Nikolay Troshin

A simple germanium power amplifier.

Recently, there has been a noticeable increase in interest in power amplifiers based on germanium transistors. There is an opinion that the sound of such amplifiers is softer, reminiscent of “tube sound”.
I bring to your attention two simple circuits of low-frequency power amplifiers using germanium transistors, which I tested some time ago.

More modern circuit solutions are used here than those used in the 70s, when “germanium” was in use. This made it possible to obtain decent power with good sound quality.
The circuit in the figure below is a reworked version of the low-frequency amplifier for “germanium” from my article in Radio magazine No. 8, 1989 (pp. 51-55).

The output power of this amplifier is 30 W with a speaker load impedance of 4 ohms, and approximately 18 W with a load impedance of 8 ohms.
The amplifier supply voltage (U supply) is bipolar ±25 V;

A few words about the details:

When assembling an amplifier, it is advisable to use mica capacitors as constant capacitors (in addition to electrolytic ones). For example, the CSR type, such as below in the figure.

MP40A transistors can be replaced with MP21, MP25, MP26 transistors. Transistors GT402G - on GT402V; GT404G - to GT404V;
The GT806 output transistors can be assigned any letter indices. I do not recommend using lower-frequency transistors such as P210, P216, P217 in this circuit, since at frequencies above 10 kHz they work rather poorly here (distortion is noticeable), apparently due to a lack of current amplification at high frequencies.

The area of ​​radiators for output transistors must be at least 200 cm2, for pre-terminal transistors - at least 10 cm2.
For transistors of the GT402 type, it is convenient to make radiators from a copper (brass) or aluminum plate, 0.5 mm thick, 44x26.5 mm in size.

The plate is cut along the lines, then this workpiece is shaped into a tube, using for this purpose any suitable cylindrical mandrel (for example, a drill).
After this, the workpiece (1) is tightly placed on the transistor body (2) and pressed with a spring ring (3), having previously bent the side mounting ears.

The ring is made of steel wire with a diameter of 0.5-1.0 mm. Instead of a ring, you can use a copper wire bandage.
Now all that remains is to bend the side ears from below to attach the radiator to the transistor body and bend the cut feathers to the desired angle.

A similar radiator can also be made from a copper tube with a diameter of 8 mm. Cut a piece of 6...7 cm, cut the tube along the entire length on one side. Next, we cut the tube into 4 parts half the length and bend these parts in the form of petals and place them tightly on the transistor.

Since the diameter of the transistor body is about 8.2 mm, due to the slot along the entire length of the tube, it will fit tightly onto the transistor and will be held on its body due to its springy properties.
Resistors in the emitters of the output stage are either wirewound with a power of 5 W, or type MLT-2 3 Ohm, 3 pieces in parallel. I do not recommend using imported films - they burn out instantly and imperceptibly, which leads to the failure of several transistors at once.

Setting:

Setting up an amplifier correctly assembled from serviceable elements comes down to setting the quiescent current of the output stage to 100 mA using a trimming resistor (it is convenient to control the 1 Ohm emitter resistor - voltage 100 mV).
It is advisable to glue or press the VD1 diode to the heatsink of the output transistor, which promotes better thermal stabilization. However, if this is not done, the quiescent current of the output stage from cold 100mA to hot 300mA changes, in general, not catastrophically.

Important: Before turning on for the first time, you must set the trimming resistor to zero resistance.
After tuning, it is advisable to remove the trimming resistor from the circuit, measure its real resistance and replace it with a constant one.

The most scarce part for assembling an amplifier according to the above diagram is the GT806 output germanium transistors. Even in the bright Soviet times it was not so easy to acquire them, and now it is probably even more difficult. It is much easier to find germanium transistors of types P213-P217, P210.
If for some reason you cannot purchase GT806 transistors, then we offer you another amplifier circuit, where you can use the aforementioned P213-P217, P210 as output transistors.

This scheme is a modernization of the first scheme. The output power of this amplifier is 50W into a 4-ohm load and 30W into an 8-ohm load.
The supply voltage of this amplifier (U supply) is also bipolar and is ±27 V;
Operating frequency range 20Hz…20kHz:

What changes have been made to this scheme;
Added two current sources to the “voltage amplifier” and another stage to the “current amplifier”.
The use of another amplification stage on fairly high-frequency P605 transistors made it possible to somewhat unload the GT402-GT404 transistors and boost the very slow P210.

It turned out pretty good. With an input signal of 20 kHz, and with an output power of 50 W, distortion at the load is practically not noticeable (on the oscilloscope screen).
Minimal, barely noticeable distortions of the output signal shape with P210 type transistors occur only at frequencies of about 20 kHz at a power of 50 watts. At frequencies below 20 kHz and powers below 50 W, distortion is not noticeable.
In a real music signal, such powers at such high frequencies usually do not exist, so I did not notice any differences in the sound (by ear) of an amplifier with GT806 transistors and P210 transistors.
However, with transistors like GT806, if you look at it with an oscilloscope, the amplifier still works better.

