June 29, 2013 at 02:02 PM

LME49710 + LT1210CT7 composite headphone amplifier

DIY or do it yourself

A bit of lyrics

By the nature of my work, I constantly communicate with professional audio equipment, so the goal of this development was to obtain a device with high fidelity of reproduction. Therefore, circuit solutions without OOS were immediately discarded and a composite circuit was taken as a basis, which has great potential. I have seen several devices with this topology, but most of the designs used the LT1795 or AD815 devices. However, as Dmitry Andronnikov (aka Lynx) mentioned in one of his articles:

even very powerful op-amps with TOC, when connected to their output of headphones with a resistance below 50 ... 60 Ohm, operate in a sufficiently intense mode both in terms of the output current (which leads to an increase in distortion) and in terms of heat generation.

It was decided to involve "heavy artillery", but more on that below.

Initially, I intended to install a two-link LC filter using murata PLA10 and PLH10 series chokes; EMI suppression capacitors - Epcos X2 with a capacity of 0.22 μF and a varistor Epcos S20K275 with a large absorbed energy - 150 J.

But in the end, it was decided to use ready-made solutions in order to save space in the case:

M5230L devices manufactured by Mitsubishi Electronic are used as stabilizers. Quite interesting devices with very low intrinsic noise (many times lower than that of the widely used LM317 / LM337) in a wide frequency range: 12 μV RMS, 20Hz - 100 kHz, high temperature stability (0.01% / ° С). But there is one drawback - the output current of the IC itself is limited to 30 mA, so to obtain high currents, it is necessary to use external regulating transistors. The switching circuit has no peculiarities and is taken from the datasheet - High ripple rejection circuit. Regulating transistors are used 2SC4793 / 2SA1837 and are installed on a common radiator.

SMD Schottky diodes 10MQ100 are used as rectifiers. Capacitors 3300 uF Panasonic FC, the rest - Elna series Silmic and Silmic II. Initially, it was planned to install additional capacities around the LT1210, but already during the layout of the board, it was decided to abandon them, since the output capacities of the PSU were in the immediate vicinity of the consumer. When using Elna capacitors, even at the stage of designing a printed circuit board, it is necessary to clarify their dimensions on the manufacturer's website, since these capacitors have several times larger dimensions than "ordinary" ones:

Amplifier

The composite circuit has a large loop gain, the output stage on a powerful op-amp is covered by its own feedback loop, which makes it possible to obtain fairly low distortions when operating on various types of loads.

The output stage uses Linear's LT1210CT7 high-speed current-feedback op amps in a TO220 package that provides much better heat dissipation than other packages offered by the manufacturer. These op amps can supply 1.1A long-term current (2A peak), making it easy to drive a low impedance load. The gain of the output op-amp is 2, but can easily be reduced to 1 or increased. The choice of resistor values in the OS circuit of an OS with TOC is somewhat more complicated than for an OS with a voltage OS: the stability of the circuit directly depends on its value. Decreasing the Rf value increases the operating frequency band, but worsens the stability, decreasing it increases the stability and narrows the operating range. The resistors in the feedback loop (1.5 kΩ) are selected for maximum stability. Compensation capacitors (C4, C23) ensure stability when driving a capacitive load. In general, the manufacturer promises stable operation for a capacitive load of up to 10,000 pF!

A little about the thermal regime. The datasheet contains (which is not very common) examples of calculations of the thermal regime are given. The microcircuits are installed on individual HS211 heatsinks with a thermal resistance of 7.5 ° C / W. The thermal resistance of the crystal-package is 5 ° C / W. Without knowing the exact thermal resistance of the insulating pad, it was taken for calculations 2 ° C / W (according to some data found on the Internet).
With a load of 16 Ohm (although professional headphones rarely have a resistance below 50-60 Ohm) and an output signal of 4V RMS (which corresponds to 5.6V peak voltage), the microcircuit will dissipate 2W of heat (the current consumption in this mode is slightly more than 100 mA). At an ambient temperature of 25 ° C, the crystal will heat up to no more than 65 ° C.
Since our device will be in a closed case, the temperature inside can reach quite decent values of 50-60 ° C. As a result, we get a crystal temperature of the order of 100-110 ° C, which is quite acceptable. But this is on a sinusoidal signal, on a music signal (although you had to "see" phonograms with RMS -3 db :-)) the heating will be much less.

