Sensor Formulas

How to Calculate Sensor Accuracy, Drift, and Yield: A Worked Guide

The core math behind industrial sensor and instrumentation work: accuracy scoring, drift rate, first-pass yield, calibration workload, and burn-in throughput, each computed with real units.

Total measurement accuracy combines several error terms in quadrature, not by simple addition. For a pressure transmitter with span 0 to 100 bar, take nonlinearity 0.05 percent FS, hysteresis 0.03 percent, repeatability 0.02 percent, and thermal drift 0.04 percent over the operating band. Root sum square gives sqrt(0.05^2 + 0.03^2 + 0.02^2 + 0.04^2) = sqrt(0.0025 + 0.0009 + 0.0004 + 0.0016) = sqrt(0.0054) = 0.0735 percent FS, or 0.0735 bar on a 100 bar span. The Instrument Accuracy Score calculator uses this RSS method so you avoid the pessimistic worst-case sum of 0.14 percent that arithmetic addition would report.

Drift rate converts a calibration history into a projected error. Suppose a temperature probe reads 0.00 degrees C error at commissioning and 0.18 degrees C error after 400 days. Drift rate equals (0.18 minus 0.00) divided by 400, or 0.00045 degrees C per day, roughly 0.164 degrees C per year. To find days until the probe breaches a 0.25 degrees C tolerance, divide 0.25 by 0.00045 to get 555 days. The Sensor Drift Risk calculator turns that rate plus your tolerance band into a recalibration interval, so a probe drifting 0.164 degrees C per year against a 0.25 tolerance schedules near 18 months.

First-pass yield is finished-good units divided by units started, before any rework. A pressure sensor line starting 5,000 diaphragm assemblies that passes 4,725 at final electrical test runs 4,725 / 5,000 = 94.5 percent FPY. If three stations each pass 99.0 percent, 98.5 percent, and 96.9 percent, rolled throughput yield is 0.990 times 0.985 times 0.969 = 0.945, matching the 94.5 percent. The Pressure Sensor Yield calculator multiplies station yields so you can see that the 96.9 percent trim-and-seal step, not final test, is the loss driver worth 155 units per 5,000 built.

Calibration workload sizing starts from population, interval, and touch time. With 1,200 instruments on a 12-month interval, you owe 1,200 calibrations per year. At an average 2.5 technician-hours each, that is 3,000 hours annually. Divide by 1,600 productive hours per technician per year and you need 1.875 technicians, so staff two. The Sensor Calibration Workload calculator handles mixed intervals: 400 units at 6 months contribute 800 events, while 800 units at 24 months contribute 400, totaling 1,200 events even though the headcount base looks identical.

Instrumentation test time per unit drives line takt. If functional test runs 45 seconds, leak test 90 seconds, and calibration verification 120 seconds in series, cycle time is 255 seconds, or 4.25 minutes per unit. One station clears 3,600 / 255 = 14.1 units per hour. To hit 40 units per hour you need ceil(40 / 14.1) = 3 parallel stations delivering 42.4 per hour. The Instrumentation Test Time calculator sums sequential steps and applies your parallelism so you size test benches to demand rather than guessing.

Burn-in capacity depends on chamber slots, dwell time, and utilization. A chamber holding 240 sensors for a 48-hour dwell completes 240 units every 48 hours, or 5 units per hour, so 120 per day at full packing. Apply 85 percent effective utilization for load, unload, and ramp and you get 102 per day. To burn in 500 units per day at 48 hours dwell you need 500 divided by 102, so five chambers. The Sensor Burn-in Capacity calculator ties dwell, slot count, and utilization together so a shorter 24-hour dwell doubling throughput is obvious before you buy hardware.

Firmware validation effort scales with test coverage, not code size alone. For a smart sensor with 60 requirements at an average 3 test cases each, you have 180 cases. At 25 minutes per case authored and 8 minutes per automated run, initial authoring is 180 times 25 = 4,500 minutes, or 75 hours, and each regression pass costs 180 times 8 = 24 hours if run manually or minutes if automated. The Smart Sensor Firmware Workload calculator converts requirement counts and cases per requirement into engineer-hours so a 20 percent requirement growth to 72 requirements maps directly to 90 authoring hours.

Tie the numbers together with an example unit. A pressure transmitter built at 94.5 percent yield, verified at 0.0735 percent FS accuracy, drifting 0.164 degrees equivalent per year, needing 2.5 calibration hours and 48 hours burn-in, carries a full quantitative fingerprint. Feed the same inputs across the Instrument Accuracy Score, Pressure Sensor Yield, Sensor Drift Risk, Sensor Calibration Workload, and Sensor Burn-in Capacity calculators and every downstream schedule, from recalibration at 555 days to chamber count of five, falls out of consistent arithmetic rather than rule-of-thumb padding.

Published 2026-07-01.