Even when a measurement setup avoids ground loops or analog input stage saturation by following the above guidelines, the measured signal will almost inevitably include some amount of noise or unwanted signal "picked up" from the environment. This is especially true for low-level analog signals that are amplified using the onboard amplifier that is available in many data acquisition devices. To make matters worse, PC data acquisition boards generally have some digital input/output signals on the I/O connector. Consequently, any activity on these digital signals provided by or to the data acquisition board that travels across some length in close proximity to the low-level analog signals in the interconnecting cable itself can be a source of noise in the amplified signal. In order to minimize noise coupling from this and other extraneous sources, a proper cabling and shielding scheme may be necessary.
Before proceeding with a discussion of proper cabling and shielding, an understanding of the nature of the interference noise-coupling problem is required. There is no single solution to the noise-coupling problem. Moreover, an inappropriate solution might make the problem worse.
An interference or noise-coupling problem is shown in Figure 12.
Noise-Coupling Problem Block Diagram
As shown in Figure 12, there are four principal noise "pick up" or coupling mechanisms—conductive, capacitive, inductive, and radiative. Conductive coupling results from sharing currents from different circuits in a common impedance. Capacitive coupling results from time-varying electric fields in the vicinity of the signal path. Inductive or magnetically coupled noise results from time-varying magnetic fields in the area enclosed by the signal circuit. If the electromagnetic field source is far from the signal circuit, the electric and magnetic field coupling are considered combined electromagnetic or radiative coupling.
Conductively Coupled Noise
Conductively coupled noise exists because wiring conductors have finite impedance. The effect of these wiring impedances must be taken into account in designing a wiring scheme. Conductive coupling can be eliminated or minimized by breaking ground loops (if any) and providing separated ground returns for both low-level and high-level, high-power signals. A series ground-connection scheme resulting in conductive coupling is illustrated in Figure 13a.
If the resistance of the common return lead from A to B is 0.1 Ω, the measured voltage from the temperature sensor would vary by 0.1 Ω * 1 A = 100 mV, depending on whether the switch is closed or open. This translates to 10° of error in the measurement of temperature. The circuit of Figure 13b provides separate ground returns; thus, the measured temperature sensor output does not vary as the current in the heavy load circuit is turned on and off.
Conductively Coupled Noise
Capacitive and Inductive Coupling
The analytical tool required for describing the interaction of electric and magnetic fields of the noise and signal circuits is the mathematically nontrivial Maxwell’s equation. For an intuitive and qualitative understanding of these coupling channels, however, lumped circuit equivalents can be used. Figures 14 and 15 show the lumped circuit equivalent of electric and magnetic field coupling.
Capacitive Coupling between the Noise Source and Signal Circuit, Modeled by the Capacitor Cef in the Equivalent Circuit
Inductive Coupling between the Noise Source and Signal Circuit, Modeled by the Mutual Inductance M in the Equivalent Circuit
Introduction of lumped circuit equivalent models in the noise equivalent circuit handles a violation of the two underlying assumptions of electrical circuit analysis; that is, all electric fields are confined to the interior of capacitors, and all magnetic fields are confined to the interior of inductors.
The utility of the lumped circuit equivalent of coupling channels can be seen now. An electric field coupling is modeled as a capacitance between the two circuits. The equivalent capacitance Cef is directly proportional to the area of overlap and inversely proportional to the distance between the two circuits. Thus, increasing the separation or minimizing the overlap will minimize Cef and hence the capacitive coupling from the noise circuit to the signal circuit. Other characteristics of capacitive coupling can be derived from the model as well. For example, the level of capacitive coupling is directly proportional to the frequency and amplitude of the noise source and to the impedance of the receiver circuit. Thus, capacitive coupling can be reduced by reducing noise source voltage or frequency or reducing the signal circuit impedance. The equivalent capacitance Cef can also be reduced by employing capacitive shielding. Capacitive shielding works by bypassing or providing another path for the induced current so it is not carried in the signal circuit. Proper capacitive shielding requires attention to both the shield location and the shield connection. The shield must be placed between the capacitively coupled conductors and connected to ground only at the source end. Significant ground currents will be carried in the shield if it is grounded at both ends. For example, a potential difference of 1 V between grounds can force 2 A of ground current in the shield if it has a resistance of 0.5 Ω. Potential differences on the order of 1 V can exist between grounds. The effect of this potentially large ground current will be explored further in the discussion of inductively coupled noise. As a general rule, conductive metal or conductive material in the vicinity of the signal path should not be left electrically floating either, because capacitively coupled noise may be increased.
Improper Shield Termination—Ground Currents Are Carried in the Shield
Proper Shield Termination—No Ground or Signal Current Flows through the Shield
As described earlier, inductive coupling results from time-varying magnetic fields in the area enclosed by the signal circuit loop. These magnetic fields are generated by currents in nearby noise circuits. The induced voltage Vn in the signal circuit is given by the formula:
n = 2p fBACosÆ (1)
where f is the frequency of the sinusoidally varying flux density, B is the rms value of the flux density, A is the area of the signal circuit loop, and Æ is the angle between the flux density B and the area A.
