Detecting a small pressure step on a wide-range transducer: noise matters, absolute accuracy mostly cancels

Typical case: an igniter-head or flame-detection pressure rise of ~10-15 psi near 15-30 psia, on a transducer that must also survive a much higher pressure (e.g. a purge supply of several hundred psi if an outlet blocks). The instinct is to reject a wide-range sensor because its accuracy spec (say +/-0.5% of an 800 psi full scale, +/-4 psi) is a large fraction of the step. But light detection compares the reading against the value just before the event, seconds apart on the same sensor. Linearity and zero offsets are nearly identical in both readings and cancel. What matters is short-term noise and repeatability, plus thermal zero drift over the window. Quick check: step in volts = step_psi / full_scale x output_span (12 psi on 800 psi FS into 5 V is 75 mV). Compare that with the conditioner output noise (around 0.1% of full scale peak to peak is common, ~5 mV here): a 15:1 step-to-noise is easy to detect. A lower range gives proportionally more signal, but check overpressure survival (often 2x FS) against the worst blocked-line pressure first. Verify on the bench by logging the channel with the igniter off.