Common-mode noise is only harmless in theory - a real differential input rejects it well, but never perfectly. The number that says how well is the common-mode rejection ratio, or CMRR, and it appears on every serious analog input card datasheet. Understanding it tells you how much of a ground-potential difference or a common-mode disturbance will leak through as an error in your reading. This guide explains what CMRR measures, why it is written in decibels, and how to read and use the figure when you are choosing I/O for a noisy field.
CMRR in one line: Common-mode rejection ratio (CMRR) is a specification that quantifies how well a differential input suppresses common-mode voltage - a voltage present equally on both inputs relative to ground. It is the ratio of the input's response to a differential (real) signal versus its response to a common-mode signal, expressed in decibels. A higher CMRR in dB means more of the common-mode voltage is rejected and less of it leaks through as an error, which matters most when field grounds sit at different potentials.
A differential input is designed to respond only to the difference between its two terminals and to ignore any voltage they share. In an ideal world it would reject shared, common-mode voltage completely. In reality the rejection is imperfect: a small fraction of the common-mode voltage still leaks through and appears as an error in the measured signal. CMRR is the number that captures how small that fraction is. Formally, it compares how strongly the input responds to a genuine differential signal against how weakly it responds to a common-mode one - the bigger the gap between those two, the better the input rejects common-mode noise.
The practical meaning is straightforward: CMRR tells you how much of a common-mode disturbance survives as an error. If a large common-mode voltage sits on the inputs - say because the field ground and the control ground differ by several volts - a high CMRR guarantees that only a tiny sliver of that difference shows up in the reading, while a poor CMRR lets a meaningful chunk of it corrupt the value. Two input cards can both accept the same signal range yet differ enormously in how much a ground offset poisons the reading, and CMRR is the spec that reveals the difference.
It is worth being clear about what CMRR does not do. It describes rejection of common-mode noise only; it says nothing about normal-mode or series noise, which sits across the pair where the signal lives and is not helped by common-mode rejection at all. And CMRR usually degrades with frequency - an input that rejects a slow ground offset superbly may reject a fast common-mode transient much less well - so a single headline number does not tell the whole story. It is a measure of one specific, important weakness: how much shared-ground disturbance leaks into a differential reading.
CMRR is quoted in decibels because the rejection ratios involved are enormous and awkward to write as plain numbers. A good instrumentation input might reject common-mode voltage by a factor of a hundred thousand or more, and expressing that as a raw ratio is unwieldy. Decibels compress those large factors into small, comparable figures: each additional 20 dB represents a further tenfold improvement in rejection. So a CMRR of 80 dB, 100 dB, and 120 dB are not close together - each step of 20 dB is ten times better than the last, so 120 dB rejects common-mode voltage a hundred times more thoroughly than 80 dB.
This logarithmic scale is why you should read CMRR figures with the decibel math in mind. The difference between two cards rated at 90 dB and 110 dB is not a modest 20 percent - it is a factor of ten in how much common-mode error leaks through. When a datasheet advertises a high CMRR, that headline number is doing a lot of work precisely because of the scale, and a card with a substantially higher dB figure genuinely rejects ground offsets much more effectively, not just marginally. Conversely, a low CMRR should be taken seriously as a real limitation, because the leakage it permits is proportionally large.
Because CMRR falls off with frequency, a thorough datasheet gives it at more than one condition - often a high figure at DC or line frequency and a lower one at higher frequencies. When comparing cards, compare like for like: the CMRR at the frequency of the noise you actually face. A ground offset is essentially DC or line-frequency and is well served by the headline figure; a fast switching transient sits at a higher frequency where the rejection may be far weaker. Reading the frequency-dependent numbers, not just the best-case one, is how you avoid being misled by a single impressive figure.
On an analog input card datasheet, CMRR usually appears near the common-mode voltage range, and the two should be read together. The common-mode voltage range tells you how large a shared voltage the input can tolerate before it stops working at all; the CMRR tells you how well it rejects common-mode voltage within that range. A card might tolerate a wide common-mode range but reject it only modestly, or the reverse. For a field with big ground differences you want both a generous common-mode range so the input is not overwhelmed and a high CMRR so what it does see leaks through only faintly.
The reason this matters in real installations is that field grounds routinely sit at different potentials. Instruments bonded to earth at a wellpad, a control building on its own ground grid, and long cable runs between them practically guarantee a common-mode voltage on every analog signal, and that voltage grows during transients from switching or lightning. An input with strong CMRR shrugs off those offsets and delivers a clean reading; one with weak CMRR turns every ground difference into a visible measurement error. When you are speccing I/O for a site with challenging grounding, CMRR is not a fine-print detail - it is a headline selection criterion.
CMRR is the input side of the same problem that isolation attacks from the wiring side, and in practice they work together. A galvanic isolator upstream reduces the common-mode voltage the input ever sees, and a high-CMRR input rejects whatever remains - belt and suspenders for a site where grounds cannot be trusted. When those signals feed a cloud SCADA platform such as Merobix, the reading that reaches the dashboard is only as clean as the input that captured it, so a well-chosen high-CMRR card at the RTU or edge device is part of what keeps a trended value from carrying a ground-offset error. Historizing the channel then confirms the choice: a reading that stays steady while nearby grounds shift is the visible dividend of good common-mode rejection.
Higher is better, and because CMRR is on a decibel scale, each additional 20 dB is a tenfold improvement in rejection. Quality instrumentation inputs are commonly specified at 100 dB or more at DC or line frequency, which rejects common-mode voltage by a factor of a hundred thousand. The right figure depends on your grounding: a site with large ground-potential differences needs a high CMRR, while a clean, single-ground bench setup is more forgiving.
Because the rejection ratios are very large - often a hundred thousand or more - and decibels compress those into small, comparable numbers. On the decibel scale, every 20 dB represents a tenfold change, so 120 dB rejects common-mode voltage a hundred times more thoroughly than 80 dB. Reading CMRR figures with that logarithmic math in mind is essential, because a 20 dB difference between two cards is a factor of ten, not a small margin.
Not entirely - they attack the common-mode problem from two sides and work best together. A high-CMRR input rejects common-mode voltage that reaches it, but a galvanic isolator prevents much of that common-mode voltage from ever appearing, and it also breaks ground loops that CMRR alone does not address. On a site with untrustworthy grounding, using an isolator to reduce the common-mode voltage and a high-CMRR input to reject the remainder is the robust approach.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.