Have GPSDO - Now What?

[NT0Z]

Hi, gang,

Jr. metrology questions...

I am a ham experimenter with an age old conundrum: Now that I have a GPSDO (or some other high-order standard), how do I know whether it's accurate and working normally? If I connect the GPSDO output to a counter, the counter will have to be at least as accurate as the GPSDO to determine anything useful, right?

And what if I discipline the counter and then use the same counter to measure the GPSDO output? Is that a thing?

I also have a "new" (old stock) reference oscillator from the former AADE (ham radio cottage industry). It's never been powered up until a week ago and is about 6-7 years old. The builder says that the TCXO he used, which was calibrated before delivery to his NIST-traceable gear (and one in which he has used hundreds of times in his line of frequency displays), historically drifts/ages about 6 Hz over six years at 10 MHz. If I compare the output of the GPSDO to that of the TCXO, they outta be quite close -- but without a known comparison I will have to "assume" that the GPSDO, when locked, is the one that's on the money?

Attention mega-metrologists: I know that the TCXO is a low-cost trinket, so remain calm.

If I can receive it well, WWV is always there. I read years ago about a way to graphically calibrate signal sources by using a PC-based audio spectrum analyzer (SPECTRAN?) to plot WWV vs the unknown source on the waterfall while using WWV's 1000-Hz tones to guarantee a certain precision when tuning the signal. This should get me to about 0.5 Hz at 10 MHz, which is probably fine for anything I anticipate doing in the near future.

But it's an interesting conundrum! A chicken-egg metrology thing!

Also: I have about 3,000 feet of high-quality RG-6 coax. The kind with a solid copper center conductor and not copper-clad steel. I want to use this to make some BNC lab jumpers until I make some up "proper" cables. I assume this will work fine for distributing the 10-MHz ref signal from the GPSDO, but I suspect that many purists would liken that to genocide... Any real issues with using the 75-ohm stuff to get started?

More: I am also trying to learn the proper way to measure the output power of the GPSDO and the ref oscillator. I may need to terminate my scope with a 50-ohm load to get proper V readings. And I haven't yet had the courage to connect either to my spectrum analyzer, as I need to build/acquire a tap or an attenuator.

I have a 1975 HP 5328A counter. It has an internal TCXO/OCXO that seems to be nearly dead-on for a nearly 50-year-old device, as it counts the GPSDO output to many decimal places (undisciplined it thinks the GPSDO output is about 85 Hz low). It doesn't have either of its prescaler options, so right now it tops out at 100 MHz.

I don't yet know the impedance of the counter's 10-MHz external reference port, nor the recommended signal level. Gotta find out before connecting GPSDO?

I don't want to fall down this rabbit hole too far. I have no need for pico-hertz precision I just want to learn a bit about this stuff and how to use it to make my test bench more accurate in very reasonable ways. I had fun (?) learning the basics of using Lady Heather. At present I build stuff mostly for HF, with a bit of 6 and 2 meters thrown in for good measure. If my counter is off by 1 Hz at 30 MHz, no prob. I just wanna know.

As always, thanks,

--Kirk, NT0Z

My book, "Stealth Amateur Radio," is now available from www.stealthamateur.com and on the Amazon Kindle (soon)

[Circlotron]

And what if I discipline the counter and then use the same counter to measure the GPSDO output? Is that a thing?

Doing that will always falsely show the counter as being dead accurate. You are comparing a signal with itself and so it will naturally be identical in frequency, even if the actual frequency varies.

[NT0Z]

So, the EASY way to ensure accuracy.

It figures that this won't work, because it's the way I was hoping it would work...

Thanks,

--Kirk, NT0Z

[David Hess]

If I can receive it well, WWV is always there. I read years ago about a way to graphically calibrate signal sources by using a PC-based audio spectrum analyzer (SPECTRAN?) to plot WWV vs the unknown source on the waterfall while using WWV's 1000-Hz tones to guarantee a certain precision when tuning the signal. This should get me to about 0.5 Hz at 10 MHz, which is probably fine for anything I anticipate doing in the near future.

