[time-nuts] Problem with HP 83623A 20 GHz sweep generator stepping up/down 100 Hz when not wanted.

[Dr. David Kirkby (Kirkby Microwave Ltd)]

I run a test over the weekend with a fellow radio ham. I transmitted 100 mW
or so at 10368.115 MHz from an HP 83623A 10 MHz to 20 GHz sweep generator

http://www.keysight.com/en/pd-1000001862%3Aepsg%3Apro-pn-83623A/synthesized-sweeper-10-mhz-to-20-ghz-high-power?cc=GB&lc=eng

into a small horn antenna inside my lab.

He was able to receive me about 7 km away, although the signal was quite
week - it was 20 dB above the noise in a 2.9 Hz bandwidth.

What is odd, is that his transceiver + transverter combination indicates
the signal generator is shifting frequency up/down 100 Hz. This is not slow
drift, but a step change - see waterfall picture, where time is on the
vertical axis, and frequency is on the horizontal. Unfortunately I don't
know what the scale is on the vertical axis - I am trying to find out. The
frequency on the x-axis is not the true frequency, but that shown on an 2 m
amateur transceiver, so the true frequency is more than 10 GHz higher.

The step size on this HP sweeper is 1 kHz, so the 100 Hz up/down shift is
not due to a rotary encoder that might be just on the limit of two
frequencies.  Both the internal oscillator and a GPS locked frequency
standard were used during this test. Going from internal to external
reference caused a 450 Hz step in frequency, but did not change this
up/down 100 Hz behavior. So the problem is certainly not the crystal in a
10 MHz reference oscillator, as two have been tried, one of which was
locked to GPS.

Does anyone have an idea what may cause this behavior?

Does anyone have any ideas on the best way to investigate this, given the
signal is at more than 10 GHz?

What I do *not* have is any other signal generator capable of operation at
10 GHz. The only other sig gen I have is a 30 MHz Stanford Research DS345
function generator.

I don't have a TI counter at the minute, but had an offer of $300 accepted
on eBay for a 5370B a couple of days back, so should have a TI counter
soon. (Yes, I have had an 5370B and SR620 in the past, but for various
reasons no longer have them). The 5370B at $300 was a lot cheaper than an
SR620.

Other equipment I have include

* 22 GHz spectrum analyzer
* 30 MHz signal generator
* 20 GHz VNA

but no other signal generator capable of anywhere near 10 GHz.

I do have a couple of double balanced mixers which have RF and IF inputs
that will take 10 GHz, and an IF output that will go from DC to 4 GHz.
Introducing a REALLY long delay might allow the steps to be seen, as the
frequency at the LO and RF inputs of the mixers will be different. But
that's not really practical, as I'd need an awfully long bit of coax.

The 20 GHz vector network analyzer, which could be pressed into service as
a poor (rich) mans microwave signal source, but I suspect the output of
that is quite dirty, as the output is generated from a step recovery diode.
I have not yet tested it on a spectrum analyzer, but the SA has quite a few
spurious signals, so I'm never exactly confident of the SA. But one thing
to possibly is

* Set sweep generator to 10.368 GHz
* Set sweep generator to 10.378 GHz, so there's a 10 MHz difference.
* Mix the VNA + sweeper down to 10 MHz using a double balanced mixer
* Compare the 10 MHz at the output of a mixer to that of a 10 MHz crystal.
Steps of 100 Hz should then be seen I guess.

Any better suggestions?

Someone had kindly given me a key to change the step size of the signal
generator from 1 kHz to 1 Hz. I've not applied that yet, as it is quite a
complex procedure. But the fact the step size of this is 1 kHz, but it is
shifting up/down 100 Hz, does not make sense. Especially given the unit is
capable of 1 Hz resolution, but HP decided to charge extra for 1 Hz steps.
Luckily this is just a software option.

I did wonder if the signal generator was incapable of output the exact
frequency needed, so it was stepped up/down periodically so it gave the
right number of cycles over a long duration. But again, the fact it can
step 1 Hz with just a software upgrade suggest that's not the case.

The radio ham that noticed this step change in frequency is well used to
listening on 10 GHz, and hearing beacons. So I think its reasonable to
assume that the problem is not his end, but my end.

Any thoughts, which do NOT include purchasing a second expensive 20 GHz
signal generator. They are around $10,000 each, so a bit out of my price
range.

Dave
-------------- next part --------------
A non-text attachment was scrubbed...
Name: C100HZ-SHIFT-DAVE.JPG
Type: image/jpeg
Size: 43961 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20161221/d7e5f0f3/attachment.jpe>