[time-nuts] Re: Which GPSDO for an accurate, low phase noise 10 MHz?

[Ed Marciniak]

A while back I recall seeing an Australian company that offered a sapphire “whispering gallery” (in this context simply meaning the cavity equivalent of a waveguide operating in a higher order mode with lower losses) oscillator with a regenerative divider for direct 10GHz output. I assume they probably took a silver plated copper cavity and shrink fitted it onto the sapphire, or metallized the sapphire and then fitted it into a housing. I have no idea how they’d tune it.

Unfortunately, I’m going to assume it was far more costly than would work for me or my use cases.

For the rest of us, the best reasonable cost ~10GHz solution available is likely a DRO with a sampling phase detector locked to a ~100-200MHz oscillator, which succeeded the 70-80s technique of an VHF ~100MHz SC cut crystal locking an ~1.6 GHz cavity oscillator (with a sampling phase detector), followed by an SRD (Frequency west/CTI/MA-COM brick). It’s either that or tolerate synthesizer phase noise and noise sidebands.

________________________________
From: Bob Camp via time-nuts <time-nuts at lists.febo.com>
Sent: Tuesday, April 29, 2025 7:02:33 AM
To: Discussion of precise time and frequency measurement <time-nuts at lists.febo.com>
Cc: Bob Camp <kb8tq at n1k.org>
Subject: [time-nuts] Re: Which GPSDO for an accurate, low phase noise 10 MHz?

Hi

Any source at 10 MHz will have issues if multiplied directly to 10 GHz.

Your phase noise goes up be 20 log N. You are multiplying by a thousand. The
net result is phase noise that is 60 db worse at 10 GHz than at 10 MHz.

Even your “pretty good” 10811 will be suffering when you do that.

The answer is to go up to 10 GHz in stages. One of about a billion approaches:

1) lock up a quiet 100 MHz crystal oscillator to your 10 MHz source set the PLL
bandwidth to maybe 50 Hz. Outside that bandwidth the quieter 100 MHz source
“rules”.

2) Do the same thing to get to (maybe) 1 GHz. Lock up a DRO there. Same idea
with PLL bandwidth.

3) Now step up to 10 GHz and lock up a third quiet oscillator.

In each case you are making a tradeoff. The phase noise at some offset from carrier
on each of the 100, 1G, 10G sources is below that of the “multiplied up” source you
are locking to. Closer in to carrier that’s not the case.

GPSDO’s are no different.

GPS “as delivered” is a very noisy signal. The typical module adds to this noise.
This means it is not a good source to use for at all “offsets”. What gets confusing
is that we now talk about offsets with units of time rather than frequency. That’s
a pretty common thing as you go below 1 Hz.

Averaged over a second, your GPS is very poor compared to just about any crystal
oscillator you might pick. Compared to your 10811 it might be poor out to 100’s of
seconds.

The same process applies here as to the 10 GHz setup. You very much have to pick
that crossover point between the lousy GPS signal and your great local standard.

Why even do this? Eventually GPS wins. Go out long enough in time and it will out
perform anything you likely have hooked to your radio. Some GPS’s do better than
others. They will “win more” long term. It should be no big surprise that those modules
cost more than the ones that don’t perform as well.

These days, the eBay or typical home brew GPSDO uses a simple GPS module for
cost reasons. That’s not a bad idea as long as the end result meets your needs.
Don’t spend $3,000 if $100 will solve the problem ….

Bob


> On Apr 28, 2025, at 11:17 PM, Bill Katz via time-nuts <time-nuts at lists.febo.com> wrote:
>
> Hello.
> I found this group last year when I was working on my microwave rover
> station (up to 10GHz currently, 24 GHz soon) when I ran into a challenge.
>
> First I bought a small Rubidium standard, hoping that would keep my system
> accurate. (Symmetricom x72 from ebay).  It looked fine with some
> transverters, but with one brand the phase noise was so bad, one couldn't
> make out the primary tone on the spectrum analyzer.  The next step was a
> club member sent me an HP 10811A, which has great phase noise, but showed
> some drift when I fed my counter from the rubidium standard.
>
> Oh, and once I repaired the (apparently somewhat common) temperature fuse
> inside the second 10811A inside the counter, I now had two oscillators that
> almost agree.  Almost.
>
> So I'd like to get something that can run the instruments in my house, and
> allow me to calibrate the OCXO in my rover equipment.  And tweak in the
> various other OCXO standards.
>
> At the high end, there are used HP Z3801As that seem to have excellent
> phase noise, but are somewhat expensive.  At about half the price are used
> Thunderbolt units.  and there are a myriad of other options.  What are the
> pros/cons of some of these units?
>
> Thank you,
> Bill Katz
> KA1TZ
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com
_______________________________________________
time-nuts mailing list -- time-nuts at lists.febo.com
To unsubscribe send an email to time-nuts-leave at lists.febo.com