Operational Amplifier Layout: Input Range and Feedback Stability
An operational amplifier looks like the simplest part of a board. A handful of resistors, a couple of capacitors and a supply connection are usually enough to make it work on a bench. Yet the same circuit can oscillate, latch or drift once it is on a real board, because the amplifier is sensitive to things the schematic does not show: input range, capacitive loading, feedback phase and the layout of the ground it references.
Small Circuits, Long List of Failure Modes
Operational amplifiers appear everywhere in signal conditioning, in the analogue front end ahead of an analogue to digital converter, in power control loops and in instrumentation. The external component count is low, which creates a false sense of safety. Most of the problems that reach production are not caused by the amplifier itself but by assumptions about its inputs and its environment.
Three categories account for the majority of issues. The first is exceeding the specified input range. The second is adding capacitance where the amplifier cannot tolerate it, either at the output or inside the feedback path. The third is layout related, where high impedance nodes pick up noise and the reference the amplifier compares against is not actually the reference the designer intended.
Input Voltage and Common-Mode Range Limits
Every amplifier has a specified input voltage range, and exceeding it produces behaviour that no amount of gain calculation will predict. A classic example is a precision device running on plus and minus fifteen volt supplies with an input range of plus and minus 13.5 volts. Drive the input beyond that window and the output stops following the input, sometimes inverting its behaviour entirely.
Some datasheets specify the common-mode input range instead of an input voltage range. For a device powered from a single five volt rail, that figure might be zero to 3.5 volts. The two specifications describe the same physical constraint, because during normal operation the inverting and non-inverting inputs sit at essentially the same potential. Designers should therefore check the common-mode input range against the actual signal swing, including any transient that can occur at power up.
Do Not Capacitively Load the Output
Placing a capacitor directly on an amplifier output is a common mistake, usually made while trying to reduce noise on a slowly varying signal. The reasoning seems sound, since the signal is nearly direct current. The problem is dynamic: a step at the input, or simply the moment power is applied, forces the amplifier to deliver a large current into the capacitor, and the capacitor also shifts the phase of the loop.
Phase shift in the loop reduces phase margin, and once the margin disappears the circuit enters self-oscillation. The correct arrangement is an RC network: a series resistor from the output, with the capacitor placed after it. The resistor limits the instantaneous current the amplifier must supply and isolates the capacitance from the loop, so the phase behaviour stays close to the uncompensated case. The same reasoning applies when driving a cable or a long trace, which is capacitive whether or not a component is fitted.
Keep Capacitors Out of the Feedback Loop
The feedback loop is even less tolerant. A capacitor placed across the feedback path, however small, changes the phase of the feedback signal and invites the same oscillation. This error is easy to make on a direct current gain stage, where a designer adds a capacitor intending to filter noise and inadvertently creates a phase lag inside the loop that controls the amplifier.
The same principle extends to regulated supplies built around an amplifier. A capacitor connected directly to the feedback pin is a phase shift in the control loop, not a filter. Where ripple needs to be reduced, the capacitor belongs in parallel with the upper divider resistor, which strengthens the negative feedback at the ripple frequency and suppresses the disturbance without destabilising the loop.
Supply Decoupling and Reference Integrity
A decoupling capacitor at each supply pin, placed as close to the device as the layout allows, is the minimum requirement. It should connect to a solid ground reference rather than to a thin trace that meanders back to the regulator. Where the amplifier shares a board with switching circuitry, the decoupling is what prevents switching noise from entering the analogue supply rail.
The reference matters just as much. A resistive divider that sets a reference voltage will drift with temperature if the resistors are poorly matched, and it will pick up noise if the node is physically large. Keeping the divider close to the amplifier, using a low impedance node, and referencing it to the analogue ground rather than to a digital return all improve stability. Techniques for containing switching noise appear in EMI suppression design principles and DC-DC converter layout and routing.
Guarding, Bias Currents and Thermal Effects
High impedance inputs need physical protection. A guard ring driven to the same potential as the input, routed around the node on both sides of the board, intercepts leakage currents that would otherwise flow across the surface. Cleaning flux residue matters for the same reason, since ionic contamination creates a leakage path whose resistance changes with humidity.
Input bias current must also have somewhere to flow. If the direct current path from each input to ground is not balanced, the bias current develops a voltage across the source impedance and appears as an offset. Thermal gradients are another quiet source of error: dissimilar metal junctions generate small voltages, so symmetric layout and keeping heat sources away from the input stage reduce drift.
Layout Practice for Analog Sections
Keep the feedback network compact and its traces short, place the inverting input node as small as possible, and route sensitive inputs away from clock and switching nodes. Where a long connection is unavoidable, treat it as a transmission line and terminate it deliberately rather than hoping the amplifier will absorb the reflection, an approach described in microstrip and stripline routing.
Finally, separate the analogue and digital return currents and join them at one point, so that digital switching currents do not flow through the reference the amplifier depends on. None of these measures is complicated, but applying them consistently is what keeps a working prototype working in volume production.
It is worth testing for self-oscillation deliberately rather than waiting for it to appear in the field. Feeding a fast step into the input and observing the output on a scope with a short ground lead will reveal ringing that a slow measurement would hide. Capacitive loads, long cables and poorly decoupled supplies are the conditions to reproduce, and a circuit that stays clean under all three is unlikely to surprise a customer.
FAQ
Why does my op-amp oscillate even though the gain is correct? Capacitance is the usual cause. A capacitor on the output, a long cable, or a capacitor in the feedback path all shift the loop phase. Removing the capacitance or isolating it with a series resistor normally restores stability.
What is the difference between input range and common-mode range? They describe the same constraint. Because the two inputs sit at nearly equal potential during normal operation, the permissible input voltage and the permissible common-mode voltage are effectively one specification.
How do I reduce ripple without destabilising the loop? Never place the capacitor directly across the feedback path. Instead parallel it with the upper resistor of the divider, which increases negative feedback at the ripple frequency while leaving the loop phase intact.
