Audio design has for many years relied on a very small number of opamp types; the TL072 and the 5532 dominated the audio small-signal scene for many years. The TL072, with its JFET inputs, was used wherever its negligible input bias currents and low cost were important. For a long time the 5534/5532 was much more expensive than the TL072, so the latter was used wherever feasible in an audio system, despite its inferior noise, distortion, and load-driving capabilities. The 5534 was reserved for critical parts of the circuitry. Although it took many years, the price of the 5534 is now down to the point where you need a very good reason to choose any other type of opamp for audio work.
The TL072 and the 5532 are dual opamps; the single equivalents are TL071 and 5534. Dual opamps are used almost universally, as the package containing two is usually cheaper than the package containing one, simply because it is more popular.
There are, however, other opamps, some of which can be useful, and a selected range is covered here.
The opamp is today thought of as quintessentially a differential amplifier, responding to the difference of the input voltages while (hopefully) ignoring any common-mode component. The history of differential amplifiers goes back to that great man, Alan Blumlein, and his 1936 patent  for a pair of valves with their cathodes connected to ground through a common resistor. However, the first valve-based operational amplifiers, i.e. those intended to be capable of performing a mathematical operation, were in fact not differential at all, having only one input. That had to be an inverting input, of course, so you could apply negative feedback.
The first opamp to get real exposure in the UK was the Fairchild μA709, designed by the great Bob Widlar and introduced in 1965. It was a rather awkward item which required quite complicated external compensation and was devoid of output short-circuit protection. One slip of the probe, and an expensive IC was gone. It was prone to latch-up with high common-mode voltages and did not like capacitive loads. I for one found all this most discouraging and gave up on the 709 pretty quickly. If you’re going to quit, do it early, I say.
The arrival of the LM741 was a considerable relief. To my mind, it was the first really practical opamp, and it was suddenly possible to build quite complex circuitry with a good chance of it being stable, doing what it should do and not blowing up at the first shadow of an excuse. I have given some details of it in this chapter for purely historical reasons. There is also an interesting example of how to apply the LM741 appropriately in Chapter 15.
The first IC opamps opened up a huge new area of electronic applications, but after the initial enthusiasm for anything new, the audio market greeted these devices with less than enthusiasm. There were good reasons for this. The LM741 worked reliably; a snag with using it for audio was the leisurely slew rate of 0.5 V/μs, which made full output at 20 kHz impossible, but a bigger drawback was that it was obviously noisy. Up until about 1978, the only way to obtain good and cost-effective performance in a preamp was to stick with discrete transistor Class-A circuitry, and this became recognised as a mark of high quality. The advent of the TL072 and the 5532 changed this situation completely, but there is still sometimes marketing cachet to be gained from a discrete design.
An excellent and detailed history of operational amplifiers can be found in .
There is no point in regurgitating manufacturers’ data sheets, especially since they are readily available on the internet. Here I have simply ranked the opamps most commonly used for audio in order of voltage noise (Table 4.1).
The great divide is between JFET input opamps and BJT input opamps. The JFET opamps have more voltage noise but less current noise than bipolar input opamps, the TL072 being particularly noisy. If you want the lowest voltage noise, it has to be a bipolar input. The difference, however, between a modern JFET-input opamp such as the OPA2134 and the old faithful 5532 is only 4 dB; but the JFET part is a good deal more costly. The bipolar AD797 seems to be out on its own here, but it is a specialised and expensive part. The LT1028 is not suitable for audio use because its bias compensation makes it noisy. The LM741 has no noise specs on its data sheets, and the 20 nV/rtHz is from measurements.
Several important new opamps have been released recently, the OPA828, OPA2156, OPA1622, and AD8656, which give lower noise and distortion in some applications. Another addition is the OPA1671, which is very economical of supply current but is only available in a single package. They have been added to the tables. (NB, the AD8655 is the single version of the AD8656 dual opamp.) The OPA1622 is notable for being able to drive loads down to 32 Ω at very low distortion but with the maximum output limited to 2 Vrms.
Both voltage and current noise increase at 6 dB/octave below the 1/f corner frequency, which is usually around 100 Hz. The only way to minimise this effect is to choose an appropriate opamp type.
Opamps with bias cancellation circuitry are normally unsuitable for audio use due to the extra noise this creates. The amount depends on circuit impedances and is not taken into account in Table 4.1. The general noise behaviour of opamps in circuits is dealt with in Chapter 1.
|Opamp||en nV/rtHz||in pA/rtHz||Input device type||Bias cancel?|
* No current-noise spec given by manufacturer’s data sheet. Presumed negligible.
Slew rates vary more than most parameters; a range of 100:1 is shown here in Table 4.2. The slowest is the LM741, which is the only type not fast enough to give full output over the audio band. There are faster ways to handle a signal, such as current-feedback architectures, but they usually fall down on linearity. In any case, a maximum slew rate greatly in excess of what is required appears to confer no benefits whatever.
The 5532 slew rate is typically +/-9 V/μs. This version is internally compensated for unity-gain stability, not least because there are no spare pins for compensation when you put two opamps in an 8-pin dual package. The single-amp version, the 5534, can afford a couple of compensation pins, and so is made to be stable only for gains of 3x or more. The basic slew rate is therefore higher at +/-13 V/μs. Note the remarkable slew rate of the OPA828; it is technically impressive but gives no special benefits in audio use.
Compared with power-amplifier specs, which often quote 100 V/μs or more, these opamp speeds may appear rather sluggish. In fact they are not; even +/-9 V/μs is more than fast enough. Assume you are running your opamp from +/-18 V rails and that it can give a +/-17 V swing on its output. For most opamps, this is distinctly optimistic, but never mind. To produce a full-amplitude 20 kHz sine wave, you only need 2.1 V/μs, so even in the worst case, there is a safety-margin of at least 4 times. Such signals do not of course occur in actual use, as opposed to testing. More information on slew limiting is given in the section on opamp distortion.
