Push-pull cascade

The simplest push-pull cascades

Push Pull Emitter Repeater

CMOS Inverter

The simplest linear push-pull cascade - a complementary emitter follower in mode B - is formed by the counter-inclusion of two emitter followers on transistors npn- (upper arm) and pnp-structure (lower arm) ^[2] . At zero control voltage, both transistors are closed, the load current is zero ^[3] . When the threshold for turning on the transistor is exceeded, approximately +0.5 V, the top transistor (npn) smoothly opens, connecting the positive power bus to the load. With a further increase in the control voltage, the output voltage repeats the input voltage with a shift of 0.5 ... 0.8 V, the lower transistor remains closed. Similarly, at negative control voltages, the lower (pnp) transistor opens, connecting the load to the negative power bus, and the upper one remains closed ^[3] . In the field of small control voltages, when both transistors are closed, characteristic switching distortions are observed waveform in the form of a step ^[4] .

Similarly, the simplest key push-pull cascade — the CMOS logic inverter — works differently. Field-effect transistors of the inverter are turned on not in the repeater mode, but in the mode with a common source - therefore, they amplify and invert the input voltage ^[5] . The upper transistor of the p-type conductivity opens by a low logic level and transfers a high logic level to the output, the lower transistor opens by a high logic level and transfers a low level to the output, switching the load to the lower power bus ^{[6] [7]} . The switching thresholds of transistors are selected so that in the middle of the interval between high and low input levels, both transistors are guaranteed to be open - this speeds up switching at the cost of insignificant power losses during short-term passage of through current ^[6] . In stable states of logical zero and logical unit, only one of the two transistors is open, and the other is closed ^[7] . The typical load of a logic element is the gates of other logic elements, so its transistors transmit current to the load only when switching. As the load capacities recharge, the output current decays to zero, but one of the two transistors remains open ^[6] .

Alternative Definitions

Push-pull cascades can be performed according to other schemes, amplify direct or alternating voltage or current, operate on an active or reactive load , they can be inverting or non-inverting. Common to all configurations is the principle of antiphase : with increasing control voltage, the current increases only in one of the two arms of the circuit; when the control voltage drops, the current rises in the other, opposite arm ^[1] . The behavior of the circuit in static mode, in general, is not defined - only its reaction to a change in the input signal is important ^[1] . In certain branches of electronics and in historical, outdated literature, narrower particular definitions may also be found:

• A push-pull amplifier ( push-pull amplifier ) is an amplifier in which the input signals controlling transistors are out of phase and the output signals are added up, which allows you to double the output power compared to a single-cycle amplifier (USA, 2013) ^[8]
• A push-pull circuit ( push-pull circuit ) is a symmetrical circuit in which two active devices operate alternately, each in its own half of the input signal period, and together control the transfer of current to the total load. Push-pull switching reduces the level of even harmonics, but increases the level of odd harmonics (USA, 2011) ^[9] .
• A push-pull circuit is a circuit consisting of two identical [active] circuits connected in such a way that currents flow in them that are identical in magnitude but opposite in phase (USSR, 1960) ^[10] .
• A push-pull amplifier is a power amplifier in radio transmitting and receiving devices, containing two electronic lamps or two groups of lamps in one cascade, working together for a common load. The voltages on the grids of these lamps act in antiphase to each other. At the output load, the powers given by the lamps add up (USSR, 1952) ^[11] .
• A push-pull amplifier is a power amplifier in radio transmitting and receiving devices, consisting of two electronic lamps (or two groups of lamps) working together for a common load, for which the voltage on the [control] grids is out of phase (USSR, 1955) ^[12] .

Cascade concept

In lamp circuitry, the concept of an output stage literally corresponds to the concept of an “amplification cascade” (“amplification stage, a radio engineering device containing an amplifying element, a load circuit, a communication circuit with the previous or subsequent cascades” ^[13] ). In this interpretation, the only active device works in each arm of the push-pull output stage. It can be a single lamp or a group of lamps connected in parallel ^[11] , but as a rule, there was no talk of successive switching on of lamps inside the cascade. A similar approach is used in transistor circuitry of radio frequency power amplifiers.

