|Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting|
Operating current stabilization
All of these improved circuits were intended to operate in class AB, that is to say that it was necessary that, even at zero signal level, there would always be some output stage operating current, usually in the range 30-150mA. This would be chosen, on test, to give the least distortion at the crossover point, and it was essential to ensure that this chosen value was maintained during use, preferably without the need for further adjustment, if the desired low extent of residual distortion was to be sustained.
The amplified diode:
Although various combinations of diodes and negative coefficient thermistors have been proposed, undoubtedly the most popular technique remains the use of an 'amplified diode', using the circuit layout shown in FIG23. In this, a small signal transistor, Q1 is inserted in the load current path of the output stage driver transistor, Q2, between the connections to the bases of Q3 and Q4. A voltage drop will then appear across Q1 of sufficient magnitude that the proportion of it which is developed across R1 will cause Q1 to conduct. This allows the total voltage across Q1 to be adjusted by alteration of RV3. Since the forward diode potential of the base-emitter junction of Q1 will change with temperature, some measure of compensation for the temperature dependence of the output stage quiescent current can be obtained by mounting Q1 in thermal contact with the output transistors them selves. This allows the mean value of quiescent current to be maintained at a fairly stable value over a moderately wide range of ambient temperatures, which would be far from true if no such temperature compensation was used.
FIG23 Amplified diode circuit for providing forward bias on output transistors
The problem with this system is that, in operation, there will always be a time lag between any rise or fall in the junction temperature of the output devices, and a compensating change in the temperature of Q1. This could mean that on a sudden high power demand, occasioned perhaps by a sudden loud passage in the music being reproduced, the output transistor junction temperature and quiescent current would both in crease, because the transistor, Q1? in the compensating circuit, is still at the temperature it was prior to the event, so that the bias voltage which it causes to be applied to the output transistors would be too high.
Then, as the increased temperature of the output heat sink was communicated to the compensating circuit transistor, the voltage across this, and the forward bias on the output devices would be reduced. However, perhaps by this time the loud passage in the music would have passed, and the output transistor junctions will have cooled down, leaving the quiescent current setting temporarily too low. So, although very commonly used, this particular technique is still somewhat less than perfect.
The need to ensure that the output stage quiescent current remains precisely set, at some optimum value, has encouraged some manufacturers to develop quite elaborate systems for this purpose, as, for example, the Japanese manufacturer Pioneer, who employ a purpose built IC specifically for this purpose, to upgrade the performance of an otherwise relatively straightforward amplifier design.
Class A designs:
The other route to the avoidance of crossover distortion was to return to the use of class A operation, and to employ adequate heat sinks for the output transistors to allow the output transistors to remain at an acceptable temperature. Since this will be consider ably above ambient, and won’t increase with output power, the problem of thermal compensation for the output devices will be relatively easy to solve.
The adoption of class A operation for the output devices offered many advantages. Of these, the principal one was that because of the high intrinsic system linearity which would be possible, much less negative feedback would be required to achieve the same level of output distortion than would be the case with a comparable class AB layout, and this would greatly reduce the possibility of unexpected instability in use.
It would also allow a greatly improved transient response, especially since the output impedance of the transistors would be sufficiently low that no output transformer would be required to match the amplifier to the LS load. A 10 watt class A transistor power amplifier design of my own (Linsley Hood, J. L., Wireless World, April 1969, pp.148-153), is shown in FIG24.
FIG24 10 watt class-A transistor power amplifier.
At the time of this design I had an audio system based on a Williamson 15 watt valve amplifier. How ever, this was only a 'mono' system, and I was keen to take advantage of the growing number of 'stereophonic' gramophone records which were becoming available, and which offered a much greater degree of realism in the reproduced sound. I had used the Williamson amplifier for many years, and did not welcome the difficulty and expense of building another power amplifier of this type. On the other hand, side by side comparisons between the Williamson and all of the transistor power amplifier designs which had been offered up to that date (1967) showed that these amplifiers were much inferior to the Williamson in tonal quality.
The class A design of FIG24 had a very similar performance specification to the Williamson in respect of bandwidth and harmonic distortion, over a wide range of output power levels and frequencies, and had an identical tonal quality, so far as I was able to tell. However, all class A systems run hot in use, because the zero signal operating current will be the same as that at full power. An unfortunate corollary to this is that the permitted quiescent thermal dissipation will then set a limit to the maximum power output.
This constraint is a matter of little importance in the case of valve amplifiers, since these will run hot any way, and their bulk ensures that there is adequate scope for normal air convection cooling, but goes against the whole recent concept of transistor audio amplifiers, which are seen as offering high output powers from physically small, cool-running packages. All recent design work has therefore been aimed at achieving an improved performance from transistor power output stages operated in either class AB or in class B, and a variety of ingenious designs have been proposed.
Floating class A bias systems:
The idea of arranging that the quiescent current of the output stage is increased automatically as the signal level increases -- so that the amplifier can, effectively, operate in class A all the time -- though superficially attractive, and proposed by several designers in the 1960s, has proved difficult to implement in any satisfactory manner, and has therefore been abandoned, in favor of improvements in the operation of the basic class AB layout.