As the first "stage" was chosen a device manufactured by another company - LME49710 from Texas Instrument, which has excellent characteristics: low level of all types of distortion, very low noise level (0.34 μV RMS @ 20Hz - 20,000 Hz), low offset (± 0.05 mV), high open-loop gain (140 dB) and high CMRR and PSRR levels (120 and 125 dB).
However, in the process of setting up the device, one copy of this op-amp came across, when installed, the constant voltage at the output was about 14 mV; after some deliberation, another device was installed (out of 10 purchased) - everything returned to normal.

The board as a structural element

When using a high-speed op-amp like the LT1210, which has an output voltage slew rate of 900 V / μs, special attention should be paid to the topology of the printed circuit board, since in high-speed analog circuits it can significantly affect the quality of the device. The board should be designed to minimize the impact on circuit performance. You can read more about the topology of printed circuit boards in books and.
The amplifier is made on a double-sided printed circuit board, the lower layer of which is reserved for the GND polygon, of course, during the wiring process, some circuits still ended up on the lower layer, but they are short. This board topology was chosen to reduce the ground impedance as much as possible for stable operation of high-speed op amps. The vulnerability of the op-amp inverting inputs to ground capacitance was also taken into account, since even a capacitance of 1 pF can lead to an increase in the op-amp transfer coefficient at frequencies close to the maximum. The most obvious solution to this problem is to shorten the length of the conductor. Another, less obvious, is a decrease in its width. As a result, the application of 0.3 mm conductors to the inverting input of the op-amp gives a capacitance of about 0.1 pF, depending on the dielectric constant of the board material (for FR-4 from 4 to 5).

During assembly and configuration

To power the device, a transformer with four secondary windings with a voltage of 14 V and a current of 500 mA is used, which was specially ordered from the factory for this project. I used the volume control ALPS RK27 with a resistance of 10 kOhm. The case was purchased in China.

Unfortunately, at this moment nothing but a phone was at hand.

Now the measurements

After assembly and listening, measurements were carried out using the complex

An article about the creation of an amplifier, in the circuitry and design of which unconventional technical solutions are used. The project is non-commercial.

I began to get carried away with audio equipment and listen to music a very long time ago, from the end of the 80s and for a long time was firmly convinced that any UM with the label Sony, Technics, Revox, etc. much better than domestic amplifiers, and homemade products - even more so, since Western brands and technologies have the highest quality parts and experience.

Everything changed after A.M. Likhnitsky in the journal Audiomagazin No. 4 (9) 1996, which described the development and introduction into production in the 70s of the Brig-001 amplifier, of which he is the author. By chance, after a short period of time, the faulty Brig-001 from the first issues fell into my hands. Using only original domestic parts of the 70s - 80s, he brought this UM to its original state so that it was possible to assess its sound abilities as reliably as possible.

Connecting the Brig-001 amplifier instead of the Technics SU-A700 home audio system shocked me - Brig sounded much better, although the parameters were more modest and was 20 years older. It was at this moment that the idea arose to make an amplifier with his own hands, capable of replacing the standard one in the audio system. which was done in 1998, mainly on the domestic element base of military acceptance. The new device did not leave any chances for comparative listening to more eminent amplifiers such as NAD and Rotel mid-range models and was quite convincing even in comparison with their older brothers. The project was further developed in 2000, in the form of a two-block PA according to the same scheme, but with a new design and increased power consumption of the power supply. It was already compared with transistor and tube amplifiers from the price category up to several thousand US dollars, and, in many cases, surpassed them in sound quality. Then I realized one more thing - the design of the amplifier decides almost everything.

Analyzing the results of listening, especially with the participation of those amplifiers that sounded better than my two-unit PA, I came to the conclusion that more often either good tube designs or transistor ones without general OOS turned out to be at their best. Among them there were also PAs with deep LLCS, in the specifications of which very high values of the rate of rise of the output voltage - 200 V / μs and higher, often flaunted. As a rule, these devices were expensive, and their circuitry was not publicly available. My power supply also had a fairly deep OOOOS, but low speed in comparison with them - about 50 V / μs, with a comparable output voltage. Sometimes he lacked the ability to fully convey the naturalness of the timbres of musical instruments and the voices of performers, the emotions of musicians. On some compositions, the presentation of the music was simplified, part of the timbre richness was hidden behind a kind of thin gray veil. This is probably what is called the "transistor sound" inherent in the PA with feedback.

