[time-nuts] Phase noise at 1-100Hz

John Miles jmiles at pop.net
Thu Apr 7 09:58:14 UTC 2011


>
> Javier Herrero wrote:
> > Hello,
> >
> > I will need to measure the phase noise of a 35MHz oscillator in the
> > range of 1 to 100Hz (well... and also at a higher range but this is no
> > problem), and I would like to know about the different alternatives. I
> > would like not to have to mess too much with mixers and soundcard
> > sampling, if possible :)

You probably need to find someone with an HP 3048A or similar outfit.  A TSC
5125A will also do the job.  The requirement to go below 100 Hz takes a lot
of the 'easy' solutions like the 11729C out of contention.

> > I think that the HP 5372A with opt 040 is able to do that, is this
> > right and worthy? If so, opt 040 is a pure software option or requires
> > also additional hardware? Would be possible also with a 5370A and
> > software processing of the raw data? Any other alternatives?
> >
> > Thanks! Regards,
> >
> > Javier
> >
> The 5372A or 5370A will only suffice if the phase noise is sufficiently
> high.
> They are likely to be too noisy by several orders of magnitude for a
> good crystal oscillator.

That was my first thought as well, but the 5370 might be OK for measuring
TCXOs and the like.  An HP paper from the mid-1970s suggests that it's
possible to reach ~-150 dBc/Hz at 1 Hz with contemporary hardware
(5345A-based, 2-ns resolution):

http://www.hparchive.com/seminar_notes/a-114.pdf

They mix down to a low IF before measurement, which they appear to vary in
order to get a lower floor at close-in offsets.

If nothing else, it should be possible to back the PN spectrum out of an
ADEV graph plotted with data from a counter, taking the noise slope into
consideration.  It would take a lot of hacking (not to mention the effort
needed to get hundreds of measurements per second out of the counter, which
I also haven't looked into.)

There's a more user-friendly app note here:
http://www.ko4bb.com/Manuals/HP_Agilent/1%29_HP_App_Notes/HP_AN225_Measuring
_Phase_Noise_with_HP_5390A.pdf

Could be a reasonable way to measure PN at offsets below 10 or 100 Hz, at
least until the higher-order noise slopes invalidate some of the assumptions
baked into the software.  It wouldn't scale very well to broadband offsets
in any event.

-- john, KE5FX




More information about the time-nuts mailing list