The Blomley design:
A very ingenious approach to this problem was suggested by Blomley (Blomley, P., Wireless World, Feb. 1971, pp. 57-61, and March, 1971, pp. 127-131), who proposed that the output devices should be caused to operate at some fixed, optimum, non-zero, level of quiescent current, during the whole of the output volt age swing, with the input signal being divided into two, cleanly separated, halves by a preceding small signal switching stage. Blomley's circuit is shown in somewhat simplified form in FIG25.
FIG25 The Blomley amplifier
This type of operating system has not been widely adopted, presumably because it simply refers the requirement for accurate biasing of the push-pull stage back to the small-signal driver-transistor stage, which displaces the problem rather than solving it.
Super class A:
A further scheme which has been proposed, and which has, indeed, been adopted by one Japanese audio amplifier manufacturer, employs the apparently ingenious scheme of operating the output stages of the amplifier in true class A, but from a low voltage power supply, so that the power dissipation in the output transistors is kept to a relatively low level. The DC mid-point potential of this power supply is then caused to swing, in synchronism with the input signal, by means of a high-power, class B -- or AB -- amplifier, so that the class A amplifier is always able to accommodate the required voltage excursions of the signal, as seen at the output load.
This layout is shown schematically in FIG26, and has been described as 'Super A'. The fallacy with this arrangement, as can be seen from analysis of the operation of the circuit, is that the return path of the signal, from the output of the class A amplifier, through the loudspeaker, and the amplifier power sup ply, and back to the class A amplifier again, is by way of the high power class B unit, so the overall performance of the system can never be better than that of the class B amplifier.
FIG26 Floating power supply system for super class A operation
So, although the actual amplifier in question does indeed give a satisfactory performance, this circum stance cannot really be attributed to the supposed design breakthrough.
Contemporary high quality audio amplifier designs
Improved Lin type:
Transistor operated audio power amplifier designs fall, in general, into three categories. Of these, the first, which represents by far the most commonly adopted type of circuit layout, is based on straightforward developments of the basic Lin type circuit, in which a linear low-power voltage amplifier stage is followed by a pair of emitter followers, connected in push-pull, to reduce the output impedance to a level suitable for driving a loudspeaker load.
In modern circuitry the output emitter followers will nearly always be devices having fully complementary symmetry (NPN/PNP, or N-channel/P-channel), using one or other of the forms shown in FIG27, which are essentially similar in use to that used by Bailey ( FIG19).
FIG27 Various output impedance conversion layouts
It’s arguable that true equivalence between such apparently complementary devices is not really possible, because of the differences in their way of operation, but, in practice, this lack of symmetry only becomes evident at operating frequencies towards the upper end of their frequency range. Also, technical developments since this type of design first appeared, in the late 1960s, have made available much better devices, such as higher power, higher voltage, higher gain, and higher transition frequency junction transistors, or very linear and robust high voltage power MOSFETs, and the availability of these components has greatly improved solid-state amplifier performance, when used as output devices in this kind of circuitry.
The overall linearity of the amplifier has also been improved by the use of more highly developed circuitry in the input voltage amplifying stages, as, for example, that used in a design due to the author (Linsley Hood, J. L., Electronics Today International, July 1984, pp. 44-49), for use with power MOSFET output transistors, and shown in a slightly simplified form in FIG28, or in designs intended for use with conventional junction transistor outputs, in which the performance has been improved by the use of fully symmetrical driver layouts, of the kind shown, schematically, in FIG29 (Borbely, E., Wireless World, March 1983, pp. 69-75, and also in Audio Amateur (USA), Feb. 1984, pp. 13-24). FIG28 80-100 W power amplifier based on MOSFET devices
FIG29 Power amplifier stage based on fully symmetrical layout
Unfortunately, the performance of all of these circuits will still depend on the maintenance, during use, over a wide range of ambient temperatures, of an accurately chosen level of output stage quiescent cur rent, though the use of power MOSFETs will make the precise value required very much less critical.
Negative feedback and HF compensation Assuming that adequate care is taken in the circuit layout to maintain the correct level of output quiescent current, and to avoid damage to the output transistors due to inadvertent misuse, the design of such a circuit is really a quite straightforward combination of a suitable voltage gain stage with an output impedance converting emitter follower pair. Care must also be taken, in any new design, to ensure that the amplifier remains stable when overall negative feedback is applied around the system. As a general 'rule of thumb', NFB can be applied around a circuit consisting of just two stages -- an amplifier followed by an emitter follower, or two successive voltage amplifying stages -- without the system becoming unstable under normal resistive load conditions.
In circuits in which the NFB loop encloses three or more such stages, such as two voltage amplifier stages followed by an emitter follower, instability will almost certainly occur, usually at some fairly high frequency, for reasons examined in Section 6, and some arrangement to provide HF compensation will be necessary.
This will usually consist, in simple audio amplifier designs, of an output 'Zobel' network (a resistor in series with a small capacitor to ensure that the circuit sees a suitable load under open-circuit conditions), labeled R1/C1 in FIGs 28 and 30, together with some means for ensuring that the open-loop gain will decrease with increasing frequency. This will normally take the form of a simple capacitor, shown as Cy in the schematic layout of FIG30, connected to impose a dominant lag on the phase-shift/output frequency characteristics.