In transistor circuitry of audio frequency power amplifiers, by contrast, simple cascades are rare. Two-transistor bipolar output stages are operable only in relatively low-current devices, and in order to match the intermediate amplification stages with a low-impedance load, it is necessary to switch on at least two current amplification stages in series. In practice, each arm of a push-pull output stage can have from two to four “stages in a stage”. The transistors that make up these twos, triples and quadruples are covered by local feedbacks , and are usually considered in combination. The simplest cases of such complexes are Darlington pairs and Shiklai pairs . In addition to them, in practice, at least seven ^[14] bipolar “triples” are used (“triple” Quad 303, “triple” Bryston and so on), four-stage emitter repeaters and “four” Bryston ^[15] , which are protected against current overload or power by additional active circuits. These schemes as a whole are called output stages, and their internal parts, if it makes sense to distinguish them at all, are considered as steps of the output stage.

There are three basic schemes ( topologies ) that allow you to implement an output push-pull cascade. All three topologies are variants of a half-bridge circuit for connecting the load to two active devices and one or two power sources ^[16] . Symmetric and asymmetric (quasi-complementary) switching can be implemented on all types of active devices, complementary - only on pairs of transistors with opposite (complementary) types of conductivity.

Symmetric inclusion

In a symmetric circuit, two identical active devices are connected parallel to each other by direct current: the total quiescent current consumed by the cascade at zero input signal is divided into two equal parts flowing through the left and right arms of the amplifier ^[17] . The voltage of the amplified signal is supplied to the control electrode of the inverting (left in the circuit) arm, and its mirror copy formed by an external phase splitter is fed to the input of the inverting (right in the circuit) arm ^[17] . With a positive signal voltage, the inverting arm current increases, the non-inverting arm current decreases. In order to transfer these current changes to the load, active devices are included in the lower arms of the H-shaped bridge circuit, and the currents of the upper arms of the bridge are fixed in one way or another. The difference between the currents of the upper and lower shoulders of the bridge is closed through the load included by the "crossbar" of the bridge.

The role of the upper shoulders of the H-shaped bridge can be, for example, inductors , the total resistance of which in the entire operating frequency range is significantly higher than the load resistance, and the DC resistance is relatively small. It is even more convenient to use a transformer with a tap from the midpoint of the primary winding ^[18] . Transformer coupling allows matching relatively large internal resistances of real lamps and transistors with low resistances of real loads - loudspeakers , electric motors , antennas , cable lines ^[17] , but its main task is switching out-of-phase output currents into a common load ^[18] . It was the transformer circuit developed by RCA in 1923 ^[19] that was the main one in the tube circuitry, and the “symmetrical switching” was actually a synonym for the push-pull cascade ^[17] . The first transistor amplifiers were built according to this scheme, and transistor amplifiers of radio frequencies of especially high power continue to be built ^{[20] [18]} . Other advantages of the transformer circuit are a high efficiency and a high level of output power in mode B, symmetrical reproduction of positive and negative input voltages, suppression of odd harmonics, a simple device of a unipolar power source, relative insensitivity to the spread of the rest currents of two arms ^{[20] [18] [17]} . Disadvantages - limited bandwidth and phase distortion of real transformers, limiting the possibility of using feedback , and the fundamental impossibility of transmitting direct current to the load ^{[20] [18]} .

The symmetric push-pull cascade is similar to the differential voltage amplification cascade , which is also a variant of the parallel half-bridge circuit ^[21] . The total current of the two arms of the differential cascade is limited by a stable current source in the common circuit of emitters, sources, or cathodes, which eliminates the possibility of power amplification in the economy mode B.

Asymmetric (quasicomplete) inclusion

An alternative to a symmetrical bridge is a bridge in which identical active devices are included in the upper left and lower left shoulders, and power sources in the right shoulders. The total quiescent current flows through both active devices, that is, the active devices are connected in constant current in series ^[22] . The upper lamp (transistor) is connected to the load by the cathode (emitter, source) according to the cathode ( emitter , source) repeater circuit of the input signal. The lower lamp (transistor) is connected to the load by the anode (collector, source) and operates in an inverting amplifier with a common cathode ( with a common emitter , with a common source) ^[23] . The internal resistances and amplification factors of the lamps (transistors) in these modes are fundamentally different, therefore such a bridge is called asymmetric. The selection of the preliminary amplification coefficients of the input signals arriving at the upper and lower arms of the output stage compensates for this asymmetry only partially: in real amplifiers, deep negative feedback is required. The circuit is sensitive to the spread of the quiescent currents of two arms, and the arrangement of bias circuits specifying these currents is relatively complicated. In tube amplifiers, the problem is aggravated by the limitation of the maximum permissible voltage of the heater-cathode; therefore, asymmetrical switching did not take root in the tube circuitry ^{[20] [24]} .

