in search of a GPSDO

[ivo3510]

Hello, everyone, this is my first post here.
I am in a process of building a GPSDO for a home hobbyis tasks, and I've both NEC board without OCXO.
I've made a linear PSU, with all the filtering and stuff, soldered Morion MV-180 OCXO i've had, closed the board in a ordinary plastic box.
I've connected an old Trimble Pathfinder antenna, and awaited a day and a bit, so the GPSDO would stabilize.
Measurement is done with homemade timebase, I used the same model OCXO, and HP 53132a counter.
Please, check, and advise if this is good enough, or I should try and change something, if this setup is capable of reaching more stable result?
I want to use it for a timebase of a counter, like that HP I have, and for a DDS-generator, like these with LCD, awailable for something like 120 or so USD - FY6900 or similar one.
Thanks!

[edpalmer42]

Your ADEV graph looks a bit too good to be true.  The 53132A has a resolution of 150 ps., i.e. 1.5e-10.  Typically, the ADEV at 1 sec. is about the same as the resolution, but yours is showing 1e-11 !  This likely means that there's some averaging going on.  You don't want to do any averaging because it makes things look better than they really are.  How did you collect this data?

I don't have a 53132A, I usually use a Fluke PM6681.  I use square waves, never sine waves, and run the two oscillators into the A and B channels.  Then I measure the time delay from A to B and make *ONE* and only one measurement every second to ensure that there's no averaging going on.

The graph is showing the typical disciplining hump at about 70 sec., which is good, but you really need to let it run longer to see how well the oscillator is being disciplined.  The end of an ADEV graph flops around quite dramatically so you have to continue the data collection long enough for it to settle down.

GPS by itself typically measures about 1e-10 @ 100 sec. and continues down and to the right at a slope of -1, i.e. 1e-11@ 1000 sec., etc.  If your oscillator is being properly disciplined, your trace will run parallel to the GPS line.  Some units run on one side of the line, some on the other - it just depends on the design of the disciplining loop.  Eventually, your graph will deviate from the line as the noise in your system starts to take over.

In the Lady Heather screen, I see that the elevation mask is set for 5 degrees.  That seems really low.  There's a way to measure your incoming signals vs. elevation to optimize that setting, but I haven't used LH for a long time.  Maybe someone else can comment on that.  If you optimize that setting, you might slightly improve the output of your GPSDO.

If you're using this as a frequency reference, you're probably fine.  For normal frequency measurements or timebase duties you almost don't need a GPSDO.  A bare GPS receiver with a 10 MHz output is enough for many tasks.  All the items I've mentioned are just 'fine tuning'.  It looks like you've got the basics covered.

Have fun!

Ed

[Solder_Junkie]

Like edpalmer42, I'm not sure about the ADEV plot. My GPSDO use is also as a frequency reference for a counter/signal generator and to stabilise UHF amateur radio equipment.

I have a rubidium reference, a TinyPFA (phase frequency analyser) and also use TimeLab.

While using a reference oscillator is a good idea in my opinion, as opposed to "marking your own homework" with a plot of the correction data from an Arduino controller, it relies on the stability of the reference... and that is hard to determine. However, my measurements are repeatable and give consistent results within the limits of the rubidium stability.

By way of a comparison, see the attached plot comparing a Lars Walenius Arduino controlled GPSDO vs a Leo Bodnar GPSDO, the reference in both cases was a rubidium. The frequency offset has been zeroed in TimeLab as the rubidium shows a slight frequency offset against GPS, it has no discernible drift. The Lars unit uses an NEC high stability ovened oscillator and a Ublox M8T (timing) GPS module, the Leo Bodnar uses a (navigation) GPS module and a TCXO, both are used on an outdoor "mushroom" antenna.

SJ

[ivo3510]

Thanks, Ed!
I've simply measured freq. to channel 1, with internal timebase.
Should I connect ref out of the counter to ch1, and the GPSDO to ch2, and measure time interval 1 to t.i. 2, gated to 1 sec?
It seems more like what you've described as resolution, below are the plots.
The output of the GPSDO is square, it has differential buffer, connected to one transformer like these, used in cheap routers WAN inputs.
Actually, the GPSDO is with it's default settings.
Before i posted my firs measurements, I've tried and have set it to auto elevation mask - it calculates between 15 and 25 degrees, depending on what antenna I connect to it.
I have left it to def., haven't toched the antenna, and measured again to havethe plots below, in order to compare similar environments.
I will try and set it to auto again later, since the antenna is on the roof of my garage, about 3 meters above the ground, and there are
two houses near it - one on SE, and one on NW position, as per LH screen.
I see something strange in this plot - the Sun, and all sattelites are hopping several degrees, maybe these positions are wronglly calculated by the reciever from reflected signals I let sneak in, because of the low elevation mask?
 