This is simply the range over which the inputs can be expected to work as proper differential inputs. It usually covers most of the range between the rail voltages, with one notable exception. The data sheet for the TL072 shows a common-mode (CM) range that looks a bit curtailed at -12 V. This bland figure hides the deadly trap this IC contains for the unwary. Most opamps, when they hit their CM limits, simply show some sort of clipping. The TL072, however, when it hits its negative limit, promptly inverts its phase, so your circuit either latches up or shows nightmare clipping behaviour with the output bouncing between the two supply rails. The positive CM limit is, in contrast, trouble free. This behaviour can be especially troublesome when TL072s are used in high-pass Sallen and Key filters.
A perfect opamp would have its output at 0 V when the two inputs were exactly at the same voltage. Real opamps are not perfect, and a small voltage difference –usually a few millivolts –is required to zero the output. These voltages are large enough to cause switches to click and pots to rustle, and DC blocking is often required to keep them in their place.
The typical offset voltage for the 5532A is ±0.5 mV typical, ±4 mV maximum at 25°C; the 5534A has the same typical spec but a lower maximum at ±2 mV. The input offset voltage of the new LM4562 is only ±0.1 mV typical, ±4 mV maximum at 25°C.
Bipolar-input opamps not only have larger noise currents than their JFET equivalents, they also have much larger bias currents. These are the base currents taken by the input transistors. This current is much larger than the input offset current, which is the difference between the bias current for the two inputs. For example, the 5532A has a typical bias current of 200 nA, compared with a much smaller input offset current of 10 nA. The LM4562 has a lower bias current of 10 nA typical, 72 nA maximum. In the case of the 5532/4, the bias current flows into the input pins as the input transistors are NPN.
Bias currents are a considerable nuisance when they flow through variable resistors; they make them noisy when moved. They will also cause significant DC offsets when they flow through high-value resistors.
It is often recommended that the effect of bias currents can be cancelled out by making the resistance seen by each opamp input equal. Figure 4.1a shows a shunt-feedback stage with a 22 kΩ feedback resistor. When 200 nA flows through, this it will generate a DC offset of 4.4 mV, which is rather more than we would expect from the input offset voltage error.
If an extra resistance Rcompen, of the same value as the feedback resistor, is inserted into the non-inverting input circuit, then the offset will be cancelled. This strategy works well and is done almost automatically by many designers. However, there is a snag. The resistance Rcompen generates extra Johnson noise, and to prevent this, it is necessary to shunt the resistance with a capacitor, as in Figure 4.1b. This extra component costs money and takes up PCB space, so it is questionable if this technique is actually very useful for audio work. It is usually more economical to allow offsets to accumulate in a chain of opamps and then remove the DC voltage with a single output-blocking capacitor. This assumes that there are no stages with a large DC gain and that the offsets are not large enough to significantly reduce the available voltage swing. Care must also be taken if controls are involved, because even a small DC voltage across a potentiometer will cause it to become crackly, especially as it wears.
FET input opamps have very low bias current at room temperature; however, it doubles for every 10°C rise. This is pretty unlikely to cause trouble in most audio applications, but a combination of high internal temperatures and high-value pots could lead to some unexpected crackling noises.
While it may not appear on the datasheet, the price of an opamp is obviously a major factor in deciding whether or not to use it.
Table 4.3 was compiled using prices for the SOT-8 surface-mount version, aggregated from a number of vendors such as Mouser, Digikey, RS, etc. Note that the price is per package and not per opamp section, and this can make a big difference. As always, things are cheaper if you buy more of them; at the 2,500 level, prices are about half what they are for a one-off. The significance of 2,500 is that in many cases, the devices come on a reel of that size. It is obviously only a rough guide, and YMMV. Purchasing in different countries may change the rankings somewhat, but the basic look of things will not alter too much.
The OPA627 is an outlier here; it is eye-wateringly expensive. This is probably because (as far as I can determine) it is the only laser-trimmed part here. Its DC precision is not useful in audio. One thing is obvious: the 5532 is still one of the great opamp bargains of all time.
DIL versions of the older opamps are still available, but the newer opamps are only manufactured in surface-mount packages. The OPA1622 is only available as a 10-pin VSON package –home soldering will be difficult.
Relatively few discussions of opamp behaviour deal with non-linear distortion, perhaps because it is a complex business. Opamp “accuracy” is closely related, but the term is often applied only to DC operation. Accuracy here is often specified in terms of bits, so “20-bit accuracy” means errors not exceeding one part in 2 to the 20, which is -120 dB or 0.0001%. Audio signal distortion is of course a dynamic phenomenon, very sensitive to frequency, and DC specs are of no use at all in estimating it.
Distortion is always expressed as a ratio and can be quoted as a percentage, as number of decibels, or in parts per million. With the rise of digital processing, treating distortion as the quantisation error arising from the use of a given number of bits has become more popular. Figure 4.2 hopefully provides a way of keeping perspective when dealing with these different metrics.
There are several different causes of distortion in opamps. We will now examine them.
This is what might be called the basic distortion produced by the opamp you have selected. Even if you scrupulously avoid clipping, slew-limiting, and common-mode issues, opamps are not distortion free, though some types such as the 5532 and the LM4562 have very low levels. If distortion appears when the opamp is run with shunt feedback, to prevent common-mode voltages on the inputs, and with very light output loading, then it is probably wholly internal, and there is nothing to be done about it except pick a better opamp.