With an 8 Ohm load in this amplifier, it is also possible to use output transistors P216...P217, and even P213...P215. In the latter case, the amplifier supply voltage will need to be reduced to ±23V. The output power will, of course, also drop.
Increasing the power supply leads to an increase in output power, and I think that the amplifier circuit in the second option has such potential (reserve), however, I did not tempt fate with experiments.

The following radiators are required for this amplifier - for output transistors with a dissipation area of ​​at least 300 cm2, for pre-output P605 - at least 30 cm2, and even for GT402, GT404 (with a load resistance of 4 Ohms) are also needed.
For transistors GT402-404, you can do it easier;
Take copper wire (without insulation) with a diameter of 0.5-0.8, wind the wire turn to turn on a round mandrel (4-6 mm in diameter), bend the resulting winding into a ring (with an internal diameter less than the diameter of the transistor body), connect the ends by soldering and put the resulting “donut” on the transistor body.

It will be more efficient to wind the wire not on a round, but on a rectangular mandrel, since this increases the area of ​​contact of the wire with the transistor body and, accordingly, increases the efficiency of heat removal.
Also, to increase the efficiency of heat removal for the entire amplifier, you can reduce the area of ​​the radiators and use a 12V cooler from the computer for cooling, powering it with a voltage of 7...8V.

Transistors P605 can be replaced with P601...P609.
The setup of the second amplifier is similar to that described for the first circuit.
A few words about acoustic systems. It is clear that to obtain good sound they must have the appropriate power. It is also advisable, using a sound generator, to go through the entire frequency range at different powers. The sound should be clear, without wheezing or rattling. Especially, as my experience has shown, this is especially true for the high-frequency speakers of speakers like S-90.

If anyone has any questions about the design and assembly of amplifiers, ask, I will try to answer if possible.

Good luck to all of you in your creativity and all the best!

Having become fed up with designs based on lamps and modern components, lately, in a nostalgic impulse, I have been toying with designs based on germanium transistors.

Having read on forums that, supposedly, due to imperfect production technology, their parameters degrade greatly over time, to check my reserves, I even purchased an L2-54 industrial meter for the parameters of transistors and low-power diodes.

I tested more than a hundred different copies of transistors and I can note with satisfaction that not a single one was rejected - all correspond to the reference data with at least one and a half times (and most often with 2-3 times) margin. So it’s not at all a sin to employ them, especially since in my youth many of them were as desirable as they were unavailable.

And we start traditionally - with ULF construction.

A number of popular amateur radio receivers to this day, for example, are made on germanium transistors and are designed to work with high-impedance headphones, which are now in short supply. Simple emitter followers recommended there for increasing output power are capable of providing more or less decent sound only to connected low-impedance headphones (100-600 Ohms) or a low-impedance load (4-16 Ohms modern headphones or speaker), connected through a transformer with a KTP of at least 1 /5 (1/25 in resistance) and still, at low levels, step-type distortion has a strong effect. You can, of course, try to install modern ULFs on ICs there, but they require positive power supply. We can go even further and transfer the designs to modern transistors, but... the “zest”, the taste of time - “nostalgia” is lost, so this is not our way.

A power amplifier with deep feedback (Fig. 1 circled in blue), connected instead of high-impedance headphones, will help to significantly improve the sound quality for a low-impedance load and ensure loud-speaking reception.

As you can see, his scheme is almost a classic of the 60-70s. A distinctive feature is the deep (more than 32 dB) feedback on direct and alternating current (through resistor R7), which ensures high linearity of amplification (at average levels of Kg less than 0.5%, at low (less than 5 mW) and maximum power (0 .5 W) Kg reaches 2%). The somewhat unusual activation of the volume control ensures an increase in the depth of feedback when the volume is reduced, thanks to this it turned out to be possible to make the ULF more economical (the quiescent current of the entire ULF PPP is no more than 7 mA) with virtually no “step” distortion. Capacitor C6 limits the passband to approximately 3.5 kHz (without it it exceeds 40 kHz!), which also reduces the level of self-noise - the ULF is very quiet. The output noise level is approximately 1.2 mV! (with the left pin C1 grounded). The total Kus from the input (from the left pin C1) is approximately 8 thousand. the level of self-noise referred to the input is approximately 0.15 µV. When connected to a real signal source (LPF), due to the current component, the level of intrinsic noise referred to the input increases to 0.3-0.4 µV.