FIG30 Dominant-lag method of HF compensation for feedback amplifier
Although this type of HF compensation is effective in securing HF loop stability, and very widely used for this reason, it suffers from the snag that the rate at which Cy can charge or discharge is limited by the output current from Q1, and this imposes an upper limit on the speed with which the voltage output of the gain block can slew, following a sudden change in the voltage level presented at its input. This effect is known as slew rate limiting and gives rise to the phenomenon described by Otala (Otala, M. J., /. Audio Eng. Soc, 1972, No. 6, pp. 396-399), as 'transient intermodulation distortion' (or TID), in which all signals accompanying such a sudden change in the input signal voltage level will be obliterated until the circuit returns to its steady-state condition. This effect is audibly quite unpleasant, but can be lessened, though not entirely avoided, by the use of a network, R2/C2, which serves to lessen the rate at which the input voltage can change.
The use of the stabilization component, Cy, in the position shown in FIG28, avoids this snag, in that while it also serves to reduce the gain of the voltage amplifying stages with increasing frequency, and thereby impose a phase characteristic which assists loop stability, there is much less difficulty in providing the charging current from the drain of Q7, and there fore much less tendency to slew-rate limiting. This HF stabilization method allows the retention of good over all phase margins within the feedback loop, which, in turn, makes for pleasant amplifier sound quality. Regrettably, the HF compensation layout shown in FIG30 is normally used in commercial audio amplifier designs because it gives slightly better harmonic distortion figures at higher audio frequencies.
Additional HF phase correction components will almost certainly need to be used with more complex circuit layouts in order to ensure adequate overall loop stability, but their position and value will need to be determined specifically for each new design.
The Quad current dumping amplifier Because the performance of all normal class AB audio amplifiers will depend on the accuracy of setting of the quiescent current of the output stages, which will usually require some form of adjustment, with instrumental monitoring of the results, before the amplifier leaves the manufacturer's assembly line, and, because the required operating current, even when initially correctly set, may drift away from the desired value during use, the manufacturers of such units can never guarantee that the desired performance is always given.
Various attempts have therefore been made over the years to devise systems which will operate satisfactorily, with low distortion, with output stages which operate at zero quiescent current. Of these, the only design which has proved really satisfactory is that described by Walker and Albinson (Walker, P. J., Wireless World, Dec. 1975, pp. 560-562), and introduced by Quad (The Acoustical Manufacturing Company), in their '405' and subsequent power amplifier systems.
The idea employed is, in principle, very simple. The power amplifier consists of a basic class B system, with which is associated a low power, but very low distortion, voltage amplifier, and this low power amplifier is arranged to fill in those regions of the output waveform in which the output of the power amplifier departs from the ideal. Since these departures are likely to be small, probably less than 1% of the total, the output power requirements from the low distortion amplifier will also be small. However, although this idea is simple in concept, its implementation requires some subtlety in design, and the way in which it’s done is shown in a simplified schematic form in FIG31.
FIG31 The basic layout of the QUAD current dumping amplifier
The operation of this circuit can be explained by consideration of a simple amplifier of the form shown in FIG32a, in which a high gain linear amplifier (A^ is arranged to drive a load through an unbiased push-pull pair of output emitter follower transistors, Q1 and Q2. Without any negative feedback to straighten out its transfer characteristics, the input/out put transfer curve of this arrangement would have the shape shown by line 'a' in FIG33. The slope of this line would be steep from D to E, when transistor Q1 was conducting, and feeding current into the load resistor ZL. It would, however, be much less steep from E to F, when Q1 had ceased to conduct, and the load was only driven by A1? by way of R3. It would then steepen up once more, from points F to G, when transistor Q2 began to conduct if enough overall negative feedback was applied, by way of Ät, this could straighten out the kinks somewhat, to give the kind of characteristic shown in line 'b\ but this type of performance would still be unacceptable as a Hi-Fi de sign. What is required is some way of reducing the gain of the circuit during the periods when Q1 and Q2 are conducting, so that the slope of the transfer graph is the same from D to E, and from F to G as it’s from E to F. This can be done if an additional small resistor, R4, is connected between the junction of Q1 and Q2, and the load resistor, ZL, and if the negative feedback through R1 is then taken from the junction of Q1 and Q2, rather than from the junction between R4 and ZL. When the ratios between R2 and R3 and Rx and R4 are correct, the discontinuity in the amplifier transfer characteristic disappears, in spite of the fact that Qx1 and Q2 operate with zero forward bias.
FIG32 Method of operation of current dumping power output stage
In practice, R4 would be wasteful of output power, so Walker and Albinson proposed the use of a small, low resistance inductor L instead, with R2 being replaced by a suitable value of capacitor (C) to ensure that the ratios of the impedances, R1 / L1 and C1/R3, remain the same over the audio frequency band. This modification of the circuit, while entirely effective in practice, complicates the theoretical analysis of the circuit, and gave rise to a lot of subsequent discussion and analysis, of which the most detailed was that of McLoughlin (McLoughlin, M., Wireless World, Sept. 1983, pp. 39- 43, and Oct. 1983, pp. 35-41).
FIG33 Input -- output transfer curve of circuit of FIG32a
Sandman's Class S design:
A further, very interesting design innovation, having a similar intention -- making the bias of the output power transistors a less important factor in determining the amplifier performance -- was that due to Sandman (Sandman, A. M., Wireless World, Sept. 1982, pp. 38-39), which is shown in schematic form in FIG34. This design was the culmination of an investigation, by Sandman, of the possibility of reducing amplifier distortion by the use of 'error feed-forward' techniques, to supplement the existing negative feedback technology.
