in Figure 2.1. This is only an analogy, and connecting two discrete diodes in this manner will not produce a transistor, because the point where they meet must be a common junction on the same piece of silicon (or germanium) - hence (in part) the term Bipolar Junction Transistor. The 'bipolar' term means that transistors use 'charge carriers' of both polarities - positive and negative, or minority and majority.
Since the base to collector junction is reverse biased in normal operation, there will be no current flow. It is the action of injecting current into the base that causes current flow in the collector circuit. I do not intend to explain the exact conduction mechanism, as it is somewhat outside the scope of this article.
This is very convenient, because it gives us an easy way to check if a transistor is likely to be good or bad, simply by measuring the 'diodes'. Early PNP germanium devices would actually work equally well if the emitter and collector were reversed, but devices are now optimised to maximum performance, so this trick is not as successful (it does still work, but the device gain is much lower when the terminals are reversed).
To make the transistor actually do something useful, it is necessary to bias it correctly. This is done (having selected a suitable collector resistance) simply by applying enough base current to ensure that the collector is at 1/2 the supply voltage. In the same way that the plate load resistor determines the output impedance of a valve amplifier, the collector resistor determines the output impedance of a transistor amplifier. Unlike a valve, the transistor is not said to have a 'collector resistance' as in the equivalent resistance between emitter and collector, because this is not relevant to the operation of a transistor.
Figure 2.2 shows three methods of biasing a transistor, wired in 'common emitter' configuration ¹. Of these Figure 2.2a is the least usable, because there is no mechanism to ensure that the circuit will be repeatable with different devices or with temperature. Variations caused by temperature are (and always have been) a real problem, and it is necessary always to ensure that the circuit has some feedback mechanism for DC operating conditions to ensure stability. Different transistors (of the same type and even from the same manufacturing run) will have different gains, and this, too, must be compensated for.
For the three circuits below, assume that the gain of the transistor is 100 (exactly). This means that for 1mA of base current, 100mA of collector current will flow. The emitter current is the sum of the base and collector currents. To bias the transistor we need only meet this criterion (in theory), and everything will be well. With a Supply voltage (Vcc) of 20V, we want to have 10V at the collector, to allow maximum voltage swing. This will allow the voltage to go to +20V or to 0V, however the signal will be badly distorted by then.
Figure 2.2A is unusable in practice, even though it appears to satisfy the criteria for correct operation. Figure 2.2B is a simple way to achieve (acceptably) stable bias, but has some drawbacks. Because the bias resistor (Rb) is supplied from the collector circuit, it will have some of the collector current flowing in it. This will introduce negative current feedback, which at DC stabilises the circuit, but with the AC signal makes the input impedance very low, as well as reducing gain for any finite value of source impedance. This is not necessarily a drawback, however, as the feedback also reduces the distortion components.
This problem is overcome with the circuit in Figure 2.2C, with a bias divider providing a fixed voltage reference, and the emitter resistor (Re) providing stabilising feedback as we saw with the valve voltage amplifier. In the same way as with a valve, this also provides feedback, increasing linearity and reducing gain. With a transistor we get one additional effect - the input impedance is increased (more on this subject later). Again, to achieve maximum gain, it is common to place a capacitor in parallel with Re to defeat the feedback for AC signals, allowing maximum gain.
To bias a PNP device, we use exactly the same circuitry, but the supply polarity is reversed, so the collector (and base) will have a negative voltage with respect to the emitter.
One of the major differences between valves and transistors is that once we have decided on a suitable biasing circuit (or specified a gain from the amplifier), we can make device substitutions with little or no change in performance, provided the transistors have similar basic parameters. Often the same circuit will work just as well with perhaps 10 or 20 different devices, all from different manufacturers.
I shall only discuss the basic characteristics of transistors (as with valves), and there is really only one variable parameter and two fixed parameters (which are the same for every silicon transistor) to deal with. With transistors, the parameters are not as interactive as with valves, and the circuit gain is not as reliant on the device gain as with valves. In the same way as with valves, there are small signal devices (low power), working all the way up to power devices, which can have collector current ratings of 50 to over 100A for some of the very large power transistors.
As stated before, the gain of a transistor is dependent on collector current, but will normally be applicable over a fairly wide range. The gain normally falls at very low currents (compared to the device maximum), and again at high current (approaching the maximum rated collector current for a given device).
