FM tuner for Jitter analysis

A few quick notes on using a tuner for digital audio clock jitter analysis (before I forget the details). Someone here (I have forgotten who) suggested an FM tuner for monitoring jitter. After a little research and experimentation I went pretty deep into figuring this out.

First the internal clocks on ADC's and DAC's have strong harmonics into the FM band. Second, FM tuners are very sensitive to modulation to carriers. modulation and jitter are closely related. The other advantage of this is that the jitter/phase noise is multiplied by the ratio of the actual carrier and the harmonic you are looking at. E.g. a 22.5792 MHz clock becomes 90.3168 MHz with the noise amplified by a factor of 4.

This is very simplified but covers the essentials-You need a really low noise FM tuner. Sound good is not an issue. The Yamaha TX-930 and TX-950 are possibly the lowest noise tuners ever made. They can be had for around $100 on eBay. What I did was to take the output directly from the detector preamp (an inline sip dual opamp) and AC couple that directly to a BNC connector I added on the back.

I also converted one of the input RF connectors to a BNC. They are AC coupled so additional stuff is not needed. You could also leave the connector as is and use RF cabling. You couple the clock into the tuner through a small cap since the 75 Ohm loading could impact the clock system.

You tune the tuner to a multiple of the clock you want to examine, set it to narrow band (if it has junk that spills over you have real problems. . .) and you can look at the phase noise spectrum directly. If you have a spectrum analyzer you can see what its composed of..

The spectrum does not map directly to the familiar plots and calibration becomes an impossible challenge. First, you are looking at frequency not phase deviation. Second there is the multiplication factor and third the sensitivity of the tuner to modulation. The last factor should be pretty consistent if the tuner is properly aligned. However converting from single sideband phase noise levels to FM deviation was more than I could effectively work out.