FIG34 The Sandman class S design
The particular circuit design employed was based on the observation that the performance of most amplifiers is better with a high-impedance than with a low-impedance load, and the circuit is arranged so that the input amplifier, A1? sees the load as if it were a much higher impedance than it really is. This is achieved by the use of a high gain error amplifier, A2, which is connected to sense the difference between the output voltage and that from an unbiased emitter-follower pair (Ch/Cb), which also is connected so that it feeds power into the load, and to drive C^/C^ so that this difference is made as small as possible, thereby increasing the apparent impedance of the load, as seen by Av Once again, as in the current dumping amplifier, the small power amplifier, Ax, is able to fill in the deficiencies of the higher power circuit, and if the ratios between R3 and R4, and R5 and R6 are correctly chosen, the crossover-type distortion, due to the unbiased power transistors (Ch/C^), will disappear.
The basic philosophy of this system has been adopted in several Japanese audio amplifier designs, and a representative example of this approach, used by Technics, in their SE-A100 amplifier, is shown in FIG35.
FIG35 Simplified layout of Technics SE-A100 power amplifier
Whether or not it’s practicable to make a sufficient improvement to the best of the existing audio power amplifiers, that an impartial listener would be able to detect any difference in the final output sound quality, is debatable. Certainly, a number of comparative trials have been set up, over the years, with the specific aim of trying to establish whether critical listeners can distinguish, on the basis of 'sound quality', between different, high quality, amplifiers, as is commonly claimed in the Hi-Fi press, during equipment reviews.
Indeed, it’s this claim which provides the reason for the existence of the periodicals in question.
In order to ensure fairness, some of these trials were conducted with audience panels which were deliberately chosen to include those journalists and other critics who were convinced that all amplifiers sound different, and whose own 'in-house' tests had shown that they could hear obvious differences between known systems. They were also permitted to select the program material and the ancillary equipment --loudspeakers, turntables, etc. -- which was to be used during these tests. However, in the event, with tests based on statistically valid sampling techniques, and
'double-blind' methodology (in which neither the listening panel, nor the operators in charge of the test were aware of which amplifier was being auditioned at the time), all of these trials have failed to show that the audience was able to distinguish between one unit and another, more consistently than would have been the case for a purely random response.
For myself, I think that there are still some small remaining differences in sound quality between different power amplifier circuit designs, and that there is still, therefore, some purpose in attempting to improve on existing system performance, quite apart from the normal engineering search for ways of making such equipment simpler, more efficient, more reliable, and less costly.
The problem with any kind of multiple comparison test which can be conducted is that the sampling duration may either be individually too brief to allow the listener to pick up various small acoustic effects, which might allow a preference to be formed, or, alternatively, it may be so long that the memory of comparative sound characteristics will fade, where the differences sought are, of themselves, relatively small.
On the other hand, a more leisurely side-by-side comparison between a pair of systems, conducted over a period of time, can allow the identification of particular, and individual, sound effects which could very easily be overlooked in such group tests.
An example, in point, is the difference between a good, 38cm/second analog tape recording, and a '16 bit' encoded digital recording made of the same performance. Both of these can offer very high quality sound images, of which the principal difference lies in the nature of the background hiss. Given time, it’s possible to distinguish between one and another, but in a multiple, short duration, test it may prove impossible to tell them apart.
Since, at the moment, most integrated circuit components are restricted to power supply line voltages less than +/-20V, and this is too low to allow enough output voltage swing at the LS load of a power amplifier, the great majority of audio power amplifier de signs are based on discrete component circuitry, and the circuitry used is specific to each individual equipment manufacturer. This situation is bound to change with time, as IC technology is improved, to allow higher output voltages and powers, and when this happens, the choice between most audio power amplifiers will probably become simply the choice between one IC and another.
The purpose of an audio power amplifier is to take a relatively low level input signal and to amplify it to an output voltage and power level which is suitable for driving loudspeakers or some other type of load. The purpose of a pre-amplifier -- which may be either a separate unit, or a piece of circuitry incorporated in the same box as the power amplifier with which it will be used -- is to select input signals from one or more program sources, and to amplify and modify these, as necessary, so that they are of the size and form needed to drive the power amplifier. This will usually require some adjustment of both the signal level and the line impedance. For example, the program sources may work best with load impedances which may be anywhere between 100 ohms and 100 kilohms, while the output signal levels from these sources may be between 50 microvolts -- in the case of a low-output moving coil gramophone pick-up -- and 2 volts, RMS, in the case of a typical compact disk (CD) player. On the other hand, the power amplifier may require an input signal in the range 300 millivolts up to 6 volts, RMS, or more, to provide its full power output, and may have a fixed input impedance anywhere between 1k and 100kohms.
With any audio system it will nearly always be necessary to provide some means of adjusting the overall signal level, usually called a 'gain' or 'volume' control, and in a stereophonic system it will also be desirable to have a 'balance' control somewhere in the signal chain, to adjust the relative gain levels of the two stereo channels so that the final sound image is correctly placed. These controls will usually be placed somewhere within that part of the signal conditioning circuit called the preamplifier.
In the earlier years of High Fidelity audio amplifying equipment, it was often considered necessary to provide 'tone controls' and bandwidth limiting filters to modify the frequency response of the system, but, with improved signal sources, these are less often thought to be necessary in modern equipment.
