The heart of the system is the differential amplifier stage (see Figure 14). The schematic symbol is the same as an op amp. Like the operational amplifier, a differential amplifier rejects the input common-mode signal and only amplifies the voltage difference between the two inputs. Unlike the op amp, the differential amplifier has a known, finite gain. In some configurations, the gain is user-selectable. The output is singled-ended and referenced to ground. The inputs are often FETs to give very high impedance. The input signal may pass through a high-impedance attenuator to reduce larger signals to a range the amplifier can handle. The demands on the attenuator are much greater than those in a single-ended amplifier. Both sides must have identical DC and AC attenuation. The amount of mismatch has a first-order effect on the CMRR. For example, to maintain a 100,000:1 CMRR specification, the attenuators must match to better than one part in 100,000 (0.001%); this leaves no margin for error in the differential amplifier! Of course, this match needs to be maintained all the way from the signal source.

 

Figure 14. Simplified schematic of a differential amplifier with attenuator.

Input Connections

Interconnecting the differential amplifier or probe to the signal source is generally the greatest source of error. To maintain the input match, both paths should be as identical as possible. Any cabling should be of the same length for both inputs. If probes are used, they should be the same model and length. When measuring low-frequency signals with large common-mode voltages, avoid the use of attenuating probes. At high gains, they simply cannot be used as it is impossible to precisely balance their attenuation. When attenuation is needed for high-voltage or high-frequency applications, special passive probes designed specifically for differential applications should be used. These probes have provisions for precisely trimming DC attenuation and AC compensation. To get the best performance, a set of probes should be dedicated to each specific amplifier and calibrated with that amplifier using the procedure included with the probes.

Figure 16. With the input leads twisted together, the loop area is very small, hence less field passes through it. Any induced voltage tends to be in the VCM path which is rejected by the differential amplifier.

It's common practice to twist the + and - input cables together in a pair. This reduces line frequency and other noise pick up. Input cabling that is spread apart (see Figure 15) acts as a transformer winding. Any AC magnetic field passing through the loop induces a voltage which appears to the amplifier input as differential and will be faithfully summed into the output! With the input leads twisted together (Figure 16), any induced voltage tends to be in the VCM path, which is rejected by the differential amplifier.

 

Figure 15. Time varying magnetic fields passing through the open leads induce a voltage as in a transformer winding. This voltage appears as a differential component to the amplifier and is summed into the true Vdm signal.

 

Figure 16. With the input leads twisted together, the loop area is very small, hence less field passes through it. Any induced voltage tends to be in the VCM path which is rejected by the differential amplifier.

High-frequency measurements subject to excessive common-mode can be improved by winding both input leads through a ferrite torroid. This attenuates high-frequency signals which are common to both inputs. Because differential signals pass through the core in both directions, they are unaffected.

Grounding

The input connectors of most differential amplifiers are BNC connectors with the shell grounded. When using probes or coaxial input connections, there's always a question of what to do with the grounds. Because the measurement application varies, there are no hard and fast rules.

When measuring low-level signals at low frequencies, it's generally best to connect the grounds only at the amplifier end and leave both unconnected at the input end. This provides a return path for any currents induced into the shield, but doesn't create a ground loop which may upset the measurement or the device-under-test.

At higher frequencies, the probe input capacitance, along with the lead inductance, forms a series resonant "tank" circuit which may ring. In single-ended measurements, this effect can be minimized by using the shortest possible ground lead. This lowers the inductance, effectively moving the resonating frequency higher, hopefully beyond the bandwidth of the amplifier. Differential measurements are made between two probe tips, and the concept of ground does not enter into the measurement. However, if the ring is generated from a fast rise of the common-mode component, using a short ground lead reduces the inductance in the resonant circuit, thus reducing the ring component. In some situations, a ring resulting from fast differential signals may also be reduced by attaching the ground lead. This is the case if the common-mode source has very low impedance to ground at high frequencies, i.e. is bypassed with capacitors. If this is not the case, attaching the ground lead may make the situation worse! If this happens, try grounding the probes together at the input ends. This lowers the effective inductance through the shield.

Of course, connecting the probe ground to the circuit may generate a ground loop. This usually doesn't cause a problem when measuring higher-frequency signals. The best advice when measuring high frequencies is to try making the measurement with and without the ground lead; then use the setup which gives the best results. When connecting the probe ground lead to the circuit, remember to connect it to ground! It's easy to forget where the ground connection is when using differential amplifiers since they can probe anywhere in the circuit without the risk of damage.