The output stage uses inexpensive and reliable GT403. The ULF is capable of delivering high power (up to 2.5 W at a 4 Ohm load), but then you will need to install transistors on radiators and/or use a more powerful one (P213, P214, etc.), but in my opinion look, 0.5 W and modern sensitive dynamics are enough “for the eyes” even when listening to music. Almost any germanium low-frequency transistors of the corresponding structure and at least 40 N21e transistors (T2, T3, T4 - MP13-16, MP39-42, and T5 - MP9-11, MP35-38) are suitable for a low-frequency amplifier. If you plan to use this ULF in PPP, then T1 needs to be low-noise (P27A, P28, MP39B). For the output stage, it is advisable to select pairs T4, T5 and T6, T7 with close (no worse than +-10%) H21e values.

Due to the deep DC OOS, ULF modes are set automatically. When you turn it on for the first time, check the quiescent current (5-7 mA) and, if necessary, achieve the required one by selecting a more successful diode. You can simplify this procedure if you use a Chinese multimeter. In diode testing mode, it passes a current of approximately 1 mA through the diode. We need a specimen with a voltage drop of about 310-320 mV.

For testing a powerful ULF was chosen diagram of a simple dual-band PPP RA3AAE. I’ve been wanting to try it for a long time, but somehow I never got around to it, but here’s the opportunity (hi!).

I immediately made minor adjustments to the circuit (see Fig. 3), which I will describe here. Everything else, incl. and the setup process, see the book.

As a two-stage low-pass filter, I have traditionally used a universal tape head, which ensured increased selectivity over the adjacent channel. The low-pass filter coil has a fairly large own capacity, so it significantly loads the GPA, especially if it is wound not with PELSHO, but with a simple wire such as PEV, PEL (including tape recorder GUs). In this case, the coil’s own capacitance is so large that it is very problematic to run a GPA with a normal amplitude on diodes - many colleagues have encountered this. That is why it is better to remove the VFO signal not from the output of the coil, but from the communication coil, which eliminates all these problems and at the same time completely eliminates the contact of the VFO voltage with the ULF input. In order not to bother with winding, I found suitable ready-made coils and went ahead to testing the PPP and unexpectedly came across a serious “rake” - when switching to the 40m range, the amplitude of the VFO signal on the communication coil decreases by 2 times! Okay, I thought, maybe I have grenades, that is, coils, of the wrong system (hi!). I found the frames and rewound them strictly according to the author (see photo)

and here we must pay tribute to Vladimir Timofeevich - without additional movements he immediately fell into the indicated frequency ranges - both the input circuits and the GPA.

But... the problem remains, which means that it is impossible to optimally configure the mixer on both ranges - if you set the optimal amplitude on one, then on the other the diodes will either be closed or almost constantly open. Only a certain average, compromise option for setting the amplitude of the VFO is possible, when the mixer will more or less work on both ranges, but with increased losses (up to 6-10 dB). The solution to the problem turned out to be straightforward - use a free switching group in the toggle switch to switch the emitter resistor, which will be used to set the optimal amplitude of the GPA on each range. To control and adjust the optimal amplitude of the GPA, we use the same method as in.

To do this, switch the left (see Fig. 3) output of diode D1 to the auxiliary capacitor 0C1. The result is a classic GPA voltage doubling rectifier. This kind of “built-in RF voltmeter” gives us the opportunity to actually directly measure the operating modes of specific diodes from a specific GPA directly in a working circuit. By connecting a multimeter to 0C1 for monitoring in the DC voltage measurement mode, selecting emitter resistors (starting with R3 on the 40m range, then R5 on the 80m range) we achieve a voltage of +0.8...+1 V - this will be the optimal voltage for diodes 1N4148, KD522, 521, etc. Here's the whole setup. We solder the diode lead back into place and remove the auxiliary circuit. Now, with an optimal operating mixer, you can optimize (increase) its connection to the input circuit (the tap is made not from 5, but from 10 turns of L2), thereby increasing the sensitivity by 6-10 dB on both ranges.

Large voltage ripples are possible along the power circuit of a powerful push-pull ULF, especially when powered by batteries. Therefore, to power the GPA, an economical parametric voltage stabilizer was used on T4, where the reverse-biased emitter junction KT315 (which was on hand) was used as a zener diode. The output voltage of the stabilizer is selected on the order of -6..-6.5V, which ensures a stable tuning frequency when the battery is discharged up to 7V. Due to the reduced supply voltage of the GPA, the number of turns of the L3 communication coil is increased to 8 turns. But with KT315 the spread in the breakdown voltage of the emitter junction is quite large - the first one that came across gave 7.5V - a bit too much, the second one gave 7V (see graphs from)

– that’s already good, using silicon KT209v as T4 I got the required -6.3V. If you don’t want to bother with selection, you can use KT316 as T5, then T4 should be germanium (MP39-42). Then it makes sense for unification and to install KT316 in the GPA (see Fig. 4), which will have a positive effect on the stability of the GPA frequency. This is exactly the option that works for me now.