If you are within range for direct reception of WWV that might work, but like most if you pick it up through bounce off of the ionosphere, then movement of the ionosphere causes Doppler shift.

[NT0Z]

David,

Thanks for that. Just read an interesting paper or two about that based on your info. Looks like a typical Doppler variation for WWV  is on the order of 0.2 Hz, with a practical (occasional) max of 2 Hz (10 MHz). That's probably within my casual zone of tolerance.

Regards,

Kirk, NT0Z

[rfclown]

I guy with one watch always knows what time it is. A guy with two watches is never quite sure.

I'm not a metrology guy; but if you don't have a known reference, you can't assess anything. It is fairly safe to assume that if you have a GPSDO that has a lock indicator, and it says that it's locked, then it is better than the 0.5 Hz you are looking for... unless it's broken. I started with a eBay purchase of 2 rubidiums. They were both close... but what if they are both way off in the same direction? Didn't really matter to me, but when I got a GPSDO years later I had a third thing to compare with (but none that I knew for sure... like you). Then I got a second GPSDO. When I saw that the GPSDOs didn't slip a cycle over time compared to each other, I knew I had the long term stability of the cesiums in the sky. This isn't close to metrology stuff, just having some assurance that I'm pretty close (I won't be late for work).

[Conrad Hoffman]

You can do a phase comparison with WWVB, which might be better. Easiest thing is to get another GPSDO. Or, do like I do, "Heck, it's a GPSDO so how far off could it reasonably be? Plenty good enough for me."

[xrunner]

I guy with one watch always knows what time it is. A guy with two watches is never quite sure.

That's why you need three! 

[NT0Z]

You can do a phase comparison with WWVB, which might be better. Easiest thing is to get another GPSDO. Or, do like I do, "Heck, it's a GPSDO so how far off could it reasonably be? Plenty good enough for me."


This, from the author of the Mini Metrology Lab series?  Scandalous!

Thanks for that, BTW. It brings back good memories.

Regards,

Kirk, NT0Z

[Conrad Hoffman]

WWVB carrier is 1 in 10^14 so if you phase compare long enough...
Just watch out for the periodic phase shift they introduced and only compare on the same phase.
Obviously, use a TRF receiver, not a super-het!

[GigaJoe]

https://www.nist.gov/pml/time-and-frequency-division/time-distribution/radio-station-wwvb
The frequency uncertainty of the WWVB signal as transmitted is less than 1 part in 10+12. If the path delay is removed, WWVB can provide UTC with an uncertainty of about 100 microseconds. The variations in path delay are minor compared to those of WWV and WWVH. When proper receiving and averaging techniques are used, the uncertainty of the received signal should be nearly as small as the uncertainty of the transmitted signal.

assuming need high cost receiver and antenna to reach ideal -12 as they stated, in my understanding even salvaged gps board will do a similar performance  ; consider of your counter resolution.


 

[Conrad Hoffman]

Hmm... You're right. The first reference I was was -14, but NIST is pretty specific about -12.

[GigaJoe]

basically type of cable doesn't matter,  im not certain it 10M sineway or 10M meander on output  ,  it doesn't matter, an impedance doesn't matter as far as reflection didn't kill the signal  )

due to: it all digital, even if the wayform of the signal distorted it has to be periodic with as low jitter as possible,  a form of the signal really indifference.
HP input usually 5Vp-p, i connect directly from gps

same for scope - hook it directly you will see a signal ...

the gps antenna i use this one with very good result
https://www.aliexpress.com/item/1005001566415350.html
feeder RG6 regardless of 50 ohm suppose to be, but it indifferent,  the matter is - cable loss on 1.6Ghz (gps signal) less loss - better signal

as soon as it lock it good for direct measure for your HP meter, need approx 1 week of running to minimize jitter and precision (close to -12 -13)
consider an additional thermal isolation of OCXO and minimal airflow on OCXO to minimize signal jitter.

[NT0Z]

Thanks for the info. I appreciate it.

I have tons of 75-ohm coax on hand, so I will start with that to distribute the 10-MHz ref signal and see how it goes.

I have a few devices that have external ref inputs, so I'm gonna start with a CATV splitter and an amplifier, probably a single high-frequency op-amp or a MMIC (have a bunch in the drawer) while I scrounge a video distribution amp or a BG7-style unit.

