A better topology and a non-inverting version
Before reading this article, I strongly recommend the reader to have a look at Rod Elliot's excellent article on the subject of volume controls in general.
Volume controls have been around for as long as electronic amplifier based audio reproduction has existed, but surprisingly little attention seems to have been paid to their design. They are certainly the most often used control on any audio system, and while they seem to be becoming more and more confined to the digital domain in most modern systems, a potentiometer based analogue volume control, in the opinion of the author, is still the best solution in terms of performance and lack of complication.
An active volume control (here upon to be referred to as an AVC) is essentially a variable gain amplifier with a minimum gain of zero and a maximum gain usually in the region of 20dB or so in a normal preamp. To qualify as an AVC, the variable gain must be implemented within a feedback loop as opposed to a static gain amplifier with a variable attenuator on the input. Such volume controls offer many benefits compared to the latter more common configuration. A logarithmic control is also inherent to the inverting 'see-saw' feedback design, far lower noise is easily achieved, and if designed correctly, the component count can actually be lower their rather crude alternatives, requiring just one op-amp.
In order to achieve full attenuation of the input, it is important that the amplifier at the heart of the AVC has a very low output impedance in comparison to the potentiometer used in situ. This makes op-amps ideal candidates for the active part of the volume control.
The very first AVCs were based around valve output stages, usually in high quality FM (known as VHF during the valve era) receivers. At the time designers had to make the best use of all the gain available to them, having a limited number of comparatively expensive, low gain, and rather non-linear amplifying devices to work with. This gain was a valuable commodity, and had to be used as efficiently as possible. Simply using an attenuator based volume control would be very wasteful of the limited amount of gain available in the audio stages, usually a common cathode triode amplifier followed by a single ended pentode output stage. The gain wasted by an attenuate and amplify volume control could otherwise be used to greatly enhance the poor frequency response and relatively high harmonic distortion of the audio amplifier to a level acceptable for high quality FM reception through the use of negative feedback.
Figure 1. A valve output stage based AVC
Figure 1 shows the solution to this historical problem. The first AVC! With its low output impedance of an ohm or two coupled with the necessity for negative feedback, a single ended pentode output stage is a suitable amplifier to use as the core of an AVC. By configuring the polarity of the low impedance side of the output transformer so that the stage acts as an inverting amplifier. Feedback can then be applied through the standard non-inverting network, in this case a variable one made up of a simple potentiometer, R1, one side being connected to the output, the other to the input and the wiper to the the output valve's grid, which behaves as a rather non-ideal virtual earth, giving a gain just under unity at its centre position. A high value potentiometer on the order of 1MΩ was necessary so as not to overload the pre-amplifier stage and also to give an acceptable degree of attenuation at the minimum setting. There was no need for any other components to limit the maximum gain which was set by the relatively low open loop gain of the simple valve output stage. R3 and R2 were often left out due to economic reasons, but this often caused crackle as the potentiometer was turned as the set aged.
This configuration allowed strong signal levels from FM detection as well as strong stations in the AM band to be amplified with more negative feedback and hence lower distortion along with greater frequency response. When tuning to weaker AM stations, the open loop gain was employed more so for amplification as opposed to negative feedback, providing a less ideal amplifier, but one with enough gain to bring the weak station to an acceptable level. These effects usually didn't matter as the weaker stations were already somewhat polluted with noise and other undesirables, meaning that the higher distortion and narrower frequency response of the audio amplifier stage was less noticeable.
It is important to note that while the noise performance of valve output stages was improved using this topology, the real reason for its use was to take advantage of the otherwise wasted gain for negative feedback, thus improving the quality of the amplification provided.
Let's fast forward to the 21st century, where high quality op-amps and very low noise digital audio sources are readily available. Now the onus is not so much on reducing distortion, which when using such devices is already vanishingly low, but more so on diminishing noise levels to take full advantage of the seemingly unappreciated dynamic range of digital audio. With this in mind, it becomes immediately apparent that the standard attenuator, followed by an amplification stage of some 20dB, yields woefully inadequate noise performance at all settings other than those at or close to 'fully open', as the input noise of the amplifier stage is also amplified by 20dB. Another common approach is the amplifier stage followed by an attenuator, but now the headroom of the stage is profoundly compromised. A practical circuit using this layout will not be able to cope with the output of a CD player or other digital audio source without the stage clipping at least 3dB below maximum output of the ADC.
