[time-nuts] HP phase-noise measurement seminar coursebook

John Miles jmiles at pop.net
Sun Sep 23 15:52:57 EDT 2007

> When making automated mixer spur measurements using the HP 8566 spectrum
> analyzer and HP 8350B sweep generators, both under HP-IB control
> the phase
> noise of the 8350B was so bad you could not make a measurement.
> With the RBW
> set very side you see just a line but as the RBW is narrowed down
> you start to
> see the filter shape and at narrower settings you see gibberish
> because the
> signal has FM noise.  Switching to synthesized sources allowed
> the RBW to be
> narrow enough to get the noise floor down where it needed to be
> to see the
> spurs.  That eye opener was my introduction to phase noise.

Yes, trying to measure anything with a sweeper is a good way to bring all
those abstract graphs and concepts into real-world focus. :)

> In the Phase Noise paper you'll notice that for a number of
> measurements the
> 11729B is needed in addition to the 3047 and that when this is
> done often the
> automatic features of the 3047 are not used.  I.e. you don't need
> the 3047 for
> those measurements.  I ended up rolling my own automated phase
> noise systems
> based on the information in the 11729B manual that had an 11729B
> (usually with
> all the optional bands installed just to avoid chasing down the
> one that was
> needed) and one of the good phase noise HP signal generators
> along with the
> Rocky Mountain Basic computer.  All this would fit a 4 foot tall rack on
> wheels.  Made a number of them.

When I first got my 11729C, I intended to add a lot of code to PN.EXE to
drive it and the 8662A alongside the spectrum analyzer, but it didn't take
long to decide that I'd rather use the 11729 and 8662 manually.  The
quadrature-locking process is twitchy enough that automating it is probably
more trouble than it's worth.

> When talking about the straight spectrum analyzer method of
> measuring phase
> noise they discount the non synthesized analog spectrum analyzers
> and focus on
> the synthesized analog SAs.  But what's not mentioned are the synthesized
> digital spectrum analyzers like the HP 4395 and 4396.  These have
> true 1 Hz RBW
> settings and also have true RMS signal detection that avoids the
> corrections
> that are needed when an analog SA is used BUT they are not
> mentioned in the
> paper.  I think this is because they are made in the Kobe, Japan
> division of HP
> and are not U.S. made products.  Kobe is also where the LCR meters and
> impedance analyzers come from.

When did the 4390 analyzers come out, though?  They weren't in the '86
catalog, so they wouldn't have been mentioned in the 1985 seminar.

There are at least one or two errors in those notes, by the way... in at
least two places, they show graphs claiming the 8568A's noise floor is much
worse than it actually was.  It looks like the baseline traces that they
showed for the 3585A are actually the ones that apply to the 8568A.  I
corrected one graph by hand before I scanned it, but noticed another one

> I've downloaded your Phase Noise software but have not yet tried
> it out on my
> 4395A.  http://www.prc68.com/I/4395A.shtml
> I think all that's needed is to enter 0 dB into the Noise Response window
> instead of the +2.5 db for a peak detector and -0.5 db for RBW correction
> needed for analog SAs.

Sounds likely, if that's how it works.  (If you use PN.EXE with an 11729B or
11729C test set, the Noise Response factor needs to be -46 dB plus the
analyzer correction factor if any; the external-LO field should equal the
carrier being tested; and the external IF is 0 Hz.  That should make all the
numbers come out right.)

Please send me a baseline .PNP file if the SCPI-acquisition code actually
works on the 4395A!  Would be good to add it to the collection.

-- john, KE5FX

More information about the time-nuts mailing list