The Classroom

Operational Amplifiers

...Revisited

Bob Harper

Issue 13, July 2018

In 2018, the µA741 is celebrating its 50th birthday!.

Operational amplifiers (op-amps) have been used in several projects within DIYODE – we’ve even published information on how they work. However, operational amplifiers are one of the most used integrated circuits (IC), even if micro-controllers are catching up.

Op-amps pre-date ICs, and even the transistor, by virtue of the fact they were originally produced in valve technology as mathematical elements for analog computers, around 1942, and that’s why they’re called ‘operational’ amplifiers.

old op-amp
IMAGE CREDIT: computerhistory.org

When op-amps became transistor-based, around 1963, an operational amplifier could fit on a single PCB card. If that doesn’t impress you then think what your kids will say about ICs in 50 years: ‘IC? Is that what came before the iPhone?’

Then the versatility of the op-amp was noticed by instrumentation engineers who reacted by basically powering a change from mechanical instruments to electronic instruments through the 1960s and 1970s.

In 1967, almost the beginning of the IC age, Dave Fullagar, an English engineer working for Fairchild Semiconductors, designed an improved integrated version of the op-amp, to be named the “µA741”. The µA741 was not the original IC op-amp however. The µA702, also by Fairchild was introduced in 1963, and was followed by several models including the µA709, with each an improvement on the previous.

It was however, the µA741 that was so well designed for the time, that it has survived to present time; a 50-year-old design that still does the job today for which it was originally created – impressive!

So this edition of The Classroom is all about operational amplifiers as a class of devices, but it seems appropriate to honour the µA741 as a well known representative of that technology, and compare what others have to add to the story.

A NOTE ABOUT "741": “‘741” is a part copied, under licence, by many manufacturers who apply their own prefix. National Semiconductor call their version an “LM741”. We have used ‘741 throughout this article.

OPERATIONAL AMPLIFIER

An op-amp is one of those “unobtanium” items engineers asked for. When engineers are confronted with a complex problem that defies current technology, they may bypass that problem by specifying “unobtanium”, which refers to a material or technology not presently available. NASA is famous for using the term, and “unobtanium” resulted in development of the ceramic tiles that protect the Space Shuttle on re-entry, velcro tape, as just a few well known examples.

Operational amplifiers were the “ideal” or “universal amplifier” when needed for building analogue computers that could calculate ballistic trajectories of navel artillery in WWII. The engineers drew a triangle, a standard symbol already representing an amplifier as a block or stage, and inside the triangle they noted what it was meant to do.

UNIVERSAL AMPLIFIER

The original specification, the unobtainium model, would have been something like:

“A small and light amplifier with a predictable performance, with an infinite input impedance, zero output impedance, infinite voltage and current gain, infinite bandwidth, perfect stability, zero drift, rail to rail output swing, low current draw, and differential inputs; inverting and non-inverting.”

Possibly the specification would have also begged for differential output as well.

Then as one group of engineers developed the analogue computer, another group set about inventing the technology to make it all work.

REAL AMPLIFIER

Yes, we wanted a Porsche but could only manage a Volkswagen; but boy what a Volkswagen we got! So let’s compare the unobtainium shopping list with the products of the IC age, particularly referring to the µA741, but noting other more recent achievements.

Size

The most dramatic change from the early valve op-amps is that about 100 of the ‘741 ICs would fit in the space required by a single valve of the type used. Additionally, the circuit required two of the valves and a number of ancillary parts. So maybe 250 of the ‘741 and similar devices or perhaps 1,000 surface mount devices could occupy the same space.

POWER SUPPLY

The ‘741 will operate at up to +/-18VDC on the rails, or 36VDC from rail-to-rail. The LM741E can handle +/-22VDC. In each case however, the input voltages are +/-15V.

Either can manage a case dissipation of 500mW, and an extended short circuit output of 25mA and up to 35mA.

Remember, these were never designed to be stereo power amplifiers for boom boxes, but the monoblock audio amplifiers of today are a direct descendant of the ‘741 op-amp.