“It’s been a while since I picked up checkers...” Or rather, I wanted to say that I haven’t assembled transistor amplifiers for a long time. All lamps, yes lamps, you know. And then, thanks to our friendly team and participation, I purchased a couple of boards for assembly. Payments are separate.

The payments arrived quickly. Igor (Datagor) promptly sent documentation with a diagram, description of the assembly and configuration of the amplifier. The kit is good for everyone, the scheme is classic, tried and tested. But I was overcome by greed. 4.5 watts per channel will not be enough. I want at least 10 W, and not because I listen to music loudly (with my acoustics sensitivity of 90 dB and 2 W is enough), but... so that it is.

Power amplifier circuit

This is what my final amplifier circuit looks like. Changed denominations are indicated in red.

Not a single lawyer has yet been able to circumvent the Ohm and Joule-Lenz laws, and in order to increase the output power of the UMZCH, it is necessary to increase its supply voltage. Let's do it at least twice, up to 30 Volts. You won't be able to do this right away. Transistors P416 and MP39B, which are used in the original circuit, have a maximum permissible voltage of 15 Volts.

I had to take the old 1978 edition of the Amateur Radio Handbook off the shelf and delve into the study of the parameters of germanium transistors of the MP and GT series, while simultaneously excavating the boxes with parts.

I was looking for transistors with parameters close to those used in the circuit, but with a maximum permissible voltage of at least 30 Volts.

After carrying out this exciting research work, the necessary candidates were found. For the input, instead of P416, the main contender was the GT321D transistor.
It was decided to replace the MP39B + MP37A pair with a similar MP14A + MP10B pair. Germanium transistors of the MP series with numbers from 9 to 16 are “military equipment”, transistors for special-purpose equipment. In contrast to their analogues with numbers from 35 to 42, which are intended for equipment of wide application.

At the output, I decided to use high-frequency GT906A transistors. There were several reasons for this, the main one being the presence of a supply of these transistors in my bedside table. The second reason is the high current transfer coefficient. During operation, the transistors of the preliminary stage will “strain” less to drive the output transistors, which should reduce their heating and have a positive effect on the level of distortion of the amplifier.

The next step, which is also important, is the selection of transistors in pairs according to the current transfer coefficient h21e. At first I tried to do this using a regular Chinese tester, but the measurement results seemed somewhat strange and clearly overestimated. In addition, the Chinese tester clearly could not cope with measuring the parameters of powerful transistors.

I had to take out the good old Soviet-era PPT device from the shelf.

With its help, a pair of GT321D transistors with h21e = 120 and two pairs MP10B + MP14A with h21e about 40 were selected. From a dozen 1T906A transistors, we managed to select 3 pcs. with beta 76 and a couple with beta 78. Still, the 1T series went through a more serious selection for parameters during manufacturing.

After selecting the transistors, assembling the printed circuit boards according to the Datagor instructions did not take much time. We also need to pay attention to the voltage of electrolytic capacitors. It must be no less than the selected amplifier supply voltage.
I used 35 volt capacitors.

Since I planned to get more power from the amplifier, it was necessary to increase the capacity of the output coupling capacitor by at least two times. A capacitor of this rating could no longer fit on the board. Instead, I soldered a couple of screw terminals so that I could connect any capacitor I liked on the wires, regardless of its size.

Another important problem was the organization of cooling of the output transistors. I found a pair of identical, rather large radiators, but they were designed to accommodate modern transistors in the TO-220 housing.
I found a solution in old burnt computer power supplies. A pair of radiators made of thick 4 mm aluminum, onto which I attached GT906 transistors through insulating gaskets, and these radiators themselves, with a wide end, were screwed through thermal paste to large radiators.

The amplifier boards were also attached to the same radiators using metal corners. Between the fins of the computer heatsink, near the output transistors, a D310 diode is conveniently placed, which ensures the thermal stability of the amplifier. Without hesitation, I filled it with Chinese hot melt glue.

First turn on, setting up the amplifier

It's time to turn on and test the assembled amplifiers for the first time. I did this using a laboratory power supply with current limitation.

At first I set it up at a supply voltage of 15 Volts. I set the quiescent current of the amplifier to 100 mA, balanced the output so that it had exactly half the supply voltage, then gradually began to raise the supply voltage to the required 30 Volts.

During this operation, it was necessary to slightly change the values ​​of some resistors, because... As the supply voltage increased, the quiescent current began to increase sharply. Without a current-limiting power supply, I would probably have lost more than one pair of output transistors. But everything worked out fine.

