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

John Miles jmiles at pop.net
Thu Aug 14 21:59:25 EDT 2008

> Yes, but kind of puzzles me a bit since I would
> be expecting phase noises more than 10x worst on a
> SA covering DC to 1GHz (+/-)
> (since the LO for this is an YIG oscillator circa 3GHz locked to a
> reference)
> comparing with an FFT analyzer that uses a few tens MHz sample rate.

Phase noise is slightly better on the FFT-based analyzers.  The RF signal
still has to be downconverted to baseband in the FFT analyzers, and they
still use YTO synthesizers to do that as far as I'm aware.  The difference,
with an FFT-based analyzer, is that the synthesizer does not have to be
designed with compromises like 1 Hz tuning resolution.  It's easier and
cheaper to build a clean microwave LO if its tuning resolution is 10 MHz or
so.  But there are still no free lunches to be had.

> Assuming similar 10MHz reference oscillator the SA
> gets it multiplied by 300 while a low freq SA (preferably FFT)
> gets it multiplied by 10 maximum.
> How can they claim similar performance ?!

Typically there is an intermediate synthesis stage in the first LO, where
the 10 MHz reference is multiplied to a few hundred MHz in a very clean
loop.  The first LO itself tunes from 2 GHz to (whatever), using a
relatively-small 'N' factor.  Again this is still true of FFT-based designs;
it's just that the 1st LO's intermediate synthesizer can be designed without
compromises like noisy subloops or spur-prone DDSs, or a tunable 2nd LO.

Looking at figure 1 in the PSA brochure:

... it looks like they are still using a high first IF, at 2 GHz or so,
because the LO noise profile still looks like a conventional spectrum
analyzer.  They don't show a second conversion in their PSA series block
diagram here:

http://cp.literature.agilent.com/litweb/pdf/5980-1283E.pdf (page 15)

... but you can tell it's still there, because there's a 321.4 MHz IF output
on the back panel, just like their older analog models.  They are either
using a traditional multiconversion receiver or an undersampling ADC.  My
bet would be on the former topology because it still offers the best
digitizing performance.

-- john, KE5FX

More information about the time-nuts mailing list