Jitter

Measuring and specifying jitter

Measuring and specifying jitter using a phase noise plot for a crystal oscillator that contains only random jitter sources has several advantages over conventional time domain methods using high-speed digital sampling oscilloscopes (DSO’s).

Disadvantages of DSOs:

The jitter performance of the oscillator is often much better than the combined jitter uncertainty of the DSO’s internal sampling oscillator, the DSO’s trigger point uncertainty and some questionable software techniques.
The measurement bandwidth is unknown with a DSO. For example Belcore specifications define any ‘jitter’ below 10 Hz as wander, not jitter.
The DSO can only give a total jitter figure. It cannot split the jitter content into particular areas of interest.

Advantages of the phase noise plot:

There is only a ±2 dB absolute uncertainty using the HP phase noise measurement system. What you measure is only phase noise (jitter), not measurement induced errors.
The measurement bandwidth is specified as part of the measurement.
The contribution to the overall jitter figure can be specified for any particular band of frequencies.

Converting phase noise to jitter

The measured phase noise plot is broken down into areas of constant slope (see Figure 32, the idealized phase noise plot).

These areas are integrated and summed to give an equivalent single sideband at the maximum frequency of integration. Now think of the oscillator as a perfect source with this noise vector rotating about its end.

Jitter RMS in degrees is then calculated as the maximum angle between the resulting vector and the carrier vector. To convert to time express as a fraction of 360 degrees and multiply by the period of the carrier frequency.

Jitter RMS in pico secs = angle/360 * T

Using the Gaussian distribution as defined in Figure 33 and the appropriate mathematical manipulation, the RMS jitter can be converted to peak-peak jitter for a specified sampling time (say) or an expected bit error rate (BER) for a given sample size etc. Figure 34 is an example of a typical phase noise plot converted to time jitter.

This conversion of a real 63.8976 MHz non-multiplied oscillator shows 99% of the contribution to the total jitter is generated by the close in phase noise between 10 Hz and 100 Hz. It would make no
sense, therefore, for the specification to demand –150 dBc at 1 kHz as this would make absolutely no difference to the overall jitter performance.