When a Valve is better than a Transistor

 

For many years, manufacturers transistor amplifiers audiophile led by the nose by offering a plausible explanation for why should an old amp model is replaced by a new one. Briefly remind these explanations:

– Too high harmonic distortion (in new models distortion amplifiers reduced to 0.0001%);

– Small damping coefficient (damping ratio reached 1000);

– Not wide enough band Frequency (band has been extended to 5 MHz);

– Amplifiers limit the speed of signal change (on the front panel of new amplifiers appeared the inscription “High Speed Amplifier “);

– Speaker requires more power (and that’s a lot of output transistors connected in parallel provide a current output amplifier 100 A).

This list could be continued.

So far managed to maintain adiofilov have faith in all these technical “troubles”, reigned revival in the amplifier market. However, a gold mine, it seems, is still dried up. Many became clear that to what is written in the advertisement should be treated with caution and can only believe his own ears, or at least the ears of qualified experts.

After the games in the era sverhparametry subjectivity. By the mid 80-ies in vogue modest “by the numbers” amplifiers lamps. Please note that none of the above sounds good recipes in tube amplifiers have never been implemented. Perhaps that is why many audiophiles arose suspicion that transistors bring on “damage” to the sound of music, and that even a simple home-tube amplifier transistor is able to remove the “evil eye” and “clean” from his transistor made ​​previously, and even digital recording.

Is it really? Does the lamp advantages over the transistor, and if so, they are shown under what conditions? The answers you will find in this article.

In the difficult moments of the presentation – and of course they will – I will now familiar to the reader objectivist.

To just not into mysticism, look at the physical level, the difference between the lamp, field and bipolar transistors.

Lamp (take as an example a triode) can be regarded as a “guide”, which consists of carefully purified oxygen electrodes – an anode and cathode – as well as the vacuum gap between them is filled with charge carriers – energy excited free electrons. Portable vacuum through the free electrons of the anode current is controlled by the voltage between the grid and the cathode.

Amplifying transistor properties can be characterized by the steepness of the characteristics, that is, the current increment to the increment ratio of anode voltage “grid – cathode” (at a constant voltage at the anode). The independence of the steepness of the electric lamp switch is an indicator of its linearity. It is particularly important that the slope of the transistor depends little on the anode current (in most cases, it is proportional to approximately the top of the 3rd degree of the magnitude of this current). The influence of the characteristics of the input valve linearity can be neglected, since the negative bias current in the grid mode, it is not a chain.

Interelectrode capacitance are constant and independent of the electric lamp modes. It is also important that the main parameters of the lamp is not dependent on the temperature of the anode or otherwise, from the power allocated to it. Another important advantage is the transistor – low internal resistance, which is at optimum mode of lamp use smaller load resistance is approximately twice.

The dependence of the internal resistance based on the reverse slope of the anode current of the transistor of this current, therefore, in a state in which the load resistance is greater than the internal resistance, increased lamp is virtually independent of the anode current.

The field effect transistor can also be viewed as a “conductor”. Conductive part of the transistor is a channel in the ultrapure silicon crystals, the conductivity type of which (p or n) is given an insignificant admixture of indium and arsenic. Depending on the type of the transistor in the channel move charge carriers:., Free electrons or “holes” (non-filled space in the crystal lattice) As the lamp transistor, the output current of the FET (drain current) is controlled by the voltage between the gate and source.

The reinforcing properties of a field effect transistor (like lamps) can be characterized by a slope (ie, the current flow ratio of increment to the increment of voltage “gate – source”).

The field effect transistor has a pronounced non-linearity than a lamp. Almost all types of FETs steepness increases proportionally to the square root of the drain current.

Like the lamps, the current control circuit (gate circuit) is not available, so non-linearity of the input characteristic of the FET can be ignored.

Several worse interelectrode capacitances. The most important capacity “drain – gate” is dependent on the current voltage between the electrodes.

