Newbie GPSDO and OCXO questions

[Squiddaddy]

I have done a lot of reading, but I think I'm still confused on terminology and expectations on frequency references.
I had a bright idea to read up and possibly build a GPSDO for a reference in my shop.
So I bought a uBlox GPS module, and a Bliley OCXO.
Was just playing around with testing the OCXO with an Arduino and a 12 bit DAC to understand adjustability.
Using my Racal Dana 1992 counter, it seemed like I couldn't keep up with the drift.
I also lucked into a TruePosition GPSDO board, so I thought I'd try to understand that more before playing with a DIY project.
I have had the GPSDO and the counter powered up for several hours now.
But the counter keeps creeping downward ever so slightly over time compared to the GPSDO.
Am I seeing the drift of the OCXO in the counter, drift in the GPSDO, or both?
Measuring the GPSDO, I am reading 9,999,999.954 using a 10 second gate time. So, I'm drifting in the millihertz range, but it just keeps dropping. I read the 1992's OCXO's were pretty good for stability.
The counter has not been calibrated yet. But should it continuously drift? I think that's the part I don't understand.
I was thinking about leaving it go all night and see where it is in the morning.
What is a realistic expectation on what I should be seeing?
Sorry for the long post, just trying to learn.
Thanks!

[DaJMasta]

The GPSDO can drift in the short term, but over time its drift should be zero (at least compared against the GPS timing standard source).  That means differences in the short term will probably be something like heating cycles in the OCXO or adjustment steps of the GPSDO or both, but if you look at it over time, the frequencies read should converge to exactly 10MHz (if your counter is referenced to a GPSDO and your source is a GPSDO).  Otherwise, even very good crystals will drift and age, so you should expect to see it fluctuate slightly.  The idea of disciplinin the oscillator is to nullify the aging and longer time period drift associated with crystal oscillators.


I don't know if I entirely understand your setup, but if you measure:
your counter alone measuring the GPSDO output, you see some drift in both from normal adjustment operations and then the drift of the undisciplined counter oscillator
your counter referenced to the GPSDO measuring another source, you see some drift in both from normal adjustment operations and then over time you approach an 'absolute' averaged measurement, given GPS time as a reference

This means that if your counter is running on its own timebase and you are measuring the GPSDO's output as being low (and given that it's warmed up and locked), then your counter's oscillator is actually running too fast, so it reads the better GPSDO output as being slightly slow.

In the moment, you will measure some difference, but when referenced to a disciplined source, that randomness can be eliminated when averaged over time.  It also means you never have a moment where you have an aged OCXO just running at a different frequency and messing with your measurements - you'll be seeing primarily the short term stability spec (10^-8 or 10^-9 ballpark) as the noise in your measurement, instead of the 10^-6 or 10^-7 spec including aging and thermal drift and such of an OXCO that is not disciplined.

Of course this can be better with a better oscillator, but the disciplined reference feeding the counter should give you at least a couple more digits of base accuracy than using the internal reference in whatever shape it's in, especially given the age of the counter (it will have drifted from its initial calibration, definitely).

[texaspyro]

An OCXO that has been powered down for a long time can take weeks to "stop" drifting. 

[thermistor-guy]

...
I have had the GPSDO and the counter powered up for several hours now.
But the counter keeps creeping downward ever so slightly over time compared to the GPSDO.
Am I seeing the drift of the OCXO in the counter, drift in the GPSDO, or both?
Measuring the GPSDO, I am reading 9,999,999.954 using a 10 second gate time. So, I'm drifting in the millihertz range, but it just keeps dropping. I read the 1992's OCXO's were pretty good for stability.
The counter has not been calibrated yet. But should it continuously drift? I think that's the part I don't understand.
...

Leave the GPSDO and counter powered on. Check after 24 and 48 hours. It's normal for a frequency counter's internal OCXO to take 24 hours or more to stabilize, especially if it has been powered off for some time (days or weeks).

Your GPSDO may take less time to stabilize. My GPS-disciplined Rubidium reference usually takes 12 hours to lock.