If the distortion is higher than expected, the cause may be internal instability provoked by putting a capacitative load directly on the output or neglecting the supply decoupling. The classic example of the latter effect is the 5532, which shows high distortion if there is not a capacitor across the supply rails close to the package; 100 nF is usually adequate. No actual HF oscillation is visible on the output with a general-purpose oscilloscope, so the problem may be instability in one of the intermediate gain stages.
While this is essentially an overload condition, it is wholly the designer’s responsibility. If users whack up the gain until the signal is within a hair of clipping, they should still be able to assume that slew limiting will never occur, even with aggressive material full of high frequencies.
Arranging this is not too much of a problem. If the rails are set at the usual maximum voltage, i.e. ±18 V, then the maximum possible signal amplitude is 12.7 Vrms, ignoring the saturation voltages of the output stage. To reproduce this level cleanly at 20 kHz requires a minimum slew rate of only 2.3 V/μs. Most opamps can do much better than this, though with the OP27 (2.8 V/μs), you are sailing rather close to the wind. The old LM741 looks as though it would be quite unusable, as its very limited 0.5 V/μs slew rate allows a full output swing only up to 4.4 kHz.
Horrific as it may now appear, audio paths full of LM741s were quite common in the early 1970s. Entire mixers were built with no other active devices, and what complaints there were tended to be about noise rather than distortion. The reason for this is that full-level signals at 20 kHz simply do not occur in reality; the energy at the HF end of the audio spectrum is well known to be much lower than that at the bass end.
This assumes that slew-limiting has an abrupt onset as level increases, rather like clipping. This is in general the case. As the input frequency rises and an opamp gets closer to slew limiting, the input stage is working harder to supply the demands of the compensation capacitance. There is an absolute limit to the amount of current this stage can supply, and when you hit it, the distortion shoots up, much as it does when you hit the supply rails and induce voltage clipping. Before you reach this point, the linearity may be degraded, but usually only slightly until you get close to the limit. It is not normally necessary to keep big margins of safety when dealing with slew limiting. If you are employing the Usual Suspects in the audio opamp world –the TL072, the 5532, and the LM4562, with maximal slew rates of 13, 9, and 20 V/μs, respectively –you are most unlikely to suffer any slew-rate non-linearity.
Output stage distortion is always worse with heavy output loading because the increased currents flowing exacerbate the gain changes in the Class-B output stage. These output stages are not individually trimmed for optimal quiescent conditions (as are audio power amplifiers), and so the crossover distortion produced by opamps tends to be higher and can be more variable between different specimens of the same chip. Distortion increases with loading in different ways for different opamps. It may rise only at the high-frequency end (e.g. the OP2277), or there may be a general rise at all frequencies. Often both effects occur, as in the TL072.
The lowest load that a given opamp can be allowed to drive is an important design decision. It will typically be a compromise between the distortion performance required and opposing factors such as number of opamps in the circuit, cost of load-capable opamps, and so on. It even affects noise performance, for the lower the load resistance an amplifier can drive, the lower the resistance values in the negative feedback can be, and hence the lower the Johnson noise they generate. There are limits to what can be done in noise reduction by this method, because Johnson noise is proportional to the square root of circuit resistance and so improves only slowly as opamp loading is increased.
Thermal distortion is that caused by cyclic variation of the properties of the amplifier components due to the periodic release of heat in the output stage. The result is a rapid rise in distortion at low frequencies, which gets worse as the loading becomes heavier.
Those who have read my work on audio power amplifiers will be aware that I am highly sceptical –in fact, totally sceptical –about the existence of thermal distortion in amplifiers built from discrete components.  The power devices are too massive to experience per-cycle parameter variations, and there is no direct thermal path from the output stage to the input devices. There is no rise, rapid or otherwise, in distortion at low frequencies in a properly designed discrete power amplifier.
The situation is quite different in opamps, where the output transistors have much less thermal inertia and are also on the same substrate as the input devices. Nonetheless, opamps do not normally suffer from thermal distortion; there is generally no rise in low-frequency distortion, even with heavy output loading. Integrated-circuit power amplifiers are another matter, and the much greater amounts of heat liberated on the substrate do appear to cause serious thermal distortion, rising at 12 dB/octave below 50 Hz. I have never seen anything resembling this in any normal opamp.
This is the general term for extra distortion that appears when there is a large signal voltage on both the opamp inputs. The voltage difference between these two inputs will be very small, assuming the opamp is in its linear region, but the common-mode (CM) voltage can be a large proportion of the available swing between the rails.
It appears to be by far the least understood mechanism and gets little or no attention in opamp textbooks, but it is actually one of the most important influences on opamp distortion. It is simple to separate this effect from the basic forward-path distortion by comparing THD performance in series and shunt feedback modes; this should be done at the same noise gain. The distortion is usually a good deal lower for the shunt feedback case where there is no common-mode voltage. Bipolar and JFET input opamps show different behaviour, and they are treated separately in what follows. The test circuits used are shown in Figure 4.3.
Figure 4.4 shows the distortion from a 5532 working in shunt mode with low-value resistors of 1 kΩ and 2k2 setting, a gain of 2.2 times, at an output level of 5 Vrms. This is the circuit of Figure 4.3a with Rs set to zero; there is no CM voltage. The distortion is well below 0.0005% up to 20 kHz; this underlines what a superlative bargain the 5532 is.
Figure 4.5 shows the same situation but with the output increased to 10 Vrms (the clipping level on ±18 V rails is about 12 Vrms), and there is now significant distortion above 10 kHz, though it only exceeds 0.001% at 18 kHz.
This remains the case when Rs in Figure 4.3a is increased to 10 kΩ and 47 kΩ –the noise floor is higher, but there is no real change in the distortion behaviour. The significance of this will be seen in a moment.
