Guide to Data Acquisition: Signal conditioning (part 2)

Noise and interference

Definition of noise and interference

Noise, by definition, is the presence of an unwanted electrical signal in a circuit. Interference is the undesirable effect of noise. Where a noise voltage causes improper operation of a circuit, or its relative magnitude is of the same order as the desired electrical signal, then it's interference. Noise itself can't be totally eliminated but only reduced in magnitude until it no longer causes interference. This is especially true in data acquisition systems where the analog signal levels from transducers measuring a physical quantity can be very small. Compounding this in many instances is the physical cable distance over which these signals must be transmitted and the effect that noise may have on this extended circuitry.

Sources and types of noise

Before considering the cabling and shielding requirements of data acquisition systems, it's important to understand the nature and source of interference caused by the coupling of noise into data acquisition systems.

C-29 illustrates that there are three components involved in any noise-induced problem:

• A noise source (AC power cables, high voltage or high current AC or switching circuitry)
• A coupling channel (common impedance, capacitance, mutual inductance)
• A receiver (the circuitry that is susceptible to the induced noise)

C-29 Noise coupling between a noise source and a receiver

The mechanisms for coupling noise most common to data acquisition and control applications are as follows:

• Conductive coupling
• Capacitive coupling
• Inductive coupling

Conductive coupling

Conductive coupling occurs where two or more circuits share a common signal return. In such cases, return current from one circuit, flowing through the finite impedance of the common signal return, results in variations in the ground potential seen by the other circuits. A series ground connection scheme resulting in conductive coupling is shown in Figr. C-30. If the resistance of the common return lead is 0.1 Ohm and the return current from all other circuits is 1A, then the voltage measured from the temperature sensor, (VT), would vary by 0.1 Ohm × 1A = 100 mV, corresponding to 10 degrees error in the temperature measured.

C-30 Series ground connections resulting in conductive coupling

Capacitive coupling

Electrical fields occur in the vicinity of voltage-varying sources. Capacitive coupling is the transmission of external noise through mutual and stray capacitances between a noise source and receiving circuit. This is sometimes referred to as electrostatic coupling, although this is a misnomer, since the electrical fields are not static. Since cables tend to be the longest circuit elements, capacitive coupling is best demonstrated by considering a signal circuit connecting a signal source to a measurement system by a pair of long signal-carrying conductors.

The physical representation of electric field coupling between a noise source and such a signal circuit is shown.

(top) Physical representation of an electrical field coupling into a signal circuit

The equivalent circuit representation for this system is shown.

(above, bottom) Equivalent circuit representation of an electric field coupling into a signal circuit

Where the source resistance (RS) is much less than the load resistance (RL) and also much lower than the impedance of the stray capacitances (C12 and C2G) (i.e. RS << 1/j ω[C12 + C2G]), then Vn = j ω Rs C12 Vn

The preceding equation clearly shows that the capacitively coupled noise voltage is directly proportional to the frequency and amplitude of the external noise source, the resistance to ground of the signal circuit, which in this case is RS, and the mutual capacitance between them. Where the signal source resistance is comparable in magnitude to the load resistance, and their combined resistances to ground are much larger than the impedance of the stray capacitances (C12 and C2G) (i.e. RS 1/j ω[C12 + C2G]), then it can be shown that the capacitively-coupled noise voltage, is independent of the frequency of the noise source, and is much greater than in the case where the same resistance is relatively small.

This equation shows that the capacitively-coupled noise voltage is independent of the frequency of the noise source and is much greater in magnitude than in the case where the source resistance is relatively small. Where the amplitude and the frequency of the noise source can't be altered, the only means for reducing capacitive coupling into the signal circuit is to reduce the equivalent signal circuit resistance to ground or reduce the mutual stray capacitance. The mutual stray capacitance can be reduced by increasing the relative distance of the signal wires from the noise source, correct orientation of the conductors, or by shielding.

Magnetic field coupling

Magnetic field coupling or inductive coupling is the mechanism by which time-varying magnetic fields produced by changing currents in a noise source, link with current loops of receiving circuits. The physical representation of magnetic field coupling between a noise source and a signal circuit is shown.

