Analog Electronics

Chopper-Stabilized Amplifier: Modulating DC Offset to Near Zero

A garden-variety op amp drifts its input offset by roughly 1 to 5 microvolts for every degree Celsius; a chopper-stabilized amplifier holds that drift below 0.05 µV/°C — a hundredfold improvement — and squeezes the raw offset itself from millivolts down to a few microvolts. It does this with a deceptively simple trick: it refuses to amplify DC at DC. Instead it chops the incoming signal into an AC square wave, amplifies that AC, and then reconstructs the DC — leaving the amplifier's own low-frequency errors behind.

A chopper-stabilized amplifier is a precision op amp that uses switched modulation (the "chopper") to translate the DC input up to a carrier frequency, amplify it there, then synchronously demodulate it back to baseband. Because the amplifier's own DC offset and 1/f (flicker) noise are not chopped, they end up modulated away from the signal band and filtered out, yielding near-zero effective offset and drift.

  • TypePrecision (zero-drift) operational amplifier
  • InventedE. A. Goldberg, 1948 (RCA, US Patent 2,684,999)
  • Used inBridge/strain-gauge front ends, thermocouples, current sensing, DC instrumentation
  • Typical offset / drift0.5–5 µV offset, 0.005–0.05 µV/°C drift
  • Chopping frequency~1 kHz (mechanical) to 4.8 MHz (modern CMOS, e.g. ADA4522)
  • Key ideaModulate DC to AC carrier; offset & 1/f noise fall outside signal band

Interactive visualization

Press play, or step through manually. The visualization is yours to drive — try it before reading on.

Open visualization fullscreen ↗

Watch the 60-second explainer

A condensed visual walkthrough — narrated, captioned, under a minute.

What It Is and Where It's Used

A chopper-stabilized amplifier — marketed today as a zero-drift op amp — is a precision amplifier whose input offset voltage and low-frequency (1/f) noise are continuously cancelled by a switched modulation scheme rather than by careful device matching alone. The result is an amplifier whose DC behavior is set by the accuracy of a clock and a pair of switches, not by transistor mismatch, so offsets of a few microvolts and drifts below 0.05 µV/°C are routine.

These amplifiers dominate wherever a tiny DC signal must be read accurately against a large gain:

  • Strain-gauge and load-cell bridges, where full-scale output may be only 10–20 mV and offset must not masquerade as weight.
  • Thermocouples and RTDs, with ~40 µV/°C sensitivities that demand microvolt-stable front ends.
  • High-side current sensing across milliohm shunts, where offset directly corrupts the measured current.
  • Medical instrumentation (ECG/EEG) and precision integrators.

Landmark parts include the LTC1050, LTC2050, MAX44251, and Analog Devices' ADA4522 and ADA4523 families.

How It Works: Modulate, Amplify, Demodulate

The core insight is that a DC amplifier's errors — offset voltage V_OS and 1/f noise — live at and near DC, exactly where they contaminate the signal. Chopping moves the wanted signal out of that error band.

The signal path is three stages:

  • Input chopper (modulator): a switch network reverses the input polarity at the chopping frequency f_chop, converting the DC input V_in into a square-wave AC signal at f_chop.
  • AC amplifier: the amplifier's own offset V_OS and flicker noise add at DC, but the signal is now at f_chop — the two are frequency-separated.
  • Output chopper (demodulator): a second switch, driven by the same clock, demodulates the amplified signal back to DC. Crucially, this second chop shifts the amplifier's offset and 1/f noise up to f_chop, turning them into an AC ripple that a low-pass filter removes.

So the signal is chopped twice (up then down = back to DC) while the offset is chopped once (DC up to f_chop). A ripple-reduction or notch filter, or a nested auto-zero loop, then attenuates the residual carrier ripple.

Key Quantities and a Worked Example

The residual offset of an ideal chopper is limited by charge injection and clock feedthrough from the MOS switches, plus finite filtering of the ripple. The governing relation for effective offset is roughly:

V_OS(eff) ≈ (V_OS · f_signal / f_chop) + V_ci

where V_OS is the amplifier's native offset, f_chop the chopping frequency, f_signal the input bandwidth of interest, and V_ci the switch charge-injection residue. Higher f_chop pushes the offset error down and moves ripple further from the signal band.

Worked example (ADA4522): native input offset is trimmed to about 5 µV maximum, chopping runs at f_chop = 4.8 MHz, and a separate offset/ripple correction loop operates at 800 kHz. Drift is 22 nV/°C max. Compare a plain precision bipolar op amp at 60 µV offset and 0.3 µV/°C: over a 100 °C swing it drifts 30 µV, while the chopper drifts about 2.2 µV — an order of magnitude better. The 1/f corner effectively vanishes, so the noise spectrum is flat down to DC at roughly 5.8 nV/√Hz.

Designing and Operating With Them in Practice

Chopper amplifiers behave like ordinary op amps at the pins, but several practical rules apply:

  • Watch the ripple. The output carries a small residual at f_chop (and its harmonics). Modern parts hide it above the audio band with MHz chopping and internal notch filters; older kHz-chopped parts need external low-pass filtering.
  • Respect bandwidth. Usable closed-loop bandwidth must sit well below f_chop; a Nyquist-style limit means signal content near f_chop/2 aliases. Zero-drift parts are precision devices, not RF amplifiers — typical GBW is 0.5–10 MHz.
  • Input bias current spikes. Switch charge injection produces small current transients; keep source impedances balanced and moderate (kΩ range) to avoid converting them into offset.
  • Filter the supply and avoid injecting the chopping clock into sensitive nodes.
  • Use for DC/low-frequency measurement, integration, and level-shifting — not for signals whose bandwidth rivals the carrier.

