Output stages of power amplifiers using transistors. Output power amplifiers. Measuring the output transistors

25171

Comparison of the sizes of the original (large) and counterfeit (small) 2n3055 transistor crystals

Common board for two JLH2005 amplifier channels and two voltage regulator channel boards

Testing the JLH1969 amplifier from a switching power supply

Test of the JLH1969 amplifier from an analog power supply with a 120 W transformer

Selection of transistors for the JLH amplifier

Output transistors

• Old copies that were made using mesa-planar technology (2N3055), which was replaced by modern epitaxial-palar technology (MJE3055) - are very musical transistors.
• Despite the frequency response, the sound of the 2n3055 is louder and more transparent, but the sound of the 2sc3281 is more muffled and tube-like, or something. Apparently, the distribution of harmonics affects
• The best and most stable in this unit still turned out to be MJ15024, MJ15003, 2N2773. The BAT of the output stage transistors at a 4 Ohm load should be at least 120.
• Super transistors - MJ15026, 15027 for $27 one, in the States $7.

Well, the Motorola clone 2SC3281 is MJL3281A, it is generally a record holder in terms of Kus linearity. Almost a straight “shelf”, and the decline in beta starts from 5-6 Amps!!! In terms of sound, the leaders are MJL3281A (NPN) MJL1302A (PNP) as the most integrally linear powerful bipolar transistors for AF.

A very good result is obtained by parallel connection at the output of 2 3 medium power transistors 2sc5707, pre-selected by betta (they have a very high one - up to 560). We solder 2-3 transistors onto a common copper plate, and then attach it to the radiator through a gasket; it is better to solder with low-melting solder pos-61.

In plastic (TO-247) you can install MJE21193, 2CS5200, KT8101 (in order of quality deterioration); In metal (TO-3) you can use MJ15003, MJ15024, 2N3055, KT819VM, GM (in the same order); Of ours - KT908, KT903, KT808, KT805, KT803 (KT908 is head and shoulders above all others, they are the best among domestic ones).

Do not use MJL21294, these transistors are not for this amplifier. Especially with a 4 ohm load. This is where they belong in Igor Semynin’s single-cycle repeater or amplifiers with composite transistors at the output. In an amplifier according to the JLH circuit, the higher the voltage of the output transistors and the pre-output transistor, the better. MJL-21194 is now the best for sound but not for Hood, JLH can use MJ15003, but their body is uncomfortable, like 2N3055

I looked at the characteristics of the device on this set of transistors: High-frequency output 2sc5200 + driver stage at VS550bp, input transistor bc109b. The distortion turned out to be 0.02...0.03% with an excellent meander. Under the same conditions, low-frequency motorolas with a low beta give distortion of 0.08-0.1% with a strongly blocked meander front.

The output must necessarily be corrected from excitation by installing capacitors between the base and collector of the driver transistor of the order of 10-15 pF and a capacitor with a capacity of 22-60 pF in parallel with the OOS resistor R5 2.7 kOhm. If the OOS capacitor has a rating of 470-680 μF, then the OOS divider 2.7 kOhm/240 Ohm is better reduced to 1.2 kOhm/120 Ohm, which will give less distortion and greater stability.

Modern transistors are inferior to vintage ones in terms of bass reproduction quality. I find the 2SA1943, 2SC5200 to provide better sound than the MJ15003, 15004 or MJ15024, 25.

MJL21194 combine the advantages: a flat, easy-to-install case and a narrow band of 4-6.5 MHz. True, they have two “minuses” - high cost and low gain. It is not recommended to install powerful modern transistors with ft>30MHz - it will excite. Old low-frequency transistors behave better than new high-frequency transistors. In this sense, it’s worth trying our Kt805-Kt819

For transistors of the series: MJ, MJL, MJW - 21193, 21194, 21195, 21196... copper metallization is used on the surface of the crystal to form the base terminal, which equalizes the temperature of the crystal surface, improves the current distribution over the crystal area and expands the OBR, especially in the high voltage range .

Driver transistor

I tried many transistors in the driver, the 2sc2240 showed the best results, which is natural because it has 300-700 batts, with excellent collector current linearity in the range of 1.0-50 mA and a small capacitance of 3 pF, glue a copper plate to it and you get an excellent medium-power driver = Ibuki

If you have output transistors with a large betta, then the current from the driver transistor is not very large, 15-25 mA, so there is no need to put a stupid horse transistor there. Of the Soviet ones, KT602B is not bad, but it needs to be selected with beta at a current of 20-30 mA of at least 200.

The low-power pre-output transistor shows much better results in terms of meander quality and distortion than the BD139 and the same “medium-power” ones due to more linear characteristics at currents of 10-30 mA, high h21e and small interelectrode capacitances. The quality increase in the classic 1969 scheme is especially good.

The best driver cascade is: 2sc5706, 2 sc5707 with beta 300-400, worse than 2sc2120 (these need to be glued to the radiator), even worse than 2sc5171, bd139. Try it 2sc5707 for a powerful version of the amplifier, two in parallel (IMHO the best for this circuit), you just need competent installation, such as RF devices and correction. You need to assemble a JLH breadboard, leave transistor T2 without a heatsink, measure the output stage current after a while, and then heat transistor T2 with a soldering iron and measure it again.

As a driver, there is one tricky super transistor with a beta of under 1000 2sd2165.

Instead of a bipolar transistor, you can try installing a mosfet with a small input capacitance (for example, irf510) into the circuit. Now the voltage at the collector of the first transistor is less than 2 V, but with a mosfet it will be more than 5 V, which will reduce distortion. Plus - the gain of the first transistor will increase due to the higher input resistance of the mosfet, just do not forget to put a resistor with a nominal value of about 150 Ohms in the field gate

Input transistor

The input transistor must have a low reverse collector current, high betta and low noise figure, which allows it to operate at a meager collector current of 100-300 µA. In the first stage, low-power transistors with a collector capacitance of less than 30 pF and a beta of more than 250 performed well. The first transistor has a small quiescent current of 0.3 mA; there should be a transistor with a beta of 500-700 types bc560c, 2sa970.

Flipping the diagram toP-N-P

Several times, both on our forums and on foreign resources, I have come across the statement that the amplifier is based on the JLH circuit on output transistors P-N-P structures sounds much better than n-p-n. Also, some local gurus were seen offering fleeting praise pnp transistors at the exit and more. Not long ago on the forums I started asking questions about this and reached giants like A. Nikitin, Lynx and Alex. But I didn’t receive any clear answers, like “guess it yourself” or “everyone already knows this”, something like that. Foreign comrades turned out to be simpler, but they didn’t bother justifying the fact - they just took it and turned it over, and it turned out to be better and that’s all!

Many foreigners on forums report that with PNP transistors The output sound is much better. It’s quite possible to try putting almost everyone’s favorite MJ15003 NPN conductors at the output and compare them with 15024. Then reverse the power supply and put PNP MJ21193 at the output, and MAT-12 from the AD assembly at the input, half for each channel. Or carry out full-scale

There were already publications on Habré about DIY tube amplifiers, which were very interesting to read. There is no doubt that their sound is wonderful, but for everyday use it is easier to use a device with transistors. Transistors are more convenient because they do not require warming up before operation and are more durable. And not everyone will risk starting a tube saga with anode potentials of 400 V, but transistor transformers of a couple of tens of volts are much safer and simply more accessible.

As a circuit for reproduction, I chose a circuit from John Linsley Hood from 1969, taking the author’s parameters based on the impedance of my 8 Ohm speakers.

The classic circuit from a British engineer, published almost 50 years ago, is still one of the most reproducible and collects information about itself exclusively. positive reviews. There are many explanations for this:
- the minimum number of elements simplifies installation. It is also believed that than simpler design, those better sound;
- despite the fact that there are two output transistors, they do not need to be sorted into complementary pairs;
- an output of 10 Watts is sufficient for ordinary human dwellings, and an input sensitivity of 0.5-1 Volts agrees very well with the output of most sound cards or players;
- class A - it is also class A in Africa, if we are talking about good sound. Comparison with other classes will be discussed below.

