Hi

I think you will find that they put a SAW based oscillator into the option 003
units. It certainly is more than just a “select the best” process.

One would guess that since this is phase noise and not aging, the factory
number still probably do matter.

Bob

Bob  Camp kb8tq@n1k.org writes:

Thanks for that info Bob.

There was another thread here recently entitled "Phase noise of HP8663A
640 MHz reference?" that contains a lot of relevant info on this
topic. Jim Muehlberg posted a plot/data of a non-003 unit vs a 003 unit
and, at least for those units, the phase noise of the non-003 unit
seemed to be in line with the 003 spec+typ and what John Miles measured
in his units.

I have one 003 and one non-003 here. I may see if I can do a similar
measurement. Will post back with results if so.

Matt

Option 003 was aimed at customers who needed to use the synthesizer with a 3048A system equipped with an 11729B/C downconverter.  AFAIK all of them had 640 MHz available on an SMC jack under the top cover, but it seems that only the -003 units brought it out to the rear panel.

Many of the 8662As were built without the rear-panel jack, but it can be added easily enough.  I've never seen an 8663A that didn't have it.

Whether a given unit met the PN requirements for option 003 depended on not only the 10811 OCXO but on the rest of the reference multiplier chain as well, including the 40 and 160 MHz crystal filters.  It's unlikely that all of them met the spec, given how tight the 100 Hz limit is in particular.  So there would have been some selection involved at the factory.  It might have come down to something as banal as fan blade balance.

With regard to the SAW oscillator, none of the 8662/8663 generators included one.  But if you didn't have option 003 -- or if you didn't have an 8662/8663 at all -- the SAW filter stage at the 640 MHz input on the 11729B/C could be configured as a standalone oscillator.  The filter was necessary in any case because the 640 MHz output from the generator was full of spurs from the other stages of the reference multiplier chain.

-- john

On Sat, 22 Apr 2023 at 22:39, John Miles via time-nuts
time-nuts@lists.febo.com wrote:

I seem to vaguely recollect Opt 003 also provided external EFC to the
timebase reference to allow the phase of that 640MHz signal to be
tweaked until in quadrature for phase noise measuring purposes.
Although I am always amused by the fact that for checking the SSB PN
of the 640MHz signal one needs another 8662 whose PN on the 640MHz
output is known. This feels very Alice in Wonderland.

All of mine don't have it, but at least there's a blanking plug, not a
need for a new rear panel (as with moving from front to rear inputs on
some HP products).

I'm assuming that in fact the PN measurement is of the incoherent sum
of the reference and DUT PN and what really matters is having the ref
PN as low as possible so that the measurement provides a tight upper
bound (possibly they even go as far as to assume 50-50 split of noise
power between generators?). But I'd be very interested to know if
there is more to it than that.  The manual also suggests that really
using a 3047 would make this measurement much easier and since this
would seem to be the reason for owning an opt 003 why not.

Which brings me to the question of the interfaces (I have none of
them, but do have two 8662s and would be interested in making some PN
measurements). The 3047 used the spectrum analyser interface 35601A
which I know almost nothing about and the 3048 used the 11848A for
baseband or 11729B/C for mixed down measurements. Are they all equally
useful today, possibly with a more modern FFT analyser as some of
those HP dynamic signal analysers have very low numbers of bits in the
ADC even compared to a soundcard!

Alan

Given that a 3048A system isn't too expensive to put together, it's probably
not worthwhile to assemble a 3047A system today unless you already have a
35601A and 3582A ready to go.

Unfortunately the various FFT analyzers made by HP are completely
incompatible with each other due to different GPIB command sets.  Only the
3582A can be used with the 3047A software, and only the 3561A can be used
with the 3048A.

Bit depth is not a concern for two reasons, one being that the interface
boxes have a number of variable gain stages to make the most of the
available dynamic range, and the other being that you're going to be
measuring at 1 Hz resolution anyway, so the effective bit depth is higher
than what the hardware ADC provides.

-- john

Hi

The other (possibly obvious) thing with the 3048 is that you have taken
out the carrier by the quadrature downconversion process. That happens
before you ever get to the spectrum analyzer(s).

Indeed the HP analyzers from “back in the day” had pretty limited bit depth.
That was as good as it got for the era.

Even without the 3048 hardware, you can breadboard up a quadrature setup
for < $100. Run it into a good sound card and you will have phase noise
data. There will be ground loops to deal with. There’s also some math to get
the calibration right. The ground loop part also comes in if you run the 3048.

Software wise, there are a number of free “spectrum analyzer” programs out
there to get your sound card data captured and (in some cases) normalized
to 1 Hz bandwidth.

Bob

Matt,

FWIW, here is the phase noise of an opt 003 8663 v. another of the same
at 640 MHz.  I did not bother to check if it met spec...it's Friday, and
apparently, I am needed at home for my birthday party.  Go figure.

Jim

On 2023-04-20 12:24 PM, Matt Huszagh via time-nuts wrote:

--

Jim Muehlberg
Senior Engineer

National Radio Astronomy Observatory
ngVLA Local Oscillator Lead

1180 Boxwood Estates Rd B-111
Charlottesville, VA 22903-4602
P 434.296.0270
C 434.422.2017