The most disappointing fact is necessary to recognize the high sensitivity of the drain current and transconductance of the FET to changes in the temperature of the crystal. This sensitivity is due to increased carrier mobility with an increase in temperature, and is usually characterized by a temperature coefficient of voltage ‘the gate – source “(that is, increment of the gate voltage which is necessary to maintain a constant drain current of the transistor by increasing its temperature by one degree the crystal). Depending on the mode in which FET is used, the temperature coefficient may have a value of from 2 to -3 mV / ° C [1] .

Worst of all, that the temperature of the transistor chip, although with inertia (determined by the thermal time constant of the transistor), but manages almost all the changes dissipated in the transistor instantaneous power, however, the negative value of this will be discussed later.

In addition to the static induction transistor, other types of FETs have an internal resistance is much greater than the load impedance.

Bipolar transistor – as a sort of “guide”. However, the physical processes associated with the passage of current therein, are fundamentally different from those which occur in the lamps and the FET.

The first difference is that the charge carriers, and they are electrons or holes, had to overcome two barriers (p – n junction): emitter – base and base – collector, ie twice to move from the crystal lattice of one type to another type of lattice .

The second difference – in the current control principle reservoir. The magnitude of this current depends on the number “injected” into the base from the emitter so-called minority carriers for it to “wander” in it until they are drawn into the strong electric field collector reverse biased with respect to the base. Managing injection into the base of the minority carriers is carried out by moving in the forward direction (in other words, priotkryvaniya) base-emitter junction of the transistor. Amplifying properties of the bipolar transistor can be characterized as a slope (ie the ratio of the collector current increment to the increment of voltage base – emitter). In accordance with the theory of the steepness of the bipolar transistor is approximately proportional to collector current, so it has a more pronounced nonlinearity than the FET.

Unlike lamps and FET transconductance nonlinearity to be added bipolar transistor nonlinearity its input characteristics. This is understandable, since even mind it a little different from the current-voltage characteristics of the forward bias diode.

With the inter-electrode capacitances is going on here is the same as in the FET. The most important capacity of the collector – base bipolar transistor depends on the voltage of the current between the electrodes.

In the bipolar transistor, we are also faced with the high sensitivity of its parameters to change the temperature of the crystal. Namely, the temperature coefficient of voltage base – emitter voltage is -2.2 mV / ° C, and the gain of the transistor on-current is increased by 2-3% / ° C. Just as in the FET, the bipolar transistor junction temperature to the inertia (determined by the thermal time constant) lags behind the changes in dissipated instantaneous power transistor.

The internal resistance of bipolar transistors also leaves no hope – it is always greater than the load resistance.

Now group the most important and it is the objective differences between lamps, field and bipolar transistor in Table 1.

Table 1

reinforcing elements View differences	Lamp (triode)	Field-effect transistor	bipolar transistor
conductivity type	E (in a vacuum)	Electron or hole (through the channel in a silicon crystal)	Electron or hole (through two barriers: the emitter-base and base-collector)
Linearity: front           output	Absent           Is proportional to the root of the third power of the magnitude of the anode current	On HF due to the dependence capacity “drain-gate” of the voltage. At no bass     Proportional to the square root of the drain current	On HF due to the dependence capacity “collector-base” on the voltage. On LF is caused by nonlinearity of CVC “base-emitter junction.”   Proportional to the magnitude of the collector current
Temperature-sensitive settings	Missing	Drain current and transconductance depend on the instantaneous temperature of the crystal	The collector current and the current gain is dependent on the instantaneous temperature of the crystal
Output impedance	At half the size of the load resistance	More load resistance (except SIT type transistors)	More load resistance

 

Major differences between the three. Bipolar transistor is different from the lamp main parameters of thermal sensitivity, greater linearity of input and output characteristics (FET occupies an intermediate position in this series); and besides, the lamp (the transistor) is superior in terms of convenience, the transistor harmonize their internal resistance with a loudspeaker.

From my point of view, all this is enough to choose a tube amp amplifier FET, and the last – amplifier using bipolar transistors.

Objectivist: Your point of view does not sound very convincing, because against you selected transistors deficiencies have radical means – negative feedback (NFB). It is better to look for the causes of poor sound transistor amplifiers in their schemes.