Some measurements

After setting the DC modes, I connected a generator and an oscilloscope to the amplifier. He gave a signal. At the output, signal limitation (blue color) occurs at an amplitude of approximately 12 Volts at a 4-ohm load, and this corresponds to output power 18 W. Hooray!!! :yahoo:
The signal amplitude at the input (yellow) is approximately 1.5 Volts. That is, the amplifier has a sensitivity of about 1 Volt RMS.

Frequency band I was also pleased. Almost no rollover from 15 Hz to 60 kHz. If we removed the 100 pF capacitors from the feedback circuit and at the input, it would probably be even wider.

Just what you need! This exactly corresponds to the output signal level of the computer sound card, which will be used as the main signal source.

I checked what maximum current the amplifier consumes. When a rectangular signal with a frequency of 10 kHz and an amplitude of 1.5 V is applied to the input, the amplifier draws slightly less than 2 A of current from the power supply.

Now it's time for the crash test. I install 1.5 A fuses in the holders, set the maximum possible current limit on the power supply (I have 5 A) and apply a sine wave with a frequency of 10 kHz to the input. I turn the power up to maximum when the signal begins to limit. After this, I use a screwdriver to make a short circuit in the load. The fuse burns out. I replace the fuse with a new one, turn on the amplifier again - the output transistors are intact! After I blew three fuses (two on one amplifier board and one on the other), I decided that the reliability test had been passed and I could now proceed to the final assembly of the amplifier into the case.

General amplifier assembly

I do preliminary fittings and begin metalwork to secure all the parts in the body.

The power transformer is toroidal. With the terrible name BY5.702.010-02, which was intended to confuse a potential enemy. The transformer produces 20 volts at the output. I couldn’t find the current parameters of this winding, but it holds the heat of the GM-70 lamp (which is 3.5 A) without straining or overheating. So to power two channels of this amplifier, it has enough power even with a reserve.

I also used germanium D305 rectifier diodes (10 A, 50 V). Thus, it was possible to assemble an amplifier in which there is not a single silicon part. Everything is according to Feng Shui.

Filter capacitors - 2 pcs. 10000 µF each. One would have been enough, but, as I wrote at the beginning, greed took over, and besides, there was room in the building.

I installed three 1000 μF 63 V capacitors connected in parallel to the output. The capacitors are high-quality, from the Japanese Matsushita.

After all the components are securely fastened in the case, all that remains is to connect them together with wires, without getting anything mixed up. I did the installation using a copper core with a cross section of 0.5 sq mm in silicone heat-resistant insulation. I took this wire from the cable used for fire alarms. I recommend it for use. Due to the fact that the wire is rigid, it can be laid evenly and neatly in the housing without much effort.

At the end of the century before last, the German chemist K.A. Winkler discovered an element whose existence had been predicted in advance by D.I. Mendeleev. And on July 1, 1948, a short article appeared in the basement of the New York Times newspaper under the heading “The Making of the Transistor.” It reported the invention of “an electronic device capable of replacing conventional vacuum tubes in radio engineering.”

Of course, the first transistors were germanium, and it was this element that made a real revolution in radio engineering. Let's not argue whether music connoisseurs benefited from the transition from tubes to transistors - these discussions have already become rather boring. Let's better ask ourselves another, no less pressing question: did the next round of evolution benefit sound, when silicon devices replaced germanium ones? The last century was short-lived, and they did not leave behind, like lamps, a tangible sound heritage. Now germanium transistors are not produced in any country, and they are rarely remembered. But in vain. I believe that any silicon transistor, be it bipolar or field-effect, high-frequency or low-frequency, small-signal or high-power, is less suitable for high-quality sound reproduction than germanium. First, let's look at the physical properties of both elements.*

* Published by H. J. Fisher, Transistortechnik fur Den Funkamateur. Translation by A.V. Bezrukova, M., MRB, 1966.

Properties	Germanium	Silicon
Density, g/cm 3	5,323	2,330
Atomic weight	72,60	28,08
Number of atoms in 1 cm 3	4,42*10 22	4,96*10 22
Band gap, EV	0,72	1,1
Dielectric constant	16	12
Melting point, °C	937,2	1420
Thermal conductivity, cal/cm X sec X deg	0,14	0,20
Electron mobility, cm 2 /sec*V	3800	1300
Mobility of holes, cm 2 /sec*V	1800	500
Electron lifespan, μsec	100 - 1000	50 - 500
Electron mean free path, cm	0,3	0,1
Hole free path, cm	0,07 - 0,02	0,02 - 0,06