Whereas the major problems with audio power amplifiers are concerned with producing an acceptable performance at the interface between the equipment and the listener, those of the preamplifier circuitry are more related to the purely technical aspects of the design, where there still remain a number of practical difficulties, such as that of avoiding signal voltage overload in the circuitry. This is generally referred to by the term 'headroom'.
There is an absolute limit to the output voltage excursion which can be obtained from any simple electronic circuit -- and this limit is imposed by the design of the circuit and the supply line voltages. For example, if a circuit is powered from a +/-24 volt supply rail pair, the maximum amplitude of any possible output volt age swing can not be greater than 48V. In practice, even with optimally designed circuitry, it will always be somewhat less than this, say 46V, peak to peak. This is equivalent to a symmetrical sinusoidal output of 16.26V RMS. If the stage has a gain of 100, the maximum sinusoidal input which could be handled without overload, and consequent peak clipping, will be 162.6mV. The concern of the designer is therefore to make sure that the output voltage swing, which is available from any part of the circuit, is greater than the likely input signal voltage multiplied by the stage gain.
Although it seems to be fashionable in the Hi-Fi press to portray this problem as one which requires the design of all preamplifier stages preceding the gain control so that they can handle very high signal volt ages without overload, in practice most signal sources will, in any case, have a finite limit to their possible output voltage. For, example, in the case of signals derived from a vinyl gramophone record disk, the limiting factor is the ability of the pick-up (PU) stylus to track the undulations in the record groove, and the ability of the disk manufacturer to cut such lateral excursions without distortion, or breakthrough into adjacent grooves. These aspects were examined by Walton (Walton, J., Wireless World, Dec. 1967, pp. 581-588), who derived the relationships between groove radius, recorded signal frequency and maximum practicable stylus velocity shown in FIG36.
The recording levels normally found on a commercial disk are shown in curve 'd' in this drawing, and this approaches the ability of a high quality pick-up cartridge to follow the path of the groove. Allowing some margin for error, therefore, it’s sensible to expect that the pick-up output won’t exceed some 30dB, at 1-2kHz, with reference to a lcm/s recording velocity.
If the cartridge has an output of 3mV for a typical 5cm/s recorded velocity, which is typical of a high quality moving magnet (MM) cartridge type, the maximum output obtainable before acoustically un pleasant, and groove damaging, mistracking occurs will be 19mV.
30 100 1k 2k 10k 20k -- Frequency (Hz)
FIG36 Walton’s analysis of maximum recorded levels on 12" vinyl disc records
If a typical normal signal level to the power amplifier corresponds to a cartridge output of 2mV, then a preamplifier, without any facility of internal gain adjustment must be able to handle an overload factor of the order of 9.5 without clipping. For a typical transistorized power amplifier, requiring an input signal level of 0.77V RMS, such a preamplifier will require to be able to handle a peak output voltage of 7.3V RMS, which is equivalent to 21V p-p. A +/-15V supply line would therefore be adequate.
If the pre-amplifier gain control were to be positioned nearer to its input, a lesser degree of headroom would be adequate, but this would worsen the extent of the background noise level at the 'zero gain control' setting because there will be a higher overall gain, between the gain control and the amplifier output at such a 'zero level' setting, and this will amplify all the circuit noise generated by the stages following the volume control. For this reason, there is a conflict between the need to have the gain control at a circuit as near to the preamp output as possible, to keep the background hum and hiss levels low, and the desire to offer higher levels of headroom. A customary com promise is to separate out the input circuits used for various input sources and to choose optimum, or possibly pre-set, gain levels for each of them.
Although the case of a moving magnet PU cartridge was quoted, the same arguments concerning tracking capability would apply to a low output-voltage moving-coil (MC) cartridge. The MC input channel would simply need a higher gain. However, since a low output level MC cartridge will usually also require a lower load impedance than an MM type (typically 100R, rather than 47k), most normal pre-amp systems will provide a separate MC input channel, with appropriate gain and input impedance levels.
Similar constraints exist in tape, radio, and CD inputs. For example, if the preferred 'loud signal' recording level for a cassette tape is chosen by the user to be that shown as '0dB' by his recording level meters (where, typically, the 'off-tape' distortion might be 0.5%), then, at '+6dB', the tape third-harmonic distortion could be of the order of 3%, while at '+12dB' this distortion could have increased to some 15%. Since most listeners would regard such a distortion level as unacceptable, in practice they would choose to record at a peak signal level which was lower than this. So, in this case, a preamplifier input overload margin of some 12dB, (4x), would also be adequate.
In the case of an output from an FM tuner, the maximum signal level will correspond to a carrier deviation of +/-75kHz, equivalent to 100% modulation. (In the case of a stereo signal, the maximum modulation level is, in fact, limited to 90%). For a typical program level of 10-15% modulation, the headroom required would be 10x (20dB) at the most.
In the case of a CD player, typical maximum output levels are 2V RMS. Excursions above this level are precluded by the digitally encoded form of the input medium, where 2V RMS may represent the 'all Is' condition of the digital signal. So, if the power amplifier has a 0.77V RMS input requirement for a full output signal, the interface circuitry between the CD payer and the power amp input will only require to have a headroom of 2.6x (8.3dB). All of these considerations assume that the pre-amplifier gain levels have been chosen to be appropriate to the input signal source being used, either by pre-set input gain settings, or, more usually, by the selection of circuit blocks with specific applications to a certain type of input.
