Design Feature: September 26, 1996

Most oscilloscopes can perform only single-ended voltage measurements; that is, measurements of signals referenced to earth ground. Wiring within the probe connects the probe's reference lead to the shell of the BNC. When you plug the probe into the scope, the reference lead becomes electrically common with the scope's chassis. The power cord's ground conductor connects the chassis to earth ground. In many applications, the inability to make anything except single-ended measurements poses no problems. In other situations, however, the measurement you want to make either is completely impossible or is possible only with unacceptably degraded accuracy. The most obvious troublesome measurements involve voltages that are not referenced to ground. A common example is the voltage across the switching device in an off-line switching power supply. To get an accurate picture of such signals, you need to make a measurement that does not involve ground--a "differential" measurement.

Another type of signal that you must measure differentially is a balanced signal. Balanced signals are signals whose source and return paths have equal impedances. Telephone lines, read channels in magnetic-storage systems, and some digital-communication systems use balanced-signal paths.

Sometimes, you should make differential measurements even of ground-referenced signals. Because the scope-probe reference lead is grounded, attaching it to a circuit creates multiple ground paths, otherwise known as a ground loop. Magnetic fields radiate from current that passes through circuit conductors. Passing these currents through the ground loop induces circulating currents in the loop. These currents can interfere with the circuit operation and corrupt measured waveforms.

When you try to make single-ended measurements of very small-amplitude signals, interference from ground-loop currents can be particularly troublesome. The additional ground path introduced by the probe's ground lead alters the ground distribution in the circuit under test. This effect can either mask or accentuate circuit problems. You often experience similar difficulties when you connect a single-ended probe to a fast digital circuit. Using a differential-measurement technique can overcome such problems.

Unlike a conventional scope probe, a differential amplifier has an input that is only implicitly referenced to ground. As the name implies, a differential measurement produces a waveform that represents the difference in voltage between the two inputs. Ground does not enter into the measurement (Figure 1). (To supply input-bias current, many differential amplifiers do, however, include high-resistance paths from each input to ground. True floating-input amplifiers eliminate even this path to ground. Nevertheless, because of bandwidth limitations, such amplifiers are not often used with oscilloscopes.) Differential amplifiers ignore potentials that are equal in amplitude and phase and appear on both inputs. This characteristic is known as "common-mode rejection" (CMR). The other key feature of a differential amplifier is balanced input impedance. Both inputs have identical impedance to ground.

To specify the requirements on a differential amplifier, you need to understand some fundamentals. Figure 2a illustrates the classical model of a differential-measurement system. The measured voltage appears across the inputs of the amplifier and is known as the "differential-mode voltage" (V_DM). V_DM is the sum of two voltage sources, each with an amplitude of V_DM/2. The common-mode voltage (V_CM) is common to the two inputs and appears between ground and the midpoint of the sources of the two V_DM components. The amplifier's transfer function is V_O=V_DM·A_V, where A_V is the amplifier's differential gain. Because the V_DM/2 components do not exist on any real circuit nodes, the model is often simplified, as in Figure 2b. V_CM can appear on either the positive or the negative input. The accuracy of the simplified model approaches that of the classical version in situations in which V_CM>>V_DM.

An ideal differential amplifier totally rejects the common-mode component; hence, V_CM does not appear in the transfer function. Unfortunately, real implementations do not totally reject common mode; a small portion appears in the output. The measure of the amplifier's ability to reject V_CM is called "common-mode rejection ratio" (CMRR). CMRR, the ratio of differential-mode gain to common-mode gain, is important, because, at the amplifier output, the portion of the common-mode component that is not rejected is indistinguishable from a differential signal. The error increases as the CMRR decreases or as the ratio of V_CM to V_DM increases.

Suppose you attempt to use a differential amplifier to measure the emitter current of an emitter follower (Figure 3). The output swings 30V p-p, and the output current is 2A, resulting in a voltage drop of 200 mV across the emitter resistor. The differential amplifier's CMRR spec is 1000-to-1 at the frequency of the output voltage. The common-mode component appears as [1/1000]th of the 30V common-mode signal, or 30 mV, added to the 200-mV measured voltage. This error is 15%.

The CMRR of wideband differential amplifiers usually decreases as the frequency of the common-mode signal increases. Therefore, most wideband differential amplifiers specify CMRR vs frequency as a graph. Significant differences between the source impedances driving the two inputs also reduce the CMRR in many differential amplifiers.

Selecting differential amplifiers

There are several types of differential amplifiers, with topologies and features optimized for different applications. One alternative that scope users often consider is to skip the differential amplifier and "float the scope" by defeating the power cord's protective ground. Because the reference lead of a floating scope's probe is disconnected from earth ground, the reference lead, along with the entire scope chassis, can reach the common-mode potential. You should never float a scope, however, and there are several reasons for not doing so. The most obvious is safety. If the common-mode signal is a high voltage, you risk serious injury or even electrocution from accidental contact with conductive parts of the scope.

A second problem is that the shells of all of the scope's input BNCs are tied to the chassis. Therefore, with a scope's normal single-ended input circuitry, floating the scope doesn't let you view multiple waveforms referenced to different potentials.