The table shows that the mobility of electrons and holes, the lifetime of electrons, as well as the mean free path of electrons and holes are significantly higher in germanium, and the band gap is lower than in silicon. It is also known that the voltage drop across the p-n junction is 0.1 - 0.3 V, and at n-p - 0.6 - 0.7 V, from which we can conclude that germanium is a much better “conductor” than silicon, and therefore, the amplification stage on a p-n-p transistor has significantly less loss of sound energy than a similar one on n-p-n. The question arises: why was the production of germanium semiconductors stopped? First of all, because according to some criteria, Si is much preferable, since it can operate at temperatures up to 150 degrees. (Ge - 85), and its frequency properties are incomparably better. The second reason is purely economic. The reserves of silicon on the planet are practically limitless, while germanium is a rather rare element, the technology for obtaining and purifying it is much more expensive.

Meanwhile, for use in home audio equipment, the mentioned advantages of silicon are absolutely unobvious, while the properties of germanium, on the contrary, are extremely attractive. In addition, in our country there are heaps of germanium transistors, and the prices for them are simply ridiculous.**

** I foresee that after the publication of this article, prices on radio markets may jump, as has already happened with some types of lamps and microcircuits - Approx. ed.

So, let's start looking at amplifier circuits based on germanium semiconductors. But first, a few principles, adherence to which is extremely important to obtain truly high-quality sound.

1. There should not be a single silicon semiconductor in the amplifier circuit.
2. Installation is carried out in a volumetric hinged manner, with maximum use of the leads of the parts themselves. Printed circuit boards significantly degrade the sound.
3. The number of transistors in the amplifier should be as small as possible.
4. Transistors should be selected in pairs not only for the upper and lower arms of the output stage, but also for both channels. Therefore, it will be necessary to select 4 specimens with the closest possible values ​​of h21e (at least 100) and minimal Iko.
5. The core of the power transformer is made of plates Ш with a cross section of at least 15 cm 2. It is highly advisable to provide a screen winding that should be grounded.

Scheme No. 1, minimalist

The principle is not new; such circuitry was very popular in the sixties. In my opinion, this is almost the only configuration of a transformerless amplifier that corresponds to audiophile canons. Thanks to its simplicity, it allows you to achieve high sound quality at minimal cost. The author has only adapted it to modern requirements of High End Audio.

Setting up the amplifier is very simple. First, we set resistor R2 to half the supply voltage at the “minus” of capacitor C7. Then we select R13 so that the milliammeter connected to the collector circuit of the output transistors shows a quiescent current of 40 - 50 mA, no more. When applying a signal to the input, you should make sure that there is no self-excitation, although it is unlikely. If, nevertheless, signs of RF generation are noticeable on the oscilloscope screen, try increasing the capacitance of capacitor C5. For stable operation of the amplifier when the temperature changes, diodes VD1, 2 must be lubricated with thermal conductive paste and pressed to one of the output transistors. The latter are installed on heat sinks with an area of ​​at least 200 cm2.

Scheme No. 2, improved

The first circuit had a quasi-complementary output stage, since the industry 40 years ago did not produce high-power germanium transistors with an n-p-n structure. Complementary pairs GT703 (p-n-p) and GT705 (n-p-n) appeared only in the 70s, which made it possible to improve the output stage circuit. But the world is far from perfect - for the types listed above, the maximum collector current is only 3.5 A (for P217V Ik max = 7.5 A). Therefore, you can use them in the scheme only by placing two per shoulder. This, in fact, is what distinguishes No. 2, except that the polarity of the power supply is opposite. And the voltage amplifier (VT1), accordingly, is implemented on a transistor of a different conductivity.

The circuit is configured in exactly the same way, even the quiescent current of the output stage is the same.

Briefly about the power supply

To obtain high sound quality, you will have to look in the bins for 4 D305 germanium diodes. Others are strictly not recommended. We connect them with a bridge, shunt them with KSO mica at 0.01 μF, and then install 8 capacitors 1000 μF X 63 V (the same K50-29 or Philips), which we also shunt with mica. There is no need to increase the capacity - the tonal balance goes down and air is lost.

The parameters of both circuits are approximately the same: output power 20 W into a 4 Ohm load with distortion of 0.1 - 0.2%. Of course, these numbers don't say much about the sound. I’m sure of one thing - after listening to an amplifier well made using one of these circuits, you are unlikely to return to silicon transistors.

April 2003

From the editor:

We listened to Jean's prototype of the first version of the amplifier. The first impression is unusual. The sound is partly transistor (good load control, clear bass, convincing drive), partly tube (lack of harshness, air, delicacy, if you want). The amplifier turns on, but does not irritate with intrusiveness. There is enough power to drive floor-standing speakers with a sensitivity of 90 dB to unbearable volume without the slightest sign of clipping. What's interesting is that the tonal balance at different levels remains almost unchanged.

