What Do ENOB, SINAD, and Noise Floor Mean on an ADC Datasheet?

Detailed Explanation

An ADC's headline resolution — "12-bit," "16-bit," "24-bit" — is a useful first filter when selecting a converter, but it is not the number that determines whether a design will actually resolve the measurement it needs to make. That number is ENOB, and understanding how it's derived and how it should actually be used is the difference between a noise budget that holds up in the field and one that looks fine on paper but fails once real hardware and real signals are involved.

SINAD: The Foundation Measurement

SINAD (Signal-to-Noise-and-Distortion Ratio), expressed in dB, is a measured figure: apply a known, clean sine wave input to the ADC at a specified amplitude and frequency, digitise it, and compare the power of the fundamental signal against the combined power of every other component in the output spectrum — noise, harmonic distortion, and any spurious content together. A higher SINAD means the digitised output more faithfully represents the ideal input signal, with less total corruption from noise and distortion combined.

ENOB: Converting SINAD into a Bit Count

ENOB (Effective Number of Bits) converts the SINAD measurement into a number directly comparable to the ADC's nominal resolution, using the relationship:

ENOB = (SINAD − 1.76) / 6.02

The constants come from the theoretical SINAD of an ideal ADC limited only by quantisation noise — 6.02 dB per bit of resolution, plus a 1.76 dB constant term. A real ADC's measured SINAD is always lower than this ideal figure once actual thermal noise, reference noise, and nonlinearity are included, which is why ENOB is always below the converter's nominal bit count — this is expected behaviour, not a sign of a defective part. A 16-bit ADC delivering an ENOB of 13–14 bits at its datasheet's stated test conditions is a completely normal result for many precision converters.

THD: Isolating the Distortion Component

THD (Total Harmonic Distortion) separates out just the distortion contribution — the power in the harmonics of the input signal frequency — from the noise contribution that SINAD lumps in together. A converter can have excellent SINAD dominated by noise with negligible distortion, or a case where distortion is the larger contributor; THD (and the related THD+N, which adds the noise term back in for a figure closer to SINAD) tells a designer which one it is. This matters for diagnosing a noise-budget shortfall: a distortion-dominated problem often points to input signal amplitude too close to full scale or an anti-aliasing filter interaction, while a noise-dominated problem more often points to reference noise, power supply noise, or the converter's own thermal noise floor.

Noise Floor and Dynamic Range

The ADC's noise floor is the effective random-noise level below which no genuine signal can be distinguished — closely related to SINAD, since a lower noise floor directly improves it. Dynamic range describes the ratio between the largest signal the converter can represent without clipping and its noise floor, and for many converter architectures (particularly sigma-delta parts, which trade sample rate for noise performance via oversampling) dynamic range and ENOB move together: a converter with better dynamic range at a given input condition will generally show a correspondingly better ENOB there too. See how do you design active filters? for how an anti-aliasing filter's own noise and distortion contributions interact with the ADC's own noise floor in a real signal chain, and sensor signal conditioning basics for the broader analog front-end noise budget these figures feed into.

Using ENOB as a System Noise Budget

The practical application of all of this is straightforward once the terms are clear: convert the application's actual required measurement resolution into an equivalent number of effective bits at the real signal amplitude and frequency the design uses, then compare that figure against the ADC's datasheet ENOB at matching (or the closest documented) test conditions — never against the nominal bit count. A design that needs 12 effective bits of genuine measurement resolution and selects a "16-bit" ADC with a datasheet ENOB of 13 bits at the relevant test condition has real, if modest, margin; the same design built on the assumption that "16-bit" alone guarantees 16 bits of usable resolution has none, and may fail to meet its accuracy target once real hardware replaces the datasheet's controlled test setup.

Practical Examples

A 24-bit sigma-delta ADC for a load-cell or strain-gauge application commonly delivers an ENOB well below 24 — often in the high teens to low 20s depending on the specific gain, sample rate, and filter settings — which is still far more than enough resolution for the application, but only if the design was budgeted against the realistic ENOB figure rather than the nominal 24-bit resolution.

A 12-bit SAR ADC on a microcontroller sampling a sensor signal near the top of its input frequency range will typically show measurably worse SINAD (and therefore ENOB) than the same ADC sampling a slow, near-DC signal — worth checking against the datasheet's ENOB-vs-frequency curve if the application's signal bandwidth is a meaningful fraction of the ADC's rated analog input bandwidth.

Design Considerations

• Always compare a required resolution against ENOB at matching test conditions, not the nominal bit count. A datasheet's headline resolution number is a ceiling, not a guarantee, and the gap between the two can be several bits depending on the converter architecture and operating point.
• Check whether the datasheet's ENOB/SINAD figure was measured under conditions close to the application's actual operating point. A figure measured at low input frequency and the converter's ideal reference voltage may not hold at the application's actual signal frequency, sample rate, or supply noise environment.
• Budget additional margin beyond the datasheet ENOB for board-level effects the vendor's evaluation board won't reproduce — power supply noise coupling, reference noise, and layout-induced interference on a real product PCB typically degrade effective resolution further than the datasheet figure alone suggests. Zeus Design designs precision analog measurement and sensor acquisition systems, including ADC selection and system noise budgeting, for embedded products.
• Use THD alongside SINAD/ENOB when diagnosing a noise-budget shortfall, since distortion-dominated and noise-dominated problems point toward different root causes and different fixes.

Common Mistakes

• Selecting an ADC based on its nominal bit count alone, without checking the datasheet's ENOB or SINAD figures at conditions relevant to the actual application, and discovering the achievable resolution is materially lower than assumed once the design is built and tested.
• Comparing a required resolution against an ENOB figure measured at unrelated test conditions — a low-frequency ENOB figure does not necessarily apply at the application's actual signal frequency or sample rate.
• Assuming a higher nominal-bit-count ADC always outperforms a lower one. A well-designed 12-bit converter can deliver a higher ENOB at a given operating point than a poorly-matched 16-bit converter operating outside its optimal conditions — the datasheet ENOB comparison, not the nominal resolution, is the figure that actually matters.
• Ignoring board-level noise contributions and treating the datasheet ENOB as the final achievable system resolution. The datasheet figure is measured on the vendor's own evaluation board under controlled conditions; a real product PCB's power, reference, and layout noise sources typically erode some of that margin.

For ADC fundamentals — resolution, sampling rate, SAR vs sigma-delta architecture selection — see what is an ADC (analog-to-digital converter)?