Interior design

An amplifier starts with power. It is best to separate two channels for stereo using two different transformers, but I limited myself to one transformer with two secondary windings. After these windings, each channel exists on its own, so we must not forget to multiply by two everything mentioned below. On a breadboard we make bridges using Schottky diodes for the rectifier.

It is possible with ordinary diodes or even ready-made bridges, but then they need to be bypassed with capacitors, and the voltage drop across them is greater. After the bridges there are CRC filters consisting of two 33,000 uF capacitors and a 0.75 Ohm resistor between them. If you take a smaller capacitance and a resistor, the CRC filter will become cheaper and heat up less, but the ripple will increase, which is not comme il faut. These parameters, IMHO, are reasonable from a price-effect point of view. A powerful cement resistor is needed for the filter; at a quiescent current of up to 2A, it will dissipate 3 W of heat, so it is better to take it with a margin of 5-10 W. For the remaining resistors in the circuit, 2 W of power will be quite enough.

Next we move on to the amplifier board itself. Online stores sell a lot of ready-made kits, but there are no fewer complaints about the quality of Chinese components or illiterate layouts on boards. Therefore, it is better to do it yourself, at your own discretion. I made both channels on a single breadboard so that I could later attach it to the bottom of the case. Running with test elements:

Everything except the output transistors Tr1/Tr2 is on the board itself. The output transistors are mounted on radiators, more on that below. The following remarks should be made to the author’s diagram from the original article:

Not everything needs to be soldered tightly at once. It is better to first set up resistors R1, R2 and R6 as trimmers, unsolder them after all adjustments, measure their resistance and solder the final constant resistors with the same resistance. The setup comes down to the following operations. First, using R6, it is set so that the voltage between X and zero is exactly half of the voltage +V and zero. In one of the channels I didn’t have enough 100 kOhm, so it’s better to take these trimmers with a reserve. Then, using R1 and R2 (while maintaining their approximate ratio!) the quiescent current is set - set the tester to measure direct current and measure this same current at the power supply plus entry point. I had to significantly reduce the resistance of both resistors to obtain the required quiescent current. The quiescent current of an amplifier in class A is maximum and, in fact, in the absence of an input signal, all of it goes into thermal energy. For 8-ohm speakers, this current, according to the author's recommendation, should be 1.2 A at a voltage of 27 Volts, which means 32.4 Watts of heat per channel. Since setting the current can take several minutes, the output transistors must already be on cooling radiators, otherwise they will quickly overheat and die. Because they are mostly heated.

It is possible that, as an experiment, you will want to compare the sound of different transistors, so you can also leave the possibility of convenient replacement for them. I tried 2N3906, KT361 and BC557C at the input, there was a slight difference in favor of the latter. In the pre-weekend we tried KT630, BD139 and KT801, and settled on imported ones. Although all of the above transistors are very good, the difference may be rather subjective. At the output, I immediately installed 2N3055 (ST Microelectronics), since many people like them.

When adjusting and lowering the resistance of the amplifier, the low-frequency cutoff frequency may increase, so for the input capacitor it is better to use not 0.5 µF, but 1 or even 2 µF in a polymer film. There is still a Russian picture-scheme of an “Ultralinear Class A Amplifier” floating around the Internet, where this capacitor is generally proposed as 0.1 uF, which is fraught with a cutoff of all bass at 90 Hz:

They write that this circuit is not prone to self-excitation, but just in case, a Zobel circuit is placed between point X and ground: R 10 Ohm + C 0.1 μF.
- fuses, they can and should be installed both on the transformer and on the power input of the circuit.
- it would be very appropriate to use thermal paste for maximum contact between the transistor and the heatsink.

