Electrical Engineering › Section Q · Thyristors, operational amplifiers, and power conversion › Topic Q.8

Differentiators, integrators, and positive feedback

Put a capacitor in an op-amp circuit and its output starts to respond to how a signal changes over time: its rate of change, or its running total. Feed the output back to the noninverting input instead, and the op-amp gains a memory that cleans up noisy signals and makes oscillators.

Circuits that respond to time

The amplifiers so far respond only to the input's present value. Swapping a resistor for a capacitor makes the output depend on time, through the capacitor law from Charging and discharging transients:

i = C\frac{dv}{dt}

A capacitor's current depends on how fast its voltage is changing, not on the voltage itself. A 1.0 µF capacitor whose voltage climbs at 500 V/s carries 0.50 mA; hold the voltage steady and the current is zero. Two circuits exploit this, named after operations of calculus: the differentiator, whose output follows the input's rate of change, and the integrator, whose output follows its accumulated total. Do not confuse the differentiator with the differential amplifier.

The differentiator

Place a capacitor \(C\) between the input and the inverting input, and a resistor \(R\) from the output back to that input, with the noninverting input grounded. The inverting input is a virtual ground, so the full input voltage appears across the capacitor and its current is \(C\,dv_{in}/dt\). That current cannot enter the op-amp, so it flows through \(R\), and

v_{out} = -RC\frac{dv_{in}}{dt}

A steady input gives zero output; a steadily rising input gives a steady negative output. Because faster changes produce bigger outputs, a pure differentiator amplifies high-frequency noise strongly; practical versions add a small resistor in series with \(C\) to cap the gain.

Worked example: differentiating a triangle

A differentiator has \(R = 10\,\text{k}\Omega\) and \(C = 0.10\,\mu\text{F}\), so \(RC = 1.0\,\text{ms}\). Its input is a triangle wave swinging between −2 V and +2 V at 500 Hz. 1. Each rising or falling ramp lasts half a period, 1.0 ms, and covers 4 V, so the slope is \(\pm 4000\,\text{V/s}\). 2. Rising ramp: \(v_{out} = -(1.0\,\text{ms})(4000\,\text{V/s}) = -4.0\,\text{V}\). 3. Falling ramp: \(+4.0\,\text{V}\). 4. So the output is a ±4 V square wave at 500 Hz. Check: doubling the frequency doubles the slope and the output, which is the differentiator's rising response with frequency.

The integrator

Swap the parts: \(R\) from the input to the inverting input, and \(C\) from the output back to that input. Now the input current is \(v_{in}/R\), and it must all flow into the capacitor, so the output changes at a rate set by the input:

\frac{dv_{out}}{dt} = -\frac{v_{in}}{RC}

Equivalently,

v_{out}(t) = -\frac{1}{RC}\int_0^t v_{in}\,dt + v_{out}(0)

A constant positive input makes the output ramp steadily downwards; zero input leaves it wherever it was. The output therefore keeps a running total of the input, which is why integrators act as totalisers that turn a flow-rate signal into total volume, or a dose-rate signal into total dose.

Worked example: an integrator ramp

An integrator with \(R = 100\,\text{k}\Omega\) and \(C = 1.0\,\mu\text{F}\) (so \(RC = 0.10\,\text{s}\)) starts at 0 V, and a constant +0.50 V is applied at \(t = 0\). Its op-amp saturates at about −11 V on ±12 V supplies. 1. Ramp rate: \(dv_{out}/dt = -0.50/0.10 = -5.0\,\text{V/s}\). 2. After 1.0 s: \(v_{out} = -5.0\,\text{V}\). 3. It reaches the −11 V limit at \(t = 11/5.0 = 2.2\,\text{s}\) and stays there while the input remains. Check: the area under the input, \(0.50 \times 1.0 = 0.50\,\text{V s}\), divided by \(RC\) gives the 5.0 V change at 1 s, with the sign inverted.

Keeping an integrator honest

A real integrator also integrates its own errors. Any small offset voltage or input bias current, described in Practical op-amp limits, models, and data, acts like a tiny constant input, so the output creeps steadily towards a supply rail even with no signal applied. Two remedies are common. A switch across the capacitor resets the output to zero at the start of each measurement. Or a large resistor in parallel with the capacitor bleeds off slow errors, so the circuit integrates signals above a chosen frequency while acting as an ordinary inverting amplifier at dc. Inductors could also make these circuits, but capacitors are smaller, cheaper and nearer to ideal.