The general purpose of these is to alter the relative gain of the circuit at one part of the audio spectrum in relation to another, to compensate for deficiencies in the characteristics of the input signal or in the loud speakers and other hardware used with the system. A very wide variety of circuitry has been devised for this use, ranging from simple 'bass/treble' 'boost/cut' systems, through 'tilt' or 'slope' controls, to elaborate octave-band 'graphic equalizers'. All of these arrangements have their own advantages and shortcomings.
Bass/treble boost/cut systems
These were, historically, the first tone control systems to be used, and, typically, were expected to provide a frequency response curve which could be adjusted, as shown in FIG37, to allow up to 20dB lift or cut in the 'bass' or 'treble' parts of the frequency band, in relation to that given at 1kHz, or some other mid-band frequency.
FIG37 Frequency response of feedback-type tone control
FIG38 Typical passive tone control
FIG39 Frequency response of passive ton control
This frequency response adjustment could be pro vided by a simple frequency-selective attenuator circuit, of the type shown in FIG38, followed by an amplifier, to restore the mid-band, 'flat response' gain to its previous level. This type of arrangement is usually described as a 'passive' tone control circuit because it relies for its operation entirely on passive components, and uses no internal gain stage.
The actual response characteristics of the circuit shown in FIG38, with the component values indicated, are as shown in FIG39. A practical difficulty with this type of circuit is that a flat frequency response output is not provided at the mid point resistance setting of 'linear' law potentiometers, and, even with 'logarithmic' law potentiometers, the calibration of the controls is awkward.
Feedback tone controls:
This problem is avoided by the use of a circuit of the kind shown in FIG40, in which the frequency selective RC networks are housed within the negative feedback loop of a wide band amplifier. This type of tone control circuit is often called a 'Baxandall-type' tone control, after its originator (Baxandall, P. J., Wire less World, Oct. 1952, pp. 402-405). The frequency response of the circuit shown in FIG40 is as illustrated in FIG37. With close-tolerance component values this circuit will give a flat frequency response at the mid-point settings of linear boost/cut control potentiometers.
FIG40 Feedback type tone control
A simple elaboration of this circuit (Linsley Hood., J. L., Hi-Fi News and Record Review, Jan. 1973, pp. 60-63), to increase its versatility, is to allow the values of the capacitors in the feedback networks to be changed by switching, to provide differing boost/cut turnover frequencies, as shown in FIG41.
FIG41 Feedback tone control circuit with switchable lift and cut frequencies
Because it’s obvious that no simple adjustment to the relative gain of the bass or treble portions of the audio pass band will remedy all of the likely tonal errors in the reproduction of a musical signal -- particularly where these arise through unwanted peaks and troughs in the overall frequency response of the system -- various circuit layouts have been proposed which allow the gain at any part of the audio spectrum to be altered relative to the remainder.
It’s customary to arrange that the parts of the spectrum on which the gain adjustments are made are divided into eight or nine octave segments, to cover the audio band from, say, 30Hz to 20kHz. A typical circuit layout, due to Williamson (Williamson, R., Hi-Fi News and Record Review, Aug. 1973, pp. 1484 1491), is shown in FIG42. The possible response characteristics given by this circuit are shown in FIG43.
The difficulty with this type of system, apart from its complexity, and the tendency for the inductors to pick up hum because of their inadvertent interaction with stray mains- frequency magnetic fields, is that a ripple-free frequency response can only be obtained if all the controls are set to the mid-point position.
A modification of this type of system, sometimes called a 'parametric equalizer' uses only one or two such narrow-passband lift or cut circuits, but with some means of altering the operating frequency, per haps by altering the value of the capacitor in the LC segment of the circuit.
(all 10k linear law) 22mH-3.2H
FIG42 Graphic equalizer circuit due to R. Williamson
FIG43 Typical frequency response of graphic equalizer
In practice, with contemporary program sources, and associated transducers, it’s unlikely that any major errors in frequency response will be found, but, nevertheless, the user of the equipment may feel that the overall reproduction, at any given time, is too
'bright', or too 'bass heavy', and several circuits have been offered which allow some adjustment to be made to the overall slope of the frequency response across the pass band.
A circuit for this purpose was described by Bingham (Bingham, J., Hi-Fi News and Record Review, Dec. 1982, pp. 64-65). The layout used, and the possible frequency response adjustments allowed, are shown in FIGs 44 and 9.45.
FIG44 Tilt control due to J. Bingham
Gain (dB) 30 Hz 100Hz 1kHz 10kHz 20kHz Frequency
FIG45 Frequency response of Binghams tilt control
In the early days of audio, most program sources suffered from unwanted additions to the signal, such as low frequency 'rumble' type noises originating from poor quality turntable bearings, 'hiss' due to the emery powder loading of 78 RPM shellac discs, or whistles due to adjacent broadcasting stations during radio reception, and various techniques were adopted to allow the unwanted part of the signal to be filtered out. Inevitably, there will be some loss of wanted program material, so filtration is usually kept to the minimum level which is effective. Modern program sources are generally of a higher quality, and less in need of filtering, so contemporary preamplifier units offer fewer of these facilities than would have been common in earlier Hi-Fi systems.
