LEA-6T disciplined SRS PRS10 Rb

[notfaded1]

I've been working on improved house standard and experimenting with different options.  Here's my latest attempt.  I just got it hooked up and am going to let the gps complete it's thing before calling it good but the Rb locked up quickly.  Before when I plugged it in and after a few minutes pics of the interface board.  It hasn't even warmed up but looks really promising on the 53132A!  I'm hoping this might become my new standard that can beat my Thunderbolt.

Bill

[notfaded1]

I need to run it more but I think the DUT is probably better than the external reference the 53132 is using the BG7TBL GNSSDO.  I want to try the SR620 with recommendation from the PRS10 manual setup.

[jpb]

I am interested in following your progress.
I'm currently in the process of measuring my two Rb standards against each other and comparing them to my GPSDOs with a view to producing my own GPSDO at some point (time permitting).

The ADEV curve (line) though I think is just a measure of the 500ps one-shot resolution of the counter. Both the GPSDO and the disciplined Rb should be well below it (<10^-11 at 1 sec and ~10^-12 at 100 secs).
The specs for the LPFRS (the one I have) is
2x10^-11 /1s
7x10^-12/10s
2x10^-12/100s
and I'd expect your set up to be a bit better than this.
Also see:
http://www.ke5fx.com/rb.htm

[notfaded1]

Looking good so far.  I'll try longer.

[Theboel]

Hi Notfaded1,

Could You extend the measurement up to 20000 second

[testpoint1]

you can try another one, Spectratime SRO-100, I made a reference based on this.

[notfaded1]

Could You extend the measurement up to 20000 second

I'll post 12 hour test run when I get home after work today.  It's really good I saw it this morning (looked linear) before leaving for work.  I'm really impressed with how the PRS10 functions under LEA-6T control.  I wonder if Ublox 8 series would make any difference really?  That may not matter as much because the LEA-6T is doing very well.

you can try another one, Spectratime SRO-100, I made a reference based on this.

I plan to try SRO-100 as well... it's control loop algorithm seems superior from much I've read and it saves more in info in NV storage onboard between power cycles.  I've been thinking when comparing oscillators this accurate it becomes limited by the ability of the testing hardware and it's GPS disciplined frequency reference.

You're only as good as your weakest link right?  This is also where things like three corner hat measurements start to help some?

Bill

[DaJMasta]

I ended up trying a bunch of ublox modules and their PPS output was almost always the same (LEA 5T, NEO 7N, NEO M8T iirc).  Fairly stable, but with visible steps in the adjustment size of the PPS pulse.  In the modules I tested, the Trimble Resolution SMT had the best stability so I used it in my similar attempt to discipline a PRS10.  Works great, don't have a more stable reference to compare to really, but also the SMT has some glitch after about a month of uptime where the PPS pulse either stops or fails to be recognized - in the time I had it running it happened consistently, even though between the module and the PRS10 is just a level shifter.

In any case, I managed to pick up a Jackson Labs Fury since then, so when I can get my antenna situation worked out at the new place, maybe I can make some stability measurements again around a similar noise floor.  It uses a Motorola M12+ GPS receiver, which they claim is a top performer despite being a bit of an older model by today's standards.

[thinkfat]

What's the noise floor of your measurement setup? The PRS10 should be better than what your graph shows. I guess you're only showing your counter, but not the DUT.

[notfaded1]

Here's the 12H with the 53132A.  I'm going to try the SR620 next I think.

[KE5FX]

As others have suggested, you're measuring your counter. 

[Theboel]

I agree with John there is no way you have a straight line like that

[notfaded1]

As others have suggested, you're measuring your counter. 

Thanks John.  Is there anything I can do differently?  There's no way to get around the limitations of the 53132A counter is there?  I was planning on trying the SR620 setup from the PRS10 manual next.  For this measurement I used the configuration for TI on single port from the HP manual.  The line isn't straight while it's running it bends around up and down some and wiggles back and forth but over time it straightens out a lot.  I did notice one thing... if I save the raw data in the .TIM file and then load raw data into Stable32 it produces a slightly different result (very similar but not exactly the same).  I wish i had a phase station or time pod but they really don't sell in enough quantity that they are even in the used market.  I would buy a used one if I could find it.  I really like those realtime 3 corner hat measurements I saw on this youtube video:    I'd like to see 3 different Rb ran the way the BVA 8607 DOCXO is measured using two C-MAC DOCXOs.