We will now connect the 5532 in the series feedback configuration, as in Figure 4.3b; note that the stage gain is greater at 3.2 times, but the opamp is working at the same noise gain. The CM voltage is 3.1 Vrms. With a 10 Vrms output, we can see in Figure 4.6 that even with no added source resistance the distortion starts to rise from 2 kHz, though it does not exceed 0.001% until 12 kHz. But when we add some source resistance Rs, the picture is radically worse, with serious midband distortion rising at 6dB/octave and roughly proportional to the amount of resistance added. We will note it is 0.0085% at 10 kHz with Rs = 47 kΩ.
The worst case for CM distortion is the voltage-follower configuration, as in Figure 4.3c, where the CM voltage is equal to the output voltage. Figure 4.7 shows that even with a CM voltage of 10 Vrms, the distortion is no greater than for the shunt mode. However, when source resistance is inserted in series with the input, the distortion mixture of second, third, and other low-order harmonics increases markedly. It increases with output level, approximately quadrupling as level doubles. The THD is now 0.018% at 10 kHz with Rs = 47 kΩ, more than twice that of the series-feedback amplifier due to the increased CM voltage.
It would be highly inconvenient to have to stick to the shunt feedback mode, because of the phase inversion and relatively low input impedance that comes with it, so we need to find out how much source resistance we can live with. Figure 4.8 zooms in on the situation with resistance of 10 kΩ and below; when the source resistance is below 2k2, the distortion is barely distinguishable from the zero source resistance trace. This is why the low-pass Sallen and Key filters in Chapter 6 have been given series resistors that do not in total exceed this figure.
Close examination reveals the intriguing fact that a 1 kΩ source actually gives less distortion than no source resistance at all, reducing THD from 0.00065% to 0.00055% at 10 kHz. Minor resistance variations around 1 kΩ make no difference. This must be due to the cancellation of distortion from two different mechanisms. It is hard to say whether it is repeatable enough to be exploited in practice.
So, what’s going on here? Is it simply due to non-linear currents being drawn by the opamp inputs? Audio power amplifiers have discrete input stages, which are very simple compared with those of most opamps, and draw relatively large input currents. These currents show appreciable non-linearity even when the output voltage of the amplifier is virtually distortion free and, if they flow through significant source resistances, will introduce added distortion. 
If this was the case with the 5532 then the extra distortion would manifest itself whenever the opamp was fed from a significant source resistance, no matter what the circuit configuration. But we have just seen that it only occurs in series-feedback situations; increasing the source resistance in a shunt-feedback does not perceptibly increase distortion. The effect may be present, but if so, it is very small, no doubt because opamp signal input currents are also very small, and it is lost in the noise.
The only difference is that the series circuit has a CM voltage of about 3 Vrms, while the shunt circuit does not, and the conclusion is that with a bipolar input opamp, you must have both a CM voltage and a significant source resistance to see extra distortion. The input stage of a 5532 is a straightforward long-tailed pair (see Figure 4.21) with a simple tail current source, and no fancy cascoding, and I suspect that Early Effect operates on it when there is a large CM voltage, modulating the quite high input bias currents, and this is what causes the distortion. The signal input currents are much smaller due to the high open-loop gain of the opamp and, as we have seen, appear to have a negligible effect.
FET-input opamps behave differently from bipolar-input opamps. Take a look at Figure 4.9, taken from a TL072 working in shunt and in series configuration with a 5 Vrms output. The circuits are as in Figure 4.3a and 4.3b, except that the resistor values have to be scaled up to 10 kΩ and 22 kΩ, because the TL072 is nothing like so good at driving loads as the 5532. This, unfortunately, means that the inverting input is seeing a source resistance of 10k||22k = 6.9k, which introduces a lot of common-mode (CM) distortion –5 times as much at 20 kHz as for the shunt case. Adding a similar resistance in the input path cancels out this distortion, and the trace then is the same as the “Shunt” trace in Figure 4.9. Disconcertingly, the value that achieved this was not 6.9k but 9k1. That means adding -113 dBu of Johnson noise, so it’s not always appropriate.
It’s worth mentioning that the flat part of the shunt trace below 10 kHz is not noise, as it would be for the 5532; it is distortion.
A voltage-follower has no inconvenient medium-impedance feedback network, but it does have a much larger CM voltage. Figure 4.10 shows a voltage-follower working at 5 Vrms. With no source resistance, the distortion is quite low due to the 100% NFB, but as soon as a 10 kΩ source resistance is added, we are looking at 0.015% at 10 kHz.
Once again, this can be cured by inserting an equal resistance in the feedback path of the voltage-follower, as in Figure 4.3d. This gives the “Cancel” trace in Figure 4.10. Adding resistances for distortion cancellation in this way has the obvious disadvantage that they introduce extra Johnson noise into the circuit. Another point is that stages of this kind are often driven from pot wipers, so the source impedance is variable, ranging between zero and one-quarter of the pot track resistance. Setting a balancing impedance in the other opamp input to a mid-value, i.e. one-eighth of the track resistance, should reduce the average amount of input distortion, but it is inevitably a compromise.
With JFET inputs, the problem is not the operating currents of the input devices themselves, which are negligible, but the currents drawn by the non-linear junction capacitances inherent in field-effect devices. These capacitances are effectively connected to one of the supply rails. For P-channel JFETs, as used in the input stages of most JFET opamps, the important capacitances are between the input JFETs and the substrate, which is normally connected to the V-rail. See Jung. 
According to the Burr-Brown data sheet for the OPA2134, “The P-channel JFETs in the input stage exhibit a varying input capacitance with applied CM voltage”. It goes on to recommend that the input impedances should be matched if they are above 2 kΩ.