The signal transfer curve is similar to that of a valve, and is shown in Figure 2.3. There is generally less distortion in the linear part of the curve, but because of the lower operating voltage, a transistor amp must work closer to the supply rail and earth, so distortion may be higher with simple circuits such as those in Figure 2.2 than with an equivalent valve amplifier.
The major cause of distortion in small signal transistor amplifiers is the variation in the internal emitter resistance (re). Because transistors can tolerate a wider range of supply voltage and operating current than valves, it was common (when transistors were new and frightfully expensive) to try to extract as much voltage gain as possible from each device. This is no longer an issue, but the underlying problem is still there and it is necessary to take steps to prevent distortion. It's common to operate transistors at a constant current to minimise distortion. Very high gain circuits with global feedback are now the most common with transistor circuits, which renders the circuit immune from almost any variation of the device parameters, whether intrinsic (internally fixed) or manufacture dependent.
The gain of a transistor stage is approximately equal to the collector resistance divided by the emitter resistance (including the internal resistance re). So for the circuit of Figure 2.2c, the gain will be 9.75 without the emitter bypass capacitor, or about 384 with it installed. The distortion will be much higher with the emitter bypass in place, and it is uncommon to see these circuits any more.
The input impedance of a transistor voltage amplifier is low, and the output impedance is determined by the collector resistance (ignoring any feedback that may be applied from collector to base).
The input impedance is essentially determined by the gain of the device, and the value of emitter resistance (including the internal resistance), and in theory (that word again) is approximately equal to the emitter resistance multiplied by gain. The circuit of Fig 2.2a will therefore have an input impedance in the order of 2600 Ohms, Fig 2.2b will be very low because of the feedback, and 2.2c (without the bypass capacitor) will have an input impedance of 100k - but as this is shunted by the bias resistors, the impedance will actually only be about 12k.
A transistor can be thought of as a current controlled current source (CCCS).
The current amplifier is much more common in transistor circuits than with valves, and is called an emitter follower (or occasionally common-collector). The emitter follower (like the cathode follower) has a voltage gain of less than 1 (or unity), but the difference is much less. Typically, the gain of an emitter follower circuit will be about 0.95 to 0.99 - depending on the operating current. The use of feedback to lower this further is very common, and output impedances of less than 1 Ohm are quite possible.
Figure 2.4 shows a standard configuration for an emitter follower current amplifier stage. It is common to bias the base to exactly 1/2 the supply voltage, using equal value resistors. I say 'a' standard because there are many different configurations that can be (and are) used, including direct coupling, which is very common with transistor circuits.
One of the great attractions of transistors is their flexibility, which is considerably enhanced by having two polarities of device to work with. Because of this, circuits such as that shown in Figure 2.5 are common (or they were before the advent of opamps). Indeed, opamps themselves use the flexibility of transistors to the full, as can be seen if you have a look at the 'simplified equivalent circuit' often published as part of the specification sheet for many opamps.
The common base amplifier is something that you rarely see these days. It was also used in valve circuits and was sometimes called a 'grounded grid' amplifier. Input impedance is very low, and the circuit shown has an input impedance of around 50 ohms. It has high gain, and can be used at radio frequencies because there is almost no collector-base feedback (or plate-grid feedback) due to stray (or internal) capacitance. In early designs common base stages were sometimes used for low impedance microphone preamps, or for other low-Z applications. The input capacitor (Cin) needs to be large to pass audio frequencies, due to the very low input impedance. The base capacitor (Cb) connects the base to ground for all AC signals.
As shown, the circuit will have a gain of around 70 times (35dB), but that depends on the source impedance (50 ohms is assumed). It's an interesting circuit overall, but cannot compete with an opamp 'virtual earth' stage, which has an input impedance of close to zero. The common base arrangement was also used in 'cascode' amplifiers, as were common grid valve circuits - indeed, that's where the circuit came from. Cascode designs were mainly used where high gain at radio frequencies was necessary, but have re-emerged in valve audio gear because they (allegedly) sound 'better' than other circuits.
The vast majority of circuit 'blocks' used today are combinations of stages. A combined voltage and current amplifier are very common, and these can be found in IC equivalent circuits, as well as many of the older designs that were in general use before opamps took over for the majority of circuitry.
As can be seen, this amplifier uses an emitter follower for the output, is direct coupled within the circuit itself, uses both NPN and PNP devices, and has feedback to set a gain which is dependent only on the ratio of the two resistors Rfb1 and Rfb2. It is this sort of circuit that the opamp came from in the beginning, and there are still ICs (and small power amplifiers) that use similar circuitry internally. Regular readers may even recognise the basic circuit from the Projects Pages - essentially this is a discrete opamp, and will have a very high gain, which is brought back to something sensible by the feedback.