The reasons for the "transistor" sound in a UM with an LLC have been repeatedly discussed both on forums, and in books on circuitry, and in publications of magazines corresponding to this topic. One of the well-known versions, which I also adhere to, is that the low output impedance of the amplifiers covered by the general OOS, measured on a sinusoidal signal and a resistive load, does not remain so at all when playing music on the speakers, which allows back-EMF signals from the dynamic heads penetrate from the output of the amplifier through the feedback circuits to its input. These signals are not subtracted by the LLC, since they already differ in shape and have a phase shift relative to the original ones, so they are safely amplified and again get into the acoustic systems, causing additional distortion and extraneous sounds in the audio path. Methods to combat this effect are discussed periodically. Examples include the following:

1. "False" channel of the LLC, when its signal is taken from one of the parallel-connected elements of the final stage, which is not connected to the AC, but is loaded onto a resistor of a certain value.

2. Reducing the output impedance of the PA even before the coverage of the LOS.

3. Increase of speed of action inside the loop of the LLC to "cosmic" speeds.

Naturally, the most effective way to deal with LOS artifacts is to exclude it from the UM circuitry, but my attempts to build something worthwhile without LOS on transistors were unsuccessful. I considered it impractical for myself to start from scratch in the field of tube audio technology. The method from point "1" aroused many questions, so I began experiments with increasing the speed inside the feedback loop, taking into account point "2". I would like to immediately draw attention to the fact that the rate of rise of the output voltage, sufficient for the amplifier to correctly reproduce the attack of the sound of musical instruments by the amplifier, is a relatively small value, and its ultra-high values are relevant only in relation to the operation of the LLC.

It is clear that in amplifiers with a common OOS, not all problems are solved by increasing the slew rate, but the main idea was the following, all other things being equal: the higher the speed inside the LOS loop, the faster the “tails” of uncompensated feedback signals will fade and what should be some threshold of their visibility by ear, given the decrease in the duration of artifacts with an increase in performance. Moving in this direction, I very quickly faced the problem of approaching at least the 100 V / μs bar in the PA on discrete elements - in the presence of cascades on powerful transistors in the circuit, everything turned out to be much more complicated. In amplifiers with voltage feedback, my high performance did not "fit" with stability, and in a PA with TOC (with current feedback) it was not possible, without the use of an integrator, to obtain an acceptable level of constant voltage at the output, although everything was okay, and problems were solved with stability. The integrator does not change the sound for the better, in my opinion, so I really wanted to do without it.

The situation was almost a dead end, and it was not the first time that thoughts arose that if you create a power amplifier with OOS in terms of voltage, then using the topology of a preliminary or telephone amplifier, it will be much easier to make it fast, broadband, stable and without an integrator, which, in my opinion, is should have a positive effect on the sound quality. All that remained was to figure out how to implement it. For almost 10 years, there was no solution, but during this time a home "R&D" was carried out to study the effect of the rate of rise of the output voltage inside the common OOS loop on the sound quality, for which a prototype was created that allows testing various composite amplifiers on an op amp.

The results of my "R&D" were as follows:

1. The speed and bandwidth of the composite amplifier should increase from input to output.

2. Correction is only single-pole. No capacitors in the OOS circuits.

3. For an amplifier with a maximum output voltage of 8.5 V RMS, with a depth of about 60 dB, a noticeable increase in sound quality appears somewhere in the range of 40-50 V / μs, and then closer to 200 V / μs, when amplifier practically ceases to be "audible" LLC.

4. Above 200 V / μs, no noticeable improvement was observed, but for an amplifier with an output voltage of 20 V RMS, for example, it takes 500 V / μs to achieve the same result.

5. The input and output filters that limit the PA bandwidth are far from being the best in sound, even if the cutoff frequency is significantly higher than the upper limit of the audio range.

After unsuccessful experiments with PA on discrete elements, my gaze turned to high-speed op-amps and integral buffers with the highest output current. The search results were disappointing - all devices with a high output current are hopelessly "slow", and fast ones have a low allowable supply voltage and not a very high output current.

In 2008, by chance, an addition to the specification for the integrated buffer BUF634T was found on the Internet, where the developers themselves showed a composite amplifier circuit with three such buffers at the output connected in parallel (Fig. 1) - it was then that the idea came to design a PA with a large number of such buffers in the output stage.

The BUF634T is a wideband (up to 180 MHz), ultra-fast (2000 V / μs) parallel-follower buffer with an output current of 250 mA and a quiescent current of up to 20 mA. Its only drawback, one might say, is the low supply voltage (+ \ - 15 V nominal and + \ - 18 V - the maximum allowable), which imposes certain restrictions on the amplitude of the output voltage.