The lumped circuit equivalent model of inductive coupling is the mutual inductance M as shown in Figure 15(b). In terms of the mutual inductance M, Vn is given by the formula:
n = 2p fMI
where In is the rms value of the sinusoidal current in the noise circuit, and f is its frequency.
Because M is directly proportional to the area of the receiver circuit loop and inversely proportional to the distance between the noise source circuit and the signal circuit, increasing the separation or minimizing the signal loop area will minimize the inductive coupling between the two circuits. Reducing the current In in the noise circuit or reducing its frequency can also reduce the inductive coupling. The flux density B from the noise circuit can also be reduced by twisting the noise source wires. Finally, magnetic shielding can be applied either to noise source or signal circuit to minimize the coupling.
Shielding against low-frequency magnetic fields is not as easy as shielding against electric fields. The effectiveness of magnetic shielding depends on the type of material—its permeability, its thickness, and the frequencies involved. Due to its high relative permeability, steel is much more effective than aluminum and copper as a shield for low-frequency (roughly below 100 kHz) magnetic fields. At higher frequencies, however, aluminum and copper can be used as well. Absorption loss of copper and steel for two thicknesses is shown in Figure 18. The magnetic shielding properties of these metals are quite ineffective at low frequencies such as those of the power line (50 to 60 Hz), which are the principal low-frequency, magnetically-coupled noise sources in most environments. Better magnetic shields such as Mumetal can be found for low-frequency magnetic shielding, but Mumetal is very fragile and can have severe degradation of its permeability, and hence, degradation of its effectiveness as a magnetic shield by mechanical shocks.
Absorption Loss as a Function of Frequency (from Reference 1)
Because of the lack of control over the noise circuit parameters and the relative difficulty of achieving magnetic shielding, reducing the signal circuit loop area is an effective way to minimize inductive coupling. Twisted-pair wiring is beneficial because it reduces both the loop area in the signal circuit and cancels induced errors.
Formula (2) determines the effect of carrying ground-loop currents in the shield for the circuit in Figure 16. For In = 2 A; f = 60 Hz; and M = 1 µH/ft for a 10-ft cable results in the following:
n = (2)(3.142)(60)(1 ´ 10
–6 ´ 10)(2) = 7.5 mV
This noise level translates into 3.1 LSB for a 10 V range, 12-bit data acquisition system. The effectiveness of the data acquisition system is thus reduced roughly to that of a 10-bit acquisition system.
When using an E Series device with a shielded cable in differential mode, the signal circuit loop area is minimized because each pair of signal leads is configured as a twisted pair. This is not true for the single-ended mode with the same device and cable because loop areas of different sizes may be formed with different channels.
Current signal sources are more immune to this type of noise than voltage signal sources because the magnetically induced voltage appears in series with the source, as shown in Figure 19. V21 and V22 are inductively coupled noise sources, and Vc is a capacitively coupled noise source.
Circuit Model of Inductive and Capacitive Noise Voltage Coupling
(H. W. Ott, Noise Reduction Techniques in Electronic Systems, Wiley, 1976.)
The level of both inductive and capacitive coupling depends on the noise amplitude and the proximity of the noise source and the signal circuit. Thus, increasing separation from interfering circuits and reducing the noise source amplitude are beneficial. Conductive coupling results from direct contact; thus, increasing the physical separation from the noise circuit is not useful.
Radiative coupling from radiation sources such as radio and TV broadcast stations and communication channels would not normally be considered interference sources for the low-frequency (less than 100 kHz) bandwidth measurement systems. But high-frequency noise can be rectified and introduced into low-frequency circuits through a process called audio rectification. This process results from the nonlinear junctions in ICs acting as rectifiers. Simple passive R-C lowpass filters at the receiver end of long cabling can reduce audio rectification.
The ubiquitous computer terminal is a source of electric and magnetic field interference in nearby sensitive circuits. This is illustrated in Figure 20, which shows the graphs of data obtained with a data acquisition device using a gain of 500 with the onboard programmable gain amplifier. The input signal is a short circuit at the termination block. A 0.5 m unshielded interconnecting cable was used between the terminal block and the device I/O connector. For differential signal connection, the channel high and channel low inputs were tied together and to the analog system ground. For the single-ended connection, the channel input was tied to the analog system ground.
Noise Immunity of Differential Input Configuration Compared with that of RSE Configuration (DAQ board gain: 500; Cable: 0.5 m Unshielded; Noise Source: Computer Monitor)
Miscellaneous Noise Sources
Whenever motion of the interconnect cable is involved, such as in a vibrational environment, attention must be paid to the triboelectric effect, as well as to induced voltage due to the changing magnetic flux in the signal circuit loop. The triboelectric effect is caused by the charge generated on the dielectric within the cable if it does not maintain contact with the cable conductors.
Changing magnetic flux can result from a change in the signal circuit loop area caused by motion of one or both of the conductors—just another manifestation of inductive coupling. The solution is to avoid dangling wires and to clamp the cabling.
In measurement circuits dealing with very low-level circuits, attention must be paid to yet another source of measurement error—the inadvertent thermocouples formed across the junctions of dissimilar metals. Errors due to thermocouple effects do not constitute interference type errors but are worth mentioning because they can be the cause of mysterious offsets between channels in low-level signal measurements.