Author: I agree. Compare the traditional and well proven amplifier tubes (see. Diagram in Fig. 1), and simple enough for consideration of power bipolar transistors (see. Diagram in Fig. 2).

Fig. 1.	Fig.2	Fig. 3

The main thing that distinguishes tube amplifier – is the output transformer, which is used to convert low speaker impedance to the optimum load impedance output tubes. The transistor amplifier optimum adaptation is possible without using a transformer. Without transformer tube amplifier is difficult to do, if only because that it provides the symmetrical operation of the output stage in “pull-push” (push – pull). In this mode transistor amplifier can be achieved by including in series transistors of different conductivity type. The lamp with the opposite conductivity type, unfortunately, not yet invented.

OG: You see! The transistor amplifier is not complicated tube, and most importantly – there is no output transformer, so if the transistors themselves do not cause poorly amenable to objective analysis of “spoiling” the sound, the transistor amplifier must sound better lamp.

A: Do not jump to conclusions, and carefully look into both schemes (Figures 1 and 2.). The principal difference from the tube amplifier transistor is the lack of it NFB. At the same amplifier transistor each transistor and the amplifier as a whole are covered by the CAB. Indeed: T1 covered local serial CCA current through R W, T 2 covered local CAB current through R 6, the output transistors T W and 4 are covered by local CAB current through the resistors R 7 and R 8 also consistent HSE voltage through a resistance speaker; generally consistent overall power covered EP voltage divider through resistors R 4 and R H (Fig. 2).

OG: Although the use of environmental protection, I see only benefits, but is ready to offer you a transistor amplifier circuit in which the CAB is not (see Figure 3..).

A: I do not want to be petty, but each bipolar (or field) transistor has a consistent HSE current, which is formed as a result of the fall of the signal to the internal resistance of the emitter (source) of the transistor. These bonds could be neglected if the proposed scheme did not have more serious deficiencies.

First – it is an order of magnitude greater (compared with a tube amp) and unfavorable for range of non-linear distortions.

If a tube amplifier does not bring to the clipping, harmonic distortion at the output does not exceed 1-3%, and is dominated by 3rd harmonic distortion in the composition of these; the second as a result of the principle of “push-pull” is compensated, and the higher harmonics are attenuated. In the amplifier shown in Fig. 3, a combination of non-linearity of input and output characteristics of the bipolar transistor is the cause of a spectrum of harmonic, and if the composite signal – considerably higher power higher order intermodulation distortion. It is well known that an effective means to reduce higher-order harmonic distortion there. The use of CCA even worse off, because it can help lower-order distortion are converted to distortion of higher order.

The presence in the musical signal even small largest intermodulation products of higher order causes the listener a sense of “metallicity”, stiffness, roughness, turbidity sound, the sound is often referred to as simply unnatural.

The second drawback of the proposed scheme – a dependence on the parameters of the amplifier transistors instantaneous temperature crystals. This is easily seen, collecting the proposed scheme and then watching as strolling current in transistors and the voltage at the output of the amplifier, especially if a little blow on the assembled circuit. It is possible to stabilize the output of the amplifier, using for this purpose the so-called servo drive (which, by the way, is a kind of environmental protection), but how to solve the problem of distortions, which are called “heat”?

heat distortion [2] It occurs when a signal change (voltage and current) at the output transistor is accompanied by a change in it dissipated instantaneous power and, consequently, changes its instantaneous temperature of the crystal. What causes the following phenomena: in the course of music signal amplification factor of the current amplification of the output transistors smoothly (due to processes of thermal inertia) is changed to 20-30%. These changes, in turn, become a cause infrasound intermodulation distortions in the amplifier, to which the listener’s ear is extremely sensitive.

Another manifestation of thermal distortion because the voltage base – emitter voltage of the transistor depends on the crystal temperature. It turns out that the change in voltage (and current) at the transistor output, which represents the change in dissipation therein power is first converted to a transistor die temperature change, then the change in voltage base – emitter voltage, which in turn is converted back into a voltage ( and current) at the transistor output. As a result of these changes there is a nonlinear Electricity negative feedback, which, if you do not use the local CAB current causes in the low audio frequencies in each of the amplifier transistor (and particularly shown in Fig. 3) (below 150 Hz) decrease in the gain on 10-15 dB, and an increase of harmonic and intermodulation distortion, which reach 10-15%.

