Cypress宣布推出了phase jitter 低至600fs的Flexo可编程硅振荡器，我对phase jitter这个说法比较困惑，于是就给Cypress的Richard Fellows写信求教。以下是我们的书信来往。
我写给Richard的第一封信，以及他的回信。
Richard,
I noted Cypress latest flexo phase jitter low to 600fs,how do you measure it,because as we know it's so close to oscilloscope jitter noise floor.
BR
Ming
Hello Ming,
Thanks for the question.
Measuring phase jitter using an oscilloscope is problematic. The oscilloscope often has greater noise in its triggering than the device being measured. In addition, it will include jitter , drift and wander frequencies that are not of consequence to the clock use.
At Cypress, we use an Agilent E5500 or an Agilent E5052B to measure phase noise.
These are specialized instruments made just for this one measurement. Their technique is to measure in the frequency domain the clock power offset from the carrier. The output is a plot of power in dB relative to the carrier power (dBc), vs. frequency offset from the carrier frequency.
The phase noise in ps or fs is calculated. It is an integration from the power vs. offset frequency plot. Different data transmission specifications have different frequency bands over which the integration is done. SONET, for example, uses the band from 12 kHz to 20 MHz offset from the carrier for the integration.
There are tools available on the web to do this integration, here is a link
http://www.jittertime.com/resources/pncalc.shtml
Regards,
Richard
我写给Richard的第二封信。
Richard,
Thanks for your professional reply rapidly,it's great to know you.
I see jitter is in time domain and phase noise is in frequency domain,in fact, it's equivalent and can be convert each other.I just confuse the description"phase jitter", haha.
E5052B(sig source analyzer) and E5500(phase noise),both of them are professional tools for phase noise measurement,large dynamic range,lower floor noise,bandwidth offset from carrier,all of these is so different with oscilloscope.Sometimes, oscilloscope also simulate like this.
I think if TIA is a better way for oscillator characterization.
BR
Ming
相噪和抖动同样都能用来表征时钟信号，但是对于高速数字时钟，一般用抖动来描述，而晶振通常都是用相噪来描述。Richard提供的http://www.jittertime.com/是一个非常有用的网站，该网站不仅对抖动进行了深入讨论，还提供了一些非常有用的抖动转换和计算工具。比如可以计算BER，ppm,RMS to PP jitter,jitter budget,还可以将相噪转换为相对应的抖动。
Richard进一步解释了Cypress Flexo 600fs的Phase jitter是如何得到的。无独有偶，在无意google E5052的时候，发现了以下链接也在讨论相噪。http://bbs.ednchina.com/ADI/ShowTopic.aspx?id=81184
这个同样是用E5500测试的10MHz的OCXO的相噪指标一下子就雷倒了我，这是多么牛的指标啊。转换到抖动，积分后的指标是137fs，我从来都没看到这么牛的晶振。遂致电询问，才得知测试这个是采用了类似“背靠背”的测试方法，由于该晶振的噪底已超过了E5500的标配信号源-160dBm的噪底，只能拿两个10MHz晶振，一个座位参考源，一个作为待测体，取两者相较的最差指标，并对结果加以软件修正。
常见相噪测试仪器有Agilent的E5500和E5052，以及Aeroflex的PN9000.那么你是如何测试相噪的呢？
;){kind=link}