[Squiddaddy]

Thanks for the comments guys!
I left everything powered up all night. It's been 16 hours so far. Counter now shows 9,999,999.919.
I guess I have to be impressed that I am still within 1Hz of the two oscillators.
So, to understand more terminology, would this test be showing me the 'retrace' operations of the OCXO's, since they've been off for so long?
I also noticed that the GPSDO goes into holdover, and I see the frequency drift up about 3-5 mHz or so, and when it locks back up, it goes back down. So the GPS appears to be adjusting fine.

texaspyro: You posted a little while back that your Lady Heather program now supports the TruePosition. Is that available? Maybe that would help me graphically watch what's going on? Of course, that would be another learning curve. Right now I'm just watching Putty and looking at lock status. LOL

[texaspyro]

Lady Heather v6.0 Beta .exe for windows is here:

https://www.eevblog.com/forum/metrology/lady-heather-v6-beta-for-windows-exe/msg1434005/#msg1434005

[Squiddaddy]

Not sure if I'm chasing my tail here or not, but here's where I am.
The GPSDO and counter have both been powered up for about 2 weeks now.
I had the setup working to where the counter was showing 10,000,000.007 for about 3 days straight.
Then, when I got home from work to check on things again, the counter suddenly started drifting down and drifted at least 15 counts(.015Hz) Never went back up.
My counter hasn't been calibrated in a long time, and still had the sticker over it until I adjusted it this week. That thing is very touchy! If you move it ever so slightly, it drifts up and down until it settles out, and that takes an hour or so.
I feel like I'm watching paint dry trying to see what this setup is doing.
All I wanted to do was compare the counter to the GPSDO and see if I needed to calibrate the counter.
But it seems like I keep chasing drift. Does this sound like the OCXO in the counter is having issues?
Since I'm watching millihertz, am I expecting too much?
I don't want to make the GPSDO the reference to the counter and signal generator if it's not stable.
I wanted to calibrate their OCXO's first to see how stable they would be on their own, compared to the GPSDO.
Maybe I shouldn't have even touched the counter calibration yet.

[edpalmer42]

Not sure if I'm chasing my tail here or not, but here's where I am.
The GPSDO and counter have both been powered up for about 2 weeks now.
I had the setup working to where the counter was showing 10,000,000.007 for about 3 days straight.
Then, when I got home from work to check on things again, the counter suddenly started drifting down and drifted at least 15 counts(.015Hz) Never went back up.
My counter hasn't been calibrated in a long time, and still had the sticker over it until I adjusted it this week. That thing is very touchy! If you move it ever so slightly, it drifts up and down until it settles out, and that takes an hour or so.
I feel like I'm watching paint dry trying to see what this setup is doing.
All I wanted to do was compare the counter to the GPSDO and see if I needed to calibrate the counter.
But it seems like I keep chasing drift. Does this sound like the OCXO in the counter is having issues?
Since I'm watching millihertz, am I expecting too much?
I don't want to make the GPSDO the reference to the counter and signal generator if it's not stable.
I wanted to calibrate their OCXO's first to see how stable they would be on their own, compared to the GPSDO.
Maybe I shouldn't have even touched the counter calibration yet.

Which OCXO does your 1992 have?  I have a 1992 with Option 04E.  It seemed to move around more than I thought it should.  Even after being on standby for weeks, it drifted after I turned it on.  This didn't happen when I used an external reference so that's what I started doing.

Keep in mind that any OCXO, but particularly a high-quality one, can be a bit of a diva.  That's partially because we tend to expect a lot from them.  They can jump up or down in frequency, drift down for weeks and then change direction and drift up or vice-versa.  They might improve after weeks or months of continuous operation - or not.  Power them down for an hour and your aging can go back to what it was months ago.  IOW, keep them powered and use a UPS.  By the way, I found that my 1992 didn't like a typical, cheap UPS.  When power fails, a cheap UPS flips a relay to start powering from the batteries.  The brief power interruption was enough to reset my 1992 and wreck my data run!  I don't think there's anything wrong with the counter, it just doesn't like it when the power blinks.