The actual gain is almost entirely dependent on the resistor values (for gains less than about 50 or so), and may be calculated by
Av = (Rfb1 + Rfb2) / Rfb2 where Av is voltage gain (Amplification, voltage)
So to obtain a gain of 20, Rfb1 would be 22k, and Rfb2 1k2 - this is actually a gain of 19.33, representing an error of 0.3dB. This gain is so stable that a completely different set of transistors from a different manufacturer would make no difference to measured gain performance. Other factors, such as noise or distortion must vary with the quality of the active devices, but the changes will generally be very subtle, and may not be noticeable at all, depending on the similarity of the transistors.
A transistor power amplifier uses (typically) another configuration for the input stage. This is called a 'long tailed pair', (LTP) and acts as both the input stage and error amplifier (Q1 and Q2). This circuit operates in current mode, so there is little output voltage to be seen from its output.
The second stage (Q3) is a Class-A amplifier, and is responsible for a large proportion of the overall gain of the circuit. Notice the current sources that are typically used for the LTP and Class-A amp sections. These are commonly made using transistors and maintain a constant current regardless of the voltage at the collector. If the current were truly constant, this implies that the impedance is infinite (which means that the gain of the transistor stage is also infinite!), and although this is not the case in reality, it will still be remarkably high.
For more information on how current sources are constructed, see Section 5.1
The output stage (Q4 and Q5) typically is a pair of complementary emitter followers, which must be correctly biased to ensure that as the signal passes from one transistor to another, there is no discontinuity. This form of operation is known as Class-AB, since the amp operates in Class-A for very low level signals, then changes to Class-B at higher levels. Any discontinuity while passing the signal from one transistor to the other is the cause of crossover distortion, and for many years gave transistor amplifiers a bad name in the audio world. With proper biasing, and properly applied feedback, the crossover distortion can be made to go away - although never completely, but amplifiers with distortion levels of well below 0.01% are common.
The resistors at the emitters of the output transistors help to maintain a stable bias, and also introduce some local feedback to linearise the output stage. This is a simplified circuit, and in reality the output stage will usually consist of multiple transistors, commonly a driver transistor followed by the output transistor itself. This does not change the operation of the circuit, but simply gives the output stage more gain, so it does not load the Class-A driver too heavily (this will result in greatly increased distortion).
Like the previous example, the gain is entirely dependent on the ratio of Rfb1 and Rfb2. As shown, the amp in Figure 2.6 is DC coupled, meaning that it will amplify any voltage from DC up to its maximum bandwidth. Not shown on this circuit are the various components needed to stabilise the circuit to prevent oscillation at high frequencies - often in the MHz range. Such oscillation is a disaster for the sound, and will quickly overheat and destroy the output transistors.
There are also transistor amplifiers that operate in Class-A, which means that the output transistors conduct all the time, and are never turned off. This can produce distortion levels that are almost impossible to measure, but this is at the expense of efficiency, and Class-A amplifiers will get very hot while doing nothing. Unlike the more common Class-AB amplifier, they will actually get slightly cooler as they reproduce a signal, since some of the input power is then diverted to the loudspeaker.
Just as with valve amplifiers, I have only scratched the surface. Entire books are written on the subject, and range from basic texts used in technical schools, to very advanced tomes intended for university students. Since transistors are easy to work with (and safe), there is much to be gained by experimentation, and you will have the satisfaction of having designed and built a functioning amplifier.
Transistors also have their fair share of problems, and there are some things that they are just not very good at. Some of the major failings include:
Again, there are many advantages as well. Transistor amplifiers are very reliable, and can be counted on to give many years of life without requiring even a basic service ( most of the time anyway).
They are also very quiet (generally much quieter than valve amps) and do not suffer from microphony, so room vibrations are not re-introduced into the music. Efficiency is much higher, with lower voltages and no heaters (its a pity they don't look really nice, though).
Output impedances of 0.01 Ohm are achievable, so loudspeaker damping can be very high. Because transistor amps are very mechanically rugged, they can be installed in speaker boxes, so speaker lead lengths can be very short.
Typical transistor amplifiers have much wider bandwidth than valve amps, because there is no transformer, this is especially noticeable at the lowest frequencies - a transistor amp can reproduce 5Hz as easily as 500Hz.
Previous (Part 1 - Valves) Next (Part 3 - FETs)
| Main Index
Articles Index
|