[time-nuts] phase noise questions

Bruce Griffiths bruce.griffiths at xtra.co.nz
Tue Jan 22 19:43:38 EST 2008


Steve
steve the knife wrote:
> Hello,
>
> I followed with some interest a discussion about a NIST doubler circuit
> using matched FET's and I was wondering if you could get similar results
> using an analog multiplier chip from Analog Devices. 
No, there is no local RF feedback in such multipliers, consequently they
are very noisy.
Use a conventional diode, FET or BJT low phase noise doubler.
Dont use the transistor doubler circuits in the ARRL handbook they are
noisy.
Smooth switching action combined with local RF feedback to suppress
flicker noise is necessary for low phase noise.
N.B. a finite source impedance can provide sufficient RF feedback.
The input signal level should be high to maximise the conversion efficiency.
Using small input signals and exploiting the approximate square law
characteristics of devices like JFETs will degrade the output phase
noise significantly over that achievable with a large signal input
switching doubler. The output current waveform (before filtering) should
resemble a rectified sinewave for a NIST style doubler.
> It would seem that
> they take some care about device matching and have parts that work up
> to pretty high frequencies. Of course there would need to be some filtering
> employed. Oh, and I think those parts do pretty well with temperature.
>
>   
Device matching has little effect on phase noise, however it does
improve odd harmonic and fundamental suppression in the output.
If your input frequency is 10MHz you dont need particularly high ft
transistors for efficient low phase noise frequency doublers.
> Also, when using a doubler that is rated in dBc how do you apply that
> number to get an expectation from a given starting dBc oscillator. So
> if my 10 MHz clock is -125dBc and I use the NIST circuit, what would
> I see at 20 MHz in dBc?
>
>   
An ideal doubler degrades this by 6dB.
i.e to -119dBc/Hz at the (unspecified) offset from the carrier.
Real frequency doublers degrade this further by a few dB.
> thanks in advance,
> steve
>   
Bruce




More information about the time-nuts mailing list