Still I stopped my choice on BUF634T, resigning myself to the low output voltage, since all the other characteristics of the buffer and its sound properties were completely satisfactory to me, and began to design a PA with a maximum output power of 20 W / 4 Ohm.

Fig. 1

The choice of the number of elements of the output stage came down to obtaining a PA operating in a pure class A for a load of 8 Ohms and to provide the modes of the elements of the output stage in terms of current that are far from the limiting ones. The required amount was determined as 40 + 1. For the additional 41st buffer, the minimum quiescent current was set - only 1.5 mA, and it was supposed to be used in order to carry out the first launch of the structure even before installing the radiators, as well as for the purpose of carrying out some settings and experiments in more comfortable conditions. Later it turned out that it was a very good idea.

As you know, parallel connection of integrated circuits does not lead to an increase in the overall noise level and Kg, but the input impedance of such a module decreases and its input capacitance increases. The first is not critical: the input impedance of the BUF634T is 8 megohm and, accordingly, the total will not be lower than 195 kOhm, which is more than acceptable. With the input capacitance, the situation is not so rosy: 8 pF per buffer gives 328 pF total input capacitance, which is already a noticeable value and will negatively affect the operation of the swinging op-amp (Fig. 1). To globally reduce the output impedance of the output stage driver, another op-amp was introduced in front of it, covered by its own OOS loop. Thus, the circuit grew into a triple composite amplifier, but in which all the points of the results of my "R&D" were carried out. After numerous experiments, the composition of the composite amplifier UN was determined: AD843 took the place of the input op-amp, and the powerful high-speed op-amp AD811, with current feedback, was designed to serve as the output buffer of the driver stage. To ensure the required PA speed (over 200 V / μs), the gain of the AD811 was chosen equal to two, which ideally doubled the available 250 V / μs of the AD843 and gave hope that with the appropriate circuitry and good design it would be possible to maintain the required slew rate of the output voltage for the complete PA circuit. Looking ahead, I note that expectations were justified - the real value of this parameter with output buffers turned out to be more than 250 V / μs.

The general circuit of the amplifier has undergone many changes during the setup and debugging, so I’ll give you the final version right away, which includes all the corrections and improvements (Fig. 2).

Rice. 2

The structure is simple - an input selector, volume control, VN, a buffer amplifier for recording on a tape recorder, an output stage and a protection relay, which is controlled by an optoelectronic circuit for delaying the connection of speakers and protecting them from constant voltage (Fig. 3). For compactness, buffers and accompanying resistors are combined by 10 pieces, but the part numbering is preserved in full. As seen in Fig. 2, the contact group of the UM protection relay (K6) is not included in the sound path and closes the output to ground during transient processes or possible emergency situations.

Rice. 3

For BUF634T, such an inclusion is not dangerous, especially since all buffers have a 10 Ohm resistor at the output. In order to avoid loss of stability by the amplifier, due to a short to ground of the resistor OOOOS (R15), simultaneously with the actuation of relay K6, relay K5 is also closed, forming a time circuit of the OOOOS driver stage through the resistor R14. If the values of the resistors R14 and R15 are equal, then there are no extraneous clicks in the AC during operation of the protection, even if they are over 100 dB sensitive.

It is worth noting that the first year of operation, the amplifier functioned reliably both without the K5 relay and without the OOS time circuit with R14, but I was haunted by the very probability of self-excitation during the protection operation, so these additional elements were introduced. By the way, the amplifier works fine even without coverage of the final stage by the OOOOS circuit. You can remove the resistor R15, relay K5, and with the resistor R14 close the feedback in the UN, which I did as an experiment. I liked the sound less - it is possible that this is the option when we get more pluses than minuses from the use of ultra-fast feedback.

The diagram also shows that one of the 4 inputs (CD input) switches the PA to the DC amplifier (DCA) mode, and from the LP input (vinyl player) the Tape Monitor function is implemented, and without additional contact groups in the circuit signal flow. I am a fan of analog recording, so I did it for myself. If the audio system does not have analog recording devices, then the block on op-amp IC1 can be eliminated.

The diagram does not show the blocking capacitors for power supply - for convenience, they will be displayed on the power supply diagram.

The ideology of this amplifier is significantly different from the classical one and is based on the principle of current sharing - each element of the final stage operates with a low current, in a very comfortable mode, but a sufficient number of these elements connected in parallel can provide this 20-Watt amplifier with the maximum current in the load. more than 10 A continuously and up to 16 A in a pulse. Thus, the output stages are loaded during listening, on average, by no more than 5-7%. The only place in the amplifier where high currents can pass is the two copper buses on the PA board, leading to the terminals for connecting the AC, where the outputs of all BUF634Ts of each channel converge together.