Metalworking and carpentry

Now about the traditionally most difficult part in DIY - the housing. The dimensions of the case are determined by radiators, and in class A they must be large, remember about 30 watts of heat on each side. At first, I underestimated this power and made a case with average radiators of 800 cm² per channel. However, with the quiescent current set to 1.2A, they heated up to 100°C in just 5 minutes, and it became clear that something more powerful was needed. That is, you need to either install larger radiators or use coolers. I didn’t want to make a quadcopter, so I bought giant, handsome HS 135-250 with an area of ​​2500 cm² for each transistor. As practice has shown, this measure turned out to be a little excessive, but now the amplifier can be easily touched with your hands - the temperature is only 40°C even in rest mode. Drilling holes in the radiators for mounts and transistors became a bit of a problem - the initially purchased Chinese metal drills were drilled extremely slowly, each hole would have taken at least half an hour. Cobalt drills with a sharpening angle of 135° from a well-known German manufacturer came to the rescue - each hole is passed in a few seconds!

I made the body itself from plexiglass. We immediately order cut rectangles from glaziers, make the necessary holes for fastenings in them and paint them on the reverse side with black paint.

The plexiglass painted on the reverse side looks very beautiful. Now all that remains is to assemble everything and enjoy the music... oh yes, during final assembly it is also important to properly distribute the ground to minimize the background. As was discovered decades before us, C3 must be connected to the signal ground, i.e. to the minus of the input-input, and all other minuses can be sent to the “star” near the filter capacitors. If everything is done correctly, then you won’t be able to hear any background, even if you bring your ear to the speaker at maximum volume. Another “ground” feature that is typical for sound cards that are not galvanically isolated from the computer is interference from the motherboard, which can get through USB and RCA. Judging by the Internet, the problem occurs often: sounds can be heard in the speakers HDD operation, printer, mouse and background power supply system unit. In this case, the easiest way to break the ground loop is to cover the ground connection on the amplifier plug with electrical tape. There is nothing to fear here, because... There will be a second ground loop through the computer.

I didn’t make a volume control on the amplifier, because I couldn’t get any high-quality ALPS, and I didn’t like the rustling of Chinese potentiometers. Instead, a regular 47 kOhm resistor was installed between ground and the input signal. Moreover, the regulator is on the outside sound card always at hand, and every program also has a slider. Only the vinyl player does not have a volume control, so to listen to it I attached an external potentiometer to the connecting cable.

I can guess this container in 5 seconds...

Finally, you can start listening. The sound source is Foobar2000 → ASIO → external Asus Xonar U7. Microlab Pro3 speakers. The main advantage of these speakers is a separate block of its own amplifier on the LM4766 chip, which can be immediately removed somewhere away. An amplifier from a Panasonic mini-system with a proud Hi-Fi inscription or an amplifier from the Soviet Vega-109 player sounded much more interesting with this acoustics. Both of the above devices operate in class AB. JLH, presented in the article, beat all the above-mentioned comrades by one wicket, according to the results of a blind test for 3 people. Although the difference was audible to the naked ear and without any tests, the sound was clearly more detailed and transparent. It's quite easy, for example, to hear the difference between MP3 256kbps and FLAC. I used to think that the lossless effect was more like a placebo, but now my opinion has changed. Likewise, it has become much more pleasant to listen to files uncompressed from loudness war - dynamic range less than 5 dB is not ice at all. Linsley-Hood is worth the investment of time and money, because a similar brand amp will cost much more.

Material costs

Transformer 2200 rub.
Output transistors (6 pcs. with a reserve) 900 rub.
Filter capacitors (4 pcs) 2700 rub.
“Rassypukha” (resistors, small capacitors and transistors, diodes) ~ 2000 rub.
Radiators 1800 rub.
Plexiglas 650 rub.
Paint 250 rub.
Connectors 600 rub.
Boards, wires, silver solder, etc. ~1000 rub.
TOTAL ~12100 rub.