In the circuitry of transistor power amplifiers of the 1960s, on the contrary, the asymmetric Lin amplifier circuit ^{[20] [25]} dominated. On the one hand, it made it possible to abandon transformer coupling, replacing it with either capacitive coupling or direct connection to the load; on the other hand, in the 1950s, industry produced powerful transistors of only pnp structure ^[26] . In the mid-1960s, they were replaced by more powerful and more reliable silicon transistors, but npn structures, and only in the late 1960s the US industry mastered the production of pnp transistors complementary to them ^{[20] [26]} . By the end of the 1970s, designers of linear UMZCH on discrete transistors switched to a complementary circuit ^[27] , and a quasi-complementary circuit is still used in the output stages of integrated power amplifiers ( TDA7294 , LM3886 and their numerous functional analogs) and in class D amplifiers ^{[28 ]} .

Complementary inclusion

Replacing one of the active devices of an asymmetric circuit with a device of a type complementary to it turns the circuit into a complementary one. If the selected types of output transistors (“complementary lamps” do not exist ^[29] ) have the same dynamic characteristics in the entire range of operating currents, voltages, and frequencies, then such a circuit reproduces positive and negative input voltages symmetrically (in real amplifiers, asymmetry is inevitable, especially on the upper limit of the frequency range of the output transistors). The input phase splitter is no longer needed: the same alternating voltage of the signal (usually with some constant voltage shift, which sets the operation mode of the output transistors) is applied to the bases or gates of both arms ^{[30] [31]} .

Bipolar transistors of a complementary circuit can operate in any of three basic modes ( OK , OE or OB ) ^{[30] [31]} . In power amplifiers operating on a low-impedance load, bipolar transistors are usually turned on according to a common collector circuit (a complementary emitter follower , shown in the illustration), field-effect transistors according to a common drain circuit (source follower) ^[32] . Such a cascade amplifies current and power, but not voltage. The inclusion of transistors according to the scheme with a common emitter or a common source is also common - this is exactly how the CMOS logic amplifier buffers are arranged. В этом варианте комплементарный каскад усиливает и ток, и напряжение, и мощность ^[31] . В выходных каскадах операционных усилителей применяются оба варианта: повторители обеспечивают лучшее быстродействие, а схемы в режиме с общим эмиттером — наибольший размах выходного напряжения ^{[33] [34]} .

Efficiency and power consumption

The maximum theoretical efficiency (Efficiency) of a single-cycle harmonic signal amplifier in mode A , achievable only with transformer coupling with a purely active load, is 50% ^[35] . Efficiency of about 30% is achieved in real single-ended transistor amplifiers, about 20% in tube amplifiers — that is, for each watt of maximum output power, the amplifier consumes 3 ... 5 W from the source ^[36] . The actual value of the power transmitted to the load practically does not affect the power consumption: the latter begins to increase only when the cascade is overloaded ^[2] . In bezformatornye amplifiers, the efficiency is noticeably worse; in the worst case of an ordinary emitter follower with an active load, the maximum theoretical efficiency is only 6.25% ^[37] .

Replacing a single-cycle repeater with a two-cycle repeater in mode A, operating at the same quiescent current and consuming the same, approximately constant power from the power source, increases the maximum output power by four times, and the maximum efficiency to 50% ^[38] . Putting the push-pull repeater into mode B increases the ultimate theoretical efficiency to 87.5% ^{[39] [40]} . The maximum output power in mode B is limited only by the safe operation area of ​​transistors, supply voltage, and load resistance ^[2] . The power consumed by the cascade in mode B is directly proportional to the output voltage ^[41] . Theoretical efficiency of 87.5% is achieved at maximum power output; with its decrease, the efficiency gradually decreases, and the relative power losses on transistors gradually increase ^[41] . Absolute losses of power dissipated by transistors also increase and reach a flat maximum in the region of intermediate powers, when the peak value of the output voltage is about 0.4 ... 0.8 of the maximum possible ^{[41] [42]} .

In real amplifiers, the qualitative nature of the dependence is preserved, but the proportion of losses increases, and the efficiency values ​​decrease. So, the output stage of a low-frequency amplifier , designed for an output power of 100 W at a load of 8 W, dissipates about 40 W at a maximum power (efficiency of about 70%). When the output power is halved to 50 W, the power loss on the transistors increases to the same 50 W (efficiency 50%) ^[43] . A significant decrease in absolute power losses is observed only when the output power decreases below 10 W ^[43] .