Within the framework of the same ideology, the power supply unit of the UM was developed (Fig. 4) - in it, all power elements also operate with relatively small currents, but there are also many of them, and as a result, the total power of the power supply unit is 4 times higher than the maximum power consumed by the amplifier. The PSU is one of the most important parts in an amplifier, which, from my point of view, is worth considering in more detail. The amplifier is built using the "dual mono" technology and therefore contains on board two independent power supplies for signal circuits, fully stabilized, with a power of 150 W each, separate stabilizers for the voltage amplifier, as well as a power supply for providing service functions, powered by a separate network transformer 20 W. All power supply network transformers are phased among themselves - during the manufacture of transformers, the conductors of the beginning and end of the primary windings were marked.

Rice. 4

The power section of each channel is divided into 4 bipolar lines, which made it possible to reduce the load current of each stabilizer to a value of only 200 mA, and to increase the voltage drop across them to 10 V. In this mode, even simple integrated stabilizers such as LM7815 and LM7915 have proven themselves perfectly in powering sound chains. It was possible to use more "advanced" LT317 and LT337 microcircuits, but there were many original LM7815С and LM7915С from Texas Instruments, with an output of 1.5 A, which determined the choice. In total, the power supply of the amplifier signal circuits is provided with the help of twenty such integral stabilizers - 4 for VC and 16 for VC (Fig. 4). Each pair of power section stabilizers feeds 10 pcs. BUF634T. One pair of UV stabilizers is loaded with one channel AD843 + AD811 bundle. The RC circuit (R51, C137, for example) in front of the UN stabilizers has a dual purpose: it protects the rectifier from inrush current when the amplifier is powered on and forms a filter with a cutoff frequency below the edge of the audio range (about 18 Hz), which significantly reduces the amplitude of the rectified voltage ripple and the level of other interference, which is important for the input stages.

Another feature of the power supply is that the main part of all filter capacitors (160,000 uF out of 220,000 uF) are located after the stabilizers, which makes it possible to supply a large current to the load, if necessary. However, this required the introduction of a soft start system "Soft Start" to protect the stabilizers when the amplifier is turned on and the initial charge of the capacitor battery. As seen in Fig. 4, Soft Start is implemented quite simply, on one transistor (VT1), which with a delay (about 9 s) connects a low-current relay K10, which, in turn, includes 4 high-current relays K11-K14, with four groups of contacts in each, closing 16 current limiting resistors with a nominal value of 10 ohms (R20, R21, for example). That is, when the amplifier is turned on, the maximum peak current of each stabilizer is strictly limited to 1.5 A, which is its normal operating mode. I do not use "Soft Start" in the primary circuit of 220 V - in the event of a break in the current-limiting resistor or loss of contact at the soldering points of its terminals, severe consequences for the entire PA are possible.

The PSU for service functions is responsible for connecting the mains voltage to the main transformers (relay K8), powering the Soft Start system components, input selector relays, the supply voltage of which, by the way, is also stabilized. There is also a +5 V output brought out to the connector on the rear panel of the PA - this is already a kind of standard in my amplifiers for simultaneously turning on any external units. This amplifier may well work as an amplifier-switching device (pre-amplifier) for more powerful monoblocks, for example, which will turn on when a control voltage of +5 V is applied to them.

The power supply unit of the amplifier was built in the first place, since the further advancement of the development process required the presence of a full-fledged power supply unit so that the first launch, experiments and tuning were performed in a mode close to real operating conditions. After the successful start of all power circuits, an input selector, a turn-on delay and AC protection unit, as well as a composite amplifier with one BUF634T (BUF41) at the output, as a final stage, were assembled on the PA board. As mentioned above, this 41st buffer has a low quiescent current and does not require installation on a radiator, but headphones were now easily connected to the amplifier output, which made it possible to auditory control, along with measurements. At the end of the debugging of the circuit with one output buffer in each channel, all that remained was to solder the remaining 80 pieces. and see what happens. I did not have any guarantees of a positive result, and it could not be - there was no information about successfully implemented similar projects of other developers. As far as I know, there are no designs on parallel op amps with similar performance either in Russia or abroad even now.