Output stages based on "twos"

As a signal source we will use an alternating current generator with a tunable output resistance (from 100 Ohms to 10.1 kOhms) in steps of 2 kOhms (Fig. 3). Thus, when testing the VC at the maximum output resistance of the generator (10.1 kOhm), we will to some extent bring the operating mode of the tested VC closer to a circuit with an open feedback loop, and in another (100 Ohm) - to a circuit with a closed feedback loop.

The main types of composite bipolar transistors (BTs) are shown in Fig. 4. Most often in VC, a composite Darlington transistor is used (Fig. 4a) based on two transistors of the same conductivity (Darlington “double”), less often - a composite Szyklai transistor (Fig. 4b) of two transistors of different conductivity with a current negative OS, and even less often - a composite Bryston transistor (Bryston, Fig. 4 c).
The "diamond" transistor, a type of Sziklai compound transistor, is shown in Fig. 4 g. Unlike the Szyklai transistor, in this transistor, thanks to the “current mirror”, the collector current of both transistors VT 2 and VT 3 is almost the same. Sometimes the Shiklai transistor is used with a transmission coefficient greater than 1 (Fig. 4 d). In this case, K P =1+ R 2/ R 1. Similar circuits can be obtained using field-effect transistors (FETs).

1.1. Output stages based on "twos". "Deuka" is a push-pull output stage with transistors connected according to a Darlington, Szyklai circuit or a combination of them (quasi-complementary stage, Bryston, etc.). A typical push-pull output stage based on a Darlington deuce is shown in Fig. 5. If emitter resistors R3, R4 (Fig. 10) of input transistors VT 1, VT 2 are connected to opposite power buses, then these transistors will operate without current cut-off, i.e. in class A mode.

Let's see what pairing the output transistors will give for the two "Darlingt she" (Fig. 13).

In Fig. Figure 15 shows a VK circuit used in one of the professional and onal amplifiers.

The Siklai scheme is less popular in VK (Fig. 18). At the early stages of the development of circuit design for transistor UMZCHs, quasi-complementary output stages were popular, when the upper arm was performed according to the Darlington circuit, and the lower one according to the Sziklai circuit. However, in the original version, the input impedance of the VC arms is asymmetrical, which leads to additional distortion. A modified version of such a VC with a Baxandall diode, which uses the base-emitter junction of the VT 3 transistor, is shown in Fig. 20.

In addition to the considered “twos,” there is a modification of the Bryston VC, in which the input transistors control transistors of one conductivity with the emitter current, and the collector current controls transistors of a different conductivity (Fig. 22). A similar cascade can be implemented on field-effect transistors, for example, Lateral MOSFET (Fig. 24).

The hybrid output stage according to the Sziklai circuit with field-effect transistors as outputs is shown in Fig. 28. Let's consider the circuit of a parallel amplifier using field-effect transistors (Fig. 30).

As effective way To increase and stabilize the input resistance of the “two,” it is proposed to use a buffer at its input, for example, an emitter follower with a current generator in the emitter circuit (Fig. 32).

Of the “twos” considered, the worst in terms of phase deviation and bandwidth was the Szyklai VK. Let's see what using a buffer can do for such a cascade. If instead of one buffer you use two on transistors of different conductivities connected in parallel (Fig. 35), then you can expect further improvement in parameters and an increase in input resistance. Of all the considered two-stage circuits, the Szyklai circuit with field-effect transistors showed itself to be the best in terms of nonlinear distortions. Let's see what installing a parallel buffer at its input will do (Fig. 37).

The parameters of the studied output stages are summarized in Table. 1 .

Analysis of the table allows us to draw the following conclusions:
- any VC from the “twos” on the BT as a UN load is poorly suited for working in a high-fidelity UMZCH;
- the characteristics of a VC with a DC at the output depend little on the resistance of the signal source;
- a buffer stage at the input of any of the “twos” on the BT increases the input impedance, reduces the inductive component of the output, expands the bandwidth and makes the parameters independent of the output impedance of the signal source;
- VK Siklai with a DC output and a parallel buffer at the input (Fig. 37) has the highest characteristics (minimum distortion, maximum bandwidth, zero phase deviation in the audio range).

