14.2 Power, Grounding, Segregation, and EMI

A loop that is entity-correct and still chatters at 12 mA is not a classification problem. It is a power, grounding, and electromagnetic-compatibility problem. PE Control Systems items in this neighborhood ask whether the 24 V is still 24 V at the load, whether two supplies can share a bus without back-feeding, and whether the analog pair was routed as if it were another motor lead.

UPS, fusing, and diode-auctioneered 24 VDC

Control-room 120 VAC for DCS/PLC processors is usually an online (double-conversion) UPS: incoming AC is rectified and the inverter carries the load continuously, so transfer time is not a ride-through gamble. Line-interactive or standby UPS units that switch after a dropout are a poor match for switch-mode I/O supplies that already dropped out. SIS power is a separate battery-backed source when independence from the BPCS is part of the safety design; a shared UPS that dies for both systems is not “redundant” in the safety sense.

Field I/O more often lives on 24 VDC bulk with battery backup or a DC UPS. Size the battery for the documented ride-through (orderly shutdown versus continuous operation). Treat inrush of analog barriers, solenoids, and network switches as part of the current, not as a surprise after commissioning.

Fusing is coordination, not decoration. Fuse the +24 V feeder to a marshalled group so a shorted solenoid or a crushed tray takes that group, not the entire cabinet. The fuse in a zener barrier is part of the IS certificate — substituting a larger field-replaceable fuse is how a loop stops being the loop on the control drawing. Do not add a second series fuse on the IS field pair unless that fuse is on the drawing; extra resistance and extra opening energy were not in the entity file.

Diode auctioneering (diode-OR) lets two DC supplies feed one bus through steering diodes. The higher source voltage conducts; the lower sits reverse-biased. A silicon diode drops about 0.7 V; a Schottky about 0.3–0.4 V. The bus the loops actually see is Vsupply − Vf, and that drop belongs in every analog and solenoid budget. Diodes must be rated for the full bus current plus inrush; a diode that opens is a single point of failure unless you have monitored parallel paths. If Supply A is 24.5 V and Supply B is 23.8 V, only A carries load until A fails — “both supplies sharing” is not what the circuit does.

Worked numeric voltage drop on a 24 V loop

Instrument copper at about 20 °C is close to 6.5 Ω per 1000 ft for 18 AWG and 4.1 Ω per 1000 ft for 16 AWG. Use those working values unless the problem states NEC Chapter 9 Table 8 (75 °C) resistances, which are higher.

Round-trip drop for a two-conductor run of one-way length L (in kft) and resistance R' (Ω/kft) is:

Vd = I × 2 × R' × L

Worked solenoid (power loop). A 24.0 VDC bulk bus feeds a coil that draws 350 mA holding. The one-way cable length is 600 ft of 18 AWG. Neglect the auctioneering diode for the first pass.

• One-way resistance = 6.5 Ω/kft × 0.600 kft = 3.90 Ω
• Round-trip resistance = 7.80 Ω
• Vd = 0.350 A × 7.80 Ω = 2.73 V
• Voltage at the coil = 24.00 − 2.73 = 21.27 V

A coil specified 24 V ±10% has a 21.6 V floor. 21.27 V is below 21.6 V, so the valve may chatter, pull in weakly, or drop out on a warm day when copper resistance rises. Two honest fixes: 16 AWG on the same route (R_rt = 2 × 4.1 × 0.600 = 4.92 Ω, Vd = 1.72 V, Vcoil = 22.28 V) or a shorter run / local 24 V supply. If you then auctioneer through a 0.4 V Schottky, subtract that from the bus first: 23.6 V at the cabinet already, then 2.73 V of cable, 20.87 V at the coil — the diode you ignored is now the whole problem.

Same cable, analog loop. At 20 mA, that 7.80 Ω round trip drops only 0.156 V. Analog cable drop is usually tenths of a volt. Solenoid and 4-wire transmitter power conductors are where 24 V loops actually fail voltage drop. Still include analog cable in the Section 14.3 compliance budget; it is small, not zero, and it grows with thin wire and long runs.

A 10% voltage-drop rule of thumb (2.4 V on a 24 V bus) is a screening test, not a substitute for the nameplate minimum. Many 24 V solenoids and 2-wire transmitters have a minimum operating voltage tighter than 10%.

Control-signal isolation is the other half of “the 24 V arrived but the milliamp signal is garbage.” When a field device, a cable shield, and a panel analog common sit on different earths, the difference current has nowhere to go except along the signal conductor. Analog isolators, isolated analog-input cards, and galvanic IS isolators break that metallic loop. Isolation is not optional just because both ends say “24 V common.”

Single-point versus multi-point grounding, and shields

Single-point (star) grounding ties analog commons, shield drains, and the cabinet earth bar to one reference so that 50/60 Hz current has no loop area to circulate. It is the default mental model for millivolt thermocouples and quiet 4–20 mA.

Multi-point bonding ties cable shields and armor to earth at more than one place so that high-frequency noise (VFD PWM edges, radio, ESD) has a short path off the shield. At a few hundred kilohertz, a 30 m drain wire to a distant star point is an inductor, not a ground.

Plants use both, on purpose:

• Analog 4–20 mA and millivolt pairs: ground the shield at one end, almost always the control-room / I/O cabinet end. Leave the field end insulated unless the manufacturer’s control drawing says otherwise. Two-end analog shield grounds on a long pair are a 60 Hz loop around the tank farm earth gradient.
• VFD / motor leads: bond the shield (or armor) at both ends with 360° glands, not a pigtail. The standing-wave and common-mode current on a PWM motor cable will not drain through a single pigtail.
• Hybrid: some digital / fieldbus installations ground the shield at the cabinet and capacitor-couple the field end so that AC noise has a path while DC earth current does not. Follow the bus installation guide; do not invent a third earth on a spare drain.

The instrument cylinder ground and the electrical safety ground (PE) must still be bonded as a safety system. “Single-point analog” does not mean floating the cabinet. It means you do not let the analog shield be a parallel PE conductor between two buildings.

Segregation from VSD/motor leads, ferrites, and filters

A variable-speed drive’s motor leads are a fast common-mode noise source. Capacitive coupling into a parallel 4–20 mA pair looks like a wandering process value, a HART modem that will not connect, or a barrier that appears to “noise-trip.” Geometry beats firmware.

• Do not pull analog, IS, or millivolt cable in the same conduit or undivided tray as VSD output or motor leads.
• Separate with distance (a common installation practice is on the order of 300 mm) or a grounded metallic barrier / separate steel tray.
• Cross at right angles when the route must meet.
• Use symmetric, shielded VFD cable, 360° termination, and a motor-end EMC gland. Unshielded THHN bundled with analog is the commissioning failure mode.
• After the route is clean, ferrite sleeves (common-mode chokes) on the analog pair at the cabinet entry, and RC or active filters on the analog input, take residual RF. Ferrites on a still-bundled motor/analog pair treat the symptom.
• Do not hang random capacitors across an IS pair to “filter noise.” That capacitance is Ci on the entity sheet.

If the value jumped the day the VFD started, move the analog cable before you retune the PID.