Frequency accuracy

w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Frequency accuracy

Postby w9mdb » Mon Apr 27, 2020 5:20 pm

Trying to use my ANAN100 for the FMT test and I find that the Calibration Correction Factor doesn't allow getting the GPS reference signal closer than about .180Hz. Adjusting the correction factor doesn't always make any change at all and when it does the frequency will jump 0.5Hz or so. I would imagine this might be because of the digital representation of the correction factor.
Would it be possible to get more accurate frequency in Thetis or is this a firmware change or an inherent limit of the hardware?

de Mike W9MDB
Mike W9MDB
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Frequency accuracy

Postby w-u-2-o » Mon Apr 27, 2020 7:11 pm

What is an FMT test?
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Frequency accuracy

Postby w9mdb » Mon Apr 27, 2020 7:24 pm

The ARRL FMT test...for those who want a challenge and to learn about frequency measurement.
Usually twice a year but may be another one soon just for funsies...

https://fmt.arrl.org/

de Mike W9MDB
Mike W9MDB
WA2DVU
Posts: 47
Joined: Wed Nov 01, 2017 3:27 pm
Location: Cape May, NJ

Re: Frequency accuracy

Postby WA2DVU » Mon Apr 27, 2020 8:31 pm

For FMT I have been using HP 3336, HP 3836, tek 465, Tek 465, and a dart board. On ground wave sigs I can measure within .01mhz. On FMT I can hit .1 at times when doppler is not to bad. I am on the east coast and it gets rough at times.

I checked out my 7000 thousand and found the best I can measure a gps locked 15mHz sig is +5 mhz on CWU mode and - 23 mhz on CW mode.

To check out 7000 I fed 15 mHZ from Lucent gpsdo into 7000. Set rx tuning to 14.999.00, Pitch F to 500 hz, mode CWU. Drag narrow passband over carrier at 15,000.000 on left side of screen. Feed audio to Spectrum Lab set up to measure audio tones from front earphone jack. I read + 5 mhz

Change tuning to 15.001.000, Mode to CWU. Drag narrow passband over carrier at 15,000.00 on right side of screen. I read - 23 mhz.

I ran both side bands over night and the readout list did not change.

Your apache may have different errors but I think that they do not change.

If you are going to use your Apache for FMT you will have to add or subtract these errors along with the 1500 hz errors as indicated in SL.

I can see my self making many math errors!!

I believe the timing errors have something to do with the different bins that are used in the timing method used in the 7000. Would be interested in how this happens.

73,
Bill, WA2DVU
Cape May,NJ
.
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Frequency accuracy

Postby w-u-2-o » Tue Apr 28, 2020 4:14 am

w9mdb wrote:Trying to use my ANAN100 for the FMT test and I find that the Calibration Correction Factor doesn't allow getting the GPS reference signal closer than about .180Hz. Adjusting the correction factor doesn't always make any change at all and when it does the frequency will jump 0.5Hz or so. I would imagine this might be because of the digital representation of the correction factor.
Would it be possible to get more accurate frequency in Thetis or is this a firmware change or an inherent limit of the hardware?

FMT, hmmmm, interesting! I wasn't aware of that sort of contest, thanks for pointing it out.

There are a lot of moving parts in this game.

Certainly there is a resolution/precision limit to the correction factor in the software. I do not believe the developers ever envisioned needing better than 0.5Hz steps.

To work around this you can (and I do) use a GPSDO on the external 10MHz input. This will get you to within millihertz of the correct frequency. Correction factor is set to exactly 1 when you do this. If you want to win the FMT contests this would be an easy thing to do, and not terribly expensive. Cheap Chinese made GPSDOs are all over eBay.