The result was still positive. Since the amplifier was assembled on a rigid chassis made of aluminum bars, where all the commutation connectors were fixed (photo 1), it was possible to connect it to the audio system without a case. The first auditions began, but more on that later - first, I will give some parameters:

Photo 1

Output power: 20W / 4Ohm, 10W / 8Ohm (class A)

Bandwidth: 0 Hz - 5 MHz (CD input)

1.25Hz - 5 MHz (AUX, Tape, LP Inputs)

Slew rate of output voltage: more than 250 V / μs

Gain: 26dB

Output impedance: 0.004 ohm

Input impedance: 47 kOhm

Input sensitivity: 500 mV

Signal to noise ratio: 113.4 dB

Power consumption: 75W

Power supply unit power: 320 W

Overall dimensions, mm: 450x132x390 (excluding the height of the legs)

Weight: 18KG

Based on the parameters, without even looking into the circuit, it is obvious that the amplifier lacks input and output filters, as well as external frequency correction circuits. It should be noted, however, that it is stable and works great even with unshielded interconnect cables. Quite informative in this regard is the oscillogram of a meander 2 kHz 5V / div, at a load of 8 Ohms at almost the maximum level of the output voltage (Photo 2).

Photo 2

From my point of view, this is the merit of the correct wiring of the "ground" conductors, as well as their large cross-sectional area: from 4 sq. Mm. up to 10 sq. mm. (including tracks on printed circuit boards).

There are oscillograms recorded at frequencies of 10 kHz, 20 kHz and 100 kHz, but the tests at high frequencies were carried out with a low signal level, so the presence of a high-ohm volume control at the input, as well as the RC Zobel circuit at the output of the PA, which was still present at that time ( meander 100 kHz 50mV / div - photo 3).

Photo 3

At the very first listening in the home audio system, it became clear that the device was sounding and that it was time to order a case so that you could go on a "tour" with it :) More than 5 years have passed since the completion of the project and the first listening. During this time, dozens (more than 70, according to rough estimates) of comparative auditions of the amplifier with exclusive tube and transistor PAs from well-known manufacturers, as well as with high-level author's designs, were held. Based on the expert estimates obtained, we can say that the amplifier is not inferior in natural sound to most of the listened push-pull and single-ended tube and transistor amplifiers built without using negative feedback, but often significantly surpasses them in musical resolution. Many fans of tube sound and adherents of single-ended PAs without OOS noticed that in this design, the work of negative feedback is practically not "audible" and "does not betray itself" the presence of push-pull output stages in the circuit.

The amplifier was connected to various acoustics - these are speakers of famous Russian manufacturers: Alexander Klyachin (models: MBV (MBS), PM-2, N-1, Y-1), horn speakers of Alexander Knyazev, bookshelf speakers on professional speakers from Tulip Acoustics, Speakers of foreign brands in the middle and high price categories: Klipsh, Jamo, Cerwin Vega, PBN Audio, Monitor Audio, Cabasse and many others, with different sensitivity and input impedance, multiband with complex and simple crossover filters, broadband without crossover filters, speakers with different acoustic design. No particular preferences were found, but the PA is best revealed on floor acoustics with a full-fledged low-frequency range and, preferably, with a higher sensitivity, since the output power is low.

At the initial stage, the auditions were organized not for the purpose of "sports" interest - their main task was to identify any artifacts in the sound that you can try to correct. Very informative and useful from this point of view, listening was in the audio system of Alexander Klyachin, where there was a unique opportunity to evaluate the sound of the amplifier on 4 different speaker models at once, and some of these speakers (Y-1) liked so much that they soon became components of my home audio systems (Photo 4). Naturally, it was very pleasant to receive a high assessment of my product and some comments from an audio expert with vast experience.

Photo 4

The audio system of the famous master of the Russian Hi-End Yuri Anatolyevich Makarov (photo 5, UM at the listening), built in a specially equipped listening room and being a reference in all respects, made major adjustments to the design of this amplifier: the Zobel circuit was removed from the PA output and the main input made bypassing the blocking capacitor. You can hear everything and even more in this audio system, so it is difficult to overestimate its contribution and Yuri Anatolyevich's advice to the process of fine-tuning the sound of the amplifier. The composition of its audio system: a source - transport and a DAC with a separate Mark Levinson 30.6 power supply, Montana WAS speakers from PBN Audio, an uncompromising single-ended tube amplifier "Imperator" and all antiphase cables designed by Yu.A. Makarov. The lower cutoff frequency of the Montana WAS 16 Hz (-3 dB) speaker made it possible to evaluate the "contribution" of the isolating capacitor, and of sufficient quality (MKP Intertechnik Audyn CAP KP-SN), to distortion of the low frequency range of the music signal, and the highest musical resolution of the audio system - to hear the negative impact output filter, in the form of a Zobel RC circuit, which had no effect on the stability of the amplifier and was soon removed from the board. Connecting external low-ohm volume controllers from 100 Ohm to 600 Ohm (the standard WG was set to the maximum position) made it clear that even the high-quality discrete DACT 50 kOhm regulator used in my amplifier would be nice to replace with a lower value (from the external ones connected to me seemed to be the best WG 600 Ohm), but for this it would have to be redone quite a lot and it was decided to implement this and other accumulated improvements already in the new project.

