ADC / DAC SNR & Noise Budget
Start from the bit count and add what a real board brings: sampling-capacitor noise, the driver amplifier, clock jitter and the datasheet's own SNR. The budget shows which source limits your ENOB, and the alias finder shows where every tone lands.
1 · Converter
2 · Signal
3 · Clock
4 · Front end
5 · Datasheet and measurement
Result
Noise budget
Each source as a share of the total noise power in the signal band. The right column gives its rms voltage and the SNR you would get if it were the only noise.
SNR vs input frequency
At the current signal level, in dBc. Jitter falls 20 dB per decade. The other terms don't depend on frequency. The dot is your input.
Where the tones land
From bits to the noise you'll actually measure
An ideal N-bit converter has a quantization error spread evenly over ±½ LSB. That is a noise of q/√12 rms, spread flat from DC to fs/2. Against a full-scale sine, it gives the textbook ceiling:
A real board adds noise that has nothing to do with N. Each source below is turned into a noise power in the signal band, and the powers are summed. Each source also gets its own bar, so you can see which one to fix first.
Jitter differs from the other sources. Its noise is proportional to the signal's slew rate, so it is a dBc figure. Lowering the input level doesn't improve jitter SNR, and with no signal at all the jitter noise disappears. That is why the dynamic range (idle-channel noise vs full scale) can be far better than the SNR on a high-frequency tone. The jitter for ≤ 1 dB figure is the rms jitter budget that costs at most 1 dB against everything else at your input frequency. Use it to choose between a crystal oscillator, an MCU PLL output or a jitter-cleaner.
Aliasing
A sampled converter only sees 0 … fs/2. Any tone or harmonic above that folds back, and in even Nyquist zones it comes back inverted. That inversion matters for undersampling receivers. The chart and table show where the input and its first harmonics land. In DAC mode they show the zero-order-hold images instead, at k·fs ± f, together with the sin(x)/x droop: −3.92 dB at fs/2, which an inverse-sinc filter or a higher update rate corrects.
Limits
- Every source is treated as white across the Nyquist band. Close-in clock phase noise stays near the carrier, so oversampling does not remove it. Flicker (1/f) noise is not modeled. Use the integrated jitter from your clock's phase-noise plot, over the band you care about.
- The ΣΔ result is the ideal peak SQNR for a pure (1 − z⁻¹)^L noise transfer function. Real high-order modulators cannot take a full-scale input and use optimized NTFs, so treat it as an upper bound and check the part's datasheet.
- Distortion is a single THD figure at your level. SFDR, INL/DNL, and the driver's own distortion are not modeled. Results are engineering estimates; confirm on hardware with a coherent FFT measurement.
References
- W. Kester, Analog Devices tutorial MT-001, "Taking the Mystery out of the Infamous Formula, SNR = 6.02N + 1.76dB, and Why You Should Care."
- W. Kester, Analog Devices tutorial MT-003, "Understand SINAD, ENOB, SNR, THD, THD + N, and SFDR so You Don't Get Lost in the Noise Floor."
- W. Kester, Analog Devices tutorial MT-007, "Aperture Time, Aperture Jitter, Aperture Delay Time—Removing the Confusion," and MT-008, "Converting Oscillator Phase Noise to Time Jitter."
- W. Kester (ed.), The Data Conversion Handbook, Analog Devices / Newnes, 2005: sample-and-hold kT/C noise, noise-free code resolution, DAC sin(x)/x response.
- R. Schreier, G. C. Temes, Understanding Delta-Sigma Data Converters, IEEE Press / Wiley, 2005: peak SQNR of an L-th order modulator.
- Texas Instruments, Analog Applications Journal 3Q 2010 (SLYT379), "Clock jitter analyzed in the time domain"; TI application note SNAA334, "Impact of PLL Jitter to GSPS ADC's SNR" (clock and aperture jitter add root-sum-square).
- IEEE Std 1241, Terminology and Test Methods for Analog-to-Digital Converters (ENOB and SINAD definitions).
Chasing the last few bits?
Kyros designs precision and high-speed signal chains: converter selection, front-end drivers, clocking and the layout that keeps the noise budget on the board.
