Why the cable between a transmitter and its control system is never just a conductor — it is a distributed electromagnetic system that stores energy, filters signals, and generates sparks, whether the engineer accounts for it or not.
A capacitor is two conductors separated by an insulator. Apply a voltage and charge accumulates on the surfaces — positive on one plate, negative on the other. No current flows through the insulator, but something happens in the space between: an electric field builds. That field is not a mathematical abstraction. It is a physical state of the dielectric — the insulating material is electrically stressed, its molecular dipoles partially aligned, and energy is stored in that stress. The energy lives in the field itself, not in the metal. Remove the voltage source and the field persists, holding its energy until a discharge path appears. The amount of energy a given geometry can hold per volt of applied potential is its capacitance: E = ½CV².
Every current-carrying conductor also generates a magnetic field. Where the electric field stores energy in the stressed dielectric between conductors, the magnetic field stores energy in the space around them. That stored energy resists any change in current — not because of friction, but because changing the current requires changing the field, and changing the field requires energy. That resistance to change is inductance. The energy it holds is proportional to the square of the current: E = ½LI².
One is charged by applying voltage. The other is charged by sustaining current. Both are invisible, both are real, and both will release their energy the instant the conditions that created them are disturbed.
That release is where the engineering consequence lives. When a conducting path suddenly appears between the charged plates — a short circuit, a make contact, a moisture bridge — the electric field collapses, driving current through whatever resistance the path offers. The stored energy does work: it heats the conductor, ionises any gas in the gap, and can produce a spark. When the current through an inductor is suddenly interrupted — a loose terminal, a fractured wire, a break contact — the magnetic field collapses, but it cannot simply vanish. It forces the current onward, generating whatever voltage is necessary to push it across the gap. That voltage can far exceed the original supply, and the energy it delivers does the same work at the break point: heating, ionisation, arc.
In a cable carrying a steady signal, both fields are quietly charged. The electric field sits in the dielectric between the conductors. The magnetic field wraps around them. Nothing happens — until something changes.
Now look at two wires running side by side — a twisted pair in a cable tray, the signal and return of a 4–20 mA loop. You have, whether you intended it or not, two conductors separated by insulation. Every millimetre stores a little charge. The longer the run, the larger the total capacitance. For a typical instrumentation twisted pair, capacitance per unit length lands in the range of 15–60 pF/m depending on wire diameter, spacing, insulation material, and shielding construction.
But the capacitor never comes alone. Two wires carrying equal and opposite currents trap magnetic flux in the loop between them. The same geometry that creates a distributed capacitor also creates a distributed inductor. Wider spacing increases inductance — more area for the magnetic flux loop — while reducing capacitance — weaker electric field coupling. Tighter spacing does the opposite. You cannot tune one without affecting the other. The two are born together from the geometry, and the geometry is set the moment the cable is manufactured.
Stop imagining continuous wires. Slice the cable into vanishingly thin sections. Each slice contains four elements:
Connect an infinite cascade of these blocks end to end and you have the transmission line — the true distributed reality of the cable. Not a single capacitor. Not a single inductor. A continuous, position-dependent electromagnetic structure where voltage and current vary along the length and across time simultaneously.
For a single slice, the voltage drop and current leakage yield two coupled partial differential equations:
Voltage falls along the line because resistance dissipates energy and inductance opposes any change in current.
Current leaks away because imperfect insulation bleeds a trickle, and capacitance diverts charge to build the electric field.
Every signal, every reflection, every loss, and every stored joule in the cable obeys these two statements. They are the cable's constitution.
Consider a typical twisted-pair instrumentation cable: 0.8 mm conductor diameter, 1.5 mm centre-to-centre spacing, polyethylene insulation (εr ≈ 2.3).
Capacitance per metre:
C' = (π × 8.854×10⁻¹² × 2.3) / acosh(1.875) ≈ 50 pF/m
Over 1 km: 50 nF of distributed capacitance.
Inductance per metre (external plus internal):
L' ≈ 0.50 + 0.10 = 0.6 μH/m
Over 1 km: 0.6 mH of distributed inductance.
Loop resistance (both conductors, copper at 20 °C):
Rloop = 2 × ρL/A = 2 × (1.68×10⁻⁸ × 1000) / (π × 0.4² × 10⁻⁶) ≈ 67 Ω/km
Characteristic impedance:
Z₀ = √(0.6×10⁻⁶ / 50×10⁻¹²) ≈ 110 Ω
Propagation velocity:
v ≈ 1.83×10⁸ m/s — about 0.61c, slowed by the dielectric.