Photo 5

Probably, it is worth mentioning the participation of the amplifier in the Exhibition in 2011 (photo 6), as the only non-commercial project, the material about which was published in the Stereo & Video magazine in January 2012, where the UM was named “the discovery of the year”. The demonstration was carried out with Tulip Acoustics speakers, which have a sensitivity of 93 dB at an impedance of 8 ohms and, oddly enough, the available 10 W / 8 ohms was enough in a large hall with a high level of background noise. 10 watts from an amplifier in class A, in which each watt of output power is sufficiently provided by the energy intensity of the power supply, is perceived subjectively louder, according to my observations, than the sound of an amplifier with a higher output power, but with the final stages contained on a "starvation solder".

Photo 6

After the Exhibition, I received more frequent requests from those wishing to repeat the project via e-mail and personal messages from the forums, but certain difficulties arose - information support was provided to everyone, but my boards were drawn on graph paper, on both sides, and were not suitable for scanning to a file , since the paper was translucent through, and an almost unreadable drawing was obtained. Without a finished printed circuit board, repeat design became increasingly difficult and enthusiasm faded. Now, on the portal forum Vegalab. ru, an electronic version of the board is available, the author of which is Vladimir Lepekhin, a well-known PCB layout specialist from Ryazan, who is well-known on Russian-language forums. The board is freely available, there is a link to it in the first post of the topic about this amplifier. The topic is very easy to find: just type the phrase "Prophetmaster amplifier" in the search bar of Yandex or another search program. It is on this board that one of the forum participants Vegalab- Sergei from Gomel (Serg138) managed to repeat this project and get a very good result. Information about this implementation of the UM and a photo of its design can also be found in the relevant topic, following the links in the first post.

A few tips:

When choosing electrolytic capacitors, I was guided by my own measurements of ESR and leakage current, therefore, there are original Jamicons. I specially inserted the word "original", because they are very often counterfeited and many have probably already come across low-quality products under the brand of this manufacturer. In fact, these are some of the best capacitors for use in powering audio circuits.

The volume control is set to DACT 50 kOhm. Now, I would choose their lowest rating - 10 kOhm, or I would use a Nikitin relay regulator with a constant input and output impedance of 600 Ohm. RG type ALPS RK-27 will be much worse and is not recommended for use.

In the electrolyte shunts, a total of more than 90 μF film capacitors are installed. On my boards are "vintage" Evox from the 70s, which I got for the occasion, but no worse will be polypropylene Rifa PEH426, Wima MKP4, WimaMKP10.

I recommend the Finder relay in the power section, AC protection and soft start, and for the input selector, you need to use only those relays whose parameters have a minimum switching current. There are few models of such relays, but they are.

Domestic high-speed rectifier diodes KD213 (10 A) or KD2989 (20 A) in the power supply of the final stage will be better than most imported ones.

I want to note that the amplifier circuitry is quite simple, but to work with such high-speed and broadband microcircuits, you need the appropriate skills and measuring instruments - a functional generator, an oscilloscope with a bandwidth of at least 30 MHz (preferably 50 MHz).

In conclusion, I would like to say that the conclusions I made based on the results of the experiments carried out, as well as during the work on this project and its subsequent refinement, do not claim to be absolute truth. There are many ways to achieve the goal, which in this case is high-quality sound, and each of them implies a set of measures that may not give a positive result individually. Therefore, there are no simple recipes in this area.

Photos of the amplifier on the website of the Danish company DACT:

Best regards, Oleg Shamankov ( Prophetmaster)

The proposed amplifier is designed for use with headphones with an impedance of ≥ 25 Ohm. The sound signal source can be a PC sound card, CD / DVD-players, portable devices. Allows you to "unload" the source output, which has a positive effect on the sound quality (reduction of harmonic distortion, elimination of "blockage" at the low frequency when using a source with isolation capacitors at the output).