Bill

[jpb]

There are several ways of measuring ADEV with a low noise floor but none of them are very easy. It is something I'm working on at the moment for my setup.
The core of most methods is to use a mixer to amplify the effect of the noise.
The classic approach is to use two mixers (dual mixer) to take two simultaneous measurements with the mixer noise effectively cancelling.
Another approach is to frequency multiply the source and reference before mixing.
See:
https://www.wriley.com/Handbook.pdf
This is quite an interesting approach:
https://cdn.intechopen.com/pdfs/24305/InTech-High_precision_frequency_measurement_using_digital_signal_processing.pdf

It is quite a big subject and there are lots of techniques.

A simple approach that I'm currently using is I have two Phase Locked oscillators I got off ebay - one is 2GHz and the other is 2020 MHz and I mix the output to give me 20MHz. In theory this should give an improvement of around 200 times, in practice it gives me a noise floor well below 10^-12 where my counter alone is 50 psecs so around 5 x 10^-11 or more like 10^-10.

It works well with my two GPSDO measured against each other but the results I'm getting for two  Rb oscillators measured against each other are the right level but rather wavy and I can't work out if it is the oscillators I'm measuring or some interaction with my measurement system.

It is not easy in practice even though the theory seems really simple!

Here is a plot of my two GPSDO measured against each other and then scaled by sqrt(2) (assuming they are both the same). The curves are measured at different times by different methods where the noisier methods I've subtracted the noise floor from. You can see there is quite good agreement between the 2GHz mixer approach and the more direct methods with noise floor subtraction.

EDIT : Lars told me he used a mixer as a phase detector in combination with a 6.5 digit DVM to measure his various GPSDOs against each other.
Another approach is to phase lock the reference to the DUT and measure the voltage controlling the frequency:
http://www.ke5fx.com/tpll.htm

There is also the FSA3011 on ebay but it is quite expensive and I've not read any reviews from anyone to say it works in the general case.
eg :
https://www.ebay.com/itm/2020-FSA3011-OCXO-GPSDO-Atom-Clock-Frequency-stability-Analyzer/133109901703
https://www.eevblog.com/forum/metrology/turn-dynamic-systems-fsa3011-frequency-stability-analyser/msg2367576
https://www.eevblog.com/forum/metrology/bg7tbl-fa1-frequency-analyzer/50/
https://www.mail-archive.com/time-nuts@lists.febo.com/msg04082.html

[tkamiya]

You are basically doing same thing I've tried for 6 months. 

I had a PRS-10 and tried to sync it to various GPS.  I've tried two different GPS receivers and one T-bolt.  Output of pure GPS receiver were a bit jittery, even with long time constant, it made Adev of PRS-10 by factor of 1.  T-bolt on the other hand, GPS steers 10MHz and 1 pps is generated from the 10MHz.  It's much more stable, and pushed and pulled PRS-10 more gently.  It was a better fit for the purpose but ultimately, I removed all that entirely.  Neither one gave me any improvement in performance.

I suspect what SRS shows in spec sheet is a result of engineers spending months fine tuning parameters.  I certainly didn't see what they show.  Sure, in very long term, it might have been better but for my purpose, it was short term and mid term performance that was important.  I have another Rb oscillator that has been running for over a year.  It drifted few parts to 10^-12. 

I just check my calibration once a year and it's a very simple and reliable solution for me.
It would have been a nice way to SET (calibrate PRS-10) but for on going basis, I don't know if there is any merit. 

[Theboel]

Hi All,

Maybe we can try "another way" to discipline PRS-10.
what we need a GPS module / GPS receiver that can accept 10MHz external input something like ashtech
https://www.ebay.com/itm/Ashtech-Inc-Z-12-GPS-Receiver-XII-P-N-700845-6/372862079762?hash=item56d04fd312:g:mrUAAOSwsnVd5Vjo

or if We prefer a new

https://www.swiftnav.com/piksi-multi

and then use the 1PPS send back to PRS-10.
I have measured some TCXO/XO in GPS Module and I don't like it but considering the sale price. I must accept it.
anybody have try this before ?

[KE5FX]

As others have suggested, you're measuring your counter. 

Thanks John.  Is there anything I can do differently?  There's no way to get around the limitations of the 53132A counter is there?

I'd recommend using it in frequency-count mode.  It's impossible to tell what's going wrong here, but there's a lot less to go wrong in frequency mode.  Overall it's much easier to set up and run a frequency measurement when you're getting started.  The 53132A's averaging implementation is quite good for these applications (see tkamiya's results, which I think he posted a while back).

I'd like to see 3 different Rb ran the way the BVA 8607 DOCXO is measured using two C-MAC DOCXOs.

Yep, that's a pretty informative measurement, or at least it can be.  You have to keep in mind the fact that the three sources are exposed to the same environmental conditions, so they can potentially look better than they really are.

