Aperture Jitter
Understanding Aperture Jitter
An ideal ADC samples at precisely periodic intervals. In reality, the sampling clock has random timing variations caused by oscillator phase noise, clock buffer noise, and power supply coupling. Each sample is taken at a slightly different time from the ideal, creating an amplitude error equal to the signal's instantaneous slope times the timing error.
For a full-scale sine wave at frequency fin, the maximum slope is 2πfinA, where A is the peak amplitude. The rms voltage error from jitter σt is σv = 2πfinAσt/√2. This error appears as broadband noise, reducing the achievable SNR independent of the ADC's quantization resolution. At high enough frequencies, jitter noise exceeds quantization noise and becomes the bottleneck.
SNRjitter = −20 × log10(2π × fin × σt)
Total ADC SNR (combined):
1/SNRtotal² = 1/SNRquant² + 1/SNRjitter² + 1/SNRthermal²
Reference values (σt = 100 fs):
fin = 100 MHz → SNR = 84 dB (13.7 ENOB)
fin = 1 GHz → SNR = 64 dB (10.3 ENOB)
fin = 5 GHz → SNR = 50 dB (8.0 ENOB)
fin = 10 GHz → SNR = 44 dB (7.0 ENOB)
A 16-bit ADC (96 dB ideal) becomes effectively 10-bit at 1 GHz with 100 fs jitter.
Jitter Budget by Application
| Application | fin | Target ENOB | Required σt | Clock IC Example |
|---|---|---|---|---|
| Audio (192 kHz) | 96 kHz | 20 | <1 ps | Any crystal oscillator |
| IF Sampling | 200 MHz | 12 | <100 fs | LMK04832, HMC7044 |
| L-band Direct | 1.5 GHz | 11 | <40 fs | HMC7044, LMX2594+dist |
| C-band Direct | 5 GHz | 8 | <25 fs | Custom VCXO + distribution |
Frequently Asked Questions
How does aperture jitter limit ADC performance?
Jitter causes voltage errors proportional to signal slew rate. SNRjitter = −20log10(2πfinσt). At 1 GHz with 100 fs, SNR is 64 dB (10.3 ENOB) regardless of bit depth. A 16-bit ADC effectively becomes 10-bit at that frequency.
What clock jitter is needed for modern RF ADCs?
A 14-bit ADC maintaining 12 ENOB at 1 GHz needs σt below 32 fs. State-of-the-art clock ICs (TI LMK04832, ADI HMC7044) achieve 25-50 fs. The ADC's internal jitter (30-100 fs) adds in RSS. This is why RF ADC eval boards include dedicated low-jitter clocking.
Is aperture jitter the same as phase noise?
Related but not identical. Phase noise is the frequency-domain spectral density (dBc/Hz). Aperture jitter is the time-domain rms value. Convert via σt = √(2∫L(f)df) / (2πfclk). Close-in phase noise matters for narrowband; wideband phase noise matters for wideband ADC applications.