Phase noise to jitter calculator
Paste the phase noise points from a data sheet to get the RMS jitter, the phase error and the SNR limit it sets for an ADC.
6 points read, 100 Hz to 20 MHz.
ADC SNR limit
| From | To | Slope | Noise | Share |
|---|
How to use the phase noise to jitter calculator
- Type the carrier frequency of the clock or oscillator, for example 100 MHz.
- Copy the phase noise table from the data sheet and paste it into Phase noise points: one offset and one level in dBc/Hz per line. "10 kHz −140", "1e4, −140" and "10k −140" all work, and so does a table with a row of offsets above a row of levels. Lines without two numbers, such as the header, are skipped. Press Clear to paste first to start from an empty box.
- Set the integration limits, or press 12 kHz – 20 MHz. All points uses the whole curve.
- Read the RMS jitter, the integrated phase noise and the phase error. The table under the chart shows how much each part of the curve adds.
- Type the ADC input frequency and bits to see the SNR limit this clock sets, then copy the results or share a link with your points.
Worked example: a 100 MHz clock
Take a 100 MHz oscillator with this curve: −140 dBc/Hz at 10 kHz, −150 at 100 kHz and −160 at 1 MHz, flat at −160 out to 20 MHz. It is the example already loaded in the tool. We want the jitter from 12 kHz to 20 MHz.
Between 10 kHz and 1 MHz the curve falls 10 dB per decade, so the noise per hertz falls as 1 ÷ f. The area of such a part is the level (as a ratio) times the start offset times ln(end ÷ start). From 12 kHz to 100 kHz that is 10−14 × 10,000 × ln(8.33) = 2.12 × 10−10. From 100 kHz to 1 MHz it is 10−15 × 100,000 × ln(10) = 2.30 × 10−10. The flat part is just level times width: 10−16 × 19,000,000 = 1.90 × 10−9.
The total is 2.34 × 10−9, and 10 × log10 of that is −86.30 dBc. The RMS phase error is √(2 × 2.34 × 10−9) = 68.4 µrad, about 0.0039°. Divide by 2π × 100 MHz and you get 1.089 × 10−13 s: 108.9 fs of RMS jitter. The table under the chart shows the same split as the picture, so you can see which part of the oscillator is worth improving.
Why jitter limits ADCs
An ADC takes a sample each time its clock ticks. If a tick comes a little early or late, the input has already moved, and the sample is off by the slope of the signal times the time error. A fast signal has a steep slope, so the same jitter makes a bigger error. The result is noise that grows with the input frequency: SNR = −20 × log10(2π × fin × tj).
With the 108.9 fs clock above and a 100 MHz input, the limit is 83.3 dB, about 13.5 bits. A perfect 14-bit ADC has 86.04 dB of SNR, so this clock costs it 4.6 dB and it ends at 81.4 dB. A 12-bit ADC (74.0 dB) loses only about 0.5 dB. Feed the same 14-bit ADC a 10 MHz input instead and the jitter limit rises to 103.3 dB, so the clock hardly matters.
That is why fast ADC data sheets ask for clocks with jitter in femtoseconds. The ADC part of the tool shows the jitter at which clock noise equals the quantisation noise: 79.4 fs for 14 bits at 100 MHz, and about half of that, 40 fs, to lose under 1 dB. To turn SNR into bits, or to add the converter's own noise and distortion, use the ENOB calculator.
Method
The tool joins your points with straight lines on the log-frequency chart, the way data sheets draw them. A straight line in dB against log frequency is a power law, L(f) = L1 × (f ÷ f1)b, where L1 is the level at the start point as a ratio (10dBc/10) and b is the slope in dB per decade divided by 10. Each part is integrated exactly from fa to fb:
area = L1 × f1 ÷ (b + 1) × ((fb ÷ f1)b+1 − (fa ÷ f1)b+1), or L1 × f1 × ln(fb ÷ fa) when the slope is −10 dB per decade.
The areas inside your limits are added to get A in dBc. Then phase error (rad) = √(2 × 10A/10) and jitter (s) = phase error ÷ (2π × carrier frequency). The 2 counts both sidebands, because data sheets give single-sideband phase noise, L(f).
Limits
The answer is only as good as the points. Straight lines between a few points can miss a bump such as a PLL's loop peak, so paste every point the data sheet gives. Spurs, single tones next to the carrier, are not part of the dBc/Hz curve and add jitter of their own that this tool does not count. Outside your points nothing is counted unless you choose "Extend flat". The ADC numbers treat the clock jitter as random and use the ideal quantisation noise; a real converter adds its own aperture jitter and noise. For the noise of a whole receiver chain, see the noise figure calculator. All RF calculators are on the RF tools page.
FAQ
How is phase noise turned into jitter?
The noise is integrated across the offset range, then converted: jitter = √(2·10^(A/10)) ÷ (2π·f), where A is the integrated phase noise in dBc and f is the carrier frequency.
Which offset range should I use?
Use the range your system cares about; 12 kHz–20 MHz is a common telecom standard.
Why does a higher carrier mean less jitter for the same curve?
Jitter is time; the same phase error is a shorter time at a higher frequency. Double the carrier with the same curve and the jitter halves.
What jitter does a 14-bit ADC at 100 MHz need?
Under about 100 fs to avoid losing SNR; the tool shows the exact limit. At 79.4 fs the jitter noise equals the quantisation noise of a perfect 14-bit converter, which already costs 3 dB.