Goodbye, EMI: Intro to Instrumentation Amplifiers

To take accurate signal measurements, you must reduce noise. Here’s what you need to know first.

listen to this story

Separating signal from noise is an ever-present challenge. How can you focus on what really matters and ignore what doesn’t? We’re not gurus, so we can’t give you a universal answer. But what we can reveal is how you can accurately measure small electrical signals in the presence of electrical noise.

The key is instrumentation amplifiers.

An instrumentation amplifier magnifies the difference between two input signal voltages while rejecting any signals common to both inputs. In this way, it extracts small signals and delivers them in a single-ended output voltage. The instrumentation amplifier is an important frontend in precise measurement systems.

In this first part of our series on instrumentation amplifiers, we’ll examine the types of electrical noise and preview how differential amplifiers form the basis for instrumentation amplifiers.

Understanding EMI

Electromagnetic interference, or EMI, is composed of an electric field and a magnetic field. The electric field is orthogonal to the magnetic field, and both fields are orthogonal to the direction of propagation. EMI is responsible for both common-mode noise and normal-mode (differential) noise. We will elaborate on both of these shortly.

Figure 1. Used with the author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

EMI can originate from many sources, such as a switching DC voltage regulator, a power-carrying electrical conductor, or an electric motor. A varying magnetic field that links with an electrical conductor will induce a voltage in that conductor. If two parallel conductors are used to carry a signal, the conductor closest to the source will have larger induced voltage levels.

See Figure 2a for an example. The conductor closest to the EMI source (in red) is affected differently than the more distant conductor (in blue). The signal carried by the two conductors is separated by 3mV of noise by the time it reaches the amplifier input. This produces an error at the amplifier, which will respond as if it is receiving a normal differential signal—hence, this is called normal-mode noise.

If, however, the wires are twisted together as shown in Figure 2b, the EMI-induced voltages average out and the normal-mode noise is zero. A cable that contains three twisted pairs is shown in Figure 2c.

Figure 2. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

The electric field component of EMI noise can be viewed as capacitive coupling, which tends to produce a common-mode noise component. A twisted pair is susceptible to electric field coupling as shown in Figure 3 (the stray capacitances are distributed over the entire length of the wire, but lumped capacitances are shown to provide a conceptual approximation).

Figure 3. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

Recall that capacitors require two electrical conductors that are separated by an insulator. Therefore, to minimize EMI electric field coupling, we can cover the twisted pair with a conductive foil shield. This is shown in Figure 4a. Figure 4b illustrates the stray capacitances that exist inside the cable. Capacitance C1 is between Signal Hi and the Shield, capacitance C2 is between Signal Lo and the Shield, and C3 is between the two signal conductors. Figure 4c depicts the cable’s electrical schematic. In practice, only one side of the shield is usually connected to ground. Any external capacitive coupling to the shield will then be taken to ground via the drain wire.

Figure 4. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

To observe capacitive coupling effects, you may merely touch the center conductor of an oscilloscope probe. With the oscilloscope sensitivity set to a few millivolts per division, you will see a 50 or 60Hz sine wave. The AC power wiring has capacitively coupled the noise to you. An audio amplifier can also be used, and then you will hear a 50 or 60Hz hum.

The front end of an accurate measurement system usually employs a precision instrumentation amplifier. While an instrumentation amplifier tends to reject common-mode noise, it can be fooled by normal-mode noise. In general, it is good practice to minimize noise sources where practical, or at least attenuate noise that is trying to creep into a measurement system. Alternatively, we try to reduce the noise presented to the input of our instrumentation amplifier.

In many instances, we require an input amplifier that responds to a differential input voltage. To better understand instrumentation amplifiers, we must review op amp differential amplifiers.

A Review of Op Amp Differential Amplifiers

Figure 5a shows an op amp differential amplifier circuit. The reference designators R1 and R2 have been duplicated intentionally. This is done to indicate matched resistor values. Figure 5b reminds us that the gain applied to a signal at the noninverting input terminal is (1 + R2/R1).

Figure 5. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

Figure 5c reminds us that the gain applied to a signal at the inverting input is −R2/R1. Therefore, to provide equal gain on both input terminals, it is necessary to attenuate the signal applied to the noninverting input terminal. This is accomplished by using a resistive voltage divider as shown in Figure 5d.

Figure 5 (continued). Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

To analyze the op-amp differential amplifier, we use the principle of superposition. This means that we first find the output voltage (vOUT) produced by v1 acting alone. We then repeat the analysis to determine the output voltage vOUT created by v2 acting alone. Finally, we can determine the output voltage vOUT by finding the algebraic sum of vOUT and vOUT. This process is depicted in Figure 6.

In Figure 6a we set v2 to zero by replacing it with a short circuit. This creates a parallel equivalent resistance from the noninverting input to ground. However, since the inputs to an op amp draw no signal current, there will be no voltage drop across the combination. This means that the noninverting input terminal will remain at ground potential. Equivalently, we have an inverting amplifier. The voltage gain (-R2/R1) will be unaffected.

In Figure 6b, we set v1 to zero by replacing it with a short. Clearly, this results in a noninverting amplifier configuration. However, the noninverting input terminal is connected to a voltage divider. Since the noninverting input draws no current, we may find v2 by applying simple voltage division.

Figure 6. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

Figure 6c shows the superposition and allows us to derive the open-circuit (output) differential voltage gain:

Since the op amp’s output serves as the output of the differential amplifier circuit, the output resistance of the differential amplifier circuit is also zero, and the output loaded differential voltage gain of the amplifier circuit is equal to Avd(oc).

Figure 7a illustrates that the input resistance of the inverting input Rin(-) is equal to R1. However, the input resistance of the noninverting input Rin(+) is equal to R1 + R2. The imbalance in the impedance levels will cause unequal loading on identical signal sources. This will cause degradation in the circuit’s common-mode rejection. One solution is to buffer the inputs of the differential amplifier circuit with voltage followers as depicted in Figure 7b.

Figure 7. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

It is possible to use a quad operational amplifier (such as an AD713) to implement Figure 7b. This means that only one integrated circuit is required, and the fourth op amp is an uncommitted spare. Since the Routof the voltage followers is zero, the input resistance imbalance of the op amp differential amplifier is unimportant. Further, the infinite input resistance of the followers becomes the input resistance of the circuit’s inverting and non-inverting inputs.

Preview of the Classic Instrumentation Amplifier

An instrumentation amplifier is a direct-coupled differential amplifier with matched input resistances (Rin(+) is equal to Rin(-)) and a differential voltage gain that can be adjusted by varying a single resistor (RG) without degrading the amplifier’s common-mode rejection.

The constraint on the gain adjustment means that the differential amplifiers shown in Figure 7 are not instrumentation amplifiers. A gain adjustment in either circuit requires changing the values of two resistors (typically, this means changing the resistors in the R2 positions). However, the differential amplifier in Figure 7b makes Rin(+) equal to Rin(-)  and serves as a dandy segue to the instrumentation amplifier.

Figure 8 shows the classic three op amp implementation of an instrumentation amplifier.

Figure 8. Used with author’s permission from Discrete and Integrated Electronics Analysis and Design for Engineers and Engineering Technologists.

In our next article, we will analyze the common-mode voltage gain to determine the common-mode rejection ratio of this instrumentation amplifier. Later in this series, we’ll explain how the voltage gain can be adjusted by changing the gain resistor RG.