Time interval counter?

On Sun, Sep 22, 2013 at 1:30 PM, W3KL w3kl@w3kl.com wrote:

On 09/22/2013 01:30 PM, W3KL wrote:

You want to measure a drift of 4/(4E6*3600) = 278 ps. You systematic
frequency error can be at maximum 1.39E-10 relative, For your noise side
look at TDEV at tau of 4 s, multiply that number by at least three and
it should when added with peak frequency error be below 278 ps.

Cheers,
Magnus

Hi

I think he needs 16X better than that (0.1 degrees in 4 seconds) so it's 1/(44e63600).

That's right around 2x10^-11, so you would need a system that's good to maybe 2x10^-12 to do the trick. The easy approach is to build several of the same oscillator and compare them to each other. The assumption would be that the drift will be random over the 4 second period.

Bob

On Sep 22, 2013, at 7:47 AM, Magnus Danielson magnus@rubidium.dyndns.org wrote:

Magnus. Thanks.  If I understand, this reduces to a measurement of frequency
stability along a measurement of phase noise?

Jeff

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Magnus Danielson
Sent: Sunday, September 22, 2013 7:47 AM
To: time-nuts@febo.com
Subject: Re: [time-nuts] How To Measure Long Term Phase Stability Of An
Oscillator

On 09/22/2013 01:30 PM, W3KL wrote:

You want to measure a drift of 4/(4E6*3600) = 278 ps. You systematic
frequency error can be at maximum 1.39E-10 relative, For your noise side
look at TDEV at tau of 4 s, multiply that number by at least three and it
should when added with peak frequency error be below 278 ps.

Cheers,
Magnus

time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

Hi

All you really need to do is to measure the frequency stability to the ppt level. There's no real need to measure phase noise.

Bob

On Sep 22, 2013, at 8:01 AM, W3KL w3kl@w3kl.com wrote:

Jeff,

You need to measure phase with sufficient resolution and rate of time. I
was vague on the equipment side but rater noted what you needed to do in
the analysis-side.

I would prefer to measure it at least with 10 measurements a second. Bob
mentioned resolution, which is important as you don't want your
measurement being swamped by measurement noise.

I would use a TimePod, but not all of us have one, which is a pitty as
it is a good instrument suitable for exactly this.

Cheers,
Magnus

On 09/22/2013 02:01 PM, W3KL wrote:

Magnus.  Thanks.  Re-read your original post and along with your latest I
now understand what's needed.

I will come back if I have other questions.

Thanks!

Jeff

-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Magnus Danielson
Sent: Sunday, September 22, 2013 1:02 PM
To: time-nuts@febo.com
Subject: Re: [time-nuts] How To Measure Long Term Phase Stability Of An
Oscillator

Jeff,

You need to measure phase with sufficient resolution and rate of time. I was
vague on the equipment side but rater noted what you needed to do in the
analysis-side.

I would prefer to measure it at least with 10 measurements a second. Bob
mentioned resolution, which is important as you don't want your measurement
being swamped by measurement noise.

I would use a TimePod, but not all of us have one, which is a pitty as it is
a good instrument suitable for exactly this.

Cheers,
Magnus

On 09/22/2013 02:01 PM, W3KL wrote:

time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

Jeff,

Sorry if I was unclear in the first message, but I was tired.

The factor of 3 that I have in the first message is a huge
approximation, and could be a higher value for higher confidence
interval. It's an attempt to handle the chi-square distribution margin
for some probability of staying within. I used 3, as the traditional
engineering margin for the usual gaussian distribution, mostly because I
was to tired to dig up the propper value, but it doesn't really care as
it is an approximation never the less and in the right direction.

The underlying point is that you need to estimate the systematic and
random effects separately, and then fit them together with their quite
drastic different probability behaviors. You can't use either of the
tools as the single tool, unless you have a fairly good idea of the
dominant part, which I can't in this case.

I discussed this point with Dr. Allan, and he only knew of a pair of
articles even covering this aspect, but he agreed on the overall approach.

Cheers,
Magnus

On 09/22/2013 08:43 PM, W3KL wrote:

On Sun, Sep 22, 2013 at 4:30 AM, W3KL w3kl@w3kl.com wrote:

The simplest way to first to get a very much better oscillator than
the one you need to test.  You need a "refference".  Then you "add
them" and if there is any deference at all you will get a beat
frequency

If the two are "close" the beat will be very slow, slow enough you can
measure the period with a wrst watch.  Put the two on a daul channel
scope nd watch.

More sofesticaed method is to use a transformer to add the two signals
then feed the sun into a computer's audio input or otherwise record
the beat frequency.  You can do an FFT on the recording.

I sure others will have even better methods but my point is that
simple technique can work, provided you have a really god reference
oscillator.

--

Chris Albertson
Redondo Beach, California