[time-nuts] OCXO Phase Noise Measurement in Primitive Conditions

Bob Camp kb8tq at n1k.org
Tue Sep 2 22:39:22 UTC 2014


Hi

There are a lot of GPSDO ADEV plots out there. The T-Bolt has a lot of plots run on it, there are others.

Ideally the plot should start out looking like an OCXO:

1) Left hand side rising as you go left. Rate of rise 1/Tau.
2) Left hand side intersecting a flat region somewhere in the 0.1 to 1 second range.
3) Flat region running out to a few hundred or a few thousand seconds
4) Rising region past the flat area more or less straight line as you go further to the right. 

That’s what you start with before you have any GPS involved at all. The flat region could be at 1.0 x 10^-11, it could be 10X, or 100X lower than that depending on the OCXO you have. The flat region could run out to 1,000 seconds at 1.0x10^-13 if you get really lucky. One way to measure all this is to take two or three identical OCXO’s and compare them with a device capable of < 1.0 x 10^-13 accuracy at one second tau.

——

Now you have your GPS. Most modern GPS plots are pretty simple. You have an error (say 2 ns) at one second. That gives you an ADEV of 2.0 x10^-9 at a tau of 1 second. At 10 seconds the same error gives you an ADEV of 2.0 x10^-10. At 100 seconds the same error gives you an ADEV of 2.0 x10^-11. It just keeps on going down and down as tau gets longer. At some point it hits a floor and then runs flat from there on. Depending on how you do your corrections the floor is somewhere below 1.0x10^-13. 

——

Lay those two plots down on top of each other. 

An ideal GPSDO would have the best of both and only the best of both devices. You would have the close in ADEV and flat area from the OCXO. You would then hop on the falling plot of the GPS and ride it down to who knows where (let’s say 1.0x10^-14). It would magically have no peaking or odd behavior (it’s ideal). The closer you come to that ideal, the better your GPSDO is doing. 

——

In the real world we need to measure things like GPSDO’s. They can not be measured against themselves. They need to be compared against something else. Sorry, but there are no shortcuts. If you want to show how one is working, you need a comparison standard. It can be a second GPSDO. It could be a Hydrogen Maser. It might be a Cesium standard.  What ever you use to do the comparison must be up to the job. A counter with a 1 ns (or 100 ps) resolution isn’t going to show you 1 second tau data on a good OCXO.

Can you estimate what’s going on - sure. The gotcha is that you will never know if your estimates are correct without a comparison. If you have made a few hundred of these things, all done the same way, you probably can guess pretty well. If you have a new design, first time up and running, your guess may not be very good.

——

Yes, that’s the short / abridged version. The full version goes on and on for a few hundred pages. There are lots of special cases and a multitude of ways to deal with each of them. There are also a number of end goals you can target with a design. The optimum outcome will be different for each of the. 

Lots of Fun.

Bob


On Sep 2, 2014, at 12:02 PM, Bob Stewart <bob at evoria.net> wrote:

> Hi Bob,
> 
> I take it that the 3 segment ADEV pretty much defines a GPSDO.  So, is my plot an indication that I'm riding the OCXO too hard to get the results I think I should be getting?  What I think I should be getting is a flat phase line.  But is a GPS receiver, even after applying quantization error corrections, accurate enough to give that?  I see three things at play here: hardware, software, and my expectations.  I'm trying to figure out which one (or several) of these needs to be addressed.  I suppose I could hook the Rb back up and see the effect on the ADEV of the Rb vs OCXO as I loosen the gain values.
> 
> Bob
> 
> From: Bob Camp <kb8tq at n1k.org>
> To: Bob Stewart <bob at evoria.net>; Discussion of precise time and frequency measurement <time-nuts at febo.com> 
> Sent: Tuesday, September 2, 2014 6:38 AM
> Subject: Re: [time-nuts] OCXO Phase Noise Measurement in Primitive Conditions
> 
> Hi
> 
> Your ADEV is (mostly) a plot of 1-2 ns resolution out over your sample range. It’s not a “typical” three segment ADEV. It’s more likely that any departure from a straight line is random rather than systematic. Yes I’ve spent a lot of time looking at random plots that I was *sure* had significant information in them…..
> 
> Bob
> 
> 
> On Sep 1, 2014, at 9:31 PM, Bob Stewart <bob at evoria.net> wrote:
> 
> > Hi Poul,
> > 
> > I've made a few hardware changes and some software changes, and I was wondering if you (and/or others) would take a look at this plot.  The green is phase over 70,000+ seconds with ~100ps resolution, while the blue is the ADEV.  (The sharp departures in phase are from large temperature changes and not quite perfect temperature compensation software.)  I don't see your horizontal line down at the bottom, but there are some worrying horizontal lines between about 3tau and 150tau.  Should I take that to mean that I have some noise/oscillations from about 1/3Hz to 1/150Hz?  Those up at about 150tau I think I can see by eye on another plot of different data.  I've got a lot of data collected from this run (still running) including pTerm, iTerm, dTerm, and tTerm (temperature) updates to the DAC.  I hope to try to tease something out of that later, though the pTerm and iTerm are mostly +1,0,-1 stuff.  FWIW, in real terms, the dGain is at about 4.2.  The phase
> > starts to get ugly if I bring it much below that.  iGain is .01 and pGain is .05.  There are misc damping and limiting factors applied, as well.  
> > 
> > 
> > Just about ready to do a second board with updates from what I've learned, so any help is appreciated.  The plan is to release the source code at some (not too) future date and make boards available, if there's interest.  But I've still got a lot to do before that.
> > 
> > 
> > http://evoria.net/AE6RV/TIC/ADEV3.png
> > 
> > 
> > Bob
> > 
> > 
> > 
> > 
> > ________________________________
> > From: Poul-Henning Kamp <phk at phk.freebsd.dk>
> > To: Bob Stewart <bob at evoria.net>; Discussion of precise time and frequency measurement <time-nuts at febo.com> 
> > Sent: Thursday, August 28, 2014 1:04 AM
> > Subject: Re: [time-nuts] OCXO Phase Noise Measurement in Primitive Conditions
> > 
> > 
> > --------
> > In message <1409175707.2080.YahooMailNeo at web142706.mail.bf1.yahoo.com>, Bob Ste
> > wart writes:
> > 
> > This looks like a step in the right direction.
> > 
> > The "correct" allan plot will have a clearly visible horizontal
> > segment somewhere in the 100-10000 second range (depending on OCXO quality).
> > 
> > This is where the GPS long-term stability "takes over" from the OCXO's
> > better short-term stability.
> > _______________________________________________
> > time-nuts mailing list -- time-nuts at febo.com
> > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> > and follow the instructions there.
> 
> 
> 
> 
> 




More information about the Time-nuts_lists.febo.com mailing list