kb8tq@n1k.org said:

Let's change the discussion a bit.  Assuming I have a GPSDO, home built or
eBay, how can I test it with a limited budget?

There is another possible tangle in here.  What if I don't have a good antenna location?  Is there a simple way to measure/plot the goodness of an antenna?  How does the goodness of a GPSDO depend on the goodness of its antenna?

--
These are my opinions.  I hate spam.

On Sonntag, 20. Februar 2022 09:41:09 CET Hal Murray wrote:

I'd say, the "most limited budget" option is to just monitor the DAC output.
The less it changes, the better. Below that there's the output stability of
the LO. You cannot get better than that (short tau), but it's also more
difficult to assess. The more quantifiable your "measuremements" need to be,
the less budget friendly they become.

Hi Hal,

On 2022-02-20 09:41, Hal Murray wrote:

Well, it is usually hard to measure the absolute offset errors yourself,
but you can get started with stability.

So, let's assume you have a rubidium clock, which is usually available
for reasonable money for most hobbyists.

The phase and frequency will be off naturally, and we can assume that
there is a linear drift in there too. There will be both environmental
effects and random noise effects, but let's assume that we can live with
those limits to start with.

In this context, we can assume that the GPSDO is nominally tied in
frequency and drift to UTC over the GPS link, assuming we do not have a
major design-flaw which would become apparent anyway, so we can then
assign the detected frequency error and linear drift terms to the
rubidium. Similarly we assume the phase error comes from there. By doing
out measurement, then removing the quadratic trend from the measurement,
we end up with the variations of the GPSDO and the variations of the
rubidium. Having a reference trace of the rubidium alone will help to
see what is reasonably additional instability from the GPSDO. You can
view this as phase and frequency variations as well as the many ADEV
variants of your liking. Essentially, this is exactly what we do with
TimeLab in a straight setup.

Once you hit the floor of your rubidium, getting a better reference or
visit a friend with a better reference becomes worth the effort. I'd say
you can come fairly far in this approach. I strongly suggest to log all
the state of the GPSDO into a InfluxDB database and illustrate it in
Grafana. If you can include the rubidium phase measures in that, you
have a lot of useful in-loop and out-of-loop data to ponder over. Toss
in additional environmental sensor readings to help with characterizing
environmental effects.

Pulling in GPS/GNSS receiver state into the Grafana can help to identify
events happening there to deviations.

First you will find a number of bugs, some will be harder than others.
You are bound to learn a number of practicalities of implementing
real-time control systems. Most of which a rubidium would be just fine
for a long time.

So, you can learn a lot this way, for reasonable money. Once you've
covered enough of those corners, improved performance and corner cases,
that's when you need to step up testing further.

Cheers,
Magnus

Hi

If you go back in the archives, there have been a number of folks
who have headed off on a “design my own” ( = not build a kit) GPSDO
projects. Many have fiddled a bit and dropped the idea. The ones that
have kept up and actually built something they know works have all
wound up shopping for Cs standards and other similar gear.

All of the work you do in the frequency or time domain requires a
“better” standard to compare your design to. That is the unfortunate
truth of this field.

So what to do:

If you are looking at 1 to 10 second ADEV, you grab a known better
OCXO. How do you know it’s better? You bought a pile of them and
did cross comparisons. Compare A to B to C, keep the pair that looks
best, toss the other one. Step and repeat until you have sorted the group
of 50 to 100 OCXO’s. The pair that is reading X both do at least that well.
You now have a “home brew” standard for that range.

Once you get past 10 seconds you may well not find OCXO’s doing the job.
A lot depends on what you are after. You could start playing with Rb, Cs,
or a maser. Unless they are new / certified, you likely need more than one
(same process as above) to work out what you have. With the usual eBay
sourcing process, it depends a lot on how lucky you happen to get.
( = you often get one with issues / needing repair …. )

If you went with an Rb, you will start wondering about what’s what out around
a day. That’s not to say at 25 hours it collapses, only that you are not as
sure of the result. A Cs or a maser now would be more useful. Depending on
the maser, the Cs will eventually win out ( is that days / weeks …. ).

In all the cases above, the parts you want to work with likely aren’t easy to find.
You may well spend quite a bit of effort in the shopping process. Is that months
or years? Who knows. Something like a 5065 will be tough to grab and that’s
only the start.

Do you need all those standards? Do you care about all those time ranges?
How many bugs do you want to track down? That’s all very much up to you.
This is something that once you are engaged, most will want to keep up with
and find / fix the problems ….

So we’ve been comparing this to that. What did we use for the comparison?
A home brew single mixer setup is one quick / cheap way to go. With more
time and debugging, you can do a DMTD. It might speed up the OCXO sort
process a bit. With a DMTD, you then get into needing two pretty good references.
That may or may not be a good thing. Something like a TimePod  would get
you going a lot faster.

Phase noise likely needs to be checked along with ADEV and long term stability.
Your mixer setup will get you the first two. A counter with long gate times might
help on the third. If indeed you are after frequency and are looking for the elusive
99.99% ( or whatever number) of samples < some number sort of goal, then
a counter would be involved.