End-to-end delay: 1000 m / 1.83×10⁸ m/s ≈ 5.5 μs.
For a good cable above a few kilohertz, R' and G' become small relative to ωL' and ωC'. The characteristic impedance simplifies to the pure balance between the magnetic and electric siblings — a number defined entirely by geometry and dielectric, not by frequency.
The 50 pF/m used here is a moderate mid-range figure. Real instrumentation cables span roughly 15 pF/m (widely spaced, air-gap pairs) to over 150 pF/m (tightly twisted, foil-shielded, high-permittivity insulation). Everything that follows — stored energy, HART attenuation, bandwidth, reflection severity — scales directly with this number. A cable at 150 pF/m stores three times the energy and rolls off the signal three times sooner. Treat the numbers below as illustrative for a typical installation. Verify your own cable's datasheet to see where on the spectrum it falls.
For a circuit that is truly steady — no change in voltage, no change in current — the time derivatives vanish. dV/dt = 0, so the capacitance draws zero current. dI/dt = 0, so the inductance develops zero opposing voltage. The telegrapher's equations collapse to a plain resistive divider with a small leakage through G'. The siblings go dormant. The cable behaves as a simple wire.
This is the only regime where a cable can be treated as a conductor with resistance and nothing more. It is also the regime that most I&C engineers implicitly assume when they think about a 4–20 mA loop. They are correct — until something changes.
Change anything, and the full distributed structure activates.
Switch the power off: current begins to collapse. L' fights back, generating a voltage transient at the break. That transient propagates into the distributed C', which absorbs and re-releases the energy. The quiet wire becomes a resonant structure ringing at a frequency set by its own L' and C'.
A loose terminal or fractured conductor: an arc forms at the break point, sustained by the energy stored in the distributed magnetic and electric fields on either side. The cable's own electromagnetic structure feeds the spark.
A nearby lightning strike, a VFD switching, a motor starting: an external disturbance enters as a traveling wave, encountering every L' and C' segment as it propagates along the cable, reflecting at every impedance discontinuity — junction boxes, splices, unterminated ends.
A communication signal: a HART frequency tone, a Modbus bit transition, any digital edge — these are change by definition. The cable is no longer a wire. It is a 110 Ω transmission line with a 5.5 μs memory of every edge driven into it.
No change, no consequence. Any change, and the distributed ladder wakes — revealing the cable as a dynamic system of stored and trading energy.
At 24 V DC, the 50 nF of distributed cable capacitance holds:
The 0.6 mH of distributed inductance stores energy proportional to the square of the current flowing through it. For a 4–20 mA loop at full scale (20 mA):
The capacitive store dominates at low currents. The total — 14.5 μJ — sits just below the minimum ignition energy of a hydrogen-air mixture, which is approximately 20 μJ. The margin is measured in microjoules.
Raise the current to 100 mA — a powered instrument with higher consumption — and the inductive store grows to 3 μJ. Total energy reaches 17.4 μJ. The hydrogen margin has nearly vanished.
At 500 mA — a solenoid valve, a relay coil — the inductive store reaches 75 μJ and total cable energy exceeds 89 μJ. This crosses the minimum ignition energy of hydrogen (20 μJ), ethylene (80 μJ), and approaches propane (250 μJ). And this is only the cable's contribution. The solenoid coil's own inductance — typically 50 to 200 mH — stores energy measured in millijoules, dwarfing the cable entirely.
Consider what happens when a terminal comes loose or a conductor fractures. The current path is interrupted at a point. In the distributed model, this interruption launches a voltage wave from the break in both directions along the cable. The initial amplitude of that wave is set by the cable's characteristic impedance:
For 20 mA: Vwave = 0.02 × 110 = 2.2 V. Modest.
For 100 mA: Vwave = 0.1 × 110 = 11 V.
For 500 mA: Vwave = 0.5 × 110 = 55 V.
These voltage waves travel toward the cable ends at 1.83×10⁸ m/s. At an open end — which both sides of the break now are — the reflection coefficient is +1. The reflected wave doubles the voltage when it returns. The break point voltage briefly reaches VDC plus two to three times Vwave before cable resistance damps the oscillation. Each transit retains roughly 74% of its amplitude, so the ringing dies within a handful of round trips — microseconds for a 1 km cable.
