The transformer offers several advantages in coupling signals from one circuit to another. First, there is complete DC isolation. Second, impedance transformation is possible. We will also see advantages in freeing AC signals from power supply limitations. Transformer coupling is very widely used at radio frequencies, but it is also practical at audio frequencies, say 30 to 30 000 Hz. Ferromagnetic materials for the cores make transformers small and efficient, and good materials are now available for use well past 1 MHz. In this page, we shall look at small audio transformers. I managed to obtain an assortment of audio transformers from Mouser Electronics some years ago, which has proved excellent for experiments.
The most important parameter of a transformer is the turns ratio N, expressed as primary turns:secondary turns. Impedance is transformed as the square of the turns ratio. It is necessary that the reactance of a transformer winding (tested with no load) be much larger than any load or source resistance connected to it. This means that there is an indefinite lower frequency bound for the use of any transformer. Another limitation is heating, from core and I2R losses, which is seldom of importance with audio transformers.
The greatest inconvenience in the use of transformers is the effect of a DC current flowing through a winding and causing extra ampere-turns. This may cause the core to saturate, and the reactance of the windings to become small, which causes the transformer action to disappear. A transformer may be designed with this in mind, and some audio transformers will permit a limited DC current to flow in a winding. This requires a larger core and more turns, making a more expensive transformer. There is no way to tell if a transformer can take DC but to test the reactance of the transformer windings as a function of DC current.
One of the windings is usually designated the primary, if it is generally used as such, but the designation is otherwise quite arbitrary. Conditions of insulation between the windings are ample for audio use, and present no worry. This means that transformers may be used for any purpose you wish, even though the catalog may say they are for a specific use. Windings are usually center-tapped for convenience. The center tap may or may not be used. Many audio transformers have turns ratios from 6:1 to 15:1, and are intended to match a low impedance load to a collector or plate circuit. Others have turns rations close to 1:1 for coupling circuits with isolation.
A traditional low-impedance load is a moving-coil loudspeaker. We'll use a small 2" speaker in the experiments. The 2" is the diameter of the speaker cone, which is mounted elastically at its rim and at the spider of the voice coil, which moves in the strong radial magnetic field of a permanent magnet. Because the loudspeaker is not only an electrical, but also a mechanical and an acoustical device, it is rather complex in behavior. Only a few percent at most of the electrical power delivered to the voice coil appears as acoustic energy. The power rating refers to electrical power, not radiated acoustic power. The mass of the cone and the restoring force of the mounting means that there is a mechanical resonance at what is called the "bass resonant frequency" though for a 2" speaker it is not very bass. This makes a loudspeaker similar to a crystal electrically, with a parallel resonance at the bass resonance frequency and a series resonance at a slightly greater frequency. At higher frequencies, the impedance is inductive and slowly rising. For a wide interval above the bass resonant frequency, the impedance is fairly constant, and this is the frequency range in which the speaker is best used. Incidentally, the rising efficiency with frequency (caused by the increased impedance) compensated the sideband cutting of AM radio sets (caused by the finite IF bandwidth). Loudspeakers are the limiting consideration in fidelity, not the piddling irregularities of amplifiers and transformers.
A typical small loudspeaker (or earphone) has a DC resistance of 8Ω, and is called an "8-ohm" speaker, though its actual impedance is a function of frequency, rising mainly because of inductance and losses as the frequency increases. There is little difference between the impedance with the voice coil blocked and with it free to move, except near the bass resonance, where the motional impedance dominates, showing that the effect of motion is small in practical use. The working impedance is, however, quite low even for transistors, so it is usually "matched" by the use of a transformer. With transistors, rather low impedances can be driven without transformers, but with vacuum tubes, a matching transformer is necessary. Some small speakers are available with 45Ω impedance or more, to be used without transformers, and these are very useful for experiments.
A loudspeaker in the open is a very poor acoustic radiator. The air pushed out in front simply runs around behind the speaker cone, and not much else results. A loudspeaker in an enclosure designed to reduce this effect is much more satisfactory. For experiments, it is not necessary to use a speaker in an enclosure, but simply lay the speaker face down on a flat surface. The sound will come from the back of the cone, but that is perfectly all right. If you put the speaker right-side up, it will be much more feeble.
The circuit at the right shows a loudspeaker transformer coupled to a collector, at the same place a collector resistor would be expected. The transistor is biased by a voltage divider and emitter resistor feedback. The collector current is 3.2 mA, which puts the collector at 11.0 V (the transformer primary has a DC resistance of 310Ω), the base at 2.2 V, and the emitter at 1.5 V. The transformer has DC flowing in its primary, which could have a bad effect. We will find out on testing the circuit.