Physical representation of magnetic field coupling between a noise source and a signal circuit

Lenz’s law states that the voltage, Vn induced into a closed loop signal circuit of area A is proportional to the rate of change of the magnetic field coupling the circuit loop, the flux density (B) of the magnetic field and the area of the loop. This is represented by the formula:

Vn = 2 f BA cos φ(10^-4)

Where:

f = the frequency of the sinusoidal varying flux density

B = the rms value of the flux density (gauss)

A = the area of the signal circuit loop (m^2)

φ = the angle between the flux density (B) and the area (A).

This equation indicates that the noise voltage can be reduced by reducing B, A, or cos φ. The flux density (B) can be reduced by increasing the distance from the source of the field or if the field is caused by currents flowing through nearby pairs of wires, twisting those wires to reduce the net magnetic field effect to zero and or by alternating its direction. The signal circuit loop area (A) can be reduced by placing the signal wires of the receiving circuit current loop closer together. e.g., consider a signal circuit whose current carrying wires are 1 meter long and 1 centimeter apart, lying within a 10 gauss 60 Hz magnetic field, typical of fans, power wiring and transformers. The maximum voltage induced in the wires occurs for φ = 0º. Vn = (2 π × 60)(1)(1 × 10^–2)(10^–4) = 3.7 mV. If the distance between the wires is reduced to 1 mm the noise voltage is reduced tenfold to 0.37 mV. The cos φ, term can be reduced by correctly orienting the wires of the signal circuit in the magnetic field. e.g., if the signal wires were perpendicular to the magnetic field ( φ = 90°) the induced voltage could be reduced to zero, although practically this would not be possible. Running the signal wires together in the same cable as the wires carrying the noise current source would maximize the induced noise voltage. The equivalent circuit model of magnetic coupling between a noise source and a signal circuit is shown in Figr. C-34. In terms of the mutual inductance (M), Vn is given by:

Vn = 2 π f M IN

Where:

IN is the rms value of the sinusoidal current in the noise circuit and f is its frequency. The mutual inductance (M) is directly proportional to the area (A) of the signal circuit current loop and the flux density, (B).

The physical geometry of the current loop of the receiving signal circuit, specifically its area, is the key to why it's susceptible to magnetic fields and how to minimize the effect. Cables provide the longest and largest current loop. The effect of magnetic coupling is best demonstrated by considering the circuit of Figr. C-34, in which the signal cable current loop is coupled by a sinusoidal changing magnetic field with a peak flux density of B φ.

Figr. C-34 Equivalent circuit model of magnetic coupling between a noise source and a signal circuit

Ideally, the only voltage appearing across the load should be VS -- the source signal voltage. However, the magnetic flux induces a voltage in the loop that appears in series with the receiver signal circuit. The voltage appearing across the load is the sum of the source voltage and the unwanted magnetic field induced voltage (VN).

Twisting the insulated conductors of the loop together, as shown in Figr. C-35, can greatly reduce the amount of magnetic coupling into the signal lines.

Reducing magnetic noise coupling by twisting of wires

The voltage induced in each section of the loop now alternates phases; its magnitude reduced by the reduction in area of each twisted loop (i.e. 1/4). Provided there is an even number of twists in the signal conductors, the voltages due to the magnetic field cancel out and only the desired signal voltage appears across the load.

Minimizing Noise

Cable shielding and shield earthing

The effects of noise due to capacitive coupling can be greatly reduced by the use of a cylindrical metal shield placed around the signal-carrying conductor. Consider the equivalent circuit shown in Figr. C-36, in which the signal conductors are completely enclosed by the ungrounded shield.

Equivalent representation of a signal circuit, completely surrounded by a capacitive shield

Note that as the signal conductors are completely enclosed, there is no stray capacitance between the signal conductors and ground.

Where the source resistance (RS) is much less than the load resistance (RL) and also much lower than the impedance of the shield to signal conductor stray capacitance (C2S), (i.e. RS << 1 / j ωC2S), then the noise voltage capacitively coupled onto the signal line can be shown to be:

Vn = j ω RS C2S VNS

Where the shield is grounded (i.e. VNS =0), then the noise voltage induced in signal conductor is also zero.