This is the result of thoughtful design and carefully selected details. Considering that a set of transistors will cost fifty rubles (although, if you are not very lucky, selecting pairs may require several dozen, depending on what batch you come across), do not skimp on other elements, especially capacitors.

In just a couple of hours, one amplifier channel was assembled on a breadboard for circuit analysis. American germanium transistors Altec AU108 with a cutoff frequency of 3 MHz were installed at the output. At the same time, the passband at a level of 0.5 dB was 10 Hz - 27 kHz, distortion at a power of 15 W was approximately 0.2%. The 3rd harmonic dominated, but emissions of higher orders were also observed, up to the 11th. With GT-705D transistors (Fgr. = 10 kHz), the situation was somewhat different: the band narrowed to 18 kHz, but harmonics above the 5th were not visible at all on the analyzer screen. The sound also changed - it somehow warmed up, softened, but the previously sparkling “silver” faded. So the first option can be recommended for acoustics with “soft” tweeters, and the second - with titanium or piezo emitters. The nature of the distortion depends on the quality of capacitors C7 and C6 in circuits 1 and 2, respectively. But their bridging with mica and film is not very noticeable by ear.

The disadvantages of the circuit include the low input resistance (about 2 kOhm in the upper position of the volume control), which can overload the output buffer of the signal source. The second point is that the level of distortion strongly depends on the characteristics and mode of the first transistor. To increase the linearity of the input stage, it makes sense to introduce two volt boosters to power the collector and emitter circuits T1. For this, two additional independent stabilizers are made with an output voltage of 3 V. The “plus” of one is connected to the power bus - 40 V (all explanations are given for circuit 1, for the other circuit the polarity is reversed), and the “minus” is supplied to the upper pin R4 . Resistor R7 and capacitor C6 are excluded from the circuit. The second source is turned on like this: “minus” to ground, and “plus” to the lower terminals of resistors R3 and R6. Capacitor C4 remains between the emitter and ground. It may be worth experimenting with stabilized nutrition. Any changes in the power supply and the amplifier circuit itself radically affect the sound, which opens up wide opportunities for tweaking.

Table 1. Amplifier parts
Resistance
R1	10k	variable, ALPS type A
R2	68k	tuning SP4-1
R3	3k9	1/4w	BC, S1-4
R4	200	1/4w	-//-
R5	2k	1/4w	-//-
R6	100	1/4w	-//-
R7	47	1w	-//-
R8,R9	39	1w	-//-
R10, R11	1	5 w	wire, C5 - 16MV
R12	10k	1/4w	BC, S1-4
R13	20	1/4w	-//- selected during setup
Capacitors
C1		47 uF x 16 V	K50-29, Philips
C2		100 µF x 63 V	-//-
C3		1000 pF	CSR, SGM
C4		220 uF x 16 V	K50-29, Philips
C5		330 pF
C6		1000 uF x 63 V	K50-29, Philips
C7		4 x 1000 uF x 63 V	-//-
Semiconductors
VD1, VD2		D311
VT1, VT2		GT402G
VT3		GT404G
VT4, VT5		P214V

Table 2. Amplifier parts
Resistance
R1	10k	variable, ALPS type A
R2	68k	tuning, SP4-1
R3	3k9	1/4w	BC, S1-4
R4	200	1/4w	-//-
R5	2k	1/4w	-//-
R6	100	1/4w	-//-
R7	47	1w	-//-
R8	20	1/4w	-//-, selected during setup
R9	82	1w	-//-
R10 - R13	2	5 w	wire, C5 - 16MV
R14	10k	1/4w	BC, S1-4
Capacitors
C1		47 uF x 16 V	K50-29, Philips
C2		100 µF x 63 V	-//-
C3		1000 uF x 63 V	K50-29, Philips
C4		1000 pF	CSR, SGM
C5		220 uF x 16 V	K50-29, Philips
C6		4 x 1000 uF x 63 V	-//-
C7		330 pF	CSR, SGM, selected during setup
Semiconductors
VD1, VD2		D311
VT1, VT2		GT404G
VT3		GT402G
VT4, VT6		GT705D
VT5, VT7		GT703D

We make an audio amplifier using germanium transistors with our own hands.

Looking through publications on the Internet, as well as videos on YouTube, one can note a steady interest in assembling relatively simple designs of radio receivers of various types (direct conversion, regenerative and others) and audio amplifiers using transistors, including germanium ones.

Assembling structures based on germanium transistors is a kind of nostalgia, because the era of germanium transistors ended 30 years ago, in fact, as did their production. Although audiophiles still argue until they are hoarse, which is better for high fidelity sound reproduction - germanium or silicon?