Common-mode distortion can be minimised by running the opamp off the highest supply rails permitted, though the improvements are not large. In one test on a TL072, going from ±15 V to ±18 V rails reduced the distortion from 0.0045% to 0.0035% at 10 kHz.
Until recently, the 5532 was pre-eminent. It is found in almost every mixing console and in a large number of preamplifiers. Distortion is almost very low, even when driving 600 Ω loads. Noise is very low, and the balance of voltage and current noise in the input stage is well matched to moving-magnet phono cartridges; in many applications, discrete devices give no significant advantage. Large-quantity production has brought the price down to a point at which a powerful reason is required to pick any other device.
The 5532 is not, however, perfect. It suffers common-mode distortion. It has high bias and offset currents at the inputs as an inevitable result of using a bipolar input stage (for low noise) without any sort of bias-cancellation circuitry. The 5532 is not in the forefront for DC accuracy, though it’s not actually that bad. The offset voltage spec is 0.5 mV typical, 4 mV max, compared with 3 mV typical, 6 mV max for the popular TL072. I have actually used 5532s to replace TL072s when offset voltage was a problem, but the increased bias current was acceptable.
With horrible inevitability, the very popularity and excellent technical performance of the 5532 has led to it being criticised by subjectivists who have contrived to convince themselves that they can tell opamps apart by listening to music played through them. This always makes me laugh, because there is probably no music on the planet that has not passed through a hundred or more 5532s on its way to the consumer.
The LM4562 represents a real advance on the 5532. It is, however, still a good deal more expensive and is not perfect –it appears to be more easily damaged by excess common-mode voltages, and there is some evidence it is more susceptible to RF demodulation.
In some applications, such as low-cost mixing consoles, bipolar-style bias currents are a real nuisance, because keeping them out of EQ pots to prevent scratching noises requires an inappropriate number of blocking capacitors. There are plenty of JFET-input opamps around with negligible bias currents, but there is no obviously superior device that is the equivalent of the 5532. The TL072 has been used in this application for many years, but its HF linearity is not first class, and distortion across the band deteriorates badly as output loading increases. However, the opamps in many EQ sections work in the shunt-feedback configuration with no CM voltage on the inputs, and this reduces the distortion considerably. When low bias currents are needed with superior performance, the OPA2134 is often a good choice, though it is at least four times as expensive as the TL072.
The rest of this chapter looks at some opamp types and examines their performance, with the 5532 the usual basis for comparison. The parts shown here are not necessarily intended as audio opamps, though some, such as the OP275 and the OPA2134, were specifically designed as such. They have, however, all seen use, in varying numbers, in audio applications. Bipolar input opamps are dealt with first.
The LM741 is only included here for its historical interest; in its day, it was a most significant development and, to my mind, the first really practical opamp. It was introduced by Fairchild in 1968 and is considered a second-generation opamp, the 709 being first generation.
The LM741 had (and indeed has) effective short-circuit protection and internal compensation for stability at unity gain and was much easier to make work in a real circuit than its predecessors. It was clear that it was noisy compared with discrete circuitry, and you sometimes had to keep the output level down if slew limiting was to be avoided, but with care, it was usable in audio. Probably the last place the LM741 lingered was in the integrators of state-variable EQ filters, where neither indifferent noise performance nor poor slewing capability is a serious problem; see Chapter 15 for more details on this application. The LM741 is a single opamp. The dual version is the LM747.
Figure 4.11 shows a region between 100 Hz and 4 kHz at which distortion rises at 6dB/octave. This is the result of the usual dominant-pole Miller compensation scheme. When slew limiting begins, the slope increases, and THD rises rapidly with frequency.
The 5532 is a low-noise, low-distortion bipolar dual opamp with internal compensation for unity-gain stability. The 5534 is a single version internally compensated for gains down to three times, and an external compensation capacitor can be added for unity-gain stability; 22 pF is the usual value. The 5532 achieves unity-gain stability by having degeneration resistors in the emitter circuits of the input transistors to reduce the open-loop gain, and this is why it is noisier than the 5534.
The common-mode range of the inputs is a healthy +/-13 V, with no phase inversion problems if this is exceeded. It has a distinctly higher power consumption than the TL072, drawing approximately 4 mA per opamp section when quiescent. The DIL version runs perceptibly warm when quiescent on +/-17 V rails.
The 5534/5532 has bipolar transistor input devices. This means it gives low noise with low source resistances but draws a relatively high bias current through the input pins. The input devices are NPN, so the bias currents flow into the chip from the positive rail. If an input is fed through a significant resistance, then the input pin will be more negative than ground due to the voltage-drop caused by the bias current. The inputs are connected together with back-to-back diodes for reverse-voltage protection and should not be forcibly pulled to different voltages. The 5532 is intended for linear operation, and using it as a comparator is not recommended.
As can be seen from Figure 4.12, the 5532 is almost distortion free, even when driving the maximum 500 Ω load. The internal circuitry of the 5532 has never been officially explained but appears to consist of nested Miller loops that permit high levels of internal negative feedback. The 5532 is the dual of the 5534 and is much more commonly used than the single, as it is cheaper per opamp and does not require an external compensation capacitor when used at unity gain.
The 5532/5534 is made by several companies, but they are not all created equal. Those by Fairchild, JRC, and ON-Semi have significantly lower THD at 20 kHz and above, and we’re talking about a factor of two or three here.
The 5532- and 5534-type opamps require adequate supply-decoupling if they are to remain stable; otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the +ve and -ve rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimetres from the opamp; normally, one such capacitor is fitted per package as close to it as possible. It is not necessary and often not desirable to have two capacitors going to ground; every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground.
