signal conditioning

Signal conditioning is everything you do to a raw sensor signal to make it fit to measure. A sensor rarely hands you a clean, ready-to-digitize voltage: it might be microvolts, it might be a current, it might sit on a big offset, it might be buried in noise, or it might swing over the wrong range. Signal conditioning is the chain of stages between the sensor and the converter that fixes all of that — like cleaning, framing, and lighting a photograph before you print it.

A typical conditioning chain does several jobs in order. First amplify: lift a tiny or differential signal to a useful size, usually with an instrumentation amplifier when the sensor output is differential (like a bridge). Then remove offset and level-shift: slide the signal so its zero lands where the converter expects, and scale it so full-scale signal just fills the converter's input range (no more, to use all the bits; no less, to avoid clipping). Then filter: a low-pass (anti-alias) filter removes noise and any frequencies above half the sampling rate, which would otherwise fold back as false signals. Often there is also protection at the input and buffering so the sensor is not loaded down. Worked feel: a 2 mV bridge signal might be amplified by 100 to 0.2 V, level-shifted up to center it, then low-pass filtered at 1 kHz before the converter sees it.

Why this matters: this is where measurement accuracy is actually won or lost. The blunt truth, worth repeating, is that the sensor and its conditioning — not the converter — usually set the real accuracy of the whole system. A pristine converter cannot recover information that a noisy, offset-ridden, poorly-scaled front end has already thrown away. Good conditioning means matching each stage to the next (impedance, range, bandwidth), keeping noise and offset low, and never letting the signal clip or alias. Get the front end right and the rest is easy.