For a 500 mA circuit, the break-point voltage momentarily reaches approximately 24 + 2 × 55 ≈ 134 V. This exceeds the breakdown voltage of a small air gap, and an arc forms.
But the distributed model reveals something the lumped approximation misses. At the instant of the break, the current at the gap must transition from its steady value toward zero. The local capacitance immediately adjacent to the break — a few centimetres of cable, perhaps picofarads — is all that is available to absorb the collapsing current in the first nanoseconds. The full 50 nF is distributed along the kilometre and isolated from the break by its own series inductance. The local voltage spike at the gap can therefore far exceed the bulk I × Z₀ prediction for a very brief interval — long enough to initiate an arc, which then draws on the distributed energy store to sustain itself.
This is the mechanism by which a 24 V circuit can produce a spark capable of igniting a flammable atmosphere. Not because the voltage is high in steady state, but because the energy stored in the cable's distributed electromagnetic field concentrates at the break point in nanoseconds.
The intrinsic safety standards — IEC 60079-11 — exist precisely because of this. They do not merely limit supply voltage and current. They cage the cable's own L and C, because the hazard arises when the energy stored in the distributed field has nowhere to go but into a spark at a break point.
Every IS loop calculation includes four parameters from the barrier: Voc (open circuit voltage), Isc (short circuit current), Ca (maximum allowable external capacitance), and La (maximum allowable external inductance). The cable's contribution — C' × length and L' × length — counts against Ca and La. The field device's own internal Ci and Li also count. The remaining budget is what the cable is permitted to use, and that budget directly constrains maximum cable length.
The IS barrier does not just limit what the supply can deliver. It limits what the cable itself can store — because a pair of wires is an energy reservoir whether the engineer treats it as one or not.
For a cable at 50 pF/m and 0.6 μH/m, a Ca limit of 80 nF allows a maximum cable run of 1,600 m before the capacitance budget alone is exhausted — before any device capacitance is subtracted. With a typical transmitter contributing 5–10 nF of internal capacitance, the practical ceiling drops further. These are not arbitrary specification limits. They are the physics of stored energy translated into a maximum cable length.
The spark is the most dramatic consequence of the waking ladder — but it is only the loudest of many. The same interplay of R', L', C', and G' produces quieter, chronic problems that degrade signals, corrupt measurements, and confuse commissioning. Each of the following is a direct expression of the four-element model.
Each infinitesimal RC segment attenuates high frequencies. The cumulative effect over 1 km is a filter you never designed. Unlike a single lumped RC circuit, a distributed RC line attenuates more aggressively because the signal encounters resistance and capacitance continuously along its path — each segment rounding the edge a fraction further.
For our example cable at 1 km, the distributed 3 dB bandwidth is approximately:
f3dB ≈ 0.12 / (π × 0.067 × 50×10⁻¹² × 10⁶) ≈ 11 kHz
Above this frequency, the cable attenuates the signal to less than half its original amplitude. A 9600-baud Modbus RTU signal (fundamental frequency ~4.8 kHz) sees roughly 20% attenuation at 1 km — degraded but decodable. A 115.2 kbps RS-485 link (fundamental ~57.6 kHz) has passed well beyond this cutoff.
A subtlety: above approximately 18 kHz for this cable, the inductive reactance (ωL') exceeds the series resistance (R'), and the cable transitions from RC-filtered behaviour to LC-transmission-line behaviour. In the LC regime, attenuation per unit length decreases — the inductance supports wave propagation rather than simply dissipating energy. But impedance matching and reflections then become the dominant signal-integrity concern, replacing the pure filtering problem with the reflection problem described next.
The practical takeaway remains: always compare your cable's bandwidth characteristics to the fundamental frequency of your fastest signal. A cable that is invisible at DC and at 4.8 kHz becomes the defining constraint at 57.6 kHz.
The cable's 110 Ω characteristic impedance is its natural rhythm. Connect it to a 10 kΩ instrument input, and the mismatch is severe. The voltage reflection coefficient:
Γ = (10,000 − 110) / (10,000 + 110) ≈ 0.98
Ninety-eight percent of the incoming voltage amplitude reflects back toward the source. For a 5 V step, the reflected wave carries 4.9 V. The receiver momentarily sees 9.9 V — nearly double the transmitted amplitude. If the source is also mismatched, the reflected wave re-reflects, setting up a ringing pattern that persists for several round-trip times (2 × 5.5 μs = 11 μs per cycle). Even a DC power-off transient can ring up to voltages well beyond the original supply.