HIGH INPUT IMPEDANCE (ZIN)

While the minimum input impedance of the original ‘741 was only 1MΩ, 6MΩ is given as the typical value of the LM741E, a later variant. Other op-amps such as the TL071, a JFET input device which has the same pins as the ‘741, has an input impedance specified as 1e12Ω, or 1 TeraOhm – typical!

INFINITE VOLTAGE GAIN

Remembering that the original op-amps were valve technology, infinite was never going to be an option; however, the ‘741 specifies a gain of 200V/mV, or 200,000:1, or 106dB. In valve days, that was a very big ask. Interestingly, the TL071 has the same figure.

INFINITE BANDWIDTH

“Bandwidth” refers to the frequency range over which the amplifier is useful. At the bottom end, it works on DC; so the bandwidth begins at DC. The upper frequency limit is a function of frequency and gain, called the “gain bandwidth product”.

However, in a simple text answer, there is another value known as the ‘unity gain bandwidth’, or the highest frequency that will deliver an output at the same level as the input. For a typical ‘741 that is 1MHz, although that also depends on signal in and out, as both can reduce the bandwidth.

The TL071 has a more useful 3MHz, meaning that at 1MHz you will get some amplification – a gain of 3; however, most data sheets have a graph or several, showing gain bandwidth product. They may use an alternative name such as ‘large signal differential voltage amplification’.

The simple calculation, when you know the GBWP (also sometimes called “GBP”) is:

Gain = GBWP/Frequency (maximum frequency).

So the ‘741 at 1MHz has gain of 1, but at 100kHz, G = 1M/100k = 10. (i.e. 20dB.)

What this means is to avoid loss of gain over frequency, the feedback circuit needs to limit the amplifier gain to 10. If you need a gain of 10 at 1MHz, the GBWP needs to be 10MHz or greater.

frequency

With dB gain shown on the vertical axis, and frequency shown on the horizontal axis, you’ll notice there are two distinct lines. The horizontal line is the maximum possible gain of the circuit. The slanting line shows the gain reducing at a rate of 20dB/decade, meaning that the gain at 100kHz is 80dB.

ZERO OUTPUT IMPEDANCE

Should the op-amp manage zero output impedance, the output could supply as much current as the op-amp can take from the power supply. In reality, output impedance figures are not quoted, but maximum output current and output voltages are.

If the maximum output voltage is +/-15V, and the output current is around 20mA, depending on which data sheet you read, the output impedance can be estimated at around R = V/I = 15/ 0.02 = 750Ω. However, the ‘741 and other ICs have internal current limiting built in, so that is not an actual value.

PERFECT STABILITY

Perfect stability and zero drift go hand-in-hand. Every data sheet has at least one graph of how gain and other values change as a result of the device’s temperature, whether caused by power dissipation, or external climate.

The greatest challenge to electronics arises when an amplifier attempts to provide a DC value to a chart recorder, or other long-term stabilised reading. The recorded value may rise up and down, but the instrument tech has to attempt to discern whether the drift is real, or a result of drift in the electronics.

While the ‘741 is a good amplifier in general use, there are much better instrument amplifiers available.

RAIL-TO-RAIL OUTPUT SWING

Most amplifiers, and other circuits would benefit from the ability to set the output voltage from maximum negative to maximum positive voltage, but few circuits actually need it; and even fewer ever get it without replacing the op-amp supply with a higher voltage supply. The ‘741 typically manages +/-15V out for a supply of +/-18V.

Attempting to reach the rails in an analogue device invites clipping of the signal, which causes distortion and harmonic generation; so rail-to-rail is not essential.

INTERNAL NOISE

All electronic equipment has noise. Electronics work due to the movement of electrons, and electrons need to be plucked off an atom to enter the current. That sudden change in current, however slight, can be amplified, resulting in “shot noise” in the electronics

While the '741 has been used in a lot of audio gear over its life, equipment requiring high fidelity sound, low noise and better high frequency performance generally seek out op-amps with better specifications.