The external-trigger function, a powerful tool in many switch-mode power-supply measurements, can also create problems when you try to float a scope. The signal that
drives the trigger input must be referenced to the scope chassis. Unfortunately, the reference for the external-trigger signal and the signals you want to view can be different. If you connect the scope chassis to a point in the circuit under test whose reference potential differs from that of the trigger signal, external triggering often does not work. Similarly, if the scope chassis is not at the same potential as the chassis of external devices that communicate via the parallel, RS-232C, or IEEE-488 ports, the device or the scope may be damaged. Moreover, with the ground severed, the scope may radiate excessive EMI, adding noise to the measurement or to the circuit under test.

A less well-known problem with using a floating scope is the high likelihood of corrupting the measurement. When you float a scope, the resistance that loads the circuit under test approaches infinity, but the capacitive reactance does not. The positive input (probe tip) presents a very low capacitive load. In contrast, the negative input (probe reference lead) connects directly to the oscilloscope chassis and can present a huge capacitive load. Even when you add a high-quality isolation transformer to the scope's power cable, the reference-side capacitance to ground is typically 100 to 200 pF. This capacitance acts with the source impedance of the circuit under test to attenuate only one side of the signal, converting a significant fraction of the common-mode voltage to a differential signal.

The inductance of the probe lead also acts with the high capacitance to form a series-LC resonator, which generates ringing in the waveform from the input signal's V_DM and V_CM components. Finally, capacitive loads can damage or destroy some circuits, such as totem-pole gate drivers. A suitable differential amplifier overcomes all of these difficulties.

High-voltage differential probes are a safe, low-cost alternative to floating the scope for performing measurements in power devices, such as off-line, switch-mode power supplies; electronic lamp ballasts; and motor drives. High-voltage differential probes have two ranges, optimized for measuring the saturated and off-state waveforms of switching semiconductors. Unlike conventional oscilloscope-input stages, which use switched attenuators for range changing, these probes change ranges by changing amplifier gain. In amplifiers with switchable input attenuators, the maximum common-mode-voltage range decreases at the higher gain ranges. Because they do not switch input attenuation, high- voltage differential probes maintain full common-mode range in both low and high gain settings.

You can also use high-gain differential amplifiers as differential-input devices. These devices enable conventional oscilloscopes to measure signals as small as a few microvolts. These amplifiers usually incorporate selectable-bandwidth filtering to reduce out-of-band noise. Additional features include differential offset-to-null potentials caused by thermocouples that form in input-wiring and instrumentation- amplifier configurations, which eliminates loading effects when measuring circuits that have high source impedance.

A third type of differential input device for oscilloscopes is the high-bandwidth differential probe. These probes locate a differential amplifier near the probe tips. This construction provides excellent CMRR even at high frequencies. These probes are the best choice for measurements in read-channel circuits of disk drives and video and audio magnetic-storage systems. Other applications include telecommunication systems and general measurements in high-speed digital circuits.

Differential comparators are a related type of oscilloscope-input device. These wideband differential amplifiers contain a precise, built-in, adjustable voltage reference. You can switch the reference into one input and make single-ended measurements with adjustable calibrated offset. The input stage includes special clamp circuitry that enables the amplifier to recover from overloads very rapidly. These features allow such special amplifiers to measure the dynamic saturation characteristics of power semiconductors and to perform accurate settling-time measurements on amplifiers and D/A converters.

Optimize differential-amplifier performance

In any differential measurement, the relative gain match of both inputs has a first-order effect on the CMRR. Therefore, the slightest input mismatch seriously degrades the CMR. When you use probes with a differential amplifier, be sure to use the same probe type and cable length on both inputs. If possible, use 1× probes.

Oscilloscope manufacturers offer special attenuating probe sets for use with differential amplifiers when you need low loading. These probes incorporate adjustments that allow you to match the dc attenuation and ac compensation of both probes to the amplifier. Carefully follow the adjustment instructions, and readjust the probes whenever you move them to a different amplifier or interchange the channels. Differences in source impedance affect the channels' attenuation match. Try to connect the probes to circuit points whose source impedance is similar and, preferably, low.

You can improve CMR by twisting together the input-probe leads. When the leads are spread apart, any magnetic field passing through the loop induces a differential voltage that is faithfully amplified. Twisting the leads tends to cancel the induced voltage. Make sure to set the oscilloscope input impedance to the value the signal conditioner is designed to drive (often, 50[ohm]). Finally, remember to factor the gain of any external signal conditioner into the scope's scale-factor readout, scalar-measurement results, math-generated waveforms, and trigger level. (See box, "Intelligent probe interface eliminates interpretation errors.") The on-screen readout of input coupling, bandwidth limiting, and offset may not represent the true settings of the differential amplifier.

Steve Sekel is a program manager for new-product development at Tektronix Inc, Beaverton, OR, where his recent focus has been on power measurement and differential acquisition solutions. Sekel has been with Tektronix since completing his BSEE degree in 1978. His outside interests include woodworking, photography, and cross-country skiing.