ADC Quantization Noise Calculator

Calculate theoretical SNR (signal-to-noise ratio from quantization only), LSB size, RMS quantization noise voltage, and noise spectral density for an N-bit ADC. Includes oversampling SNR gain.

Calculator Electronics Updated Oct 15, 2026
How to Use
  1. Enter ADC resolution in bits (typically 8–24).
  2. Enter the full-scale voltage range (unipolar = Vref; bipolar = 2·Vref).
  3. Optionally enter sampling rate and oversampling ratio (OSR) for noise density calculation.
  4. Results: theoretical SNR in dB, LSB size, RMS quantization noise, noise spectral density, and oversampling gain.
  5. Compare the SNR with the datasheet's measured SNR or SINAD: the gap shows how much noise the real converter adds on top of quantization.
Input
V
Presets
Quantization
SNR
—
LSB
—
RMS noise
—
Levels
—

Worked Example

A 12-bit ADC with a 3.3 V full scale, sampling at 1 MHz. LSB = 3.3 ÷ 4096 = 0.806 mV. The quantization error is spread evenly over ±½ LSB, so its RMS value is LSB ÷ √12 = 0.233 mV.

SNR: 6.02 × 12 + 1.76 = 74.0 dB for a full-scale sine. Spread over the 500 kHz Nyquist band, the noise density is 0.233 mV ÷ √500,000 = 0.33 µV/√Hz.

Oversampling 16× and averaging adds 10 × log₁₀(16) = 12.0 dB, so 86.0 dB, about two extra bits, as long as the signal carries enough noise to keep the error random.

The common mistake: expecting 4× oversampling to add two bits. Each extra bit needs 4× more samples, so two bits need 16× and three bits need 64×.

Show Work

Enter values to see the noise calculation.

Formulas

Theoretical SNR
SNR = 6.02·N + 1.76 dB
Full-scale sine, quantization only.
LSB Size
LSB = Vfs / 2N
Quantization step.
RMS Noise
σq = LSB / √12
Uniform distribution over ±½ LSB.
Oversampling Gain
+10·log(OSR) dB
4× OSR = +6 dB = +1 ENOB.
ENOB
ENOB = (SINAD − 1.76) / 6.02
Real-world effective bits.
Noise Spectral Density
NSD = σq² / (fs/2)
V²/Hz across Nyquist band.

History of Quantization Theory

Bernard Widrow's 1956 MIT doctoral thesis "A Study of Rough Amplitude Quantization by Means of Nyquist Sampling Theory" formalized the statistical treatment of quantization noise, establishing that for "busy" signals the error behaves as white noise uniformly distributed over ±½ LSB. The famous 6.02·N + 1.76 dB formula falls directly out of this analysis — it's the ratio of signal RMS to quantization noise RMS expressed in dB.

Oversampling and decimation as a resolution-enhancement technique was formalized by Inose and Yasuda in 1962, who built the first Σ-Δ (sigma-delta) modulator. Their 1-bit noise-shaping converter at high oversampling ratios produced effective resolution far beyond any flash or SAR ADC of the era. Commercial sigma-delta converters through the 1980s and 1990s made 18- to 24-bit resolution practical for instrumentation and audio.

Low-resolution converters now come close to their theoretical SNR, while high-resolution ones are limited by analog noise long before quantization: a 24-bit converter's 146 dB ceiling is out of reach of any real front end. That is why datasheets quote measured SNR, SINAD and ENOB rather than just the bit count.

About This Calculator

Enter ADC resolution, full-scale voltage, and optionally sample rate and oversampling ratio. The tool computes theoretical SNR (quantization-only), LSB size, RMS quantization noise voltage, and noise spectral density. If an OSR is provided, it adds the oversampling SNR gain.

These are theoretical limits. Converters up to about 12 bits often come within a few dB of them; at 16 bits and above, thermal noise, reference noise and clock jitter leave real parts well short. Actual SNR and ENOB appear in datasheets as \"Typical SINAD\" specs at rated input frequency — usually under idealized conditions (nearly full-scale sine, optimal input bandwidth). For worst-case performance, check SNR vs. input amplitude and temperature plots. Everything runs in your browser; nothing is sent to a server.

Frequently Asked Questions

Where does 6.02·N + 1.76 come from?

Quantization noise has a uniform distribution over ±½ LSB, giving RMS value LSB/√12. A full-scale sine wave has RMS value (2N·LSB)/(2√2). The SNR is 20·log(signal/noise) = 20·log(2N · √12 / (2√2)) = 20·log(2N·√(3/2)) = 6.02·N + 1.76 dB. Every added bit increases SNR by 6.02 dB.

What does oversampling do to SNR?

Oversampling spreads the quantization noise over a wider frequency range; after digital low-pass filtering at the signal band, only a fraction of the noise remains. SNR improves by 10·log(OSR): 4× OSR gives +6 dB, one extra effective bit; 16× gives +12 dB, two bits. It only works when the error is random from sample to sample, so a quiet, steady signal needs a little noise (dither) added first. Sigma-delta converters do far better by also pushing the noise up out of the signal band (noise shaping).

Difference between SNR and SNDR?

SNR includes only noise (white quantization noise in theory). SINAD (or SNDR) includes noise + distortion — real-world harmonic distortion from nonlinearities. Datasheets quote both. ENOB is derived from SINAD: ENOB = (SINAD − 1.76) / 6.02.

What's noise spectral density?

Quantization noise power (LSB²/12) spread across the Nyquist bandwidth (fs/2). NSD = (LSB²/12) / (fs/2) with units V²/Hz. Useful for comparing ADC noise to analog input noise, or estimating the noise in a specific signal bandwidth.

Why is quantization noise treated as white?

For "busy" signals (spanning many LSBs, changing sample-to-sample), the error is effectively random and uniformly distributed over ±½ LSB. For slow or near-DC signals, quantization error is correlated — periodic pattern noise, not white. That's why dithering (adding small noise) helps randomize it for low-level signal fidelity.

How do I use the ADC Quantization Noise Calculator?

Just type your numbers. The answer shows up right away — there is no button to press. Change anything and it updates by itself.

Is it free? Does it work without internet?

Yes to both. It is free with no sign-up, and once the page has loaded it keeps working even with no internet.

Where does my data go?

Nowhere — every calculation runs on your own device. Nothing you enter is uploaded, logged, or stored.

Common Use Cases

Audio ADC Selection

A 24-bit ADC has a theoretical SNR of 146 dB, but real 24-bit audio converters measure roughly 110–125 dB, because thermal noise in the analog front end, not quantization, sets the floor. That is still far beyond what any listening room can use.

MCU Built-in ADC

A microcontroller's 12-bit ADC has a 74 dB theoretical ceiling, but board noise and the converter's own errors usually leave 9 to 11 effective bits. Fine for temperature and battery monitoring; averaging 16 readings wins back about two bits.

Seismic / Geophysical Monitoring

On paper, a 24-bit converter oversampled 500× would gain 27 dB, reaching 173 dB. Real seismic recorders fall well short of that, because the analog noise floor arrives long before the quantization limit, which is why the formula is an upper bound, not a promise.

Oscilloscope ADC

An 8-bit oscilloscope ADC has a 49.9 dB theoretical SNR. At high input frequencies, real scopes typically deliver 6 to 7 effective bits: fine for seeing a waveform's shape, poor for measuring small details on a large signal.

Instrumentation DMM

A 6½-digit multimeter resolves about 2 million counts (21 bits, 128 dB on paper). It gets there with a slow integrating converter that averages over whole power-line cycles, trading speed for low noise.

Last updated: