the effective number of bits (ENOB)
/ EE-nob /
A datasheet might boast 16 bits of resolution, but how many of those bits are actually telling you the truth, and how many are just dancing on noise? The effective number of bits, ENOB, is the honest answer. It is the number of bits a perfect, noiseless converter would need to match the real performance you measured. If a 16-bit part delivers 13.2 ENOB, four bits at the bottom are essentially worthless — flickering randomness, not real information.
ENOB is worked out from the converter's measured SINAD (signal-to-noise-and-distortion ratio, the real signal compared to everything else: quantization noise, thermal noise, AND distortion harmonics, all lumped together). You feed in a clean sine wave, measure the SINAD in dB, and invert the ideal-converter formula: ENOB = (SINAD - 1.76) / 6.02. Example: a converter measured at 80 dB SINAD has ENOB = (80 - 1.76) / 6.02 = 13.0 bits. So a chip sold as 16-bit really behaves like a flawless 13-bit one in that test.
Why this matters: ENOB is the spec to trust when you compare converters, because it folds together everything that degrades a real reading — not just quantization but distortion, jitter, and noise — into one comparable number. It also changes with conditions: ENOB usually drops as the input frequency rises (clock jitter and slew limits bite harder), so a part may show 13 bits at 1 kHz but only 10 bits at 1 MHz. Always read ENOB at the frequency YOU will actually use, not just the headline figure.
A converter advertised as 14-bit measures a SINAD of 74 dB at low frequency. ENOB = (74 - 1.76) / 6.02 = 12.0 bits. Two of its 14 bits are lost to noise and distortion, so in practice it is a clean 12-bit converter.
How many of the marketed bits actually carry real information.
ENOB is meaningless without the test conditions: input frequency, amplitude, and sample rate. A part is honest only when its ENOB is quoted at the operating point you care about; the same chip can show very different ENOB across its band.