Many designs try to achieve a compromise between the restrictions of each configuration by mixing the two, for example; an amplifier stage with 10dB of gain followed by an attenuator, and then another 10dB amplifier, bringing the total gain up to 20dB with the potentiometer at its maximum setting. This still suffers from the restraints of the amplify-and-attenuate and attenuate-and-amplify based volume controls, albeit to a lesser extent, while requiring two amplifier stages.
Clearly, a better solution is required.
Figure 2. A conventional AVC
Figure 2 shows the classic active volume control that can be seen on the web. After a brief glance it becomes clearly evident that this circuit is at its heart a solid state re-hash of the valve based AVC, although using far more ideal components. U1.2 is configured as an inverting amplifier stage, while U1.1 buffers the input so as to simulate the high impedance an output pentode's grid, which allows resistors R3 and R4 to be made low in value and hence reduce thermal noise. The maximum gain is set by the gain of the inverting stage, which is typically 20dB, so as to bring a minimum 100mV line voltage to the 1V required to drive most power amplifiers to their fullest extent. C1 and C2 are needed if the op-amp is a bipolar input type so as to avoid crackling noises as the control is turned, caused by the changing resistance that the potentiometer's wiper meets as it is rotated across the granular resistive track. Capacitor C3 is also required to prevent instability. This component is non-negotiable. Leaving it out will most certainly cause oscillation due to the extra phase shift induced by the buffer stage.
Even in this implementation, the AVC offers great benefits, with noise performance up to 14dB or so (theoretically 17dB with noiseless resistors, assuming a maximum gain of 20dB and the same amplifying devices are used) quieter than an attenuate-and-amplify based volume control. There is also the added advantage, as in all AVCs, that a linear potentiometer can be used to obtain a logarithmic characteristic in response to the rotation of the control that is far superior to the standard 'logarithmic' potentiometers. These devices usually consist of either 2, or 3 in the case of more expensive types, resistive sections of track joined in the centre position to form a rather crude approximation of a logarithmic characteristic which results in a disconcerting jump in gain as the the control is turned past it's centre point. Similarly to the valve based circuit, the gain is just under unity at the centre position, due to the limited gain of the amplifier stage.
The AVC shown in Figure 2 is inherently an inverting type, which is more a benefit than a problem as the full audio path which AVCs typically form a part of often contains a second inverting stage upstream of the AVC, usually a tone control, which allows an inverting AVC to bring the the signal back into phase again through a second inversion. However, another inverting stage may not be present in the signal path, such as in the case of some very simple pre-amplifiers which shun tone controls. Tone controls are usually left out on the dubious pretext of preserving 'signal integrity' or some other technical sounding but usefully non-definable quality that would be lost should the signal pass through a tone control circuit set to any position. There are also level control applications, a recent example of my own being a simple pre-amplifier stage ahead of a digital audio interface for making transfers off legacy formats, where an AVC gave the best noise performance but maintaining the original phase was important.
One way of keeping the output in phase with the input would be to precede the AVC with a unity gain inverter, a function which the tone controls in a conventional HiFi pre-amplifier would bear the purpose of. This stage must be placed upstream of the AVC to retain the best possible noise performance at lower gain levels and also to avoid excessive loading of the AVC amplifier itself which will already be driving fairly low impedances as a result of lowering resistor values to further decrease noise levels. This extra inverting stage will cost another op-amp, increasing complexity, burden the power supply with more quiescent current, and will also inevitably degrade noise performance.
Figure 3. A non-inverting AVC
By placing the gain of the AVC into the buffer stage instead of the inverting stage, through the addition of feedback resistors R3 and R4, a non inverting full range output can now be taken from U1.1. The inverting stage, U1.2, now reduced to unity gain, is still necessary to realise the negative feedback across the potentiometer. This requires an extra 2 resistors to refashion the buffer stage into a gain stage, but no additional op-amps are necessary to retain absolute phase. C2 and possibly C1 can be omitted should JFET input op-amps be used.