Input Impedance Effects on CMRR

Any source impedance acts to form a voltage divider with the input resistance (DC) and capacitance (AC) of the input. With single-ended measurements, the impedance effect can usually be ignored as the error seldom reaches 1%. But with differential measurements, this small error contributes to the input-gain mismatch, which reduces common-mode rejection (see Figure 17).

 

Figure 17. Effect of unequal source impedances. The + input attenuator is essentially driven from 0 ohms, however the - input attenuator is driven from something less than 3 kohms. This adds to the 900 kohms, increasing its attenuation and lowering the CMRR.

The differential amplifier CMRR specification is usually measured with both inputs driven together via a BNC tee connector. This effectively gives zero impedance difference looking into the inputs. Ideally, the real-life signal source would also have identical driving impedance. However, they seldom do. As such, the real CMRR performance will be significantly less than the amplifier specification.

If the amplifier's input impedance, attenuation ratio, and source impedances are all known, it's possible to determine the actual CMRR by calculating the actual divider ratios in each input arm. However, it's easier just to make a subjective judgment of the measurement performance.

Many high-gain amplifiers have provisions for configuring them as instrumentation amplifiers. An instrumentation amplifier has no input attenuator. The input resistance is essentially infinite (>1012 ohms). This mode greatly enhances low-frequency CMR when the source impedance is rather high, such as physiological experiments. While instrumentation amplifiers have infinite input resistance, they still have input capacitance. The CMR improvement with high source impedances will quickly degrade as the common-mode frequency increases. Because instrumentation amplifiers don't have input attenuators, they have limited common-mode and differential-mode dynamic ranges.

Common-Mode Range

Any amplifier can be overdriven, causing the output to "clip". The same effect occurs in a differential amplifier when the input differential-mode signal is large enough to force the amplifier beyond its output dynamic range. Differential amplifiers are also subject to another overload condition - exceeding the input common-mode range. This condition occurs when the voltage that the desired signal is riding on (VCM) exceeds the amplifiers input common-mode range.

Because the common-mode signal is rejected by the amplifier, the dynamic range is limited by the input stage rather than the output swing. Amplifiers with input attenuators have a greater common-mode range than differential-mode range. Because the common-mode component is (hopefully) not seen in the measurement, common-mode range overload may not be obvious to the user. This is especially true when the common-mode component is DC. Some amplifier topologies will still produce an approximate rendition of the differential signal with a significant gain error when the VCM range is exceeded. Because the waveform appears correct, many users have been fooled by this erroneous measurement.

Some amplifiers have overload indicators to warn the user of a common-mode overload condition. It's a good practice to verify that the common mode is within specified range before making critical measurements. This is easily done by moving one of the input connections to ground and measuring the common-mode component with the amplifier itself. The procedure is then repeated with the other input.

Measuring Totally Floating Signals

 

Figure 18. VCM in a consumer audio electronic component. These devices usually have a two-wire power cord with their chassis and circuitry floating.

Signal sources which are totally floating, having no connection whatsoever to ground, pose a special problem when being measured with a differential amplifier. Common examples include battery-operated electronic equipment, consumer audio components, and experimental physiological specimens. Because there's no shunting impedance to ground, any AC fields in the area will be capacitively coupled into the device being measured (see Figure 18). Line-frequency fields, radiated from fluorescent lighting and building wiring, are common in this measurement environment. When coupled into the DUT, the line-frequency field produces a common-mode voltage. With sufficient coupling and high input impedance of the amplifier, it's possible to inadvertently exceed the common-mode range of the amplifier. This is especially true with amplifiers configured as instrumentation amplifiers, since the load impedance at line frequencies approaches infinity.

The overload situation can be avoided by providing a shunting impedance to ground, reducing the capacitive coupling, or reducing the field strength. Adding a shunt path to ground is the easiest approach. It need not be a direct short, often a 10 kohm resistor is sufficient. If adding the shunt impedance upsets the device being measured or the measurement, try reducing the capacitive coupling by enclosing the DUT with a metal screen which is tied to ground. This effectively adds a Faraday shield which provides a shunt path to ground for AC fields. A final approach is to try to minimize the field strengths. Substituting incandescent lighting for fluorescent, and maximizing the distance between line-connected wiring and the DUT are good starting points.

Bandwidth

Differential amplifiers, like single-ended scope amplifiers, often include a bandwidth limiting control. High-gain amplifiers may offer a choice of low-pass frequencies. Bandwidth limiting reduces high-frequency noise components with minimal degradation on lower frequencies. The bandwidth limiting filters are located after the input signal has been transformed to single ended. Therefore, their use will not increase input common-mode range at higher frequencies.