Figure 1: The differentiator (a) and the integrator (b), and what each does to a 1 kHz square wave (c, d). Same two parts, two opposite jobs. (Schematic, not to scale.)
The differentiator — spikes at every edge
In Figure 1(a) the capacitor is in series with the signal and the resistor goes to ground. A capacitor blocks steady voltage but passes sudden changes, so when the input jumps, the full step appears across the resistor for an instant — then the capacitor charges through the resistor and the output decays exponentially with the time constant:
Ï„ = R × C = 10 kΩ × 10 nF = 100 µs
Feed in a 0–5 V square wave at 1 kHz (period 1 ms, half-period 500 µs). Each rising edge produces a +5 V spike that decays as 5 × e−t/Ï„; each falling edge produces a −5 V spike. After one time constant the spike has fallen to:
vout(Ï„) = 5 / e = 1.84 V
After the 500 µs half-period — five time constants — it has fallen to 5 × e−5 = 0.034 V, effectively zero, before the next edge arrives. That is the condition for a clean differentiator: Ï„ must be much smaller than the half-period. Here Ï„ = 100 µs is one-tenth of the 1 ms period, so each spike dies away cleanly, as in Figure 1(c). If Ï„ were close to the half-period, the spikes would run into each other and the output would look like a distorted copy of the input instead.
The integrator — a square wave becomes a ramp
In Figure 1(b) the parts are swapped: the resistor is in series and the capacitor goes to ground. Now the capacitor charges through the resistor at a nearly constant rate while the input stays steady, so the output is a ramp whose slope is:
dvout/dt = Vin / (R × C)
Feed in a ±5 V square wave at 1 kHz with R = 10 kΩ and C = 1 µF, so Ï„ = R × C = 10 ms — ten times the period, so the capacitor never gets near full charge. The slope is:
dvout/dt = 5 V / 10 ms = 500 V/s
During each 500 µs half-period the output moves by:
ΔV = 500 V/s × 500 µs = 0.25 V
So the output is a triangle wave swinging ±0.125 V (0.25 V peak-to-peak), as in Figure 1(d). The input's sharp corners have become a smooth ramp — the circuit has integrated the square wave, just as integrating a constant gives a straight line in maths. The condition is the mirror image of the differentiator's: Ï„ must be much larger than the half-period, here 10 ms against 0.5 ms, a factor of 20.
Design — a 1 V triangle from a 5 V square wave
Suppose you want a 1 V peak-to-peak triangle from the same ±5 V, 1 kHz square wave. The output must rise 1 V in each 0.5 ms half-period, so the time constant must be:
RC = Vin × (T/2) / ΔV = 5 × 0.5 ms / 1 V = 2.5 ms
Keeping R = 10 kΩ, the capacitor is C = 2.5 ms / 10 kΩ = 0.25 µF. Check the slope: 5 V / 2.5 ms = 2000 V/s, and 2000 V/s × 0.5 ms = 1 V, exactly the wanted rise. In practice a passive integrator like this only works well for small outputs — a real triangle-wave generator uses an op-amp integrator, where a resistor across the capacitor also stops slow drift from building up. And a real differentiator has the opposite practical problem: it amplifies high-frequency noise, so a small resistor is usually added in series with the capacitor to limit the gain.
Key points to remember
- The differentiator (C in series, R to ground) answers only to changes: a square wave comes out as sharp spikes at each edge.
- Worked example: R = 10 kΩ, C = 10 nF give Ï„ = 100 µs; a 0–5 V square wave at 1 kHz gives +5 V and −5 V spikes that fall to 1.84 V after one Ï„ and to 0.034 V after the 500 µs half-period.
- The integrator (R in series, C to ground) answers to the average: a square wave comes out as a ramp or triangle wave.
- Worked example: R = 10 kΩ, C = 1 µF give Ï„ = 10 ms; a ±5 V square wave at 1 kHz rises at 500 V/s, moving 0.25 V per half-period, so the output is a triangle swinging ±0.125 V.
- Design example: for a 1 V peak-to-peak triangle from a 5 V square wave at 1 kHz, RC must be 2.5 ms — with R = 10 kΩ, C = 0.25 µF.
- The rule to remember: a differentiator needs Ï„ much smaller than the half-period; an integrator needs Ï„ much larger than the half-period.
