Common Mode Rejection: How a Biosignal Front End Hears a Whisper at a Concert

Mostafa Dhouib·September 4, 2026·7 min read

Common Mode Rejection: How a Biosignal Front End Hears a Whisper at a Concert

The short answer. When you measure off a body you are never measuring only your signal. You are measuring your signal plus the person plus the room, summed together by the time it reaches your amplifier. The thing you want is microvolts. The mains hum coupling into the person can be a thousand times larger. The entire job of the front end is to pull the small thing out from under the large thing, and it does that by throwing away everything that shows up equally on both inputs and keeping only the difference. Your front end lives or dies on that rejection.

Every discussion of biosignal quality eventually reduces to one number, and it is worth understanding what that number is actually doing before you trust anything downstream of it.

What arrives at your amplifier

Put an electrode on a person and consider honestly what is now connected to your circuit.

The biopotential you want, which is microvolts. The person, who is a large conductive object standing in an electric field. The room, including the mains wiring in the walls, every switching supply nearby, and the radio environment. Motion, as electrodes shift against skin. And the electrochemistry of the electrode interface itself.

All of it sums. There is no separate wire carrying only the thing you care about, and there never will be. What reaches your amplifier is one voltage containing everything, and the interference is routinely three orders of magnitude larger than the signal.

Stated plainly: you are trying to hear a whisper at a concert. Not a quiet room with some background noise. A concert.

The trick that makes it possible

The interference has one exploitable property. It arrives at both of your measurement electrodes at roughly the same amplitude and phase, because the whole body is being driven by it more or less uniformly. The signal you want does not, because it originates from a specific place and reaches your two electrodes differently.

So you take two measurements and subtract them.

Everything common to both, the hum, the field, most of the environment, cancels. What survives is the difference, and the difference is your signal. That is the common mode, and rejecting it is the whole design.

The measure of how well an amplifier does this is common mode rejection ratio: how much larger the common signal has to be before it contributes as much to the output as the differential signal does. It is expressed in decibels, and the numbers on a good instrumentation amplifier look enormous, which is exactly why they lull people.

Why the amplifier's number is not your number

This is the part that matters and the part specification sheets do not tell you.

Your system's rejection is not the amplifier's rejection. It is the rejection of the amplifier plus the electrodes plus the skin plus the cabling, and the weakest element sets the result.

Electrode impedance mismatch destroys it. The subtraction only cancels the common signal if it arrives identically on both inputs. Your two electrodes have different contact impedances, because skin is not uniform, gel dries unevenly, and one electrode is pressed harder than the other. That difference converts part of the common mode into a differential signal before your amplifier ever sees it. The amplifier then faithfully amplifies it, because at that point it is a real difference.

This is why the practical advice is always to match electrode impedances and keep them low, and why a device that works on one person and not another is usually an impedance story rather than a physiology story.

The input has to stay in range. Rejection is only defined while the amplifier is operating linearly. If the common mode voltage drifts outside the input range, the amplifier is no longer subtracting anything, it is saturated, and the rejection figure becomes irrelevant. A specification of 120 dB tells you nothing about what happens when the input leaves the window.

Cabling asymmetry matters. Different lead lengths or different stray capacitance on the two inputs is another route from common to differential, for the same reason as impedance mismatch.

Which is why there is a third electrode

Passive rejection depends on symmetry you do not fully control. So most serious front ends do something active instead.

Rather than hoping the body's common mode stays inside the input range on its own, add a third electrode and an amplifier, sense the common mode, and drive the body against it, holding the body's potential right where the front end can see it cleanly.

This is a genuinely elegant idea, and it is the reason a front end has a rejection figure worth being proud of. It converts a hope about symmetry into an actively regulated quantity.

It also introduces the failure that this pillar keeps returning to, and it is worth flagging here rather than leaving as a surprise. That third electrode is, by design, a low-impedance connection between your circuit and the body. Which means when another part of your system injects current into that body, the lowest-impedance path home runs straight through it. The electrode that makes your rejection good is the node an actuator's return current attacks, and the amplifier holding your common mode acquires a second job it was never specified for.

The hero and the victim are the same electrode. Whether your amplifier survives that is one multiplication and one comparison, and the return-path margin calculator will do it on your numbers.

What to check

Measure your rejection, do not read it. Apply a known common mode signal to both inputs through impedances that mimic real electrodes, including a deliberate mismatch, and measure what comes out. The number you get is your system's, and it will be well below the amplifier's.

Measure electrode impedance continuously, not once. It changes over a session as gel dries and the wearer moves, and a device that meets specification at minute zero can fail at minute forty. Many front ends can measure it in situ, and the measurement is worth logging alongside the signal, because it tells you which recordings to trust.

Watch the common mode voltage itself as a channel. Not just the differential output. It is the fastest diagnostic you have: if the signal degrades at the same moment the common mode leaves the input range, you have your answer, and the answer is not noise.

Test with something else in the system running. A front end characterised in isolation on a quiet bench is a different circuit from the same front end with a radio transmitting and an actuator firing.

FAQ

What is common mode rejection in a biosignal front end? The ability to discard everything that appears equally on both measurement electrodes and keep only the difference. Interference from the mains and the environment reaches both inputs roughly identically, while the biopotential you want does not, so subtracting the two cancels the interference and preserves the signal.

Why is my system's rejection much worse than the amplifier's specification? Because system rejection is set by the amplifier plus the electrodes plus the skin plus the cabling, and the weakest element governs. Mismatched electrode impedance converts part of the common mode into a genuine differential signal before the amplifier sees it, and the amplifier then correctly amplifies it.

What is the driven reference or third electrode for? It senses the body's common mode and actively drives the body against it, holding the body's potential where the front end can see it cleanly, rather than hoping passive symmetry keeps it in range. It is the main reason a front end has a rejection figure worth quoting.

Why does my front end work on one person and not another? Usually electrode impedance rather than physiology. Skin condition, hair, gel application, and contact pressure all change contact impedance, and a mismatch between the two measurement electrodes degrades rejection directly. Measure impedance in situ and log it alongside the signal.

Does a high CMRR specification mean my signal will be clean? No. Rejection is only defined while the amplifier operates linearly. If the common mode drifts outside the input range the amplifier is saturated and is not subtracting anything, at which point the specification is irrelevant regardless of how large the number is.