Output stages based on "triples"

In high-quality UMZCHs, three-stage structures are more often used: Darlington triplets, Shiklai with Darlington output transistors, Shiklai with Bryston output transistors and other combinations. One of the most popular output stages at present is a VC based on a composite Darlington transistor of three transistors (Fig. 39). In Fig. Figure 41 shows a VC with cascade branching: the input repeaters simultaneously operate on two stages, which, in turn, also operate on two stages each, and the third stage is connected to the common output. As a result, quad transistors operate at the output of such a VC.

The VC circuit, in which composite Darlington transistors are used as output transistors, is shown in Fig. 43. The parameters of the VC in Fig. 43 can be significantly improved if you include at its input a parallel buffer cascade that has proven itself well with “twos” (Fig. 44).

Variant of VK Siklai according to the diagram in Fig. 4 g using composite Bryston transistors is shown in Fig. 46. In Fig. Figure 48 shows a variant of the VK on Sziklai transistors (Fig. 4e) with a transmission coefficient of about 5, in which the input transistors operate in class A (thermostat circuits are not shown).

In Fig. Figure 51 shows the VC according to the structure of the previous circuit with only a unit transmission coefficient. The review will be incomplete if we do not dwell on the output stage circuit with Hawksford nonlinearity correction, shown in Fig. 53. Transistors VT 5 and VT 6 are composite Darlington transistors.

Let's replace the output transistors with field-effect transistors of the Lateral type (Fig. 57

Anti-saturation circuits of output transistors contribute to increasing the reliability of amplifiers by eliminating through currents, which are especially dangerous when clipping high-frequency signals. Variants of such solutions are shown in Fig. 58. Through the upper diodes, excess base current is discharged into the collector of the transistor when approaching the saturation voltage. The saturation voltage of power transistors is usually in the range of 0.5...1.5 V, which approximately coincides with the voltage drop across the base-emitter junction. In the first option (Fig. 58 a), due to the additional diode in the base circuit, the emitter-collector voltage does not reach the saturation voltage by approximately 0.6 V (voltage drop across the diode). The second circuit (Fig. 58b) requires the selection of resistors R 1 and R 2. The lower diodes in the circuits are designed to quickly turn off the transistors during pulse signals. Similar solutions are used in power switches.

Often, to improve the quality, UMZCHs are equipped with separate power supply, increased by 10...15 V for the input stage and voltage amplifier and decreased for the output stage. In this case, in order to avoid failure of the output transistors and reduce the overload of the pre-output transistors, it is necessary to use protective diodes. Let's consider this option using the example of modification of the circuit in Fig. 39. If the input voltage increases above the supply voltage of the output transistors, additional diodes VD 1, VD 2 open (Fig. 59), and the excess base current of transistors VT 1, VT 2 is dumped onto the power buses of the final transistors. In this case, the input voltage is not allowed to increase above the supply levels for the output stage of the VC and the collector current of transistors VT 1, VT 2 is reduced.

Bias circuits

Previously, for the purpose of simplicity, instead of a bias circuit in the UMZCH, a separate voltage source was used. Many of the considered circuits, in particular, output stages with a parallel follower at the input, do not require bias circuits, which is their additional advantage. Now let's look at typical displacement schemes, which are shown in Fig. 60, 61.

Stable current generators. Modern UMZCHs widely use a number of standard schemes: differential cascade (DC), current reflector ("current mirror"), level shift circuit, cascode (with serial and parallel power supply, the latter is also called a "broken cascode"), stable current generator (GCT), etc. correct use can significantly increase specifications UMZCH. We will estimate the parameters of the main GTS circuits (Fig. 62 - 6 6) using modeling. We will assume that the GTS is a load of the UN and is connected in parallel with the VC. We study its properties using a technique similar to the study of VC.