After that you have the problem of making the actual measurement. The resolution of the spectral display may not be sufficient for your needs. I haven't played with it enough to know what a single pixel is worth. However, limitations in this area may be why other folks will route the audio for input into a separate audio spectrum analyzer, either hardware or software, and make the measurement that way.
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Frequency accuracy

Postby w9mdb » Wed Jul 15, 2020 9:56 pm

FLDigi has been nice enough to add an FMT analysis capability when you have a secondary reference that you can put next to your unknown.
http://www.w1hkj.com/alpha/fldigi/

I have 2 Bodnar Mini GPS systems for example...so I use one as the 10MHz reference on my ANAN 100 and then measure the other one (since these are programmable frequencies makes it easy to do and I placed the unknown 250Hz above 10Mhz).

In this plot it shows measuring the "unknown" Bodnar is actually more accurate than measuring the reference itself. Since these two GPS's are highly correlated that makes some sense but was an unexpected result.
3 sigma on the unknown is 9 milliHertz and 21 milliHertz on the 10Mhz reference in this 5-minute data plot.
BodnarTest.png
BodnarTest.png (93.05 KiB) Viewed 6598 times
Mike W9MDB
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Frequency accuracy

Postby w-u-2-o » Thu Jul 16, 2020 1:09 am

Well that should put you consistently in the <0.1Hz category!
User avatar
FM5GB
Posts: 130
Joined: Sun Apr 09, 2017 10:03 pm

Re: Frequency accuracy

Postby FM5GB » Thu Jul 16, 2020 12:23 pm

Hi all !

Just for fun but no practical ham use ! When running my 10 GHz transverter chain
I'm well within accuracy for QSO's when using my GPSDO. I can understand people contesting for highest
frequency accuracy, trying to get close to the 133 Cesium standard (10 to the power of minus 16
second per second deviation). This is one tenth of a femtoHertz...
No bubble no trouble.

73s Phil
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Frequency accuracy

Postby w9mdb » Thu Jul 16, 2020 12:44 pm

Actually it's more practical than you think....
Many people don't understand thermal variations or the stability of their own oscillator.
And one of the goals of FMT is also to understand propagation which means very accurate timing is important.

For daily ops yeah...makes no difference practical difference.

I'm also uncovering bugs in software doing this type of thing as any source of variation needs to be understood and currently seeing small audio dropouts that I'm trying to find out where it's happening next on my to-do list.

Mike W9MDB
Mike W9MDB
User avatar
FM5GB
Posts: 130
Joined: Sun Apr 09, 2017 10:03 pm

Re: Frequency accuracy

Postby FM5GB » Thu Jul 16, 2020 12:49 pm

Hi Mike

a low phase noise seems a more important spec to me for an
oscillator.

Phil FM5GB
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Frequency accuracy

Postby w9mdb » Thu Jul 16, 2020 12:58 pm

Phase noise is typically much less than thermal variation.
All the specs you see on oscillators are based on temperature-stabilized measurements.

Mike W9MDB
Mike W9MDB
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Frequency accuracy

Postby w9mdb » Sat Jul 25, 2020 2:59 pm

Was playing around some more and found a problem with Thetis's VAC1 Audio Monitor. When you turn off "Force" the frequency oscillations are not stable.
You can see where I was running with Force checked and then turned it off causing a +/- 1Hz variation in the measured frequency. This oscillation also shows up in FLDigi's WWV calibration.
If I could get access to the git repository for the latest beta Thetis I can work on fixing this.

Mike W9MDB

Untitled.png
Untitled.png (17.52 KiB) Viewed 6300 times
Mike W9MDB
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Frequency accuracy

Postby w-u-2-o » Sat Jul 25, 2020 7:24 pm

Mike,

The problem you write about has nothing to do with the MON function, and everything to do with the variable resampler built into VAC.

If you take a look at this topic: viewtopic.php?f=6&t=3078&p=11298

You will see that for a properly adjusted set of VAC parameters the frequency instability you observe does not exist even with the resampler turned on.

There is a problem, however, in not knowing what constitutes a properly adjusted set of VAC parameters, and there is also a known issue with the resampler where it sometimes goes unstable and runs wild, mostly on startup.

At any rate, the code is located in the GitHub TAPR repository: https://github.com/TAPR/OpenHPSDR-Thetis

73,

Scott

Return to “Thetis”