BR

Ivo

[ivo3510]

Thank you, SJ!
It makes sence, I and not sure I did mu measurements correct.
Could you please describe how I need to connect the equipment, to measure correct ADEV plots?
And the difference between these GPSDO's is really big.
I have looked at the project from Lars, it looks really good.
I'll try and construct it, maybe it'll be better than this NEC tower ref.

Have a nice day!

Ivo

[Solder_Junkie]

Could you please describe how I need to connect the equipment, to measure correct ADEV plots?
And the difference between these GPSDO's is really big.
I have looked at the project from Lars, it looks really good.
I'll try and construct it, maybe it'll be better than this NEC tower ref.

Ivo

Ivo, I’m not sure of the best method of plotting ADEV. There are lots of pages describing Allan Deviation in detail, but (and it’s a big but), I’m not sure it is relevant to what you and I are using our GPSDOs for.

Allan Deviation seems to be a favourite plot of builders who have no practical use for the GPSDO they built… plots over hundreds, or thousands, of seconds should show very accurate long term stability, what use is that if in the short term you have drift and/or noise/jitter?

You will note the “poor” performance of my Leo Bodnar unit, they are described as low jitter units. The advantage of the cigarette packet sized LB unit is that it is stable within a minute or two, it can be set to any frequency from 400 Hz to 810 MHz and can be powered from USB. The frequency excursions shown in my graph are roughly +/- 4 parts in 10^10, which is +/-4 Hz in 10 GHz… that is not enough to worry about given the low cost and convenience of the unit. My LB GPSDO is in use every day, that is not the case with my “much better” Lars unit.

SJ

[ivo3510]

Agree, usability is the main thing in my mind too, that's why I've put this simple one into a box, and made it simple as I can.
Well, for timebase for a counter, DDS, and RF gen. I will use to repair, and align FM tuners, sometimes 3-5 hundred MHz RF modules, and so on relatively LF stuff, it will do, as I understand.
I'll use HP5385A for these tasks, I was wandering if the GPSDO works as intended, and maybe do some minor alignments, if they're needed.
Thank you for your time, and have a nice evening!

Ivo

[edpalmer42]

Thanks, Ed!
I've simply measured freq. to channel 1, with internal timebase.
Should I connect ref out of the counter to ch1, and the GPSDO to ch2, and measure time interval 1 to t.i. 2, gated to 1 sec?
It seems more like what you've described as resolution, below are the plots.

Yes, that should give better results.  Is the ref out a square wave or a sine wave?  Remember that square waves are better, but doing a good sine to square conversion is not trivial when you're at these levels so don't worry about it for now.  I see that your ADEV results at 1 sec. are now 3e-10 instead of the 1e-10 that I would expect to see.  You might have some jitter or noise in the system.  Check the trigger level of both inputs on your counter and compare that to the waveforms on a scope.  Make sure that the trigger level is at a nice quiet, fast rise time part of the signal.  And why did the disciplining hump disappear?  Don't worry, it's still there, it's just hidden under the noise.  You should run the data collection longer to better confirm the disciplining.  There's something really satisfying about seeing the graph running straight down into the e-13 region! 

I see something strange in this plot - the Sun, and all sattelites are hopping several degrees, maybe these positions are wronglly calculated by the reciever from reflected signals I let sneak in, because of the low elevation mask?

Are they actually hopping, or does that just mean that LH only maps their position every few minutes?  LH has many things that it can show in that top, right area and graph across the bottom.  You should explore those options.

Solder_Junkie correctly stated that ADEV is mostly overkill for just a frequency reference.  It's the Time-Nuts version of doing a 0-60 mph or quarter-mile run in your car.  Does anyone really need a car that can go from 0-60 in 3 seconds?  No!  Do we want one?  Hell, yes!    It also helps confirm that the sources are operating at a much higher level of precision/accuracy than we need so we can rely on them for all our needs.

Ed

[ivo3510]

Thank you, Ed!
I beliewe the GPSDO has it's time stabilizing it's output, as LH now states "holdover perf good"- I'll leave it like it is.
Output is just squarewave, as it is from it's output, I just get it from LAN transformer, so I can have several outputs decoupled from each other.
Now the plot shows somethig more realistic, compared to the firstone, I've closed it in a plastic box, no termal issues for the last 3 days, it's stable enough.
I'll add some schmidt-triggers to the PPS output, and it's done.
Thank you for your help, I see I did it well enough to use it as it is.

BR

Ivo