The noise performance is slightly worse than the circuit depicted in Figure 2 at sub unity gain as inverting stage and the noise that it generates is now in the feedback loop relative to the output. The penalty is no more than approximately 2dB in a practical application, which will vary according to your choice of op-amp. Like the classic circuit shown in Figure 2, the gain is still shy of unity at the centre position. The input impedance also drops down to 900Ω when the potentiometer is rotated fully clockwise, meaning that a buffer will be vital should the circuit need to be connected directly to a line level input.
Thus far both of the solid state AVCs in this article have been derived, more so in the case of the former than the latter, from the valve output stage circuit used in the 1950s valve era. In doing so, they need to use 2 op-amps, a buffer followed by an inverting amplifier, to effectuate a gain limited, low noise inverting amplifier with a high input impedance. By starting again from scratch, it is possible to design an AVC that only uses only 1 op-amp, which brings about further reductions in complexity and noise.
Figure 4. Improved AVC topology
Figure 4 unveils the product of this labour, a single op-amp AVC. Instead of using the gain of the inverting amplifier stage, limited to the maximum desired gain, to determine the greatest level of amplification that the AVC can provide, a series resistance ahead of the potentiometer in the form of R2 is now used. The potentiometer is then simply connected as the feedback network of a single op-amp inverting amplifier stage.
A few extra components are necessary to ensure nothing nasty happens should any real-world problems present themselves, along with the usual DC decoupling capacitors C1 and C2, which again can be ignored if a JFET input op-amp is used. Resistor R4 prevents the output shooting up to the supply rails should the potentiometer's wiper go open circuit, caused by the voltage generated by the bias current, or offset voltage in the case of a JFET input op-amp being amplified by the op-amp's full open loop gain. This happens more than one would think, due to dirt particles accumulating over time on the resistive track as well as general wear as the potentiometer ages, causing a series of ear-splitting clicks and bangs as an old or dirty potentiometer is turned. When this condition occurs, R4 resistively couples the output onto the inverting input, turning the AVC into a follower referenced to ground, eliminating this highly undesirable effect. As the resistance of R4 is much greater than that of the potentiometer, it does not interfere significantly with the normal operation of the AVC.
By now the reader may have observed that the effects of R2 in series with the input side of the potentiometer, and R4 in parallel with the output side, will conspire to reduce the maximum gain of the AVC to 19.2dB, while also bringing the gain at centre position to 2dB less than unity. Along with the circuit's inherent inversion of absolute phase, when used in a full pre-amplifier design, this is not necessarily a problem as there is the possibility of a stage ahead of the AVC, such as a balance control, exhibiting a gain higher than unity and making up for this. However, if a such a stage is not going to be present, then it is possible to compensate for this, if only for aesthetics. R3 fulfills this role, by adding a parallel resistance across the input side of the potentiometer, it decreases the effective resistance of the first half of the 'see-saw', increasing the gain. The value is adjusted so that unity gain is achieved at the potentiometer's centre position, but the maximum gain is still limited to 19.2dB. The author does not consider this to be a crucial problem.
In using just one op-amp for amplification, the amplifier stage's contribution to the noise floor is reduced by over 3dB in comparison to the classic topology of Figure 2, if the same devices are used. Eliminating the resistor noise of the gain limiting inverting feedback stage, resistors R3 and R4 in Figure 2, also helps to lower noise further, while simultaneously reducing the load on the driving op-amp. The sole nit-picking disadvantage of the new topology is that the maximum gain is not independent from the absolute value of the potentiometer. The effect of an out of tolerance potentiometer only becomes apparent once the control has passed the 9 'o clock position, causing little difference at centre position. In any case the factors that affect this, such as wear over time and batch tolerances, will be almost equal in each section of a dual gang potentiometer used for a stereo pre-amplifier. It is clear that the benefits of the new topology outweigh this minuscule disadvantage.