A third disadvantage of the amplifier circuit shown in Fig. 3 – is its unacceptably high output impedance. If the output impedance of the amplifier is greater than the resistance of the speaker, the last sound is characterized by increased gulkost and tightening bass. In this regard, the international standards in the field of hi – fi as a “minimum requirement” recommend that the output impedance of the amplifier does not exceed 1/3 of the value of the loudspeaker impedance.

To provide the required output impedance of the amplifier transistors (except when used as output FETs static induction) can be, if you include the parallel resistor or speaker cover off a cascade of negative voltage feedback.

I think that in the transistor amplifier without feedback can not do, because in order to provide even a modest value of harmonic distortion and acceptable output impedance, power transistors must at least have deep local CAB.

OG: Why with such persistence you declare undesirable application in amplifiers CCA, while even willing to put up with such an anachronism tube amplifiers as the output transformer? What can you say bad about the CAB after 60 years of its successful application in many fields of technology?

A: To learn how to behave in DUS amplifiers later. First, consider some of the effects caused by the passage of the music signal through the conductors [3] Including capacitors, transistors and lamp. As I told you in our recent dispute ( “AM” number 4 (5) 95, p. 5), the signal in the conductor is split into several components, which are distributed through it at different speeds.

Fig. 4

There is a so-called multipath propagation of the signal. Signal flow through the conductor can be represented by the signal graph (see. Fig. 4a). It A – signal transmission speed of light and with a virtually unchanged amplitude; B, C, D – Signal transmission delay and with varying different factors, for example, two orders of magnitude smaller than the path A.

Unlikely as it passes through a conductor a sinusoidal signal can notice any changes at the output. Changes could be found on the music signal, but such measurements have not yet learned to do.

OG: I wonder why the wire reacts to the music signal and does not respond to a sine?

A: The reason is that music signals are much different from the sine greater variability [4] ^.

Sudden bursts of amplitude, dynamic transitions from one to another harmonic structure are the most important in the perception of music. They give sound liveliness and vigor. When multipath propagation of the signal in the conductors, in particular in the amplifier circuits, the signal portions with high variability destroyed their phase occurs destructurization. In this sense we can say that there are not marked on the filter amplifier circuit, limiting or convert signal variability. A special case is indicated in the diagram of the filter is a conventional filter variability limits the high frequencies. Yet, if the filter is invisible and thus does not affect the spectral content of the signal, detect it by conventional measurement methods are not so simple.

If standard methods are not ready to “digest” the musical cues, our brain copes with this task quite satisfactorily. When you compare the sound cable, the first thing to notice a difference in the clarity of transmission parts, intonation pattern and dynamics. The frequency range in which the variability of the signal transmitted to bad sound sluggish, expressionless, silent, and the area in which the variability is well heard, begins to dominate, although the apparent rise in the frequency response and can not be detected.

OG: I think you are again enthusiastic cable testing. Better tell me what’s harmful OOS?

A: Yes, perhaps. But it made ​​me considerations help us to understand this question. We represent a hypothetical amplifier with negative feedback in the form of a graph of the signal (Fig. 46). This direct amplifier branch has two signal path, one has instant A, phase inverted, and with the transmission in the other (where | B | << | A |) signal is delayed for a time . Feedback to transfer signal returns immediately. The gain through this hypothetical power short pulse at time t = 0 can be represented by the relation

and when for the same pulse – the relation

.

In this process of environmental protection in the loop does not end there. The fact that the final response through NFB returns to the amplifier input and causes a further response when :

After receiving additional delay signal again reaches the input of the amplifier, and so on. As a result, two vmete response to one pulse on the output of the amplifier with the CCA will have an infinite number of them with time-decaying amplitude:

It turns out that an infinite number of rounds loop signal rather significantly eroded over time. In the case of total loop NFB we can talk about the erosion of time from 100 ms or more, and therefore the most visible consequence of the effect on the overall sound of the CAB is the deterioration of the dynamics and the weakening of powerful sound of music.