Positive feedback: a two-state memory

Now return the output to the noninverting input instead. If the output starts slightly positive, it pushes the + input up, which drives the output further positive until it saturates; a negative start saturates it negative. The circuit is bistable: stable only at one extreme or the other, and it stays there, like the latching thyristors of Hysteresis, gas discharge tubes, the Shockley diode, and the DIAC. Negative feedback seeks balance; positive feedback runs away from it. On its own that sounds useless, but a little positive feedback gives a comparator hysteresis, which cures a common fault: a noisy signal crossing a single threshold makes a plain comparator chatter, switching many times instead of once.

The Schmitt trigger

Make an inverting comparator with the signal on the inverting input, and feed the output to the noninverting input through \(R_1\), with \(R_2\) from that input to ground. The + input then sits at \(B v_{out}\), where \(B = R_2/(R_1 + R_2)\). With the output saturated at \(+V_{sat}\), the threshold is \(+B V_{sat}\); once the input rises past it, the output snaps to \(-V_{sat}\) and the threshold jumps to \(-B V_{sat}\). The input must now fall all the way to that lower threshold to switch back. The gap between them, \(2B V_{sat}\), is the hysteresis. This comparator with hysteresis is the Schmitt trigger.

Worked example: Schmitt thresholds

A Schmitt trigger has \(R_1 = 47\,\text{k}\Omega\), \(R_2 = 10\,\text{k}\Omega\) and an output that saturates at ±10 V. 1. Feedback fraction: \(B = 10/57 = 0.175\). 2. Thresholds: \(\pm 0.175 \times 10 = \pm 1.75\,\text{V}\). 3. Hysteresis: \(2 \times 1.75 = 3.51\,\text{V}\). 4. A slow input carrying 1 V peak-to-peak of noise crosses the upper threshold and the output falls; the noise can lower the input by at most 1 V, nowhere near the −1.75 V needed to switch back. Check: chatter returns only if the noise exceeds the 3.51 V hysteresis band.

The relaxation oscillator

Add an RC network to a Schmitt trigger: \(R\) from the output to the inverting input and \(C\) from that input to ground. The capacitor charges towards the output voltage until it crosses the threshold \(B V_{sat}\); the output flips, and the capacitor heads the other way until it crosses \(-B V_{sat}\). The circuit is astable, with no stable state, and produces a square wave. It oscillates because the positive feedback acts instantly while the negative feedback through \(RC\) is delayed. With equal saturation voltages,

T = 2RC\ln\frac{1 + B}{1 - B}

For \(R_1 = R_2\), \(B = 0.5\) and \(T = 2RC\ln 3\): with 10 kΩ and 0.10 µF, 2.20 ms, or 455 Hz.

Common mistakes to avoid

First, forgetting the minus signs: both the differentiator and the integrator invert. Second, swapping the circuits: a capacitor at the input differentiates, a capacitor in the feedback path integrates. Third, expecting an integrator with zero input to return to 0 V: it holds its last value, or drifts. Fourth, confusing feedback types: positive feedback to the + input makes a latch, not an amplifier. Fifth, using \(R_2/R_1\) for the Schmitt fraction: it is \(R_2/(R_1 + R_2)\).

Key takeaways

Capacitor current is i = C dv/dt. A differentiator (C in, R feedback) gives v_out = −RC dv_in/dt and amplifies fast changes and noise. An integrator (R in, C feedback) gives dv_out/dt = −v_in/RC, a running total, and needs a reset or a bleed resistor against drift. Positive feedback makes a bistable latch; a Schmitt trigger has thresholds ±BV_sat with B = R₂/(R₁ + R₂) and hysteresis 2BV_sat, which stops chatter. Adding RC timing makes an astable oscillator with T = 2RC ln((1 + B)/(1 − B)).

Flashcards

Flashcard

i = C dv/dt

Capacitor current depends on how fast its voltage changes

Steady voltage, zero current.

Test your understanding

Question 1 of 5

A differentiator has R = 22 kilohms and C = 0.047 µF. Its input ramps upwards at a steady 1000 V/s. What is its output?

Question 2 of 5

An integrator with R = 47 kilohms and C = 0.47 µF starts at 0 V. A constant −0.20 V is applied. What is the output after 0.50 s, assuming it does not saturate?

Question 3 of 5

A Schmitt trigger feeds its output to the + input through R₁ = 100 kilohms, with R₂ = 20 kilohms from that input to ground. The output saturates at ±12 V. What are its thresholds?

Question 4 of 5

An op-amp relaxation oscillator has R₁ = R₂ (so B = 0.5), R = 47 kilohms and C = 0.010 µF. About what is its frequency?

Question 5 of 5

Why does adding positive feedback to a comparator stop its output chattering on a slow, noisy input?

← Previous topicNext topic →