Completely surrounding the signal carrying conductors is not practical in most instances, since conductors will extend beyond their shield. Also, in the case of a braided shield there is a small stray capacitance due to the holes in the braiding. Where signal conductors extend beyond the shield, coupling capacitance between the signal conductor and the noise source (C12) and between the conductors and ground (C2G) will still exist, although they will be much smaller. This is shown.

Figr. C-37 Equivalent representation of a practical circuit in which the capacitive shield does not completely surround the signal circuit

Where the source resistance (RS) is much less than the load resistance (RL) and also much lower than the impedance of the stray capacitances (C12 and C2G) (i.e. RS << 1/j ω[C12 + C2G + C2S]), then the noise voltage induced by external noise source onto the signal conductor is given by:

Vn = j ω RS C12 VN

This is the same as for the unshielded conductor, however the mutual stray capacitance (C12) will be much less because of the shield.

The value of C12 depends on the length of the signal conductor extending beyond the shield.

Capacitive shielding works by bypassing or providing another path for induced noise currents to flow, so that they are not carried in the signal circuits. The rules of shielding are as follows:

• For a shield to be effective it should be well grounded and the length of conductors extending beyond the end of the shield minimized. The screen continuity should be maintained at each termination point.
• The screens of individually screened cores in the same cable should be electrically isolated from each other, but continuous for each line through terminal junctions.

Grounding cable shields

To be fully effective, capacitive shielding also requires attention to the number and location of shield earths. In the way that the grounding of signal lines at both ends of a circuit may cause significant ground currents to flow, the same is also true for cable shields. e.g., a potential difference of only 1 V between the grounds at either end of a circuit will drive a current of 2A around the current loop if its resistance is 0.5 Ohm . Where the current flow is significant, and the ground loop created by earthing of the shield has a large area, shield currents may inductively couple unequal voltages into the signal cables and be a source of interference. Where possible, shields should be earthed at one end only. The placement of shield earths depends on the grounding of the signal source and the type of measurement system used. Figr. C-38 shows the preferred shield grounding when measuring an ungrounded signal source, using a measurement system where the signal lines are referenced to the amplifier common. it's assumed that amplifier common, although normally connected to ground may have a potential (ΔVg1) relative to ground potential. ΔVg2 represents the difference in ground potential. The circuit equivalent for this system shows that in this configuration neither of the noise voltages (ΔVg1 or ΔVg2) appears across the input terminals of the amplifier. Instead, if the shield was earthed at point B, then the noise voltage across the input terminals of the amplifier would be the voltage across the impedance of C2 as part of the voltage divider formed with C1.

(top) Shield grounding when measuring an ungrounded source with a grounded measurement system

When an ungrounded (differential) measurement system is used to measure a grounded source the preferred cable shielding is shown in Figr. C-39. The voltage Vg1 represents the potential of the source common above earth ground potential.

(above, bottom) Shield grounding when measuring a grounded source with an ungrounded measurement system

The equivalent circuit for this measurement system again shows that the noise voltage appearing across the input terminals of the amplifier, is zero. If the shield was grounded at the other end of the cable at point D, then the noise voltage across the input terminals of the amplifier would be the voltage across the impedance of C2 as part of the voltage divider formed with C1. Where the signal circuit is required to be grounded at both ends, the difference in ground potential and the susceptibility of the ground loop to inductive coupling deter mines the amount of noise in the circuit. The preferred shield grounding configuration when there is no other alternative is shown, in which a portion of the ground loop current is bypassed through the lower impedance shield.

Preferred shield grounding when measuring a grounded source with a grounded measurement system

Breaking the ground loop on the signal lines using transformers or optical couplers can provide additional noise reduction. The rules of shield grounding are as follows:

• Where possible, cable shields should be earthed at one end only.
• Where the source is ungrounded and the signal amplifier is grounded, the input shield should always be connected to the amplifier common terminal, even if this point is not at earth ground.
• Where the source is grounded and the signal amplifier is ungrounded, the input shield should be connected to the source common terminal, even if this point is not at earth ground.

Grounding the shield has additional benefits such as providing a path for RF currents and preventing the build-up of static charge by providing a discharge path to ground.

