2026-05-15
Every op-amp has an input offset voltage (VOS) — a small DC error that appears between the input pins even when they should be equal. For a jellybean part like the LM358, VOS can be 7 mV with a tempco of 7 µV/°C. When you're amplifying a 100 µV thermocouple signal at a gain of 1000, that offset becomes 7 V of error at the output — completely swamping your signal. Chopper stabilization is the technique that drives VOS down to the nanovolt range and nearly eliminates drift.
The core trick: modulate the DC input up to a higher frequency, amplify it there, then demodulate back to DC. Op-amp offset and 1/f noise don't get modulated — they live at DC. So when you demodulate the signal back down, the offset gets shifted up to the chopper frequency, where you filter it out.
The classic architecture uses two MOSFET switch pairs forming input and output choppers, clocked at fchop (typically 10 kHz to 1 MHz). The input chopper alternately swaps the (+) and (−) inputs at fchop. The signal becomes a square wave at fchop. The amplifier amplifies this AC signal. The output chopper, synchronized with the input, demodulates it back to DC. Any DC error introduced by the amplifier itself gets chopped up to fchop and is removed by a low-pass filter.
Real-world example: The Analog Devices LTC2057 is a chopper op-amp with VOS = 4 µV max and drift of 0.015 µV/°C — roughly 500× better than the LM358. In a strain-gauge load cell amplifier (4 mV full-scale output at 2 kg load), using a standard op-amp gives you ±2% error from offset alone. Drop in an LTC2057 and the offset error becomes 0.1% — measurement-grade.
Rule of thumb: The signal bandwidth must be well below fchop/2 (Nyquist for the chopping). For a 10 kHz chopper, keep your input signal under ~1 kHz. Above that, you get aliasing of chopper artifacts back into your signal band.
Watch out for:
Use choppers when you have small DC signals (thermocouples, bridges, current shunts) and you can't afford offset trim or auto-zero routines in firmware.