To the best of my knowledge, virtually nothing has been published about the internal operation of the 5532. This is surprising given its unique usefulness as a high-quality audio opamp. I believe the secret of the 5532’s superb linearity is the use of nested negative feedback inside the circuit, in the form of traditional Miller compensation.
Figure 4.13 shows the only diagram of the internal circuitry that has been released; the component and node numbers are mine. It has been in the public domain for at least 20 years, so I hope no one is going to object to my impertinent comments on it. The circuit initially looks like a confusing sea of transistors, and there is even a solitary JFET lurking in there, but it breaks down fairly easily. There are three voltage-gain stages, plus a unity-gain output stage to increase drive capability. This has current-sensing overload protection. There is also a fairly complex bias generator, which establishes the operating currents in the various stages.
In all conventional opamps, there are two differential input signals which have to be subtracted to create a single output signal, and the node at which this occurs is called the “phase summing point”.
Q1, Q2 make up the input differential amplifier. They are protected against reverse-biasing by the diode-connected transistors across the input pins. Note there are no emitter degeneration resistors, which would line arise the input pair at the expense of degrading noise. Presumably high open-loop gain (note there are three gain stages, whereas a power amplifier normally only has two) means that the input pair is handling very small signal levels, so its distortion is not a problem.
Q3, Q4 make up the second differential amplifier; emitter degeneration is now present. Phase summing occurs at the output of this stage at Node 2. C1 is the Miller capacitor around this stage, from Node 2 to Node 1. Q5, Q6, Q7 are a Wilson current mirror, which provides a driven current source as the collector load of Q4. The function of C4 is obscure but it appears to balance C1 in some way.
The third voltage-amplifier stage is basically Q9 with split-collector transistor Q15 as its current-source load. Q8 increases the basic transconductance of the stage, and C3 is the Miller capacitor around it, feeding from Node 3 to Node 2 –note that this Miller loop does not include the output stage. Things are a bit more complicated here, as it appears that Q9 is also the sink half of the Class-B output stage. Q14 looks very mysterious, as it seems to be sending the output of the third stage back to the input; possibly it’s some sort of clamp to ensure clean clipping, but to be honest, I haven’t a clue. Q10 plus associated diode generates the bias for the Class-B output stage, just as in a power amplifier.
The most interesting signal path is the semi-local Miller loop through C2, from Node 3 to Node 1, which encloses both the second and third voltage amplifiers; each of these has its own local Miller feedback, so there are two nested layers of internal feedback. This is probably the secret of the 5532’s low distortion.
Q11 is the source side of the output stage, and as mentioned, Q9 appears to be the sink. Q12, Q13 implement overcurrent protection. When the voltage drop across the 15 Ω resistor becomes too great, Q12 turns on and shunts base drive away from Q11. In the negative half-cycle, Q13 is turned on, which in turn activates Q17 to shunt drive away from Q8.
The biasing circuit shows an interesting point. Bipolar bias circuits tend not to be self-starting; no current flows anywhere until some flows somewhere, so to speak. Relying on leakage currents for starting is unwise, so here the depletion-mode JFET provides a circuit element that is fully on until you bias it off and can be relied upon to conduct as the power rails come up from zero.
More information on the internals of the 5534 can be found in .
The LM4562 is a new opamp, which first became freely available at the beginning of 2007. It is a National Semiconductor product. It is a dual opamp – there is no single or quad version. It costs about 10 times as much as a 5532.
The input noise voltage is typically 2.7 nV/√Hz, which is substantially lower than the 4 nV/√Hz of the 5532. For suitable applications with low source impedances, this translates into a useful noise advantage of 3.4 dB. The bias current is 10 nA typical, which is very low and would normally imply that bias cancellation, with its attendant noise problems, was being used. However, in my testing, I have seen no sign of excess noise, and the data sheet is silent on the subject. No details of the internal circuitry have been released so far and quite probably never will be.
It is not fussy about decoupling, and as with the 5532, 100 nF across the supply rails close to the package should ensure HF stability. The slew rate is typically +/-20 V/μs, more than twice as quick as the 5532.
The first THD plot in Figure 4.14 shows the LM4562 working at a closed-loop gain of 2.2x in shunt feedback mode at a high level of 10 Vrms. The top of the THD scale is 0.001%, something you will see with no other opamp in this survey. The no-load trace is barely distinguishable from the AP SYS-2702 gen-mon output, and even with a heavy 500 Ω load driven at 10 Vrms, there is only a very small amount of extra THD, reaching 0.0007% at 20 kHz.
Figure 4.15 shows the LM4562 working at a gain of 3.2x in series feedback mode, both modes having a noise gain of 3.2x. There is little extra distortion from 500 Ω loading.
For Figures 4.14 and 4.15, the feedback resistances were 2k2 and 1 kΩ, so the minimum source resistance presented to the inverting input is 687 Ω. In Figure 4.16, extra source resistances were then put in series with the input path (as was done with the 5532 in the section on common-mode distortion), and this revealed a remarkable property of the LM4562: it is much more resistant to common-mode distortion than the 5532. At 10 Vrms and 10 kHz, with a 10 kΩ source resistance, the 5532 generates 0.0014% THD (see Figure 4.6), but the LM4562 gives only 0.00046% under the same conditions. I strongly suspect that the LM4562 has a more sophisticated input stage than the 5532, probably incorporating cascoding to minimise the effects of common-mode voltages.
Note that only the rising curves to the right represent actual distortion. The raised levels of the horizontal traces at the LF end is due to Johnson noise from the extra series resistance.
It has taken an unbelievably long time –nearly 30 years –for a better audio opamp than the 5532 to come along, but at last, it has happened. The LM4562 is superior in just about every parameter, but it has much higher current noise. At present, it also has a much higher price, but hopefully that will change.