The critical length threshold: when the cable's one-way propagation delay exceeds approximately one-sixth of the signal's rise time, the cable is electrically long and must be treated as a transmission line. For a fast digital output with a 1 μs rise time, the threshold is crossed at roughly 30 m. For a signal with a 10 μs edge, it extends to about 300 m. The faster the edge, the shorter the cable at which reflections begin to matter.
Every time a digital output switches, it must charge or discharge the full 50 nF of distributed capacitance. At a 10 kHz switching rate, the average charging current is:
Icharge = 50×10⁻⁹ × 5 × 10×10³ = 2.5 mA
At 100 kHz, it reaches 25 mA. An output driver rated for 20 mA may fail not because of the connected instrument, but because the cable demands more charging current than the driver can supply. The cable is a hidden reactive load that scales with both frequency and voltage swing.
In a multi-pair cable, adjacent pairs share mutual capacitance (Cm) and mutual inductance (M). For closely packed unshielded twisted pairs, Cm is typically 5–15% of the self-capacitance — roughly 3–8 pF/m.
Consider a 200 m multi-pair cable. A 24 V relay switching on Pair 1 produces a voltage step with a rise time of approximately 10 μs (shaped by the cable's own bandwidth). The mutual capacitance between Pair 1 and the adjacent Pair 2 totals 1 nF over 200 m. The coupled current into Pair 2:
Icoupled = 1×10⁻⁹ × (24 / 10×10⁻⁶) = 2.4 mA
Across a 250 Ω sense resistor, this produces a 600 mV transient spike on the victim pair. For a 4–20 mA loop reading a 1–5 V signal, 600 mV is a 15% disturbance — enough to cause a false alarm or a spurious control action. For a thermocouple circuit reading in millivolts, it is catastrophic.
Mutual inductance adds a second coupling path that scales with the rate of current change rather than voltage change. The two mechanisms superimpose, and in closely packed cables their effects are additive.
Copper resistance rises with temperature at approximately 0.393% per degree Celsius. Capacitance and inductance are almost flat — they depend on geometry and dielectric properties, which change negligibly across operating temperatures. The cable's AC behaviour — characteristic impedance, propagation delay, high-frequency loss profile — stays nearly constant. But its DC behaviour drifts substantially.
For the example cable at 67 Ω per km at 20 °C:
At −20 °C: Rloop ≈ 56 Ω. IR drop for a 20 mA loop: 1,127 mV.
At +60 °C: Rloop ≈ 77 Ω. IR drop for a 20 mA loop: 1,547 mV.
The IR drop shifts by 420 mV across the operating temperature range — in a signal span of 4 V (1–5 V across 250 Ω), that represents a 10.5% error if the supply voltage was sized without margin. A commissioning engineer who expects a single, stable cable behaviour across both AC and DC regimes finds one that splits with the ambient temperature: the distributed transmission-line properties hold steady while the resistive loss drifts underneath.
A 4–20 mA loop is DC at heart, but HART superimposes frequency-shift-keyed tones at 1,200 Hz (mark) and 2,200 Hz (space). The cable's distributed capacitance provides a shunt path for these AC components. At every point along the cable, the local C' diverts a fraction of the FSK current away from the signal path, and the cumulative loss grows with length.
At 2,200 Hz, the shunt impedance of the full 50 nF cable capacitance is approximately 1,450 Ω — nearly six times the 250 Ω sense resistor. At 1 km, the distributed RC attenuation absorbs roughly 14% of the space-frequency signal and 11% of the mark-frequency signal. Modest, and the modem handles it easily.
But both R and C grow with cable length — and the attenuation scales exponentially with their product. At 2 km, the space-frequency loss reaches 26%. At 3 km, it reaches 37%. The mark and space frequencies are attenuated unequally, distorting the FSK waveform and narrowing the modem's noise margin from both sides: the signal shrinks while the distortion grows.
The HART specification captures this through a time-constant rule: the product of total loop resistance and total cable capacitance must remain below 65 μs.