It is now possible to understand why the local CAB better behave on the sound than the general. The shorter the signal path in the loop, as a consequence less than the delay – as a result of a shorter period of signal dilution. But we should not delude ourselves: because in this case the signal is destroyed, and suffer its most volatile areas.

These phenomena are exacerbated when above a certain frequency (the so-called dominant pole) loop gain starts falling with a slope of 6 dB / oct [5] ^. Let me remind you that the frequency of the dominant pole for most amplifiers with deep HSE is in the frequency range from 2 to 5 kHz, and operational amplifiers on chips from 50 to 300 Hz.

The fact that the reduction in the frequency of the dominant pole and getting it into the audible range has a bad effect on the sound, first noticed Ya Lohstroh and M. Otala (1973), but they explain this phenomenon is the formation of T I M-distortion [6] More than a decade misled not only audiophile, but developers audio amplifier.

In fact, this phenomenon can be explained in the light of representations about the erosion of the signal loop NFB. The fact that the worst of environmental protection starts to operate, when the loop gain depends on the frequency of the signal can be represented graphically, having considered the work loaded on the real speaker amplifier output impedance is simulated by the CAB.

Fig. 5

If the input of the amplifier to apply tone flash (Fig. 5a) with a carrier frequency equal to the frequency of the fundamental resonance of the speaker is trying to reproduce it on the output power it generates live EP loop path signal (control currents of the output transistors), which has the form shown in Fig. 5 B. In this figure, a circle circled jump drive signal phase, which occurs in response to the speaker attempted to continue their oscillations at the resonant frequency.

Comparing Fig. 5a and Fig. 5b, it can be seen that the control signal variability (inside loop NFB) must be greater than the signal acting on the input of the amplifier.

And it would not be so bad if the signal variability is not limited to “invisible” filters inside NFB loop. However, precisely because of this restriction at the time of termination of flash tone control amplifier is broken, and the signal at the output will depend on the memory stored in the CAB loop, but previously passed signals. This zone is not entirely predictable behavior of the amplifier circled in Fig. 5c.

Lack of clarity and cleanliness of the bass – is that in the end is the owner of the amplifier (whether tube or transistor), output impedance is simulated by the CAB.

OG: I’ve been listening to you and understand, finally, that you are a true objectivist. After all, first you explain and then tell how it should sound.

1. Yes, many years ago I had to watch the phenomena described in the layout of the amplifier, “Brig.”Then I could not understand why the end of the tone flash in the zero distortion, but it was at the time I made ​​the first step towards subjectivism.I trusted my feelings when I heard, as the inclusion of parallel output of the amplifier (and at the same time the speaker) Ohm resistor 4, contrary to the then dominant ideas, improve the sound of the bass.To understand what was going on, I was only 20 years later.

• · ·

In conclusion, I would like to dwell on issues that seem to have forgotten to ask objectivist.

Is there still a difference in the perceived sound tubes and transistors?

How to explain that a simple home amplifier tubes “corrects” the studio recording, made ​​at the transistor equipment?

These questions seem to be long to hover in the air and do not get a full answer.

Firstly, it is very difficult to put a clean experiment in which no circuitry changes the lamp can be replaced by a transistor. In addition, the lamp bulb strife (the same can probably be said of the transistors). For example, a lamp of the same type that are identical in design but manufactured by different companies, they sound differently. It is unlikely that the vacuum in the GDR is worse than in other countries, but it is a lamp, manufactured by “RFT”, sounds almost the same as transistors. I think the reason is that the poor are cleansed from impurities materials used for the manufacture of the lamp electrodes. Let them say that I fantasize, but lamps with a large-sized anode has a more attractive sound than a small, sound character last tougher and shrill. This is easily seen, replacing the lamp in the preamp ESS83 to another of the same type, but with more (or less), the anode [7] .

The little lamp anode – a small crystal transistor: perhaps the clue to the phenomenon? Or the reason is “poisoned” by arsenic or indium silicon crystal? A p – n junctions in the crystal, presumably also mean something? But all of this – from the field of speculation. Reliable data on the specificity of the sound tubes and transistors, as well as explanations of it, yet.