Yes, the DMTD / single mixer has a bunch of gear associated with it. Tossing in
phase noise drags in decisions and gear. It all assumes this is on a single frequency
and you just care about that one output.

So no, none of this is undoable. A number of list members have plenty of this
stuff in their basements. A smaller subset have fully working examples of this
and that in their basements. It’s a rabbit hole. Like most rabbit holes it can just
keep on going and going ….

Bob

That's a very good question, Hal.  If you're talking about stability
testing, GPSDOs are a challenge because they (can) work well over a wide
range of measurement intervals -- you can have the short term stability
of a good OXCO, the mid-term stability of a mid-range Rb, and the long
term stability of a Cesium standard.

When we measure one quantity (the DUT) relative to another (the
reference), you want the reference to be significantly better than the
DUT, or you don't know what you're actually measuring.  A rule of thumb
seems to be that the reference should be 10 dB better than the DUT.  Any
smaller difference increases the reference's contribution to the noise.

So, ignoring for a minute the resolution/noise of the measuring device
(I'll come back to that), to measure the short term stability you need
an OCXO.  Since a good GPSDO doesn't compromise the short term
performance of its internal OCXO, that means if the GPSDO has a pretty
good OCXO, to measure it you need a very good OCXO as reference,
ideally with ADEV an order of magnitude better.

For midterm stability, the GPSDO is at its worst because of the "hump"
where the drift of the OCXO meets the noise of the GPS.  The worst ADEV
in a typical GPSDO occurs around 1000 seconds, and the ADEV might be in
the high parts in 1e12.  The telecom Rbs are in the low 12's to high
13's at 1000 seconds, so they will work reasonably well.  Or a really
good OCXO may still be good enough at that tau.

Long term stability is actually the easiest measurement because you can
use a GPS PPS signal and time interval counter, or even oscilloscope, to
get good results at tau >10K seconds -- it's easy to compare the GPSDO
PPS with the GPS PPS.

At short and mid term, the capability of the measurement device is a
factor.  Few time interval and timestamping counters have a
resolution/noise floor better than high parts in 1e11 at one second,
improving at one order of magnitude for each order of magnitude tau.  So
unless the GPSDO is pretty awful, the time interval counter method is
the limiting factor for tau out to to somewhere between 100 and 1K seconds.

To do better than that, you can buy an expensive test set like the
TimePod, or build a "dual mixer frequency difference" test set or the
"tight PLL" measurement system that's been described here.

So the bottom line is that to fully characterize a GPSDO's stability,
you probably need three references -- OCXO, Rb, Cs -- (unless you have a
maser) and two measurement systems -- one for short term, the other for
mid and long term.

Or... if you build two GPSDO that perform very similarly, you can
measure one against the other and take the square root of two to get the
average performance between them.  That might or might not be absolutely
correct, but as long as the two have similar stability, it will get you
into the ballpark.

What you can't do, though, is rely on any internal measurement, like
plotting the EFC voltage in the correction loop, to get absolute
results.  How much the EFC wiggles is meaningless if the wiggles keep
the frequency where it should be.  It's the output that counts and you
have to measure that against an external source.

And all this is what happens when you become a time-nut. :-)

John

On 2/20/22 3:41 AM, Hal Murray wrote:

Dunno if this will do what you want ...

Get hold of some PLL Fract-N synthesizers, like the ubiquitous ADF4351, so
you can multiply up from two sources. Mix the outputs and look at the
frequency deviation over time.
My tests using this technique can be seen at
http://g4jnt.com/10MHz_Reference_Source_Stability.pdf    and
http://g4jnt.com/ShortTermStabilityLeoBodnarGPSDO.pdf

I used synths I had to hand, but anything that will multiply up your test
and reference signals to a pair of frequencies separated by a few kHz, so
you can look at the beat note using soundcard software.

Andy
www.g4jnt.com

On Sun, 20 Feb 2022 at 17:10, Erik Kaashoek erik@kaashoek.com wrote:

Andy,
Thanks for these references
Using two 5GHz PLL's, one hard referenced to the in-house standard and
the other to the TCXO under test and the outputs of the PLL's send into
a mixer producing a 10kHz output send into a PC for FFT analysis made
the performance of the TCXO very visible.
Can I assume 1 Hz variation of the audio signal is 2e-10 variation on
the 10MHz signal?
My concern with the high resolution FFT is the length of the FFT, does
this not hide fast frequency variations? Or will these variation become
visible through blurring of the peak and as long as the peak is nice an
sharp one can assume the signal is stable over the FFT length?
Erik.
On 20-2-2022 18:38, Andy Talbot wrote:

Yes, 1Hz observed is 1Hz on the multiplied up signal, which at 5GHz is
2E-10 on the reference causing it

The FFT process does indeed smear out faster transitions, so you need to
flip between different  FFT sizes to see fully what is happening.

If you use a prog like SPECTRAN to view the audio beat note, it's quite
feasible to have several instances of that software open at the same time,
all fed from the same soundcard source.  Set each instance to a different
FFT resolution, one wide, one narrow and others as you like, and watch what
happens

Andy
www.g4jnt.com

On Mon, 21 Feb 2022 at 08:38, Erik Kaashoek erik@kaashoek.com wrote: