[time-nuts] Designing and building an OCXO and GPSDO

Bruce Griffiths bruce.griffiths at xtra.co.nz
Wed Aug 13 18:13:32 EDT 2008


SAIDJACK at aol.com wrote:
>    Thanks Bruce,
>
>
>
>    it would be interesting to see how the different topologies affect
>    phase noise and stability etc, and what kind of performance can be
>    achieved.
>
>
>
>    bye,
>
>    Said
>   
Said

When the output signal is filtered by the crystal, the phase noise floor 
is independent of the particular oscillator design.
The phase noise floor only depends on the phase noise of the buffer 
amplifier.
Techniques for designing low phase noise buffer amplifiers are well 
established.
Basically the idea is to keep the dc and low frequency noise at the 
buffer amplifier output low and use RF negative feedback to stabilise 
the buffer amplifier gain (reduces flicker phase noise).
Thus low power supply noise is required (it modulates amplifier phase 
shift and thus contributes to phase noise).

For offsets within the crystal bandwidth, intrinsic crystal flicker 
noise contributes to the phase noise as does the phase noise of the 
oscillator active components.
The oscillator can be run in class A if AGC is used, RF negative 
feedback can then be used to reduce the sustaining circuit phase noise.
A Colpitts crystal oscillator typically operates in Class C with a small 
duty cycle, the transistor should never saturate.

Thus with carefully optimised buffer amplifiers and a fixed crystal 
current and the same crystal, the various circuits should only differ 
substantially in their flicker phase noise characteristics.
In a well designed circuit the stability will be largely determined by 
component quality, amplitude stability and residual temperature 
fluctuations. Without some stabilising RF negative feedback the flicker 
phase noise will be relatively high, however the transistor is on for 
part of the cycle which tends to reduce the phase noise.

Using a 100k resistor from the FET gate to ground isnt exactly conducive 
to a low phase noise floor unless the amplitude of the signal at the 
gate is large.

It would certainly be useful to evaluate the phase noise of the various 
oscillator circuits particularly in the flicker phase noise region as 
well as ADEV as a function of time.

Bruce



More information about the time-nuts mailing list