[Test-Equipment] "No spectrum analyzer" phase noise test setup of a 15.6 MHz TCXO?

[David]

On Mon, 28 Dec 2015 18:07:05 -0500, you wrote:

>The Kenwood TS-590 S (G) list server on yahoo has an interesting 
>discussion going on about the fairly cheap Chinese TXCO's that are being 
>sold as an optional replacement for the factory supplied 15.6 MHz 
>oscillator in the TS-590S(G) series transceivers.  It seems there has 
>been reports of phase noise issues on some Chinese replacements.  This 
>can result in unnecessary and unwanted noise being transmitted or 
>received +/- 10 to 15 KHz from the operating frequency.
>
>I am aware that today's modern procedures would use a fairly expensive 
>spectrum analyzer to measure the phase noise of such an oscillator.  
>However, most hams today do not have access to a SA, Therefore in true 
>ham fashion (meaning cheap), surely someone knows of an inexpensive way 
>to bench check the phase noise of an oscillator.
>
>Has anyone any experience in doing so?  I surmise building a good 15.6 
>MHz notch filter would be required.  In addition, a way would be needed 
>to measure 80 to 90+ db down at 10 KHz to 5 KHz away from the 
>fundamental.  Some of us have good HP power meters that have dynamic 
>range of that or a little more.  That would require the notch filter to 
>be very good (not sure one could be made that good even in a good ham lab).
>
>Anyone care to share their insight on what it would take to make such 
>measurements without a SA?
>Your attention is appreciated.
>Ray, W4BYG

I would try a procedure like this using a sound card and PC for the
spectrum analyzer:

http://wenzel.com/documents/measuringphasenoise.htm

It should be sufficient for comparison to use a pair of the Kenwood
TCXO's and a pair of Chinese TXCO's, manually adjust them for equal
frequency while looking at the output of the mixer, and then record
the low frequency spectrum of the two Kenwoods mixed together and then
the two Chinese (or suspected noisy TXCO's) mixed together.

We do not care about very low frequencies so the spectrum analyzer can
be AC coupled with a high pass filter to removed the low frequency
beat below 10 Hz or so to preserve dynamic range.

The more tradition test compares three of the same oscillator to
produce 3 results which can be used to determine the noise of each
oscillator.

A single oscillator can be measured by using a delay line to produce
two signals into the mixer but to get down into the audio range I
think the delay line would have to be impractically long.