The unbypassed emitter resistor reduces the voltage gain, but has other beneficial effects. If you bypass it with a 100 μF electrolytic (+ lead to emitter), the gain will be much larger. In fact, it will be inconveniently large for our purposes. However, it is good to remember that the voltage gain is there, should it be needed. Apply a 1 kHz signal to the input through the coupling capacitor, and look at the collector voltage with the oscilloscope. Try to find the maximum output voltage without flattening either the tops or bottoms of the waveform. You will certainly hear the speaker! I found a maximum peak-to-peak amplitude of 16 V, while the input was 2.6 V, giving a gain G = -6.15. From the gain, we can estimate the effective impedance in the collector, using our old rule of the ratio of collector and emitter resistances. For the collector current of 3.2 mA, re = 7.8ω, so the impedance presented to the collector is 6.15 x 478 = 2940Ω. The 15:1 turns ratio multiplies impedance by 225, which for 8Ω means 1800Ω. At 1 kHz, the impedance of the speaker is apparently larger than the DC value of 8Ω, and the transformer is operating properly as well, in spite of the DC current through it. I do not know if my transformer was designed to carry DC, but the lack of center taps on its windings is a hint that it was, as well as its excellent behavior in this circuit.
You may have been surprised to find the voltage amplitude at the collector to be greater than the supply voltage of 12V. This is the result of the transformer, which carries the collector as much higher than 11 V as it does below it. The base voltage rises to 2.2 + 1.3 = 3.5 V, and the difference between this and 11 V is 7.5 V. The collector can sink only about 0.5 V farther before the transistor saturates, so this is the limit of the ouput excursion, exactly what I measured. The transformer makes the 12 V supply act more like a 24 V supply as far as output is concerned, which is a great advantage of transformer coupling.
If the matching transformer could not handle a DC current, an alternative would be the shunt-feed circuit shown at the left. This is exactly the same circuit as used above, except that the transformer is now capacitor-coupled to the collector to block any DC through it. The collector must be supplied by a resistor as in the normal resistance-load amplifier. The collector resistor is chosen to bring the collector to a voltage halfway between the supply and the base, to allow the maximum swing. In this case, the collector cannot rise higher than the supply voltage, so we already may sense that this circuit will not be as good as the series-feed circuit above. The measured gain was G = -2.0, with an input of 3 V peak-to-peak, and an output of 6 V. This makes the power ouput only about 1/7 that of the series-feed amplifier. It is challenging to try various values for the resistors in an attempt to improve this value, but one is always bumping against a ceiling (12 V) or a cellar (the base) that flattens the waveform.
An amplifier in which collector current flows for the complete range of the input is called a Class A amplifier. If the average bias current is I, and Vcc the supply voltage, then power IVcc is constantly supplied to the circuit, whether there is a signal or not. If the amplifier is operated in a linear region, the power supplied does not change when a signal is applied and useful output obtained. With a collector resistor as the load, the maximum signal voltage swing is Vcc/2, while the current swing is equal to I. Since this is sinusoidal, the signal power is IVcc/4, or 25% of the power from the power supply. The other 75% has to be dissipated in the transistor.
The series-feed amplifier shows that the signal swing can be doubled with the use of a transformer, which makes the signal power twice as large, or 50% of the total power supplied. The efficiency of an amplifier is the ratio of the useful power to the total power supplied. We have seen that the maximum efficiency of a Class A amplifier with a resistive load is 25%, and with a transformer-coupled load 50%. These calculations assumed that the collector voltage could go all the way to ground (it can only go as far as the base), so they are as optimistic as possible. In any actual case, the efficiency will be less. Now, power for electronic circuits does not cost much, so this is not the reason to look for higher efficiency. The stimulus is that the power that is not effectively used must be dissipated in the amplifier, and this requires added expense that is significant.