At this point, all of the circuits presented have been shown only in isolation, but reference has been made several times to their use as part of a larger pre-amplifier. In the interests of demonstrating the correct application of the circuits described so far in this article, I have designed a simple yet very practical stereo pre-amplifier using the single op-amp AVC as the last stage in a 3 stage pre-amplifier. This design showcases what can be achieved through the efficient use of op-amps, with each device offering a control function, including a balance, bass, treble, and of course volume control. Although it is shown for demonstration purposes in this article, the pre-amplifier below will give very good real world performance, with very low noise (being an example, I have not had time to build and test it to check the noise level).
Figure 5. A full preamp using the improved AVC topology
Starting with the input stage, R1 and R2 combine in parallel to produce an input impedance of 50kΩ at audio frequency, a nice round figure that will agree with virtually all available equipment bar some poorly designed valve sources. The DC resistance of the line input is the value of R1, 100kΩ. C1 and R2 perform non-polarised DC decoupling with a time constant of 100mS, the lowest de-coupling constant in the circuit. It is important that this constant is set using a high quality polypropylene or polystyrene, but not polyester and certainly not electrolytic, dielectric capacitor to minimise non-linear distortion near and at the cut-off frequency. Polyester capacitors can exhibit small amounts, and electrolytic capacitors can generate quite dire levels, of harmonic distortion in this region. Resistor R3 forms a somewhat crude, but very effective RF filter with the input capacitance of U1.1. A value of 1kΩ gains a good compromise, retaining a good level of rejection of RF hash, even in the worst of cases, while still being low enough to avoid contributing excessive thermal noise to the line input.
While buffering the line input so as to drive the following tone control stage, U.1.1 also functions as an active stereo balance control. A dual potentiometer, R5, connected in an opposite direction on each channel and placed in the feedback loop of a non-inverting amplifier makes a fine balance control, with the addition of resistors R4 and R6 to place a limit on the maximum gain of the control at its extremes. Like all the resistors which could potentially contribute noise in the preamp, these resistors must be kept low in value in order to preserve the noise performance of the stage. As a non-inverting amplifier, it operates at a gain of greater than unity, of 2dB with the balance control at centre position. This little bit of extra gain, is helpful in that it elevates the signal level slightly and thus improves the effective signal to noise ratio of the following stages without overly compromising headroom, which will be more than adequate with ±15V supply rails. The astute reader may also notice another benefit which will become useful further down the signal path…
Adjusting the balance control to either left or right will increase the gain in that respective channel by up to, as opposed to cutting the opposing channel, due to the fact that non inverting op-amp configurations are unable to realise a gain of less than unity. This feature is more useful for dealing with real world balance issues, rare though they may be outside the realms of analogue tape, which are almost always caused by a weak channel, rather than an overly strong one. A total range of +9dB is available, with the cut channel reducing to unity, which is more than enough as any balance issue requiring more would be indicative of a serious fault somewhere along the signal chain, requiring correction through repair and not a balance control.
Next up is the tone control stage, using the classic active 'Baxandall' topology with U2.1 as the active element. There are plenty of good explanations of how this configuration works elsewhere on the web and I won't bore the reader with my own, especially in the context of an application circuit for an AVC. This common feedback based tone control has an inverting function and unity gain with the controls set to flat. It is shown here in its simplest form, using one capacitor for the treble and bass time constants, whose adjustment is set by R8 and R12 respectively. A JFET input op-amp must be used for U2.1 to make certain that the controls don't modulate the bias current and therefore crackle disconcertingly when they are turned. Adding DC decoupling networks is impractical for this topology and would bring serious complication to the design in this case. C4 furnishes the tone control with HF compensation to prevent any potential instability.
The tone control's bass and treble turnover points are not set at the usual 1kHz which tends to be of little help in practical situations, such as baffle step correction, or adjusting for different playback compensation on pre-RIAA records, but at at a much more useful 480Hz and 2.1kHz. A range of ±12dB is permissible, in contrast to the more common ±20dB, as too much range limits the accuracy of the adjustment via the panel control and also places a higher driving load on the op-amp. Higher loading would have to be compensated for by increasing resistor values and therefore affecting the noise performance of the stage. In any case, echoing a similar sentiment to that expressed in the description for the balance control, a defect which needs 20dB of correction is one that needs to be cured with practical repair.