Now about the “cleaning lamp” transistor sound recordings. I think that there is an effect of the harmonization path “record – play”. Power on lamps with a large anode often highlights and even the “ennobling” frequency range of 400 to 600 Hz, thus returning the listener to “lost” deck violin, foundation singing voice, and even saturation tutti orchestra. There is a path on the harmonization of the tonal balance.

Another feature of tube amplifier – it is a nice integration of sound, thanks to which there is a kind of transistor cleaning records of listeners annoying roughness.

This effect can indeed be sometimes observed, but also fine details and finishing touches are disappearing from the record. From my perspective, expressive sound is simply replaced by a greater comfort. But it is quite natural that the listener wants to have a balance between the study of fine details and integration of sound, that musicians often referred to as “connection” (coherency). By the way, if the “live” sound of music such balance is achieved by specially selected ratio of direct sound and reverberation in the hall. But why change something in the natural balance? As it turned out, the transistor records really bring to the music reproduced additional rigidity and roughness, especially in the upper register. The reason, as we have shown, are harmonic and intermodulation distortion of higher order. It is not surprising, therefore, that the listener prefers a more integrated than in the “live” performance, the character of the sound recordings.

Yet for us considered “clean” transistor sound recordings is not necessary to use a lamp. From my personal experience, it follows that a similar effect can be obtained by a skilful selection of interconnects and output cables. Rather convincing results obtained by using OFC wire type. However, if there are no entries in the dynamics, clarity, spaciousness and naturalness, no lights, no cables will not help. I listed features in a good sound recording, apparently lost forever.

conclusions

So is there really a phenomenon of the transistor and tube sound? I think that on an intuitive level, you can give preference to the lamp as a vacuum conductor, before having a crystal structure of a transistor. However, apart from listening results unconvincing data that transistor when used properly sounds worse than the lamp, no.

At the same time among developers “High-End” amplifiers gradually formed the opinion that it is not in transistors, and environmental protection, without which no cost, no one amplifier transistor. It became clear that the CAB is destroying the music signal.

This explanation came when it became clear that in the loop NFB circulate copies of the delayed signal. The reason for the formation of these copies have become complex physical phenomena in the conductors and other elements used in the amplifier.

Paradoxically, the damage from the NFB amplifiers were more than the application to them of such anachronisms as the output transformer.

If the lamps on the amplifier circuitry to perform close to the transistor amplifiers (ie start applying CCA), the advantage of lamps in front of the transistors will be nullified.

An indirect confirmation of this can be considered as the gradual disappearance from Audio tube amps complex type “OTL” (with a transformerless output) at a price of 4 to 10 thousand dollars and the simultaneous appearance on it unpretentious amplifiers with transformer output transistors at a price of more than 200 thousand dollars.

I think that if a transistor amplifier will be able to ever get rid of environmental protection, lamp barrier will be overcome.

1. Likhnitsky

AM number 1/1996

[1] The FET may be placed in a mode in which the temperature coefficient is zero.

[2] Details of the thermal distortions see. My article in the magazine “Cinema and Television Technology”, 1987, № 6,

from. 10-17.

[3] It should not even talk about the conductor, and an almost invisible but physically existing structural barriers in the conductor, as well as the barriers in the place of connection of conductors.

[4] For experts: the signal variability has a strict mathematical interpretation and is currently an integral module of the second time derivative of the signal being viewed through a sliding “window of time” subjective perception (on this, see .: S. Mason, G. Zimmerman electronic signal circuit and system M.. ., 1963, p. 246-250, etc.).

[5] Smooth (with a slope of 6 dB / oct) attenuation loop gain at high frequencies is necessary in order to ensure stable operation of the amplifier with negative feedback.

[6] T I M (Transient Intermodulation distortion) are the result of “hard” limit signal rate of change in the area of environmental protection to the forward branch level, forming in it the dominant pole. Musical signals whose speed is so great that it may be limited to a standard amplifier with negative feedback, in nature do not exist.

[7] Double triode ESS83 produced in three variants design, with the anode height of 11, 14 and 16mm.