Let's leave lofty matters and move on to practice...

There are plans to repeat a couple of designs of simple radio receivers (direct conversion and regenerative) for reception in the short wave range. As you know, an AF amplifier is an essential component of any radio receiver. Therefore, it was decided to manufacture the ultrasonic sounder first.

The low-frequency (or audio, as you wish) amplifier will be manufactured as a separate unit, so to speak, for all occasions...

We will assemble the ultrasonic transistors using germanium transistors produced in the USSR, fortunately I have probably hundreds of different types of them. Apparently it's time to give them a second life.

For a radio receiver, a large ULF output power is not needed; up to several hundred milliwatts is enough. The search for a suitable circuit led to this design.

This scheme comes in handy. Output power -0.5 W, all transistors are germanium, and are also available, frequency response is optimized for radio receivers (limited above by a frequency of 3.5 kHz), fairly high gain.

Schematic diagram of the amplifier.

All parts necessary for assembling the amplifier are not in short supply. Transistors MP37, MP39, MP41 took the first ones that came to hand. It is recommended to select the GT403 output transistors according to their gain, but I didn’t do this - I had a couple of new ones from the same batch, so I took them. The input MP28 turned out to be a single copy, but serviceable.

All transistors were checked with an ohmmeter for serviceability. As it turned out, this is not a guarantee against malfunctions, but more on that below... I used imported electrolytic capacitors, C1-film, C5-ceramic.

In the SprintLayout program we create the PCB layout. View from the side of the printed conductors.

Actually, the printed circuit board is manufactured using LUT and etched in ferric chloride.

We solder all the necessary parts. The board of the assembled amplifier looks like this.

Since the output power of the amplifier is small, radiators for the output transistors are not needed. They are barely warm when working.

Amplifier settings.

The assembled amplifier needs some tuning.

After supplying 9V power, we measure the voltage at the control points, which are indicated in the diagram above. At the collector of transistor VT2, the voltage was minus 2.5 V with the required -3...4 V.

By selecting resistor R2 we set the required voltage.

With the pre-amplification stage on transistors VT1 and VT2 there were no problems in setting up. The situation is different with the output stage. Measuring the voltage at the midpoint (the connection point between the emitter VT6 and the collector VT7) showed a value of minus 6 V. An attempt to change the voltage by selecting resistors R7 or R8 did not lead to the desired results.

In addition, the total quiescent current of the amplifier was reduced - 4 mA instead of 5...7 mA. The culprit of the malfunction turned out to be transistor VT3. Although it was checked by the ohmmeter as working, it refused to work in the circuit. After replacing it, all modes of the amplifier transistors were set automatically according to those indicated in the diagram. The voltages on the electrodes of the transistors in my amplifier at a supply voltage of 9V are indicated in the table. The voltages were measured with a DT830B tester relative to the common wire.

The quiescent current of the amplifier is set by selecting a diode D2 of type D9. With the first diode I came across, I got a quiescent current of 5.2 mA, i.e. exactly what is needed.

To check the functionality, we apply a sinusoidal voltage of 0.3 mV with a frequency of 1000 Hz from the G3-106 audio frequency generator.
In the photo, the output voltage level is approximately 0.3V according to the dial gauge. The signal is additionally attenuated by 60 dB (1000 times) by a divider at the generator output.

We connect a load to the output of the amplifier – a resistor MON-2 with a resistance of 5.6 Ohms. We connect the oscilloscope probes in parallel to the load resistor. We observe a clean, distortion-free sinusoid.

On the oscilloscope screen, the vertical division price is -1V/div. Therefore the voltage swing is 5V. The effective voltage is 1.77V. Having these numbers we can calculate the voltage gain: The output power at a frequency of 1 kHz was:

We see that the parameters of the amplifier correspond to the declared ones.

It is clear that these measurements are not entirely accurate, because the oscilloscope does not allow you to measure voltage with high accuracy (this is not its task), but for amateur radio purposes this is not so important.

The amplifier has a high sensitivity, so when the input is not connected anywhere, noise and the background of alternating voltage can be heard quietly in the speaker.

When the input is short-circuited, all extraneous noise disappears.

Oscillogram of noise voltage at the amplifier output with a shorted input:

The vertical division value is -20 mV/div. The noise and background voltage swing is about 30 mV. Effective noise voltage is 10mV.

In other words, the amplifier is quite quiet. Although the author's article indicates a noise level of -1.2 mV. Perhaps, in my case, the not entirely successful layout of the printed circuit board played a role.

By supplying an alternating voltage of different frequencies to the input of the amplifier at a constant level and monitoring the output voltage across the load with an oscilloscope, we can take a graph of the amplitude-frequency response of a given ULF.