The AD797 (Analog Devices) is a single opamp with very low voltage noise and distortion. It appears to have been developed primarily for the cost-no-object application of submarine sonar, but it works very effectively with normal audio –if you can afford to use it. The cost is something like 20 times that of a 5532. No dual version is available, so the cost ratio per opamp section is 40 times.
This is a remarkably quiet device in terms of voltage noise, but current noise is correspondingly high due to the high currents in the input devices. Early versions appeared to be rather difficult to stabilise at HF, but the current product is no harder to apply than the 5532. Possibly there has been a design tweak, or on the other hand, my impression may be wholly mistaken.
The AD797 incorporates an ingenious feature for internal distortion cancellation. This is described on the manufacturer’s data sheet. Figure 4.17 shows that it works effectively.
The OP27 from Analog Devices is a bipolar-input single opamp primarily designed for low noise and DC precision. It was not intended for audio use, but in spite of this, it is frequently recommended for such applications as RIAA and tape head preamps. This is unfortunate, because while at first sight it appears that the OP27 is quieter than the 5534/5532, as the en is 3.2 nV/rtHz compared with 4 nV/rtHz for the 5534, in practice, it is usually slightly noisier. This is because the OP27 is in fact optimised for DC characteristics and so has input bias-current cancellation circuitry that generates common-mode noise. When the impedances on the two inputs are very different –which is the case in RIAA preamps –the CM noise does not cancel, and this appears to degrade the overall noise performance significantly.
For a bipolar input opamp, there appears to be a high level of common-mode input distortion, enough to bury the output distortion caused by loading; see Figures 4.18 and 4.19. It is likely that this too is related to the bias-cancellation circuitry, as it does not occur in the 5532.
The maximum slew rate is low compared with other opamps, being typically 2.8 V/μs. However, this is not the problem it may appear. This slew rate would allow a maximum amplitude at 20 kHz of 16 Vrms if the supply rails permitted it. I have never encountered any particular difficulties with decoupling or stability of the OP27.
The OP270 from Analog Devices is a dual opamp, intended as a “very low noise precision operational amplifier”, in other words combining low noise with great DC accuracy. The input offset voltage is an impressively low 75 μV maximum. It has bipolar inputs with a bias current cancellation system; the presence of this is shown by the 15 nA bias current spec, which is 30 times less than the 500 nA taken by the 5534, which lacks this feature. It will degrade the noise performance with unequal source resistances, as it does in the OP27. The input transistors are protected by back-to-back diodes.
The OP270 distortion performance suffers badly when driving even modest loads. See Figures 4.20 and 4.21. The slew rate is a rather limited 2.4 V/μs, which is only just enough for a full output swing at 20 kHz. Note also that this is an expensive opamp, costing something like 25 times as much as a 5532; precision costs money. Unless you have a real need for DC accuracy, this part is not recommended.
The Analog Devices OP275 is one of the few opamps specifically marketed as an audio device. Its most interesting characteristic is the Butler input stage, which combines bipolar and JFET devices. The idea is that the bipolars give accuracy and low noise, while the JFETs give speed and “the sound quality of JFETs”. That final phrase is not a happy thing to see on a datasheet from a major manufacturer; the sound of JFETs (if any) would be the sound of high distortion. Just give us the facts, please.
The OP275 is a dual opamp; no single version is available. It is quite expensive, about six times the price of a 5532, and its performance in most respects is inferior. It is noisier, has higher distortion, and does not like heavy loads. See Figures 4.22 and 4.23. The CM range is only about two-thirds of the voltage between the supply rails, and Ibias is high due to the BJT part of the input stage. Unless you think there is something magical about the BJT/JFET input stage –and I am quite sure there is not –it is probably best avoided.
The THD at 10 kHz with a 600 Ω load is 0.0025% for shunt and 0.009% for series feedback; there is significant CM distortion in the input stage, which is almost certainly coming from the JFETs. (I appreciate the output levels are not the same, but I think this only accounts for a small part of the THD difference.) Far from adding magical properties to the input stage, the JFETs seem to be just making it worse.
Opamps with JFET inputs tend to have higher voltage noise and lower current noise than BJT-input types and therefore give a better noise performance with high source resistances. Their very low bias currents often allow circuitry to be simplified.
The TL072 is one of the most popular opamps, having very high-impedance inputs, with effectively zero bias and offset currents. The JFET input devices give their best noise performance at medium impedances, in the range 1 kΩ–10 kΩ. It has a modest power consumption at typically 1.4 mA per opamp section, which is significantly less than the 5532. The slew rate is higher than for the 5532, at 13 V/μs against 9 V/μs. The TL072 is a dual opamp. There is a single version called the TL071 which has offset null pins.
However, the TL072 is not THD-free in the way the 5532 is. In audio usage, distortion depends primarily upon how heavily the output is loaded. The maximum loading is a trade-off between quality and circuit economy, and I would put 2 kΩ as the lower limit. This opamp is not the first choice for audio use unless the near-zero bias currents (which allow circuit economies by making blocking capacitors unnecessary) the low price, or the modest power consumption are dominant factors.
It is a quirk of this device that the input common-mode range does not extend all the way between the rails. If the common mode voltage gets to within a couple of volts of the V-rail, the opamp suffers phase reversal and the inputs swap their polarities. There may be really horrible clipping, where the output hits the bottom rail and then shoots up to hit the top one, or the stage may simply latch up until the power is turned off.
TL072s are relatively relaxed about supply rail decoupling, though they will sometimes show very visible oscillation if they are at the end of long, thin supply tracks. One or two rail-to-rail decoupling capacitors (e.g. 100 nF) per few centimetres is usually sufficient to deal with this, but normal practice is to not take chances and allow one capacitor per package as with other opamps.
