[time-nuts] Phase noise of HP8663A 640 MHz reference?

[ed breya]

Tobias wrote

"Hi Ed, could you share some info about your own 640 MHz source?
I would be very interested.
I do have an 8663A, but I have no idea about how good it actually is and
currently, I don't have a possibility to test it because I still lack an
11729C carrier noise test set (but I would like to make something 
similar myself some day)."


The plan was to do the doubling and amplifying similar to the 
8662A/8663A (which I think are about the same in this respect), except 
that I don't have anything like the mentioned 40 MHz and 160 MHz crystal 
filters. Starting from the higher OCXO frequency should help some, 
depending on its noise performance.

The filters I do have can't fix anything close-in, but should work very 
well on spurious content from almost any multiplier scheme. The first 
one especially is a single unit 160 MHz BPF made from two tubular 
coaxial ones cascaded. I didn't make it - it came this way as a 
commercial product. It has enormous stop-band suppression maybe 200 dB 
(theoretical) by +/- 10 MHz away from fc, but large insertion loss about 
10 dB. The plan was to double the 80 MHz OCXO into this filter, which 
should eradicate all spurious from below. The 320 and 640 MHz BPFs are 
more conventional.

The 80 MHz OCXO was apparently quite common around 20-30 years ago. 
Despite this, I couldn't find any specs twelve years ago, and can't find 
any today, even though there seems to be plenty of them still around and 
for sale. The ones I have are Vectron 229-9237, and 229-5657-1, 
apparently the same except for mechanical construction.

I have a bunch of similar units, mostly oddball frequencies in the 100 
MHz range, and I did have to take some apart (soldered shut cans) over 
the years to modify for particular projects. I found they all used 
half-frequency crystals and built in doublers. The 80 MHz is no 
exception - a quick look on the SA shows it's a 40 MHz OCXO that's 
doubled up internally, so it's really only starting with a four times 
frequency versus multiplying a 10 MHz reference.

Another thing I noticed is that the 640 MHz SAW BPF in the 11729C may 
not be for closer in spurious content, but mostly wide cleanup, and 
optimized to form a good oscillator when used for that mode. The manual 
says that the purpose of the filter is to reduce 120 MHz, 520 MHz, and 
760 MHz spurs. These and others naturally come from the rather 
complicated 8662A/8663A reference generator/multiplier system, and the 
640 MHz output does not appear to have very much isolation from all this 
activity.

So, if you make your own 640 MHz "clean and simple" by direct 
multiplication, with no side deals for other frequencies, the result 
will not include the extra stuff that would be coming from the 
generator. If you also start with a good HF OCXO with known specs, and 
do careful multiplication, filtering, and PLLing, I think it can beat 
the noise performance of the 8662A/8663A's 640 MHz source. How much? I 
dunno, but suspect that the hump from around 10 Hz to 10 kHz may be due 
in part to all the reference making and synthesizing action going on in 
there, that's somehow included in the 640 MHz output. That is, presuming 
the 10 MHz internal reference has no such hump. If it does, then it 
could be simply the result of the multiplication factor, and unavoidable.

Ed