2026-07-04
Here's a dirty secret of precision measurement: your voltage reference is probably the weakest link in your signal chain. A 10-bit ADC with a reference that drifts 50 ppm/°C will happily give you readings that wander with the weather. Ratiometric measurement sidesteps this entirely by making the reference cancel itself out of the math.
The trick is elegantly simple: power your sensor from the same voltage that serves as your ADC reference. If both the excitation voltage and the ADC reference move together, the ratio between them stays constant. The ADC output becomes purely a function of the sensor's transfer characteristic — not the absolute voltage feeding the whole system.
Classic example — a resistive divider sensor: Suppose you're reading a 10 kΩ thermistor in series with a 10 kΩ fixed resistor, both powered by VREF. The voltage at the tap is:
Your ADC code is:
Notice VREF vanished. If VREF drifts from 5.0 V to 4.95 V because your LDO is warming up, V_tap drops proportionally, and the ADC's reference drops proportionally too. The code doesn't move.
Where this shines:
Rule of thumb: Ratiometric operation trades reference accuracy for reference stability. You still need the excitation voltage to be quiet (low noise, low ripple), because the ADC samples one instant at a time. Slow drift cancels; fast noise doesn't. A cheap 1% LDO with 20 µV/√Hz noise usually outperforms an expensive precision reference with a noisy switching regulator upstream.
Gotchas: The ADC must actually accept an external reference tied to your sensor rail. Many microcontrollers force you to use an internal 1.024 V or 2.5 V bandgap — that kills ratiometric operation. Check the datasheet before committing. Also, load current on the excitation rail must be stable; if your MCU wakes up and yanks 30 mA during a conversion, you'll get a settling error on VREF that the ratiometric magic can't fix.