Explain comparator operation and why feedback is absent.
Describe output saturation and clipping.
Use the gain-bandwidth product to find a stage's bandwidth.
Identify slew rate, offset voltage, and single-supply constraints.
Remove the feedback and Lesson 1's first golden rule vanishes with it. The op-amp reverts to what it physically is — an enormous-gain difference detector — and its output slams to whichever rail the sign of (v+−v−) demands. That is a comparator: not a broken amplifier but a deliberate one-bit analogue-to-digital converter, answering the question "is this input above that reference?"
A comparator: no feedback path, a divider setting the reference. The output can only be at one rail or the other.
The reference divider sets the threshold; the input crosses it; the output flips. Two cautions, both common in practice. First, a comparator built from a general-purpose op-amp is slow — its output must traverse the entire supply range, and it was never designed to be driven into saturation. Dedicated comparator ICs are far faster, and some have open-collector outputs suited to logic. Second, with a slowly-moving input and any noise at all, the output will chatter around the threshold — the cure is hysteresis (a little positive feedback), which is the Schmitt trigger you will meet in ME-5.
Now the ideal assumptions' bill. OUTPUT SATURATION: the output cannot reach its supply rails, typically stopping one to two volts short on older parts, or within millivolts on modern rail-to-rail types. Ask for more and the output clips.
Clipping: the amplifier reproduces the signal faithfully until the output reaches its saturation limits.
FINITE GAIN-BANDWIDTH: an op-amp's open-loop gain falls with frequency, and the product of closed-loop gain and bandwidth is roughly constant — the gain-bandwidth product, quoted on every datasheet. Gain is not free at speed.
Av×BW≈GBWthe gain-bandwidth trade
SLEW RATE: the maximum rate at which the output can move, in volts per microsecond, set by internal currents charging an internal capacitance — CA-5's i=Cdv/dt once more, now as a datasheet limit. A large fast signal can be slew-limited even when it is well within the bandwidth, and the symptom is a sine wave turning into a triangle. INPUT OFFSET VOLTAGE: the real input difference required for zero output, typically a few millivolts, which every DC-coupled circuit amplifies along with the signal. SINGLE SUPPLY: with the negative rail at ground, the output cannot go below zero, so a signal that swings both ways needs a mid-rail reference — which is why ME-2's 1.65 V appeared, and why sensor circuits on a Pico so often build a half-rail node.
WORKED EXAMPLE
Bandwidth from GBW
An op-amp has a GBW of 1 MHz. A non-inverting stage is designed for a gain of 100.
Bandwidth: BW≈GBW/Av=106/100=10 kHz.
Adequate for audio at gain 100? Not really — 10 kHz cuts the top of the band. Either choose a faster part, or split the gain across two stages of 10 each, giving 100 kHz per stage.
Cascading for bandwidth is standard practice, and it is a direct consequence of this one equation.
10 kHz — and two stages of gain 10 would give ten times the bandwidth.WORKED EXAMPLE
Slew rate versus bandwidth
An op-amp specifies a slew rate of 0.5 V/µs. Required output: a 10 V peak sine at 50 kHz.
A sine's steepest slope is 2πfVpeak=2π×50,000×10≈3.14 V/µs.
That is more than six times the available slew rate, so the output will be a triangle, not a sine — even if the part's small-signal bandwidth is several megahertz.
Bandwidth and slew rate are separate limits: one governs small signals, the other large ones. Check both.
Slew-limited — the part cannot do it at any bandwidth specification.
PRACTICE
Q1
An op-amp with GBW =4 MHz is used at a closed-loop gain of 20. What bandwidth do you get?
Q2
A comparator's reference divider uses Ra=10 kΩ (top) and Rb=10 kΩ from a 5 V rail. At what input voltage does the output flip?
Q3
An op-amp with 3 mV of input offset voltage is used in a non-inverting stage with a gain of 200. What output error does the offset alone produce?
Q4
Two conceptual checks: (a) why does a comparator have no feedback, and what breaks if you add some? (b) a circuit's output is a triangle when a sine was expected, yet the frequency is well inside the part's bandwidth. What is the likely limit?