Compare your results to the specs to decide if you're expecting too much.  If you've got the 04E oscillator, it's rated for 5e-10 aging per day and 7e-9 over 0 to 50 C (1.4 e-10 per degree C).  That works out to .005 Hz per day for aging and .0014 Hz for every degree C.  Normally, you wouldn't notice something like that.  It's only when you've got a GPSDO/Rubidium/Cesium standard that you can see it.

As you say, making these measurements is like watching paint dry.  I quickly decided that I wanted to let my computer do the dirty work.  The 1992 only has a GPIB connector (if you've got that option - I do) so I set up my lab for GPIB.  This makes it a lot easier to measure drift, jitter and Allan Deviation (if you're a Time-Nut) and, of course, is useful for lots of other equipment.

Ed

[Squiddaddy]

Thanks Ed!
I do have the 1992/02M with the 4E OCXO, and GPIB. Might need to flip the jumper to get it out of MATE/CIL mode.
I think I was getting confused as to what to expect out of the counter and the GPSDO, so I can get close to calibration.
Since you mention temperature, my equipment is in the garage, so it has wide temp swings, since I don't leave the heat on.
Since I'm running it in 10 second gate mode and watching millihertz, I should just get it to where it swings in the +- 10mHz range and call it good. I eventually want to possibly use the GPSDO as the reference since it has 2 outputs, one for the 1992 and one for the HP8656b signal generator. I was referencing the HP off of the Racal.
I was looking at how close the OCXO's were in both, because the HP didn't agree with the Racal either.
Since the GPSDO is an Ebay refurb, I need to see how good it is before I make it the standard.
Sounds like I need to get the workbench set up better, in a room that doesn't change temp as much.
I don't want to buy a Rubidium if I don't have to, because they are getting expensive again.
Just want to keep my counter and signal gen in good calibration.
Thanks for the tip on the UPS. I better get a dual conversion one.(I think that's what it's called.)
Ted

[edpalmer42]

I've never tried using a GPSDO as a reference.  Depending on the quality of the OCXO, it might be okay.  I use a Rb as my house reference and occasionally compare it to a good GPSDO (HP Z3801A) that has a DOCXO and runs 24/7.

Temperature won't be a big issue for the GPSDO since the system automatically compensates for changes in the OCXO frequency.  However, it wouldn't hurt to protect it from cold drafts or blasts of heat.  Watch out for sunbeams - they're surprisingly warm.  Don't go overboard on the protection - you don't want to cook the little beast!  A simple cardboard box with 3 or 4 inches of space on each side of the GPSDO is all you'd need.  Come to think of it, that might improve the GPSDO performance, but you'd need another reference to compare it to before you'd see the difference.

Check the specs on the 1992 and the 8656B.  The external reference input on a lot of equipment is high impedance.  This lets you daisy-chain a single line to multiple pieces of equipment rather than having one cable run for each piece.  A terminator at the end of the run would be a good idea - just like an old 10Base2 (Thinnet) Ethernet run if you know what that was.  If one piece of equipment happens to have a 50 ohm reference input, that's an obvious candidate for the end of the run.

I've seen those UPS called dual conversion or online.

Ed

[bitseeker]

Posting to follow along. One of my many pending projects is to do something similar with a uBlox module that I have and a 1992-04E. My goal is just to have the 1992 be a basic reference that I manually tune up from time to time with the GPS and a scope. I don't yet know if my OCXO drifts a lot, so I'll be reading with interest what your experiences are like.

[Squiddaddy]

I bought a uBlox and a Bliley OCXO to build a GPSDO also.
But I also lucked into a Telco TruePosition board for a good price, so I want to get that working first.
And make sure my 1992 is on the mark.
Then I can play with the DIY build and see how close I can get.
But I am at the very beginnings also.
Do a search for Lars' GPSDO. I think I'm going to build that one, after I get all of this resolved with my equipment. Or maybe the James Miller one.

