[time-nuts] Setting Osc Frequencies

[Tom Van Baak]

Don,

Yes, using an oscilloscope in this way is a time-honored
trick to observe slow changes in relative phase over time.

It gets even better if you have access to a storage 'scope
or variable persistence because then you can get a rough
sense of short-term jitter as well.

Either way, the math is the same. Frequency error is just
phase change over time. You have two things working for
you: 1) you don't even have to wait for full cycles, and 2)
there is no limit on how long you can wait.

If you have a fast sweep of 10 ns/div you can easily eyeball
fractions of a 100 ns cycle. High-end 'scopes allow ns and
sub-ns resolution. Add that to long observation times (many
minutes, even hours) and you've get quite good frequency
resolution.

Again, as an example, if you observe that it takes about
3 hours for it to move 1/10 a cycle (10 ns), that's about
1e-8/1e4, or 1e-12. You see, even using the old CRO
method, you get fine results. The downside is that you
have to keep an eye on it now and then to make sure the
relative phase changes continue to be gradual and that
you aren't cycle slipping.

Now you ask if there is a better method. If you intend to
automate it, to collect a lot of data, then the CRO method
is maybe not ideal. For example, to see how frequency
changes over time, from week to week, or month to month,
so that you can compute the aging rate, you might want
something more digital and less eyeball. Or if you want to
measure oscillator stability (rather than accuracy) you'll
want to collect a series of periodically spaced readings.

Your 5245L would partially help here. The trouble is they
don't have a modern digital output; it's a parallel BCD
interface if I recall.

Does anyone else have ideas for Don?

/tvb

----- Original Message ----- 
From: "VK3YV" <vk3yv at netspace.net.au>
To: "Tom Van Baak" <tvb at leapsecond.com>; "Discussion of precise time and frequency measurement" <time-nuts at febo.com>
Sent: Friday, March 02, 2007 19:45
Subject: Re: [time-nuts] Setting Osc Frequencies


> HI Tom, I realised when I read your reply that I had not said how I was 
> observing the result. I am feeding these signals into a 100 MHZ CRO not a 
> counter and visually observing the 2 signals in relationship to each other. 
> 1 being the Trimble which I believe to be good to 3 parts in10^12 and the 
> Sulser up in the 10^11 range..Have a couple of  HP5245L counters but I have 
> not arranged an external reference for them yet.
> I hope that clears up what I am doing.
> Regards,  Don  VK3YV....
> 
> 
> ----- Original Message ----- 
> From: "Tom Van Baak" <tvb at leapsecond.com>
> To: "Discussion of precise time and frequency measurement" 
> <time-nuts at febo.com>
> Sent: Saturday, March 03, 2007 1:40 PM
> Subject: Re: [time-nuts] Setting Osc Frequencies
> 
> 
> >> HI All , first time request. I wish to know how I can accurately set some
> >> oscillators on frequency. I have 3 Sulzer oscillators .. 1 x 2.5A and 2 x
> >> 2.5C's and 1 matching P/S with dead nicads.
> >
> >> My main reference is a  Trimble Thunderbird which was the second source 
> >> to
> >> the HP unit was used in cellular phone installations in the US a while 
> >> ago I
> >> think that they maybe Nortel pulls like the HP's that were around.
> >
> >> I have tried in the past by putting the main reference into input "A" and
> >> triggering 1 cycle of 10 MHZ and feeding one of the other oscillators 
> >> into
> >> input "B" and adjusting the Sulzer to be totally overlaid with no 
> >> movement
> >> over many hours.. All Sulzers had been on for 36 hours before the 
> >> test.and
> >> the Trimble has been on for 2 years.
> >
> >> What accuracy can I expect from this method, and is there a better method
> >> with the equipment I have.
> >> Regards, Don     VK3YV.......
> >
> > It would depend on the specs of the time interval or frequency
> > counter you're using. You mention, 1 cycle of 10 MHz; that's
> > 100 ns. And you mention "many hours". The upper limit based
> > on that alone is 100 ns over, say, 10 000 seconds which gives
> > you a resolution of 1e-7 / 1e4 seconds = 1 part in 10 to the 11th.
> >
> > That's than enough accuracy, since, during that couple of
> > hours a 36-hour warmed-up Sulzer will likely have drifted in
> > frequency by about that amount.
> >
> > /tvb
> >
> >
> > _______________________________________________
> > time-nuts mailing list
> > time-nuts at febo.com
> > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts 
>