Current reflectors

The considered GTS circuits are a variant of a dynamic load for a single-cycle UN. In an UMZCH with one differential cascade (DC), to organize a counter dynamic load in the UN, they use the structure of a “current mirror” or, as it is also called, a “current reflector” (OT). This structure of the UMZCH was characteristic of the amplifiers of Holton, Hafler, and others. The main circuits of the current reflectors are shown in Fig. 67. They can be either with a unity transmission coefficient (more precisely, close to 1), or with a greater or lesser unit (scale current reflectors). In a voltage amplifier, the OT current is in the range of 3...20 mA: Therefore, we will test all OTs at a current of, for example, about 10 mA according to the diagram in Fig. 68.

The test results are given in table. 3.

As an example of a real amplifier, the S. BOCK power amplifier circuit, published in the journal Radiomir, 201 1, No. 1, p. 5 - 7; No. 2, p. 5 - 7 Radiotechnika No. 11, 12/06

The author's goal was to build a power amplifier suitable for both sounding "space" during festive events and for discos. Of course, I wanted it to fit in a relatively small-sized case and be easily transported. Another requirement for it is the easy availability of components. In an effort to achieve Hi-Fi quality, I chose a complementary-symmetrical output stage circuit. The maximum output power of the amplifier was set at 300 W (into a 4 ohm load). With this power, the output voltage is approximately 35 V. Therefore, the UMZCH requires a bipolar supply voltage within 2x60 V. The amplifier circuit is shown in Fig. 1 . The UMZCH has an asymmetrical input. The input stage is formed by two differential amplifiers.

A. PETROV, Radiomir, 201 1, No. 4 - 12

Class A amplifier.

Operates in linear mode: both transistors operate in the same modes. This providesminimum distortion , but as a result of this low efficiency (15-30%), i.e. This class is uneconomical in terms of energy consumption and heating. Power consumption does not depend on the output power.

Class B amplifier

This class mainly includes amplifiers with output transistors of the same conductivity. Each of the transistors operates in key mode, i.e. amplifies only its half-wave signal in linear mode (for example, positive if transistors with N-P-N conductivity are used). In order to amplify the negative half-wave of the signal, a phase inverter is used on another transistor. It's like two separate A classes (one for each half-wave). An amplifier of this class has a high efficiency (about 70%). The power consumption of the amplifier is proportional to the output power; in the absence of a signal at the input, it is zero. Amplifiers of this class are rare among modern amplifiers.

Class AB amplifier

The most common type of amplifier. This class combines the qualities of Class A and Class B amplifiers, i.e. high efficiency class B and low level nonlinear distortion class A. A cutoff angle of more than 90 degrees is used here, i.e. the operating point is selected at the beginning of the linear section of the current-voltage characteristic. Due to this, in the absence of a signal at the inputthe amplifying elements are not switched off and some current flows through them (the so-called "quiescent current") , sometimes significant. And here there is a need to regulate and stabilize this current so that the transistors operate in the same modes without overloading each other. Incorrect installation quiescent current will lead to overheating of the transistors and their failure.

So: for the output stage there are two very important parameters (and especially for class AB):

quiescent current and quiescent voltage

If the transistors had an ideal characteristic (which in fact does not happen), then the quiescent current could be considered equal to zero. In reality, the collector current can increase both due to the scatter in the characteristics of the transistors and their temperature. Moreover: an increase in temperature can lead to avalanche-like overheating and thermal breakdown of the transistor. The fact is that as the temperature increases, the collector current only increases, and therefore the heating of the transistor increases.

resting voltage: constant voltage at the point of connection of the transistors (output to the load). It should be equal to "0" when the output stage is supplied bipolarly, or half the supply voltage when it is unipolarly supplied. In other words: both transistors of the output stage must have the same base bias, that is, they are open evenly, compensating each other.

These two parameters must be stabilized, and first of all, their temperature dependence must be eliminated.

For this purpose, amplifiers use an additional transistor, connected in a ballast manner to the base circuits of the output transistors (and most often it is placed directly on the radiator next to the output transistors, thereby controlling their temperature).