[Stray Electron]

Which OCXO does your 1992 have?  I have a 1992 with Option 04E.  It seemed to move around more than I thought it should.  Even after being on standby for weeks, it drifted after I turned it on.  This didn't happen when I used an external reference so that's what I started doing.

  I'm also a newbie at GPSDOs but my understanding is that they're MUCH more accurate than any realistically available frequency counter so I think it's going to be a lost cause trying to checking one with a FC.  My HP 5316B (option 004 OCXO) is spec'd at 5 x 10 -8 IIRC and a decent GPSDO is better than 1 x 10 -12 and the GPS system is capable of about 1 x 10 -14.   Or so I'm told my my friends that are T-V nuts.

   That said, I also have an RD 1992 with the same option and the display on it is a lot more jittery than my HP 5316B or HP 5343A or even my Leader LDC 8?? FC with the 0.03 PPM source.  The osc output of the 1992 is rock solid but when I use the same input on both the 1992 and any of the others, the last 2 or 3 digits on the 1992 flicker all over on it and on the others the last digit only changes by one count.  I like the 1992 but there seems to be something unstable about the triggering and/or counting circuits in it.  And no I haven't really checked it other than that.  I use the HP 5316B for most things but when I need a GOOD FC I use the HP 5343, the sensitivity of that thing is UNREAL! But it's a beast to move around.

[edpalmer42]

Which OCXO does your 1992 have?  I have a 1992 with Option 04E.  It seemed to move around more than I thought it should.  Even after being on standby for weeks, it drifted after I turned it on.  This didn't happen when I used an external reference so that's what I started doing.

  I'm also a newbie at GPSDOs but my understanding is that they're MUCH more accurate than any realistically available frequency counter so I think it's going to be a lost cause trying to checking one with a FC.  My HP 5316B (option 004 OCXO) is spec'd at 5 x 10 -8 IIRC and a decent GPSDO is better than 1 x 10 -12 and the GPS system is capable of about 1 x 10 -14.   Or so I'm told my my friends that are T-V nuts.

Yes, the frequency of a GPSDO is probably closer than your counter and, of course, the GPSDO doesn't drift.  But, you can see phase jumps when satellites are added or removed from the calculation.  Depending on how good your survey was and whether your antenna has good visibility or not, the short-term noise and jitter might be large enough to affect your measurements.  The quality of the OCXO that's inside the GPSDO determines the short-term performance of the GPSDO.  That short-term performance will directly determine the quality of your counter's measurements.  The only time that the disciplining is a factor is if you make a series of measurements over a long period of time.  It ensures that there was no drift during the entire length of your data run.

As I said, I've never used my GPSDO as a reference.  Some caution is warranted.

   That said, I also have an RD 1992 with the same option and the display on it is a lot more jittery than my HP 5316B or HP 5343A or even my Leader LDC 8?? FC with the 0.03 PPM source.  The osc output of the 1992 is rock solid but when I use the same input on both the 1992 and any of the others, the last 2 or 3 digits on the 1992 flicker all over on it and on the others the last digit only changes by one count.  I like the 1992 but there seems to be something unstable about the triggering and/or counting circuits in it.  And no I haven't really checked it other than that.  I use the HP 5316B for most things but when I need a GOOD FC I use the HP 5343, the sensitivity of that thing is UNREAL! But it's a beast to move around.

I'm not surprised that the 1992 is jumpy compared to the 5316B.  The 1992 has a resolution of 1 ns.  The 5316B has a resolution of 100 ns.  The 1992 can see things that the 5316B is blind to.  Since the 5343A is a frequency-only meter, it's hard to compare to the other two.  I'm assuming that you made sure that all three used the same gate time.  A short gate time will always seem jumpy compared to a measurement that's averaged over a long gate time.

Ed

[Squiddaddy]

I know that once I found out how to get the 1992 into 10 second gate mode, I see so much more detail in the frequency. When I first started testing, watching the display go from 9,999,999 to 10,000,000 didn't tell much about what was going on. Even overflow didn't show much more but 1 digit. But now being able to see .001 Hz really opens up that counter.

