Cardinal确定峰间的频率抖动

Cardinal确定峰值到峰值的频率抖动

Cardinal Crystal Components has determined that to accurately measure jitter of crystal oscillators, at or below 1 pico-second, the only method presently available is to measure phase noise and calculate RMS jitter. See our application note “Crystal Oscillators for Low Jitter Applications” for a full explanation. When measuring the phase noise of crystal oscillators, we use the two source method. This results in a single sideband measurement of phase in a 1 Hz bandwidth over a specific range of offset frequencies from the carrier. The single side band phase noise result is then integrated over a specified band of frequencies, usually 10 Hz to 1 MHz or 12 kHz to 20 MHz depending on the application. The SONET requirement bandwidth is 12 kHz to 20 MHz. This frequency domain integrated phase noise, in dBc/Hz, is then converted into a time domain RMS jitter in picoseconds.

Cardinal Crystal Components non-PLL oscillators, the RMS jitter has a random Gaussian distribution, because there are no phase locks loops used. This results in the lack of discrete spurious and multimodal non-Gaussian distribution. Hence, there is little or no deterministic jitter when powered by a low noise power supply. When the jitter has a Gaussian response, the RMS jitter is the standard deviation or one sigma value. 

The crystals used in the Cardinal Components oscillators have a very high Q value. The loaded Q of the oscillator resonator loop is between 10,000 to over 100,000. When the oscillator stage generates the signal, the frequency can reside anywhere within the oscillating bandwidth. However, with high Q resonators achieving a narrow oscillation bandwidth, oscillators with very low jitter can be manufactured.

With the Gaussian distribution of jitter of the Cardinal Components oscillators, the Mean value of the period is in the center of the Gaussian curve. Standard Deviation (1 sigma) is defined as the window that contains 68.26% of the total oscillation that occurs. This window is placed on one side of the mean. The 1 sigma (Standard Deviation) value is the RMS jitter of the oscillator output signal. As the number of Standard Deviations increases from the Mean, the chance of the oscillator producing a signal of that period deviation is greatly reduced. At 14.069 sigma, the probability is 1:1012. FIBRE Channel specifications requires 14 sigma reliability. 

Determining the Peak to Peak (Pk to Pk) value depends on the reliability that is desired. The more reliability desired, the greater the Pk to Pk value even though the probability of occurrences greatly reduces with higher sigma values. At 14 sigma, the probability is near 1:1012 . This is considered the standard we use to determine the Pk to Pk jitter from the RMS jitter value. If a Cardinal Components oscillator has a 2 picoseconds RMS jitter, the Pk to Pk jitter is 28 picoseconds. This is also considered the Total Jitter (TJ) . 

Cardinal确定峰值到峰值的频率抖动，基本组件已经确定，为了精确测量石英晶体振荡器的抖动，目前唯一可用的方法是测量相位噪声和计算均方根抖动。请参阅我们的应用说明“低抖动应用”以获得完整的解释。在测量晶体振荡器的相位噪声时，我们采用了双源的方法。这导致了一个单一的边带测量相位在1 Hz带宽在一个特定范围的偏移频率从载波。然后将单侧频带相位噪声结果集成在指定的频带上，通常根据应用程序为10 Hz到1 MHz或12 kHz到20 MHz。声波管要求的带宽为12kHz到20MHz石英晶振。该频域集成的相位噪声，以dBc/Hz表示，然后被转换为一个以皮秒为单位的时域均方根抖动

在基数分量非PLL振荡器中，均方根抖动具有随机高斯分布，因为没有使用相位锁循环。这导致了缺乏离散的杂散和多模态的非高斯分布。因此，当由低噪声电源供电时，很少或没有确定性的抖动。当抖动具有高斯响应时，均方根抖动是标准偏差或一个西格玛值。

在基本成分振荡器中使用的晶体具有非常高的Q值。振荡器谐振器回路的加载Q在10,000到100,000以上之间。当振荡器级产生信号时，频率可以位于振荡带宽内的任何地方。然而，随着高Q谐振器实现较窄的振荡带宽，可以制造具有非常低抖动的振荡器。

随着基本分量振荡器抖动的高斯分布，周期的平均值在高斯曲线的中心。标准差（1 sigma）被定义为包含所发生的总振荡的68.26%的窗口。这个窗口被放置在平均值的一边。1西格玛（标准偏差）值是振荡器输出信号的均方根抖动。随着标准差的数量从平均值开始增加，振荡器产生具有该周期偏差的信号的机会就会大大减少。在14.069西格玛时，概率是1：1012。光纤通道规格要求14西格玛的可靠性

确定峰值到峰值（Pk到Pk）值取决于所需的可靠性。期望的可靠性越高，Pk到Pk的值就越大，即使出现的概率随着sigma值的增加而大大降低。在14西格玛时，概率接近1：1012。这被认为是我们用来从RMS抖动值中确定Pk到Pk抖动的标准。如果基本组件振荡器具有2皮秒均方根抖动，则Pk到Pk抖动为28皮秒。这也被认为是总抖动（TJ）。