Op-Amp Circuit Design: Considerations That Prevent Oscillation

An operational amplifier looks like the simplest component on the schematic and behaves like one of the most temperamental. The external network is a handful of resistors and capacitors, yet the same topology can be stable on one board and oscillate on the next because of a capacitor that was added with good intentions, an input that sits outside its permitted range, or a feedback network that has been given a phase shift it was never meant to have.

Check the Input Voltage Range First

The most common op-amp problem is an input that is driven outside the range the device is specified for. A part running from a plus and minus fifteen volt supply may only accept inputs between plus and minus thirteen and a half volts; drive it beyond that and the amplifier will not behave as an amplifier at all. The failure is often quiet: the output stops following the input, the device draws unexpected current, or a latch-up occurs in the middle of a measurement.

Some datasheets specify an input voltage range and others specify a common-mode input range, and the distinction confuses people unnecessarily. Because a working amplifier forces its two inputs to nearly the same potential, the two specifications describe the same constraint. What matters is that the specified range is respected at both inputs, including during start-up, power-down and any transient in the preceding stage.

Do Not Connect a Capacitor Directly to the Output

Adding a decoupling capacitor directly from the output of an amplifier to ground looks harmless when the signal being amplified is DC, but it is not. The capacitor changes the phase characteristics of the loop, and the output stage sees a load whose impedance falls as frequency rises. The result is a circuit that is stable with a slow input and oscillates whenever a step appears, or that oscillates from the moment power is applied because the supply ramp itself is a step.

The correct arrangement places a resistor in series with the output before the capacitor. The resistor decouples the capacitive load from the amplifier’s output impedance, limits the instantaneous output current and leaves the phase response of the loop intact. The value is a compromise: large enough to isolate the load, small enough that the voltage drop at the load current stays acceptable.

Do Not Put a Capacitor in the Feedback Path

The same instinct applied to the feedback network is worse. A capacitor placed across a feedback resistor introduces a pole, and the phase shift associated with that pole reduces the phase margin of the loop. An amplifier that had a comfortable margin at unity gain can become marginal, and the symptom is an output that rings on every transition or breaks into oscillation when the load changes.

The rule that follows is general: nothing that changes the phase of the feedback signal belongs in the feedback path, unless it is there on purpose as part of a compensated design. The same principle extends to regulated supplies. A capacitor connected directly to the feedback pin of a regulator is an error for exactly the same reason; if the objective is to reduce output ripple, the capacitor should be placed in parallel with the upper feedback resistor, where it increases the negative feedback at the ripple frequency instead of disturbing the loop across its whole bandwidth.

<img src="https://www.gopcba.com/wp-content/uploads/2026/08/Prototype-PCB.jpg" alt="Series isolation resistor between an op-amp output and a capacitive load” />

Keep the Feedback Node Small

Layout is part of the circuit. The node where the feedback divider meets the inverting input is a high-impedance point, and it is also an aerial for anything switching nearby. Keep the traces at that node short, keep the node away from the supply switching paths, and do not route it underneath a device that switches current. A stray capacitance of a few picofarads at that node is enough to shift the response of a wide-bandwidth amplifier.

Input pins deserve the same treatment. Guard the input traces from adjacent switching nodes, and keep the input network physically close to the amplifier so the loop from the source through the input and back through the reference is as small as possible.

Decoupling and Supply Considerations

Every amplifier needs a decoupling capacitor close to its supply pins. The capacitor closes the loop between the supply and the return at the point where the current is drawn, which keeps the supply current of the device out of the rest of the board and keeps the supply impedance low at the frequencies where the amplifier still has gain. A small ceramic capacitor at the pins plus a bulk capacitor nearby is the usual arrangement.

Signal-return routing matters too. The return current of a high-impedance input is small, so the return path of the output current dominates the error the layout contributes. Reference the output and the load to the same ground area, and keep that area free of the switching currents of any digital circuit sharing the board.

Choosing the Right Device for the Application

Device selection prevents more problems than layout fixes. A general-purpose precision amplifier is the right choice for signal conditioning, analogue-to-digital converter front ends and supply control, where offset and low-frequency noise dominate. A wide-bandwidth device is the right choice when the signal has meaningful high-frequency content, but it will need more careful layout and more attention to the phase margin of the loop. Choosing a device with a large unity-gain margin and then adding feedback that reduces that margin is a common way to create an oscillating design.

Where the amplifier drives a cable or a capacitive load, the drive capability of the output stage matters as much as the gain bandwidth. An amplifier that is stable with a resistive load can oscillate into a capacitive one, which is why the series-isolation resistor discussed above is a normal part of such designs rather than a workaround.

Verification and Review

Before the design is released, check the input range against the actual signal swing, confirm that the feedback network contains nothing that alters phase unintentionally, and verify that any capacitance at the output is isolated by a resistor. These checks are cheap at the schematic stage and expensive afterwards. Our component tolerance and reliability notes cover how component variation affects a marginal loop, the EMI immunity design material explains where noise enters an analogue front end, and the design release checklist lists the review steps that catch these issues before fabrication.

FAQ

Why does my amplifier oscillate only when I connect a scope probe? The probe adds capacitance to a node that was already marginal. If the amplifier is oscillating into a capacitive load or has insufficient phase margin, a few picofarads at the output or at the feedback node can be enough to push it over. The fix is in the circuit rather than the measurement: add series isolation at the output and check that the feedback network has not introduced an unintended pole.

Can I add a capacitor at the output to reduce noise? Only with a series resistor between the amplifier output and the capacitor. A capacitor connected directly to the output changes the phase of the loop and can cause oscillation on step inputs and at power-up. With a resistor in place, the capacitor reduces noise without loading the output stage capacitively, and the resistor value is chosen from the maximum load current and the acceptable voltage drop.

What does gopcb look for in analogue layout review? We look at the feedback node geometry, the distance from the input network to the amplifier, the placement of decoupling capacitors relative to the supply pins and the separation between analogue signal returns and digital switching currents. Analogue performance problems that reach the fabrication stage are almost always layout issues rather than schematic issues, and they are much cheaper to correct before the board is manufactured.