Shielded and twisted-pair cables

Cables with copper conductors and plastic insulation are still the most common and reliable solution. This is not surprising as they combine the important elements of good electrical characteristics, low cost, mechanical flexibility, ease of installation and ease of termination. Aluminum conductors are seldom used for data communication cables because of the higher resistance and other physical limitations. The cable resistance depends on the cross-sectional area of the conductor (usually expressed in mm2) and the length of the cable. The thicker the conductor, the lower the resistance, the lower the signal volt drop, and the higher the current it can carry without excessive heating. The signal voltage drop, Vdrop =I (R+(2 π f L – 1/2 πf C)), depends on the:

• Frequency of signal
• Line current, which is dependent on the receiver input impedance, and
• Conductor resistance, which is dependent on wire size and length.

For DC voltages and low-frequency signals the resistance of the conductor is the only major concern. The voltage drop along the cable affects the magnitude of the signal volt age at the receiving end. In the presence of noise, this affects the signal-to-noise (S/N) ratio and thus the quality of the signal received. As the frequency (or data transfer rate) increases, the other characteristics of the cable, such as capacitance and series inductance, become important. Inductance and capacitance are factors that depend on the construction of the cable and on the type of insulation material. The resistance, inductance and capacitance are distributed along the length of the cable. At high frequencies they combine to present the effects of a low pass filter. The simplified electrical single-line diagram of a cable shows these electrical parameters distributed along the length of the cable and can be seen in C-41. Note, however, that a more complex model would also need to include a minor conductance factor (the inverse of resistance) in parallel across the cable.

C-41

Main parameters of a cable

To derive the optimum performance from a cable, the correct type and size must be used. The following general rules apply to most applications:

• Low data transfer rates: use low-frequency cables (e.g., twisted-pair cables)
• High data transfer rates: use high-frequency cables (e.g., coaxial cables, optical fiber; though there are some new types of twisted-pair cables that give very good high-frequency performance)
• High noise environment: use shielded copper or optic fiber cables

Twisted-pair cables

Twisted-pair cables are the most economical solution for data transmission (differential circuit). They allow for transmission rates of up to 100 Mbps on communication links of up to 300 m (or even longer distances but with lower data transfer rates). Some new types of twisted-pair cables (e.g., ‘Twistlan’) are suitable for up to 100 Mbps. Twisted-pair cables can be STP (shielded twisted-pair) or UTP (unshielded twisted-pair). Twisted-pair cables are made from two identical insulated conductors that are twisted together along their length a specified number of times per meter, typically 40 twists per meter (12 twists per foot). The wires are twisted to reduce the effect of electromagnetic and electrostatic induction. An earth screen, or shield, is often placed around them as well to reduce the capacitance-induced noise, and an insulating PVC sheath usually provides mechanical protection. As the cross-sectional area of the conductor affects IR loss, heavier conductor sizes are recommended for long distances. The capacitance of a twisted-pair is low at about 15 to 50 pF/m, allowing a reasonable bandwidth and an achievable slew rate.

For full-duplex systems using balanced differential transmission, two sets of screened twisted-pair conductors are required in one cable, with both individual and overall screens. The entire cable is covered with a protective PVC sheath.

Coaxial cables

Coaxial cables are used in applications that require high data transfer rates of up to 10 Mbps or high-frequency analog signals over long distances. Coaxial cables are more expensive than twisted-pair cables. They consist of a central conductor running through an enclosing cylinder on the same axis. This enclosing cylinder is made of a conducting material and is braided for flexibility. The insulating material separating the two conductors affects the cable capacitance, and hence the rate of signal propagation. The cable is usually covered with a protective PVC sheath, and sometimes with an additional shield as well. Several types of standard coaxial cables are manufactured; each has a different combination of electrical and mechanical characteristics to suit different applications.

The main variables are:

• Cable characteristic impedance
• DC resistance
• Power capacity
• Bandwidth
• Type of shielding
• Mechanical characteristics (such as flexibility)

Coaxial cables are more difficult to terminate than multi-core or twisted-pair cables. They are also more difficult to splice and connect to tee-offs. Special tools and connectors are required for good coaxial cable terminations. The ends of the cable should be terminated in a dead-end terminator to prevent signals reflecting from the ends of the cable. Coaxial cables can sometimes be ordered to specified lengths and with terminators already in place.

NEXT:

PREV: Signal conditioning (part 1)

All related articles   Top of Page   Home