Re: the plots looking different in Stable32, I'd need more info, likely including the original data, to say anything about that.  If the parameters match, the plots will, too... but there are quite a few of those parameters.

[notfaded1]

I'd recommend using it in frequency-count mode.  It's impossible to tell what's going wrong here, but there's a lot less to go wrong in frequency mode.  Overall it's much easier to set up and run a frequency measurement when you're getting started.  The 53132A's averaging implementation is quite good for these applications (see tkamiya's results, which I think he posted a while back).

I can try frequency.  I configured the 53132A exactly by the instructions method for single cable time interval measurement.  Another datapoint I setup the SR620 in the way it says to in the PRS10 manual.  I was able to do the fine measurement after the coarse measurement worked.  I noticed it looked like while doing the measurement the SR620 would go through a long drawn out fluctuation from around 50000 plus to cross 0 and then around -50000 and go back and forth forever basically.  This is on the digits on the right hand side of the SR620 display so it's ns and ps it's showing.  The fine measurement settings are in TIME mode on the SR620.  These gave varying results on timelab unless I just let run for long enough that it went through many of these 50000 + to - sweeps back and forth.  I can post the graphs when I get home.  I do also save all the .TIM files for everything.  Usually when load into Stable32 I take the .TIM file strip off the header lines and load the raw data into Stable32.

Bill

[notfaded1]

I started thinking maybe my BG7TBL GNSSDO wasn't stable enough to feed the 53132 or SR620 ext ref input and need to put a free running Rb on the ext inputs instead?  Ultimately I want to get my 4040A Cs working so I can use that for measurements like this.  Essentially when doing these test it's one GPS disciplined OCXO trying to measure a GPS disciplined Rb.  When I see that fluctuation of around 50000 ns back and forth across zero positive and negative slowly that makes me think one of the two aren't standing still enough but since it's very regular pattern it makes me think it's one of them and not both fluctuating.  Also would it be better to put the PRS10 into standby non disciplined free running mode to get better stability when testing it?  I suppose once I know I'm getting a good read I should just try it both ways.

Bill

[notfaded1]

Maybe we can try "another way" to discipline PRS-10.
what we need a GPS module / GPS receiver that can accept 10MHz external input something like ashtech
https://www.ebay.com/itm/Ashtech-Inc-Z-12-GPS-Receiver-XII-P-N-700845-6/372862079762?hash=item56d04fd312:g:mrUAAOSwsnVd5Vjo

I haven't tried this but I do have an old mobile Ashtech Z-12... I started to get around to getting all the special connectors for it.  I have the power cables for it so you can run it off a 12V battery.  It's an original cable that has two big battery clips like for attaching to a car or big motorcycle 12V battery.

Seems pretty old tech compared to the Ublox 6 series T.  I don't think there's anything wrong at all with the Ublox disciplining the PRS10.  It's working quite well.  I've seen it measured on a Pendulum CNT-90 and I'll have my own pretty soon to test with in a few days.  I'm hoping the CNT-90 in 53132A emulation mode will work with timelab without any issues.  I don't see why it wouldn't.

[tkamiya]

Hi,

I have number of counters but so far, I had the best result from HP53132A in frequency mode and gate time set to 1 second.  I've tried several standards.  OEM version of T-bolt made for Nortel (GPSTM) was good enough and Rubidium was also good enough to get down to low enough noise level.

The limitation of HP53132A (averaging) shows up as scale being off a bit.  It shows better than it should be.  But it will show the trend and it is above counter's noise level. 

I tried disciplining PRS-10 and kept at it for about a year.  My end result was, I could make it worse, but not better.  Pulling and yanking of PRS-10 by GPSDO makes it actually worse.  Someone told me the graph on SRS literature is a result of experts fine tuning the setup and was done in the very best of conditions.  Not something I can realize at my home lab. 

Good luck with your experiment.  If there is anything I can help you, please do let me know.

[notfaded1]

I think this should look better than this...  these were less than 1 second gate time you can see it's sub second on the samples.

[notfaded1]

I have an HP53131 too I can try.  I also have other references I can try vs. the BG7TBL but it's really handy so it's why I've been using it and it's on all the time.  I can also try my Thunderbolt E or one of 3 Rb free runners I've got as the external reference.  I should have the Pendulum by next weekend at the latest I think and it's pretty good all on it's own.  One thing I got that's kinda neat today is this to go with my SR620.

[notfaded1]

Another thing... the SRS SR620 is REALLY cool looking in the dark!

[tkamiya]

Your adev graph look very much like when I tried to gps discipline mine.  What happens if you disconnect it from GPS?