Rtotal includes the 250 Ω sense resistor plus the cable's own loop resistance. Ctotal is the cable's distributed capacitance. For the example cable at 1 km, Rtotal = 317 Ω and Ctotal = 50 nF, giving a time constant of 15.9 μs — well within the limit. At 2 km, it rises to 38 μs. At 3 km, it reaches 68 μs — the limit is breached. For this cable, the HART boundary falls at approximately 2.9 km.
The limit is not a fixed capacitance ceiling. It is a constraint on the RC time constant of the complete loop. A cable with lower resistance per metre buys more capacitance headroom. A cable with lower capacitance per metre buys more length before the time constant is exhausted. Intrinsic safety constraints — which cap capacitance independently through Ca — may tighten the ceiling further, often to well below the HART limit.
G', the shunt conductance, is typically a fraction of a nanosiemens per metre for clean, dry cable — equivalent to gigaohms of insulation resistance per kilometre. But in a flooded conduit, a water-filled junction box, or cable with moisture-wicked insulation, G' can rise by orders of magnitude. A 1 km cable with insulation resistance degraded from 10 GΩ to 100 kΩ introduces a distributed leakage path that draws 0.24 mA at 24 V.
In a 4–20 mA loop, that is over 1% of full-scale current. Worse, it is rarely stable. Humidity changes, temperature cycles, and drying winds shift G', making the error wander like instrument drift. The technician recalibrates the transmitter. The measurement returns to normal — until the next rain. The real problem is current bleeding through a damp conduit, not a faulty sensor.
| Consequence | Symptom | Root in RLCG | Key mitigation |
|---|---|---|---|
| Low-pass filtering | Rounded digital edges, bit errors | R' and C' form distributed RC | Match cable bandwidth to signal speed; shorter runs or lower-C' cable |
| Reflections | Ringing, overshoot, double transitions | Impedance mismatch excites L'–C' oscillation | Terminate with Z₀ when cable length exceeds 1/6 rise-time threshold |
| Source overloading | Driver overheating, output droop | Ctotal demands charging current (C × V × f) | Budget cable charging current into driver specification |
| Crosstalk | False triggers, noisy analogue readings | Mutual Cm and M between adjacent pairs | Shielded pairs; separate high-level and low-level signals physically |
| Thermal drift | DC offset shift with ambient temperature | R' drifts ~0.4%/°C; L' and C' are stable | Size supply for worst-case hot R'; use four-wire sensing for precision |
| HART attenuation | Modem communication failure | Ctotal shunts FSK tones past sense resistor | Keep R × C below 65 μs; lower-C' cable for long loops |
| Insulation leakage | Unexplained measurement offset, wandering zero | G' rises with moisture, draws DC current | Moisture-blocking cable; periodic insulation resistance testing |
You pull a cable schedule from the document control system. It lists cable type, size, length, and routing. The datasheet gives you C' and L' in small print — two numbers among dozens. You glance at them, confirm they meet the IS calculation, and move on.
Those two numbers encode the cable's hidden identity. They determine whether your HART modem will communicate at full distance. Whether your digital signal will arrive as a clean edge or a rounded ramp. Whether your IS loop has margin or is spending microjoules it cannot afford. Whether a broken terminal in a Zone 1 area produces a harmless open circuit or an incendive arc.
The cable between the transmitter and the marshalling cabinet is never just a conductor. It is a capacitor, an inductor, a filter, a resonant tank, a crosstalk bridge, and a leakage path — all derived from the same RLCG ladder. The datasheet gives you C' and L'. The physics writes the rest into your system whether you account for it or not.
Notice the pattern in every consequence above. The cable's DC behaviour and its AC behaviour are governed by different elements of the same ladder. At DC, only R' and G' matter — the resistive skeleton. At any frequency, L' and C' awaken and dominate. The cable has two personalities coexisting in one physical structure, and they respond to different stimuli, drift with different variables, and fail in different ways.
A temperature swing shifts the DC personality (R') by 30% while leaving the AC personality (Z₀, propagation delay) virtually unchanged. A moisture event shifts the AC-leakage personality (G') while the DC resistance stays the same. These are not different problems requiring different diagnostic frameworks. They are different expressions of the same four-element model, manifesting on different timescales.
The engineer who sees the RLCG ladder when looking at a cable datasheet does not need to memorise seven separate failure modes. They see one structure with four elements, and the consequences follow. That is the point of the model — not to add complexity, but to replace a list of memorised rules with a single physical picture that generates every rule on demand.