[time-nuts] Notes on tight-PLL performance versus TSC 5120A
didier at cox.net
Thu Jun 3 16:49:23 UTC 2010
---- Steve Rooke <sar10538 at gmail.com> wrote:
> Lets explore frequency measurement in a way that we all can
> understand. No oscillator can be measured in isolation, it has to be
> measured against another standard oscillator.
The last part of your statement is not true. An oscillator's period can be compared to a delay. A delay is the fundamental component of an oscillator, but without a number of the components that can add their own noise and instabilities.
You can test the stability of an oscillator by mixing it's output with a delayed version of it's output. You obtain a delayed version of it's output by running it through a transmission line. An integer number of wavelength plus 90 degree nominal phase shift gives you optimum phase sensitivity (and optimum amplitude noise rejection). If the temperature of the transmission line is kept stable (which may be easier said than done when dealing with very high stability oscillators, but somewhat easier to do when dealing with short measurement periods), the output from the mixer can be used to characterize the stability of the oscillator.
This method requires a mixer and a piece of coax. By adjusting the length of coax, it works for any frequency where transmission lines and mixers are available.
It is particularly useful to measure the phase noise of amplifiers, by comparing the input and output phase, using the delay line to compensate the first order phase delay through the amplifier and add 90 degrees. Using that method, the phase noise of the source can be essentially eliminated.
One significant advantage of this method over the PLL methods, or any method that requires a reference oscillator, is that it is immune from a lot of the long term noise and drift effects observed in oscillators (crystal jumps for instance).
More information about the time-nuts