My outdoor antenna is set to arrive in two days, so that should help. It's interesting to note the flat lines on Lady Heather as my GPSDO unlocks because it can't see any satellites. As soon as even a single sat pops up with even a minimal C/N indication, the system locks in a minute or so.

I'm guessing that the accuracy, even without a week's run at jitter, etc, is quite sufficient for my needs. I just kept running up against the age-old question of "how do you know whether your GPSDO is accurate, working as intended," etc.

I will find some ESR friendly foam to insulate the TCXO on my minimalist, secondary freq reference. I figured that might help it to drift even less. I keep it constantly powered.

Regards,

Kirk, NT0Z

[CatalinaWOW]

It is interesting to see how good enough has changed over the years.

There has always been a crowd that feels that good enough is the intersection of the bank account with the state of the art.

But remember when consumer radios had to be tuned?  A fifteen kilohertz AM radio band couldn't be assured from the factory so radios had continuous knobs and fine tuning controls. 

Then good crystal oscillators got cheap along with digital counters and the like and consumer products got direct tuning.  Good enough was now something like two orders of magnitude better.

Then that capability, plus the demand for bandwidth lead to re-allocation of frequency bands and the old good enough wasn't even legal to use. 

There are still relatively few real world amateur needs for time accuracy at the one part in a million millions level.  OP needs to compare his use case needs with the GPSDO performance.  I am betting that it is good enough, by orders of magnitude.

[NT0Z]

As the OP, I agree.

Much of my initial inquiry centers around "now that Lady Heather says my GPSDO is locked, how do I know if it's really working, and at what accuracy?"

My first thought of using the GPSDO ref output to discipline my freq counter -- and then using that counter to check up on the GPSDO -- was shot down. Darn

My next thought, using an audio spectrum analyzer to compare WWV to the GPSDO output (with the audio spectrum analyzer used to "lock" in on WWV's 1000-Hz audio tones to maximize accuracy), fell to Doppler shift and phase distortion (although it's still likely that the process will get me to 0.5 to 2-Hz accuracy at 10 MHz).

I have an AADE TCXO frequency standard that is likely to be within 10 Hz at 10 MHz, possibly better, but I need to use the GPSDO, which is probably working fine, to calibrate the standard.   It was calibrated to a NIST-traceable Rubidium standard when manufactured, but that was 7-8 years ago and the AADE unit is quite simple.

The seat-of-the-pants consensus -- and one that I will probably go with -- is to, unless otherwise observed, assume that the GPSDO is locked when Lady Heather says it is, that it is as accurate as GPSDOs are, and just go with it. In that condition it's likely to be the highest precision tool in the box. It's just that it's ultimately Schroedinger's GPSDO: It's accurate and its not (until it's "observed" with a tool that's known to be calibrated and capable of characterizing the GPSDO output to a higher order than Lady Heather's lock indication).

And all without falling too far down the rabbit hole, as my need for precision isn't at all extreme right now.

The rest is because I'm new to, and learning about, 10-MHz ref systems, 10-MHz input port impedances and drive levels, how to distribute, etc.

Thanks to everyone who is chiming in. Even if I'm flailing about, it's a fascinating topic.

Regards,

Kirk, NT0Z

[CatalinaWOW]

Just remember that the Shrodinger argument applies after you have found a way to verify the lady in leather.  All you really know is that it worked then.  At some point the only answers are trust and faith.

[GigaJoe]

a while ago i purchased.
https://www.ebay.com/itm/332389156868

construction:   a box about 3in styrofoam walls, ( homedepot styrofoam isolation panels) with inner space enough for:
 LT1021 (plastic)  and OCXO,  SMD caps for power filtering , multiturn outside for adjustment. all primitive
isolation gives some advance  - if OXCO open it consumed 130ma for heating,   inside the box , just 20ma -   i save the power! ,  much less jitter and deviation. less for room temperature

it run some time without attention , i guess it settle ( 2 years i think)

MAR 2022 10 000 000 006 initial set
SEP 2022   9 999 999 986
FEB 2023   9 999 999 965
so: -4x10-9 per year
i use it as ref for gen and freq. meter  it more then enough ,  sparingly GPS
( just a note as gps ocxo still under influence of the control board , stand alone crystal not, (a correction voltage permanent), UHF operators for example sense it matter.)