Finally, at the end of the signal chain, is the improved AVC, almost identical to the one in Figure 4, that this article focuses on. However, one component is now missing. In Figure 4, resistor R3 secures the centre position gain to unity, where it would otherwise be -2dB. As formerly alluded to, the balance control at the front end of the preamp exhibits an advantageous gain of +2dB at its centre setting. Due to this, the gain compensation resistor (R3 in Figure 4) can now be omitted, as the compensation is now established further up the signal chain. This also allows the whole preamp to reap an extra 2dB of gain at maximum, culminating to a healthy 21.2dB boost at 'fully open'.
A BJT input op-amp is shown as the active element of the volume control in this example, requiring the perfunctory electrolytic capacitors C5 and C6 to prevent the control from crackling as it is turned. These can be left out if a suitably low noise JFET input op-amp such as the OPA2134, or the even better OPA1642 is chosen. Using a BJT op-amp will generally improve the noise performance at the lower volume control settings, as they tend to have a lower voltage noise than JFET input op-amps, but will be noisier than a JFET amplifier at the middle and higher settings due to input current noise. The choice depends mainly on the designer's preference, as well as the noise characteristics the op-amps available and practical for his use. I have chosen to use a BJT op-amp in this example as it means that proper decoupling techniques can be also shown for the sake of education.
The pre-amplifier is completed with an unexceptionable output decoupling network. C8 and R17 form a DC decoupling network with a time constant of 1s with 10kΩ loading on the line output, a load easily driven by the AVC stage. As a rule of thumb, electrolytic capacitor based DC decoupling should use a time constant of no less than 1s at line level so as to reduce the non-linear distortion of the eletrolytic capacitor at low frequencies to negligible levels. R18 safeguards the driving stage from any instability caused by capacitive loading from of the connecting cable, while setting the output impedance to a round 100Ω.
The AVC adds a phase inversion to the signal again, after the tone control, bringing the preamp's output into absolute phase with its input. To obtain the best noise performance, it is prudent to leave the AVC as the last stage of the pre-amplifier. Doing so preserves the AVC's very low noise at lower volume settings so that it is not jeopardized by the almost certainly equal or higher noise levels of a following stage, such as a tone or balance control.
The reader may also have noticed that all 4 of the potentiometers needed for this preamp are usefully common 10kΩ dual linear types, while all the resistors occupy only 6 common values. It is the opinion of the author that it is well worth paying attention to designing with as few different component types as possible, which makes component sourcing and construction pleasingly uncomplicated, while significantly reducing the chance of any component misplacement due to human error. Although shown only for example, the pre-amplifier in Figure 5 would make a very good project.
Hopefully this article will have given the reader an adequate knowledge of AVCs and their implementation, along with a little extra, so that he can go about designing and building his own. These circuits are surprisingly versatile and can be dedicated to a great many applications with a little creative thinking. Even when using lower performance components such as the common TL072, or discrete amplifiers should the designer be so inclined, superior noise performance to conventional volume controls using much more state-of-the-art op-amps is in easy reach. It is the view of the author, after further reviewing the circuits in this article, that there is no longer any benefit in reduced complexity to using a passive volume control with gain over an AVC, let alone in terms of performance.
The importance of looking at the bigger picture is also shown here. The characteristics of the stages preceding the AVC can be taken advantage of to correct the inherent characteristics of the AVC, therefore allowing any components that would apply this correction in the AVC itself to be foregone, improving the overall simplicity of the design. Being aware of these happy coincidences as opposed to looking at each element as individual, and separate from it's adjacent stages, gives the designer great ability to obtain as best performance, and as many features, as possible with the lowest component count and therefore board space and cost. This is especially true in audio design as superfluous stages not only add unnecessary complication, but also degrade the objective performance of the whole unit.
If you enjoyed reading this article, have any constructive comments or know any more about the history of this circuit, then please feel free to drop me an e-mail. I always look forward to intelligent discussions!
Michael Fearnley 2016