ADC Averaging / Oversampling

Estimate noise reduction and effective resolution (ENOB gain) from averaging or oversampling + decimation.

How to Use

  1. Pick a mode: Averaging (noise reduction) or Oversampling (extra effective bits).
  2. Enter ADC parameters (bits + Vref) if you want LSB sizing, and enter noise in LSB or volts.
  3. Set Samples (N) or choose a target Extra bits and let the tool solve for required N.
  4. Use “Show Work” to see formulas and step-by-step math in base units.
Results Overview
Updates on input changes (no URL changes while typing).
Noise ÷
—
ENOB Gain
—
Eff. Bits
—
Output Rate
—
Status: —

Important: Oversampling improves effective resolution only when there is sufficient noise/dither and proper low-pass filtering before decimation.

Inputs & Settings
Choose what you know. The tool can work in volts or LSB.
Averaging: noise scales ~ 1/√N. Oversampling: ENOB gain ~ 0.5·log2(OSR).
Pick which value is “unknown” for the solver.
Used to compute LSB size from Vref.
LSB ≈ Vref / (2^bits). (Assumes full-scale is 0..Vref.)
Used to estimate output rate after decimation (Fs / N or Fs / OSR).
Averaging: N samples averaged. Oversampling: often N = OSR (decimation factor).
Rule of thumb: +1 bit needs ~4× samples (OSR×4). (+2 bits → 16×, etc.)
If noise is too low, oversampling may not increase ENOB without dithering.
This tool focuses on the math relationships. Filter shape impacts passband/stopband.

Tip: If you’re solving for extra bits, start with noise in LSB RMS and set N to your planned OSR/average window.

Show Work (step-by-step)
—
Work is shown using base units. Approximations are called out where used.

Reference

These are common, practical relationships used for quick engineering estimates.

  • LSB size (ideal): LSB ≈ Vref / 2^bits
  • Averaging noise reduction: σ_out ≈ σ_in / √N
  • Oversampling ENOB gain (rule of thumb): Δbits ≈ 0.5 · log2(OSR)
  • Samples required for Δbits (rule of thumb): OSR ≈ 4^(Δbits)
  • Decimated output rate (typical): F_out ≈ Fs / N
Real ENOB depends on quantization noise, analog noise, distortion, and filtering. This tool is an estimator.

FAQ

Does averaging always improve resolution?

It reduces random noise (roughly by √N). If your dominant error is DC offset, drift, or quantization with insufficient dither/noise, the improvement may be limited.

Why does oversampling need noise?

Oversampling + decimation relies on noise/dither to spread quantization error so averaging can reduce it. With a perfectly static signal and too little noise, extra “bits” won’t appear.

What’s the quick rule for extra bits?

Roughly +1 bit → 4× samples, +2 bits → 16×, +3 bits → 64×. That’s the OSR ≈ 4^(Δbits) relationship.

Is “Fs/N” always the output rate?

For a simple block average / decimation by N, yes. For more complex filters or overlapping windows, it depends. This page uses the common decimation estimate unless you implement something else in firmware.

Tool Info

Last updated:

Updates may include filter options, edge-case handling, and export formats.