Specifications:
Output signal level: 3V (RMS);
Frequency response (in the range 40 Hz - 15 kHz), dB +0.09, -0.20
Noise level, dB (A) -89.2
Dynamic range, dB (A) 87.0
Harmonic distortion,% 0.0015
Harmonic distortion + noise, dB (A) -80.8
Intermodulation distortion + noise,% 0.013
Interpenetration of channels, dB -86.6
Intermodulation at 10 kHz,% 0.014
* Measured using SB Audigy sound card (SB0507)

Amplifier and power supply circuit

The output stage of the amplifier is covered by the local NOS, and the amplifier as a whole by the NOS. A low-pass filter is installed at the input of the amplifier, a Zobel circuit and a choke are installed at the output. In the OOS, RC circuits are introduced to reduce the amplification at the HF. The gain is three.

The amplifier uses a stabilized ± 6V bipolar power supply. LM317T microcircuits are used as stabilizers.

Construction and details

Structurally, the amplifier is assembled on two printed circuit boards. The amplifier submodule is made on a double-sided printed circuit board.

This submodule is installed on the main single-sided printed circuit board, which also houses the power supply, connectors and controls.

Most of the components used are designed for surface mounting: resistors and capacitors of size 1206. The power supply uses a HAHN BVUI 3020161 transformer. Isolating capacitors MKP. NP0 type ceramic capacitors in audio circuits. ALPS RK09 double variable volume control resistor. The output chokes are wound over the MLT-2 resistors with a 0.5mm PEV-2 wire until 1 layer is filled. Instead of AD8066, it is possible to use AD823, OPA2132, OPA2134. The factory case GAINTA G738A is used.

The amplifier is ideally suited to work with any type of dynamic headphone, and can also be used as a pre-amplifier or current / voltage converter.
Load resistance range from 16 Ohm at + -12V power supply, from 8 Ohm at + -5V power supply. The amplifier has built-in high-speed current protection and protection against the appearance of a constant component at the output with an LED indicator. The amplifier should be powered from a stabilized power supply with a rated voltage of + -12V and a maximum allowable voltage of + -14V. In the overwhelming majority of cases, a stabilizer based on a pair of LM317 / LM337 will provide the required quality of power supply. Quiescent current consumption 70 mA. "Tikhonya" is equipped with an LRCL-type output filter, which, on the one hand, provides a good stability margin when operating on a complex load, on the other hand, the minimum possible output impedance, not exceeding 0.1 Ohm.

Main characteristics:

- Gain in inverting connection 12 dB (4 times)
- Total Harmonic Distortion (THD) is less than - 120 dB (100 Ohm load, 20 Vp-p), harmonic spectrum is short, decaying
- Intermodulation distortion less than -120 dB
- Output impedance less than 0.1 ohm
- Symmetrical rise / fall speed, not less than 200V / μs
- The signal-to-noise ratio in the audio frequency band is not less than 110 dB
- Peak (short-term) output current not less than 1 A

For ease of use, the pitch of the Tikhoni leads is 2.54 mm. Thus, it can be installed on a piece of a breadboard by assembling rectifiers, filters and voltage stabilizers on it.

"Quiet" is supplied only as a fully assembled device.
This is due, first of all, to the great technological complexity of the printed circuit board and high sensitivity to the types and ratings of some components, as well as to the quality of installation. "Quiet" does not forgive arbitrary, ill-conceived changes in the nominal values of elements (for example, changing the gain by replacing only one resistor), however, with the recommended ratings and in the recommended inclusion, it provides ultra-high performance. These limitations are associated with the tremendous performance of a part of the circuit and the subtlety of finding the optimal frequency correction.
However, the "Tikhonya" untouched by a soldering iron works great in a very mediocre "environment" - the most important and critical moments of operation are taken into account. An uncomplicated rectifier with cheap stabilizers and capacitors will provide the necessary minimum of life.

Specifications:
Output signal level: 3V (RMS);
Frequency response (in the range 40 Hz - 15 kHz), dB +0.09, -0.20
Noise level, dB (A) -89.2
Dynamic range, dB (A) 87.0
Harmonic distortion,% 0.0015
Harmonic distortion + noise, dB (A) -80.8
Intermodulation distortion + noise,% 0.013
Interpenetration of channels, dB -86.6
Intermodulation at 10 kHz,% 0.014
* Measured using SB Audigy sound card (SB0507)

Amplifier and power supply circuit