The TL071 has been used instead for many years, having very low noise figures, but more modern op-amps have reduced the noise even more. Unfortunately, those who love listening to loud music will never know the difference, as it’s their ears that are deficient! Their dog may however enjoy the music!

COMMON MODE REJECTION RATIO

Common mode rejection ratio (CMRR) is a measure of how well an amplifier manages to reject noise compared to the real signal. Most serious audio and instrument signals are differential, in that a signal is taken from some source, and applied to a pair of wires; so one wire has the opposite polarity signal to the other.

description

Often both are within a third conductor (i.e. a braided or foil sheath that is connected to ground). As the cable returns to the amplifier, perhaps passing through magnetic or even electrostatic fields, the fields can affect both signal wires. Each wire is connected to the op-amp – one to the inverting input and one to the non-inverting input. Such an amplifier is called a “differential amplifier”.

CMRR is a measure of how well the intended signal is amplified, compared to the interfering signal, as the interference will cause both input voltages to rise and fall together. The CMRR for a ‘741 is specified as 90dB, ~32,000 times lower amplification of the interfering voltage.

POWER SUPPLY REJECTION RATIO

Without explanation, you can probably guess what PSRR is, and how it’s measured. Ripple on the power supply should not affect the output signal, and the ‘741 has a PSRR of 96dB. That means that a 1V ripple on the power supply will be 15µV on the signal.

Let’s have a quick refresh of decibels: Gain(dB) = 20Log(V/v), where “V” is the large value and “v” is the small value of the voltage signals. Therefore:

v = V/10(96/20) = 15µV.

PRACTICALITIES

Note: the following refer to the ‘741 devices but in most cases will apply to all op-amp families.

Every IC has a purpose, and generally we delve into its true purpose in life, without considering the two most important pins on the device: the power supply pins.

Many circuits need nothing more than a battery connected across those pins, and even a 9V “transistor” battery will power a ‘741. However, an IC in a circuit supplied by a car battery, or either a linear or switch mode power supply, can require a little personal attention. For example, the circuit may need extra isolation, even though the PSRR is quite impressive. High frequency noise from other stages or blocks can become a problem if not filtered out from the common supply lines.

POWER SUPPLY PIN FILTERING

Each circuit has a different tolerance to noise, so there is no “one solution fits all” that we can offer to you. The simplest tactic is to add a capacitor or two across the power pins of each device, or from power pins to ground in a dual supply device. Usually a small electrolytic will handle fluctuating power draw of the device itself, and a smaller value capacitor might be added to bypass higher frequencies around the IC.

description

If the power supply is large compared to the IC demand, then no capacitor at the IC may be fine. But on battery powered circuits, and small power supplies, a suitable capacitor for a ‘741 can be estimated by C = IT/V (as for power supply filter capacitors), where “I” is the load current, for a ‘741 at 35-50mA; “T” is the time for the power supply to catch up, (i.e. how long you think the voltage will dip); and “V” is the voltage dip you are trying to avoid.

You really need to find, measure and react to these issues at the prototype stage. However, an example might be that a square wave generated by the IC causes the supply voltage to drop by 1V for roughly 2mS. Then the capacitance should be C = IT/V = 50mA x 2mS / 2V = 50µF. In fact, the current will most likely be under 25mA, and as such I have often used 10µF, and adjusted the value if I needed more. Electrolytic capacitors are normally used, but I have often used tantalum capacitors due to their small size.

HIGH FREQUENCY FILTER

The higher frequency capacitor depends on how high that frequency is, and what the “impedance” of the IC is. An IC taking 50mA from a 9V supply has a power impedance of 180Ω. A filter capacitor should have 10 to 20 times less impedance than the IC at the frequency of interference.

If you think the interference is at 10kHz then a 884nF capacitor is needed (~1µF). However, if the high frequency interference is 1MHz, then 8.84nF capacitor is needed (~10nF or 0.01µF).

Remember, if the only impedance is the reactance of the capacitor, then Z = Xc = 1/(2πfC) therefore C = 1/(2πfXc).