Because they need only a clock and switches, these amplifiers are almost always monolithic CMOS today, trimmed at wafer test.

Chopper vs. Auto-Zero vs. Standard Op Amps

Two sampled-error techniques compete: chopping (continuous modulation) and auto-zeroing (periodic sample-and-cancel of the offset). Both kill DC offset and 1/f noise, but their noise signatures differ:

  • Chopper: continuous, so it does not sample-and-hold the noise. In-band broadband noise is preserved (no folding), but the offset reappears as a periodic ripple at f_chop that must be filtered.
  • Auto-zero: samples the offset onto a capacitor, then subtracts it. This folds/aliases broadband noise back into the baseband, raising in-band noise, but it produces less output ripple.

The best modern parts are hybrids: they chop to eliminate 1/f noise and add a nested auto-zero or ripple-correction loop to suppress the chopper ripple — capturing low in-band noise and low ripple. Versus a standard op amp, both approaches trade a modest increase in supply current, some switching artifacts, and limited bandwidth for a 10–100× improvement in DC precision.

Failure Modes, Trade-offs, and Significance

The chopper trick is not free. Its characteristic drawbacks are:

  • Chopping ripple and glitches: incomplete filtering leaves a residual tone at f_chop, and switch transitions inject glitches — problematic for downstream ADCs sampling near the carrier.
  • Intermodulation: input signals near f_chop or its harmonics can mix down into the signal band, producing spurious DC errors. Keep f_signal ≪ f_chop.
  • Limited bandwidth and higher current than a comparable non-chopped op amp.
  • EMI susceptibility: the internal clock can radiate or couple, so layout matters.

Historically, the concept is foundational: Edwin A. Goldberg invented the chopper-stabilized amplifier at RCA in 1948 (US Patent 2,684,999, filed 28 April 1948), using a vibrating mechanical chopper to make vacuum-tube analog computers stable. The modern CMOS descendants — trimmed to microvolt offsets and MHz chopping — remain the go-to solution whenever a designer needs an amplifier that treats DC almost as accurately as an ideal component, which is why "zero-drift" is now a standard catalog category.

Precision amplifier techniques: how they attack DC offset and 1/f noise
TechniqueTypical offsetOffset driftKey limitation
Standard bipolar op amp50–500 µV1–5 µV/°CLarge 1/f noise, drifts with temperature
Standard CMOS op amp1–10 mV5–20 µV/°CVery high offset, worst 1/f corner
Chopper-stabilized0.5–5 µV0.005–0.05 µV/°CChopping ripple at carrier frequency
Auto-zero (sampled)1–6 µV0.01–0.05 µV/°CNoise folding/aliasing raises in-band noise
Chopper + auto-zero hybrid0.5–2 µV0.005–0.02 µV/°CCircuit complexity, higher supply current

Frequently asked questions

Why doesn't the chopper cancel the signal along with the offset?

The signal is chopped twice — up to the carrier by the input chopper and back down to DC by the output chopper — so it returns to baseband intact. The amplifier's offset and 1/f noise are added between the two choppers, so they are chopped only once and end up shifted UP to the carrier frequency, where a filter removes them.

What is the difference between chopper-stabilized and auto-zero amplifiers?

Chopping continuously modulates the signal and offset, preserving in-band broadband noise but leaving a ripple at the chopping frequency. Auto-zeroing periodically samples and subtracts the offset, which folds (aliases) broadband noise into the signal band but produces less ripple. Many modern parts combine both to get low noise and low ripple.

What limits the residual offset if chopping is so effective?

Charge injection and clock feedthrough from the MOSFET switches leave a small, temperature-dependent residue that isn't fully cancelled, plus imperfect filtering of the carrier ripple. This is why real zero-drift parts specify a few microvolts of offset rather than literally zero — typically 0.5–5 µV.

What is a typical chopping frequency, and does higher help?

Early mechanical choppers ran near 60–400 Hz; monolithic CMOS parts run from a few kHz up to about 4.8 MHz (e.g. the ADA4522). Higher f_chop pushes ripple further above the signal band and lowers effective offset error, making it easier to filter, but it also raises switching artifacts and current, so designers balance the two.

Can I use a chopper-stabilized op amp for AC or RF signals?

No — they are DC/low-frequency precision devices. The usable signal bandwidth must sit well below the chopping frequency (roughly below f_chop/2) to avoid aliasing and intermodulation. For wideband or RF work, use a standard low-noise amplifier; for stable DC and slow signals, the chopper wins.

Who invented the chopper-stabilized amplifier?

Edwin A. Goldberg developed it at RCA in 1948 (US Patent 2,684,999, filed 28 April 1948). He used a vibrating mechanical chopper to convert DC to AC so vacuum-tube analog-computer amplifiers would stop drifting. The same modulate-amplify-demodulate principle underlies today's monolithic CMOS zero-drift op amps.