Ed, you were right about the temperature. I checked my experiment yesterday when I left for work and when I came home. Temp was consistent, and the display showed 10,000,000.006 both times. After I heated up the garage and changed temp by at least 20 degrees, the display dropped about 15 counts to 9,999,999.089. Probably affected both units.

I am going to set up in another room with consistent temperature. The garage will have me chasing consistency. Time to set up a new lab.

[jcbottorff]

If you're playing with new GPSDO design you might look into using a 10Mhz OXCO with a one of the recent uBlox GPS modules. I've been really impressed with the features and sensitivity of the uBlox module. If you can get the correct model, it can spit out a nanoseconds error value (in the serial data stream) between the internal PPS and an input pulse you feed it. It looked like a lot of the ebay uBlox module had the right chip, but the module PCB didn't export the extra pins. 

A software command can also set the PPS rate to a MUCH higher number (use the uBlox configuration software) and you can pick something that's still a multiple of the internal clock like 1 Mhz (think it was 48Mhz, so not a nice multiple of 10Mhz). In theory, you could then use a 10x divider on the 10 Mhz OXCO and a simple PLL chip with the 1 Mhz GPS reference and use the PLL charge pump for OXCO tuning, no MCU required (although a mcu design does give some nice data output) This PLL method should lock quite quickly.

A year ago I fiddled with a Ebay OXCO+uBlox+AtTiny85=GPSDO, all on a temporary breadboard. It took quite a while (overnight) to become stable, but for $20 of parts I was pretty impressed you could to get better than a ppb pretty easily. Precise 10 Mhz/OXCO PWM voltage control on a breadboard was also fragile to movement/temperature. I suppose if you set the PPS to 100Hz, a design with a mcu will have a lot more edges to compare, so will coarse lock a lot faster. I also saw the idea of using a tri-state buffer to essentially do phase error integration, and the mcu reads the integrated value with it's ADC. As I remember this took an xor gate (pps and clock mismatch signal) and a tristate buffer to toggle current through a resistor into an integrator capacitor. Normally the mcu can't tell it's out of phase until half cycle mismatches, so for smaller phase errors it takes many seconds to notice the error and correct the VCO.

I eventually bought an ebay assembled Trimble derived GPSDO for $150, which outputs both sine and square waves. For RF work, the sine wave output make less interference because there are no sharp edge odd harmonics, and my signal/generator frequency counter doesn't seem to care if the refclock is sine or square.

DIY GPSDO's are great fun to play with, although for real use it's was way easier to buy a finished one on ebay.

[edpalmer42]

I know that once I found out how to get the 1992 into 10 second gate mode, I see so much more detail in the frequency. When I first started testing, watching the display go from 9,999,999 to 10,000,000 didn't tell much about what was going on. Even overflow didn't show much more but 1 digit. But now being able to see .001 Hz really opens up that counter.

I remember when I stumbled onto that ONE note in the manual that mentioned that there was a 10 second gate time (really?) and how to enable it.  Geez guys, way to hide the features!   

Ed, you were right about the temperature. I checked my experiment yesterday when I left for work and when I came home. Temp was consistent, and the display showed 10,000,000.006 both times. After I heated up the garage and changed temp by at least 20 degrees, the display dropped about 15 counts to 9,999,999.089. Probably affected both units.

I am going to set up in another room with consistent temperature. The garage will have me chasing consistency. Time to set up a new lab.

Well, when the temperature changes by 20 C, I think any piece of precision equipment is going to notice.  Your drop was in the range of half what the spec was so that's actually pretty good.  I found that when I used an external reference there was no change in reading from turn-on to full warm up.  This meant that the change was totally due to the oscillator rather than anything else in the counter.

Ed

[Squiddaddy]

Ed, I didnt notice you were canadian. It went about 10c, from 59f to about 78f, 15c to 26c.
But still a big jump.
I am working on setting up in another room that i can keep around 20 to 22c.
Ted