The circuits which can be used for this purpose are shown in Section 8, and the only point which needs to be made, in a specific audio context, is that the slope of the attenuation characteristic which is used needs to be chosen with care, since too steep an attenuation slope can cause audible coloration with wide-band program material. This is not usually a problem with high-pass rumble filters, since room resonances and cabinet or driver unit resonances in the LS systems will mask such effects, but, at higher frequencies, low-pass treble filters having too steep an attenuation rate may give a noticeable tonality to wide-band white noise. Filters having a variable slope, as well as an adjustable turn-over frequency, provide the best approach.
Magnetic tape and gramophone record replay equalization
Magnetic tape replay equalization:
In the early days of sound recording on magnetic tape, it became clear that there was an unavoidable loss of higher recorded frequencies due to the need for a finite gap in the pole pieces of the record head, shown schematically in FIG46, across which any point on the magnetic tape will take a finite time to pass, and because the effects of magnetic cancellation within the magnetic dipoles in the tape coating become more evident as the recorded wavelength is shortened. Both of these effects cause a high frequency loss which increases as the tape speed is reduced, and some replay frequency compensation will be needed to offset this loss.
FIG46 Action of magnetic recording head
In order to make tapes recorded on one machine compatible in replay on another, various standards have been proposed, which are listed in TBL. 1, which make the assumption that the relative remanent magnetic flux on the tape, at various record/replay tape speeds, will be as shown in FIG47. These mainly concern HF signal loss, but in the case of the '???' proposals, and also in the case of all of the 4.76cm/s (cassette tape) recording systems, some degree of LF pre-emphasis is also proposed so that replay de emphasis, either electronically introduced or inherent in the replay head characteristics, will reduce the ex tent of hum pick-up and improve the LF signal to noise (s/n) ratio.
TBL. 1 Frequency response equalization standards Tape speed -- Standard -- Time constants, \\S -3dB points
FIG47 Assumed remanent flux on record in tape equalized for various tape speeds
a. Record pre-emphasis; b. Replay de-emphasis
FIG48 Tape replay equalization network
The actual resistor-capacitor networks which will be necessary to provide the required record and replay frequency response will be similar to those shown in FIG48, though the actual component values required will depend on the characteristics of the record and replay heads, and on the tape type and equalization time constants chosen -- bearing in mind that the design requirement is that the final record/replay frequency response of the system shall be as flat as possible.
Gramophone record replay equalization:
In early shellac (78 RPM) gramophone records at tempts were usually made to lessen the obtrusiveness of the replay hiss, due to the emery powder loading of the disc, by reducing the HF response of the replay amplifier. The record company RCA, in the USA, therefore proposed that some degree of pre-emphasis should be applied, at the recording stage, so that the HF response of the recording could be preserved in spite of this replay roll-off, and records having this characteristic were marketed under their 'Dyna groove' label.
With the development of vinyl long playing records in the early 1950s, this basic concept of recording pre-emphasis, to allow HF roll-off on replay to reduce surface noise, coupled with an LF de-emphasis to reduce the width of groove-to-groove spacing, was adopted by the Radio Industries Association of America (RIAA), who published a proposal, now accepted on a world-wide basis, for a recording frequency response which would require a replay characteristic -- when reproduced by a velocity sensitive pick-up cartridge -- of the type shown in FIG49. The time- constants and 3dB points associated with this are 3180ns (50.05Hz); 318µ8 (500.5Hz); and 75µß (2122Hz), and one or other of the passive or active networks shown in FIG50 can be used to generate this type of replay frequency response. (Note, It has recently been proposed that the RIAA replay response characteristic should be modified to include a further low-frequency roll-off, to reduce the effect of replay turntable rumble, having a-3dB point at 20Hz (7950|iis), but this is not included in the networks shown in FIGs 48-9.51).
FIG49 RIAA equalized gramophone record replay response
FIG50 RIAA equalization networks
To achieve the required frequency response curve, in either of the arrangements of 50a or 50b, …, while in the circuit layouts of FIGs 50c and 50d, RaRb = 12.40, Ca Ra = 297ns and C^b = 81.21|ns. In practice, component values from the preferred numerical series may often be found which will provide close approximations to the required ratios.
It is, of course, also possible to generate the required frequency response characteristics by the use of a series of separate CR networks, each chosen to generate a portion of the curve, and each isolated from the others by a suitable buffer stage, as shown in FIG51.
FIG51 Multiple stage RIAA equalization circuit
Low noise circuitry:
The reproduction of signals from vinyl (LP or EP) discs can provide particular problems in attaining a desirably high signal to noise (s/n), ratio since many gramophone pick-up cartridges have very low output voltage levels. A normal good quality moving magnet pick-up cartridge, such as the Shure M75E, may have an output of 5mV for a groove modulation of 5cm/s at 1kHz, but the higher quality Shure V15/5 cartridge has an output of only 3m V for the same groove modulation velocity. This input signal level may need to be amplified to 0.77V RMS in order to drive the audio power amplifier fully, and a s/n ratio of better than +65dB is typically sought. This requires that the total mains induced hum and noise, from all sources, must not exceed 1.7 microvolts, referred to the input. This requires that the screening of the input leads shall be adequate, that care will be taken to avoid hum inducing earth loops, and also that the input circuitry is chosen so that the minimum amount of circuit and component noise is introduced.
Understandably, this problem is aggravated when very low output voltage pick-up cartridge types are used, such a those using moving-coil generator systems, which can give output voltages as low as 70 microvolts/cm/s groove velocity, where a 65dB s/n ratio would require an input hum and noise level of less than some 40 nanovolts.
