REAL-WORLD LIMITS
Three ways a reading goes wrong
Drift
A slow, one-way creep of the output even at constant pH
Hysteresis
The reading depends on which pH the sensor saw before
Noise
Random fluctuations that set the smallest change you can detect
DRIFT
A slow creep in the threshold
Water slowly hydrates the top of the insulator and ions diffuse in, so the threshold creeps one way even in a fixed buffer. A few millivolts per hour is common for bare SiO₂ or Si₃N₄, and denser oxides drift less. At 2 mV per hour and 53 mV per pH unit, the error grows about 0.04 pH per hour.
HYSTERESIS
The sensor remembers
TEMPERATURE
The slope depends on temperature
The Nernst slope grows with kT/q. Body temperature gives 61.5 mV per pH unit, against 59.2 at 25 °C. The threshold and mobility also shift with temperature, so measurements should be calibrated at, or corrected to, the temperature of use.
TAMING THE LIMITS
Practical fixes
ISFET–REFET pair
A twin transistor with a pH-insensitive surface. Subtracting it cancels shared drift and temperature effects
Two-point calibration
Two known buffers fix both the slope and the offset
Stable insulators
Al₂O₃, Ta₂O₅, and HfO₂ drift less than bare SiO₂
UNIT 13 STUDY COMPLETE
Ready for the Fab Challenge?
You've covered the real-world limits that separate a lab demo from a dependable sensor.