← all notes

4–20 mA is not obsolete, and the reason is boring

2026-07-30 · instrumentation

Every few years someone points out that a current loop carries one value at a handful of updates per second, while a digital bus carries dozens of values plus diagnostics, and concludes that analogue signalling is a legacy problem waiting to be solved. The conclusion does not survive contact with a plant, and the reason has nothing to do with bandwidth.

The live zero does the work

The signal range starts at 4 mA, not 0. That offset is the whole design. A broken wire, a corroded terminal, a transmitter that lost power — all of them produce 0 mA, which is outside the signal range and therefore unambiguously a fault rather than a legitimate reading of zero. Compare with a 0–10 V input, where a failed sensor reads as a perfectly plausible bottom-of-scale measurement and the operator has no way to tell.

Current beats voltage on long runs

The current in a series loop is the same everywhere in that loop, so voltage drop along 300 metres of cable does not change the measurement — it only eats into the compliance voltage budget. That budget is the one thing worth calculating before installation:

V_supply ≥ V_transmitter_min + I_max × (R_cable + R_sense + R_barrier)
  24 V supply, 12 V transmitter minimum, 20 mA, 250 Ω sense, 50 Ω cable:
  24 ≥ 12 + 0.02 × 300 = 18 V     → 6 V of margin, fine

Where it genuinely falls short

One value per pair of wires. No device identity, no diagnostics beyond the fault current, no configuration over the same wires unless you add HART on top — which is exactly why HART exists and why it rides on the analogue signal instead of replacing it. If you need twenty process variables from one instrument, a current loop is the wrong tool and nobody sensible argues otherwise.

The practical position

Use digital buses where you want device diagnostics and dense data, and where the installation is inside a building with competent grounding. Use current loops for the measurements that must keep working while everything else is having a bad day — and for anything that feeds a safety function, where a failure mode you can detect beats a data rate you cannot use.