The way forward is to look for some way to reduce the bias current to zero. Then the amplifier would dissipate no energy just sitting waiting for a signal, and we could hope that more of the energy supplied would go into the output, not into heat and hot transistors or tubes. In a Class B amplifier, the collector current is zero for a half cycle. Conditions in the collector circuits of Class B and Class A amplifiers are shown at the right. If we use two transistors, alternately off and on, they can create a complete waveform, just as in a Class A amplifier. The beauty of this is that the bias current can be zero. Then, the average current taken from the power supply is just the average current in one loop of the signal input, 2/π times the peak value. Using the same notation as above, the power supplied is Vcc(2/π)I, while the signal amplitude is Vcc, and the signal power IVcc/2. The efficiency is then 78.5%, a nice, high value. If the signal amplitude is only Vcc/2, the efficiency is reduced to 39%. These figures are to be compared with 50% and 25%, respectively. The efficiency we have been talking about was usually called the plate efficiency, referring to the anode of the amplifying vacuum tube, analogous to the collector of a transistor.
When tuned circuits are used a the load, we can go even farther, since the tuned circuits will work quite well with properly-timed kicks. Then, collector current may flow for only a short time, further reducing the average current necessary, and increasing the efficiency of the amplifier above 78%. Such amplifiers are called Class C. They are only useful when tuned circuits are involved, and where only a sine wave is desired.
A transformer-coupled Class B push-pull amplifier is shown at the left. The input transformer T1 has a turns ratio 1:1.67, the output transformer T2 a ratio of 13:1, left side to right side. Note that the DC bias currents flow in opposite directions in the transformer windings, so they do not contribute any ampere-turns. The bias current is quite low in this amplifier anyway, only 0.4 mA. The 100Ω resistors in the emitter leads help to equalize the two transistors, but also sense the collector currents for us so we can look at them with the oscilloscope. The bases are biased with a 1N4148 diode that holds them one diode drop above ground, so the transistors are on the verge of conduction. A small resistor could be used here as well, but the diode provides some temperature compensation. Note the use of the center taps of the transformer windings. With transformer coupling at the input, the annoyance of keeping the bias divider impedance high is removed. Note that the bases receive the signal with opposite phases, just what is necessary to turn each of them on for a half cycle in alternation.
Look at the collector currents of the transistors, connecting Ch 1 at "c" and Ch 2 at "d." Compare collector currents and collector voltages at "c" and "e" or "d" and "f." Note how the two transistors supplement each other on successive half cycles, and that the output waveform is smooth. This did not just happen: it had to be arranged, and the 1N4148 diode was necessary to bias the transistors just to the point of conduction. To get a better understanding of this, replace the diode by resistors of 330, 470, 560, 620 and 680 ohms successively, and look at the collector waveform. With the smallest resistor, the signal has to assume a certain value to turn the transistor on. The result in the output waveform is called crossover distortion. With 680Ω the crossover distortion is barely perceptible. The corresponding average collector current is 0.4 mA. If the bias is increased until some considerable collector current flows in the absence of a signal, the amplifier is called Class AB, because it is between the two classes.
All the amplifiers so far mentioned have used only NPN transistors. There were similar circuits for vacuum tubes, with which transformers were almost always used. With transistors, we have another alternative in PNP transistors, for which there was no vacuum-tube analogy. It is possible to use a complementary pair of transistors, one NPN, the other PNP, to drive a low impedance load in an emitter-follower circuit that avoids the use of a transformer. Such circuits have largely displaced output transformers as loudspeaker drivers, where the demands are not large.
A complementary emitter follower driving a loudspeaker is shown at the right. The transistors are biased to just conducting by the two diodes to prevent crossover distortion. For an input of 1.4 V peak-to-peak, the output was 1.1 V p-p, which drives the speaker quite well. It is not possible to get voltage gain with this circuit, so it must be preceded by voltage amplification if required. A 100Ω resistor could be put in each emitter lead to balance the transistors. The output voltage swing is limited by the maximum current that can be supplied by the transistors. An 8Ω loudspeaker is a rather low resistance, so the output voltage cannot swing far before the base drive is insufficient. Loudspeakers are available with higher impedance, up to 90Ω, and these will work better. The ±12 V supply is really not required for the collectors--± 5 V would do quite well. If the circuit were biased for a single supply, a large capacitor would be required at the output to block the DC to the speaker. With dual supplies, this is not required, but it should be checked that the circuit is balanced for DC so no appreciable DC current passes through the speaker.
To reduce the current drive requirements of a power stage, Darlington-connected transistors can be used. For a complementary emitter follower, the complement to the Darlington is the compound PNP shown at the left. High-power design is somewhat complicated, since attention must be paid to maximum power ratings, thermal effects, and overcurrent protection. Nevertheless, the basic circuits and principles have been presented here.
Audio amplifiers are well-represented in the vacuum tube experiments in that unit.
Composed by J. B. Calvert
Created 19 August 2001
Last revised 30 September 2001