FERRITE BEADS

Another method is to use ferrite beads or ferrite inductors to reject unwanted noise from your precious circuit. Sometimes a low value series resistor can be used in conjunction with the capacitor.

description

DUAL RAIL SUPPLY

In a dual rail application (e.g. +/-9V), calculate filter capacitor values to be placed between supply rails and ground as for the single rail supply. Ferrite beads are good for applications that have radio Frequency Interference (RFI), rather than simply electro-magnetic interference (EMI). Basically they are the same animal, but RFI is at radio frequencies while EMI is all frequencies, although it’s often thought of as just the lower frequencies.

ARTIFICIAL GROUND VOLTAGE DIVIDER

An op-amp can and does work from a single rail supply, but it’s designed for a split rail or dual rail supply, such as +/-9V up to +/-18V. If a single supply is used, the centre ground reference can be artificially created using two identical resistors, and preferably a filter capacitor across one of those resistors.

The value of the resistors should be equal, but otherwise it’s a matter of stability verses load on the supply. Op-amps don’t require much in the way of bias, but the bias components may. A rule of thumb for bias going back even to valve days, is that the bleed current should be 10 to 20 times the bias current. “Bleed” current is the current through the two resistors. Typically for op-amp circuits, the resistors are either two 10kΩ or two 100kΩ, depending on how many op-amps are using the artificial ground voltage divider. The figure below shows a typical setup.

description

PROTOTYPE PCB

A breadboard is a typical device for prototyping an operational circuit project. However, as op-amps can be very sensitive, and especially sensitive to dodgy connections, prototyping work is often better achieved on a prototype PCB.

Many variations exist, but a good op-amp prototype PCB should have the following features:

  • A socket for the target IC or preferable up to four ICs, and possibly one for double op-amp ICs and quad op-amp ICs.
  • Many op-amps have the same or similar pinouts and µA741, LM741, etc. have the +ve power at pin 7 and -ve power on pin 4. The PCB should allow for filtering capacitors either between +ve and -ve, or from +ve to ground and -ve to ground, or even both.
  • There should be two ground rails: one either side of the IC, and at least one power rail each side. Many times we have found a need for both power rails either side, to avoid having to pass a component over the IC and other components.
  • A prototype PCB should have space for an artificial ground using two resistors and a capacitor as described above, and/or two linear regulators such as an LM7809 and an LM7909. A good lab power supply can be used instead.

The remainder of the pins should allow for multiple components to be attached as inputs, outputs and/or feedback. Pin 2 for example, usually requires a resistor from pin 6 as feedback, and a resistor to ground or the inverting signal input. It may also require a diode or two to ground, or a capacitor in parallel with the feedback resistor. So breadboards often seem to have one too few holes requiring a jumper to another set of five holes. The PCB should have five holes as well as the IC pin hole. There needs to be two input options and one output option per op-amp. Extra breadboarding / protoboarding nodes should be used to fill otherwise uncommitted space.

The diagram below should be taken as one possible method of building a development board for the µA741, and other similar pinout Op-Amps. Some of these components may never be needed, and some would be replaced by diodes or capacitors or inductors, or other components for certain experiments. By beginning with a purpose etched PCB, the results are more predictable and better controlled than with protoboard or socket strip boards.

description

OPERATIONAL AMPLIFIER CIRCUITS

Op-amps are so popular because they are so flexible. They are like the super-transistor on its best behaviour. Beginning with the typical introductory circuits, such as inverting and non-inverting buffers and amplifiers, then differential amplifiers, and summing amplifiers and mixers, we will move on to filters and clippers, integrating and differentiating amplifiers, and perhaps PID amplifiers. Add oscillators to the list and we have a quite full agenda for the next several issues.

With most of the “component introductions” and theory over, it’s time to fire up our soldering irons and start melting metal. We will begin with simple circuits, explaining their operation as we go, and using simple math we’ll calculate the required component values. Our goal is to provide some practical implementations with a project to boot, for every topic we cover in The Classroom.