A typical contemporary gramophone pick-up input stage, using audio quality operational amplifier ICs, is illustrated in FIG52, in which the required frequency response shaping is accomplished by R3-R4 and C4-C3 in the feedback network. This method of connection is called 'series' feedback, and is usually chosen because it allows the impedance values in both the input circuit and the feedback network to be kept low, with consequently low thermal noise values. The input impedance presented to this circuit is that due to the cartridge load resistor, R1, and the pick-up coil connected in parallel with this -- which could have an inductance of 1H and a coil winding resistance of 2k ohms. At low frequencies the input circuit impedance due to these component values will be about 1k9 ohms, but as the frequency increases, so the impedance due to the pick-up coil inductance will begin to dominate, and the input impedance and the thermal noise due to this will begin to increase.
FIG52 Typical high quality input RIAA equalization circuit
In the case of the shunt feedback connection, shown in FIG53, the input impedance seen by the circuit is that due to both the pick-up coil and the load resistor, R1, in series with it, giving a higher impedance at low audio frequencies, and a higher thermal noise level in this part of the audio band. However, in this case, as the operating frequency is increased, so the gain of the circuit, in respect of the input noise component, will decrease. This leads to a difference in thermal noise characteristics between these two circuits, in this application, in which that of the shunt feedback circuit is more of a 'rustle', whereas that of the series arrangement is a higher pitched hiss.
FIG53 Shunt feedback RIAA equalization stage
Overall, the relative performance of these two circuit arrangements, in respect of thermal noise, favors the series feedback layout, below about 5kHz, and it’s therefore the preferred form in commercial use.
The somewhat inaccurate adherence of this type of circuit -- in that part of its frequency response above 5kHz -- to the frequency response characteristics defined by the RIAA specification, shown in FIG50, due to the fact that the stage gain asymptotes to unity rather than zero, at HF, can be corrected by a small additional R/C lag circuit, Rs-C5. Alternatively, the error may be minimized by ensuring that the closed-loop gain is high, by keeping the impedance of the feedback limb high in relation to R2. This aspect of circuit design is considered in greater detail in Section 6.
In general, the target of low circuit noise is achieved by keeping all of the circuit impedances as low as possible, especially at the input of the circuit; by trying to ensure that the gain of the input stage is sufficiently high that noise introduced by later stages will be small in comparison with the input noise component; and by careful choice of components.
In the case of bipolar transistors, the input noise is mainly that due to the 'base spreading resistance' -- the apparent resistance between base and emitter -- and this will depend on the input base current and the effective area of the base-emitter junction. This can be minimized by increasing the effective base junction area, either by the use of a transistor intended for use at higher power levels, or by connecting a number of small-signal transistors in parallel, as shown in FIG54.
MC PU input 10µF
FIG54 Low-noise PU head amplifier using parallel connected input transistors
There are a number of commercially available integrated circuits, such as the LM394, in which a closely matched pair of transistors is made up from a large number of parallel-connected small-signal bipolar de vices, with the specific aim of achieving a very low input base spreading resistance, and a consequent very low input noise.
The other major noise sources are those due to 'shot' noise -- effectively that due to the random arrival of electrons at the collector -- which is statistically more important as the current through the circuit is reduced, and ?/ / or 'flicker' noise, due to the random choice of current paths through the device, and this is a mechanical feature of device construction, and will vary from one type of device to another.
So, in constructing low-noise circuitry, the actual specification of the devices to be used should be examined as well as the intended circuit layout. For example, some transistors have a better noise specification, for the same operating current value, than others, and some resistors -- such as metal film -- have a lower flicker noise figure than, for example, carbon composition types, because film types have a smaller cross sectional area of conducting path.
Two design examples of low noise input stages, intended for use with low impedance moving coil gramophone pick-up cartridges, are shown, schematically, in FIGs 55 and 56.
Similar considerations apply to the design of micro-phone amplifiers, and also in the design of the input stages of cassette tape recorders, which have a very low output voltage, typically in the range 500µF to 2mV RMS max., because of the very low tape speed, and the very narrow head pole-piece gaps, necessary to reproduce high audio frequencies.
A useful design feature, employed in many commercial cassette player systems, is to arrange that the amplifier circuit is muted while the tape is stationary, so that the user is never made aware of the amplifier background noise, which will be more audible in the absence of the reproduced audio signal.
Integrated circuit systems During the past two decades, there has been a lot of activity on the part of semiconductor manufacturers aimed at the production of 'application specific' integrated circuits, intended to perform most of the functions which are required for audio preamplifier use. These have found ready application in low-cost Hi-Fi modules, but, in general, not only has the performance specification for these ICs been relatively low, but they have tended to become obsolete, and to be replaced by later, not necessarily compatible de signs, after a relatively brief period of availability.
For this reason, there has been an increasing trend among audio system designers to base their circuitry upon IC operational amplifier gain blocks, of which there are an increasing number having an exceedingly high performance specification, coupled with a relatively low unit cost. This approach allows adequate flexibility in design and also allows the system to be upgraded, with great ease, as newer, or improved performance IC gain blocks become available.
FIG55 Head amplifier using LM394 transistor array.
FIG56 Cascode input moving coil head amplifier.
Prev. | Next