Because of common-mode distortion, a TL072 in shunt configuration is always more linear. In particular, compare the results for 3k3 load in Figures 4.24 and 4.25. At heavier loadings, the difference is barely visible, because most of the distortion is coming from the output stage.
Distortion always gets worse as the loading increases. This factor together with the closed-loop NFB factor determines the THD.
TL072/71 opamps are prone to HF oscillation if faced with significant capacitance to ground on the output pin; this is particularly likely when they are used as unity-gain buffers with 100% feedback. A few inches of track can sometimes be enough. This can be cured by an isolating resistor, in the 47 to 75 Ω range, in series with the output, placed at the opamp end of the track.
The TL052 from Texas Instruments was designed to be an enhancement of the TL072 and so is naturally compared with it. Most of the improvements are in the DC specifications. The offset voltage is 0.65 mV typical, 1.5 mV max, compared with the TL072’s 3 mV typical, 10 mV max. It has half the bias current of the TL072. This is very praiseworthy but rarely of much relevance to audio.
The distortion, however, is important, and this is worse rather than better. THD-performance is rather disappointing. The unloaded THD is low, as shown in Figure 4.26, in series feedback mode. As usual, practical distortion depends very much on how heavily the output is loaded. Figure 4.27 shows that it deteriorates badly for loads of less than 4k7.
The slew rate is higher than for the TL072 (18 V/μs against 13 V/μs), but the lower figure is more than adequate for a full-range output at 20 kHz, so this enhancement is of limited interest. The power consumption is higher, typically 2.3 mA per opamp section, which is almost twice that of the TL072. Like the TL072, it is relatively relaxed about supply rail decoupling. At the time of writing (2009) the TL052 costs at least twice as much as the TL072.
The OPA2134 is a Burr-Brown product, the dual version of the OPA134. The manufacturer claims it has superior sound quality due to its JFET input stage. Regrettably but not surprisingly, no evidence is given to back up this assertion. The input noise voltage is 8 nV/√Hz, almost twice that of the 5532. The slew rate is typically +/-20 V/μs, which is ample. It does not appear to be optimised for DC precision, the typical offset voltage being ±1 mV, but this is usually good enough for audio work. I have used it many times as a DC servo in power amplifiers, the low bias currents allowing high resistor values and correspondingly small capacitors.
The two THD plots in Figures 4.28 and 4.29 show the device working at a gain of 3x in both shunt and series feedback modes. It is obvious that a problem emerges in the series plot, where the THD is higher by about 3 times at 5 Vrms and 10 kHz. This distortion increases with level, which immediately suggests common-mode distortion in the input stage. Distortion increases with even moderate loading; see Figure 4.30.
This is a relatively modern and sophisticated opamp. When you need JFET inputs (usually because significant input bias currents would be a problem), this definitely beats the TL072; it is, however, four to five times more expensive.
The OPA604 from Burr-Brown is a single JFET-input opamp which claims to be specially designed to give low distortion. The simplified internal circuit diagram in the data sheet includes an enigmatic box intriguingly labelled “Distortion Rejection Circuitry”. This apparently “linearizes the open-loop response and increases voltage gain”, but no details as to how are given; whatever is in there appears to have been patented, so it ought to be possible to track it down. However, despite this, the distortion is not very low even with no load (see Figure 4.31) and is markedly inferior to the 5532. The OPA604 is not optimised for DC precision, the typical offset voltage being ±1 mV. The OPA2604 is the dual version, which omits the offset null pins.
The data sheet includes a discussion that attempts to show that JFET inputs produce a more pleasant type of distortion than BJT inputs. This unaccountably omits the fact that the much higher transconductance of BJTs means that they can be linearised by emitter degeneration so that they produce far less distortion of whatever type than a JFET input.  Given that the OPA604 costs five times as much as a 5532, it is not very clear under what circumstances this opamp would be a good choice.
The OPA627 from Burr-Brown is a laser-trimmed JFET-input opamp with excellent DC precision, the input offset voltage being typically ±100 μV. The distortion is very low, even into a 600 Ω load, though it is slightly increased by the usual common-mode distortion when series feedback is used.
The OP627 is a single opamp, and no dual version is available. The OPA637 is a decompensated version only stable for closed-loop gains of five or more. This opamp makes a brilliant DC servo for power amplifiers if you can afford it; it costs about 50 times as much as a 5532, which is a hundred times more per opamp section and about 20 times more per opamp than the OPA2134, which is my usual choice for DC servo work.
The current noise in is very low, the lowest of any opamp examined in this book, apparently due to the use of Difet (dielectrically isolated JFET) input devices, and so it will give a good noise performance with high source resistances. Voltage noise is also very respectable at 5.2 nV/√Hz, only fractionally more than the 5532.
“Gen-mon” is the test gear output.
The series feedback case barely has more distortion than the shunt one, and only at the extreme HF end. It appears that the Difet input technology also works well to prevent input non-linearity and CM distortion. See Figures 4.32 and 4.33.
 Blumlein, Alan UK patent 482,470 (1936)
 Jung, Walt, ed Op-amp Applications Handbook. Newnes, 2006, Chapter 8
 Self, Douglas Audio Power Amplifier Design Handbook 5th edition. Focal Press, 2009, pp 186–189
 Self, Douglas Audio Power Amplifier Design Handbook 5th edition. Focal Press, 2009, p 96
 Jung, Walt, ed Op-amp Applications Handbook, Newnes, 2006, Chapter 5, p 399
 Huijsing, J. H. Operational Amplifiers: Theory & Design, Kluwer Academic, 2001, pp 300–302
 Self, Douglas Audio Power Amplifier Design 6th edition. Newnes, 2013, p 500