[tchiwam]

Monitor temperature and voltage/current drive long term. if either changes find why.

In my case I check the correction value register of the Rb disciplined oscillator and it tells me how long I can expect my reference cell has to live.

[jonpaul]

if you are far from the Russia-Ukraine War area, or Communist China (GPS Jamming, spoofing) , and have good GPS satellite réception, à good GPS DSO like the fine Leo Bodnar GPS will be most accurate.

Easiest phase comp to TCXO, Rb. Cs is analog scope lissajous figure

No need for RG6/u

our house clock 10 MHz is stupid and simple

Leo Bodnar GPS 10 MHz max vout>> Mini-Circuits splitters 50 ohm >> any 50 ohm coax > BNC

Driving 3..8 devices

enjoy

Jon

[GigaJoe]

im still wonder if someone really try Lissajous figure on 2 frequencies differ from 0.0001 Hz
like sooo interesting to watch it in a practice, when someone done it ...

the more less acceptable using scope actually watch a phase drift need at least 100M,
locking to one channel,  pointed to the cross of 2 traces on the screen, sweep to nS , and watch how the second trace slowly drifting 

you know direction , sweep timing,  duration in sec for drifting - that gives you a calculated difference between your reference and measured freq.

2-generation frequency meters with phase shifting measure  [ interpolar-enhanced reciprocal counting],  a great representatives like HP 5384A , 5A , 6A, with 11 digits give you even better performance for a reasonable cost.

[David Hess]

if you are far from the Russia-Ukraine War area, or Communist China (GPS Jamming, spoofing) , and have good GPS satellite réception, à good GPS DSO like the fine Leo Bodnar GPS will be most accurate.

And not too close to the federal government's installations in Nevada, or Nevada itself as I discovered.

[DavidKo]

im still wonder if someone really try Lissajous figure on 2 frequencies differ from 0.0001 Hz
like sooo interesting to watch it in a practice, when someone done it ...
...

Not a long time ago
https://youtu.be/xTy1kY_wtsY?t=1187

And they found that it is better to look on intersection between signals (explained later in the same video), since you get better sensitivity.

[MIS42N]

When I designed my GPSDO, I had the same question - how do you know it is accurate. So I designed it to be self monitoring. The OCXO is allowed to "free run" with a fixed control voltage for a few minutes. In that time the microprocessor detects the difference between the GPS 1 pulse per second and the supposed 10MHz signal from the OCXO. The accumulated data is used to calculate a new control voltage that aims to reduce the difference to zero. Rinse and repeat. The size of the voltage change is a measure of the frequency error. Small changes, pretty accurate. Big changes, not so much.

For the microprocessor to calculate an accurate voltage change, it has to determine the sensitivity of the OCXO fairly accurately (which it does when being commissioned). So it is possible to say within a few percent how much the frequency is changed.

So the GPSDO can measure its own accuracy. As far as I can determine, no other GPSDO works that way. It is based on the OSC5A2B02 OCXO mentioned by GigaJoe. Great value OCXO.

With a good GPS signal and reasonably stable temperature it maintains 10MHz±0.4mHz (better than 1 part in 10^-10) with a maximum phase error of ±half a cycle. With a less than ideal environment (indoor antenna) it is almost always better than 10MHz±0.01Hz (1 part in 10^-9) and will indicate if it thinks it is outside of that.

It is quite cheap. 2 ICs (74HC04 buffer for 10MHz output, PIC16F1455 microprocessor, works with low cost NEO-7 GPS module) 2 MOSFET (2N7000, BS250) and two regulators (so the OCXO has its own power supply independent of the rest). Optionally 2 4Nxx optocouplers for user interface. I am in the process of describing it in GitHub, hopefully finished in a week or so.

[NT0Z]

Sounds nifty!

If you think of it, please drop a link to your Github repo.

Thanks,

Kirk, NT0Z