Spectral composition of nonlinear distortion

A feature of all push-pull circuits is a reduced specific gravity of even harmonics in the nonlinear distortion spectrum ^[44] . In distortions generated by single transistors or vacuum triodes in quasilinear mode ^{[comm. 1]} , up to the transition to the overload mode, the second harmonic dominates ^[46] . When pushing two lamps or transistors, the second, fourth and so on harmonics generated by them mutually cancel each other ^{[44] [47]} . In ideally symmetric cascades, even harmonics are completely suppressed, distortions in the shape of the negative and positive half-waves of the signal are strictly symmetrical, and the distortion spectrum consists exclusively of odd harmonics ^[44] . In real push-pull cascades, complete symmetry cannot be achieved; therefore, even harmonics are also observed in the distortion spectra ^[44] . The distribution of harmonics can depend on the signal level and its frequency - for example, due to the difference in the boundary frequencies of the pnp and npn transistors of a complementary pair ^[48] .

The predominance of odd harmonics indicates the dependence of the cascade transfer coefficient on the amplitude of the input signal: at large amplitudes, the transmission coefficient noticeably deviates from the calculated one ^[49] . As the input signal grows, the gain can initially increase, but on large signals it inevitably decreases. The decline (compression) of the coefficient by the set value, for example, by 1 dB , serves as a criterion for the overload of the cascade ^[50] .

Switching distortion

Push-pull circuits operating in modes B and AB ^{[comm. 2]} , generate specific nonlinear switching (or combinational ^[4] ) distortions when the signal passes through zero ^[4] . In the region of low output voltages, when one transistor is disconnected from the load and the other connected to it, the linear transfer characteristic of the cascade takes the form of a broken line with two bends or fractures. In the worst case, when two transistors or two lamps ^[57] operate with zero quiescent currents, both transistors turn off in the vicinity of zero, the transmission coefficient drops to zero, and a “step” is observed on the output waveform. Negative feedback cannot effectively suppress such distortions, since in the problem area the amplifier is actually disconnected from the load ^[40] .

Switching distortion is especially undesirable when amplifying sound frequencies. The threshold of noticeability of switching distortions, expressed by the standard method for measuring the coefficient of nonlinear distortion, is only 0.0005% (5 ppm ) ^[58] . Hearing sensitivity is caused by both a special, unnatural spectrum of switching distortions and an unnatural dependence of their level on power or subjectively perceived loudness: when the output power decreases, the non-linear distortion coefficient does not decrease, but increases ^[42] .

The only way to exclude the generation of switching distortions is to transfer the cascade to pure mode A, which is usually impossible in practice ^{[59] [60]} . However, switching distortions can be significantly reduced by setting only a small constant dc quiescent current of the output stage ^[60] . The magnitude of this current should exclude the simultaneous disconnection of transistors from the load, while the area in which both transistors are connected to the load should be as narrow as possible. In practice, designers set the quiescent currents of bipolar transistors at a level of 10 to 40 mA for each device; the optimal currents of MOS transistors are noticeably higher, from 20 to 100 mA per device ^[57] . The expediency of further increasing the quiescent currents, expanding the range of the regime A, depends on the selected cascade topology ^[57] . It can be justified in cascades on bipolar transistors with a common emitter ^[57] . In push-pull emitter repeaters, on the contrary, it should be avoided: an increase in the quiescent current does not reduce, but exacerbates switching distortions ^[57] .

• Danilov, A. A. Precision low-frequency amplifiers. - M .: Hot line-Telecom, 2004 .-- 352 p. - ISBN 5935171341 .
• Stepanenko I.P. Fundamentals of the theory of transistors and transistor circuits. - 4th edition, revised and supplemented. - M .: Energy , 1977 .-- 672 p.
• Titz U., Schenk K. Semiconductor circuitry. Volume I. - 12th ed .. - M .: DMK-Press, 2008 .-- 832 p. - ISBN 5940741487 .
• Titz U., Schenk K. Semiconductor circuitry. Volume II - 12th ed .. - M .: DMK-Press, 2008 .-- 942 p. - ISBN 5940741487 .
• Tsykin, G.S. Electronic amplifiers. - 2nd ed. - M .: Svyazizdat, 1963 .-- 512 p. - 21,000 copies.
• Shkritek P. Reference manual on sound circuitry. - World, 1991 .-- ISBN 5030016031 .
• Cordell, B. Designing Audio Power Amplifiers. - McGraw-Hill, 2011 .-- ISBN 9780071640244 .
• Hood, JL Valve and Transistor Audio Amplifiers. - Newnes, 2006 .-- ISBN 0750633565 .
• Duncan B. High Performance Audio Power Amplifiers. - Newnes, 1996 .-- ISBN 9780750626293 .
• Self D. Audio Power Amplifier Design Handbook. - 3rd ed .. - Newnes, 2002 .-- ISBN 0750656360 .