A plain comparator transitions on every threshold crossing of its input, so noise at the reference produces a burst of edges. The Schmitt trigger replaces the single threshold with two state-dependent thresholds separated by a hysteresis band. After each transition the input must reverse by the full width of the band before the next transition can occur. The following sections derive the threshold pair, examine the hysteresis loop, apply the circuit to a noisy ECG channel, and consider how the band is sized for R-wave detection and contact debouncing.
The input enters the non-inverting terminal through $R_1$. A second resistor $R_2$ returns the output to the same node and supplies positive feedback. The inverting terminal is grounded, so the comparator switches whenever the non-inverting node crosses zero. Because the output takes one of two saturation values, the voltage at the non-inverting node also takes one of two values, and the input therefore sees two effective thresholds.
Superposition at the non-inverting node gives
Setting $V_+ = 0$ and substituting the two possible output values yields the threshold pair,
While $v_\text{out} = -V_\text{sat}$, a rising input must reach $V_\text{TH}$ before the output flips high. While $v_\text{out} = +V_\text{sat}$, a falling input must reach $V_\text{TL}$ before the output flips low. The width of the hysteresis band is
The band is symmetric about ground because the inverting input is tied to ground. Offsetting that input shifts both thresholds together without changing the width. The width itself depends only on the resistor ratio and the rail separation, not on the absolute resistance values.
Positive feedback is destabilising in the linear region of an op-amp and is normally avoided. The Schmitt trigger operates outside that region: the output is already at a rail before the feedback takes effect. The fraction returned to the non-inverting input reinforces the existing saturation rather than driving the output away from a quiescent point. The circuit has two stable states and no third equilibrium for the feedback to destabilise.
The non-inverting node is a summing junction for $v_\text{in}$ through $R_1$ and $v_\text{out}$ through $R_2$. The comparator switches when the junction voltage crosses zero, so each stable output state pins the junction to a different threshold of the input. The two panels on the right show this directly: the schematic carries live values for $R_1$, $R_2$ and $v_\text{in}$, and the voltage rule shows $v_\text{in}$ relative to the two thresholds and the direction it must move to flip the output.
A plain comparator transitions on every crossing of its threshold, including those produced by noise. The Schmitt trigger suppresses all crossings that lie inside the hysteresis band and produces a single transition for each genuine excursion of the underlying signal. The animation scrolls a synthetic, noise-corrupted ECG waveform through both circuits in parallel, so the difference is visible directly.
The top trace is the noisy input together with the reference $V_\text{ref}$ and, for non-zero hysteresis, the thresholds $V_\text{TH}$ and $V_\text{TL}$. The middle trace is the output of a plain comparator referenced to $V_\text{ref}$. The bottom trace is the output of a Schmitt trigger with the hysteresis set on the slider.
The width of the hysteresis band is the principal design parameter of a Schmitt-based front end. The following examples show how the band is chosen against the noise statistics and the required transition behaviour.
An ECG channel for heart-rate counting is bandpass filtered to isolate the QRS complex, amplified to a peak excursion of approximately one volt, and passed through a Schmitt trigger whose output drives a counter. The reference is placed near the centre of the expected R-wave, and the hysteresis is set to exceed the peak-to-peak noise at the comparator input. The dominant noise sources are EMG within the QRS passband, residual mains pickup that survives the instrumentation amplifier, and electrode-skin half-cell drift. A common rule selects $\Delta V$ at approximately half the expected R-wave amplitude. The detector then produces one edge per heartbeat with a timing jitter set by the slope of the QRS at the reference, not by the noise excursion across it.
If $\Delta V$ is too large, the input must climb well past the nominal reference before transitioning. This introduces a systematic delay and reduces sensitivity to low-amplitude beats. If $\Delta V$ is too small, the output retains the chatter of the plain comparator. The optimum depends on the peak-to-peak noise during the QRS interval, not on the average noise of the channel.
Mechanical switches do not transition cleanly. A button pressed against its contact bounces several times over a few milliseconds before settling, and a digital input that samples the switch directly will register multiple presses. A Schmitt trigger placed between the switch and the logic input absorbs the bounces, provided the hysteresis exceeds the residual fluctuation of the contact and the signal voltage sits well clear of each threshold during the steady portions of the press and release. Patient-worn devices with event-marker buttons commonly use this arrangement.
Stimulus-locked recordings often require the detection of evoked potentials whose amplitude exceeds a fixed threshold. A Schmitt trigger with $V_\text{ref}$ set at the desired amplitude and $\Delta V$ matched to the expected interpulse noise produces a single trigger per evoked response, independent of small fluctuations of the recorded waveform. The output edge is used to time-stamp the response or to advance an averaging buffer.