[time-nuts] Upper limit on phase noise from two oscillators.

Jeffrey Pawlan jpawlan at pawlan.com
Wed Apr 27 12:05:49 EDT 2005

On Wed, 27 Apr 2005, David Kirkby wrote:
>if one oscillator was perfect (no phase noise at all)
also wrote:
> If one oscillator was noiseless

If you or any one on this list has a lead about where to find a noiseless
oscillator, PLEASE let me know immediately.  I want one!

On Wed, 27 Apr 2005, Mike Feher wrote:
> This does not seem to make sense. No matter what you are measuring it is
> assumed that the phase noise of the item being measured is greater than the
> piece of test equipment doing the measurement, else you would not see it.

That is the correct description for using a Spectrum analyser to measure phase
noise. I have the HP phase noise measurement utility built into my spectrum
analyser and as long as the signal being measured is at least 10dB worse than
the noise floor of the spectrum analyser noise floor, the readings are
reasonably accurate.  But there is no spectrum analyser made that is good enough
to measure a precision oscillator like the 10811. Therefore the measurement is
done by mixing two of them together and one must phaselock one oscillator to the
other and drive the mixer ports at 90 degrees from each other. The resulting
output is zero frequency (DC) and an audio spectrum analyzer reads the sidebands
which is the summed phase noise of the two.
  Mike,  you forgot that the audio spectrum analyzer does not need to have the
dynamic range of the phase noise unless you are measuring quite close to the
carrier simultaneously with measuring far from the carrier. If you add high pass
filters you can make your measurements in segments so that a modest audio
spectrum analyser will work. You may have also forgotten that the carrier
becomes a DC voltage since the two oscillators are phase locked to each other.
You must use a blocking cap (minimal HPF) to not overload the inout of the
soundcard if it is DC  coupled.

I hope this helps but I need to get back to work now.


Jeffrey Pawlan

More information about the time-nuts mailing list