try these
http://tycho.usno.navy.mil/ptti/1987/Vol%2019_19.pdf
http://horology.jpl.nasa.gov/papers/FlfmSimPtti.pdf

and here
http://libra.msra.cn/Publication/50096626/a-new-time-domain-model-of-precise-clock-noise

On Thu, Nov 28, 2013 at 10:35 AM, Wolfgang Wallner
wolfgang-wallner@gmx.at wrote:

The link
http://tycho.usno.navy.mil/ptti/1987/Vol%2019_19.pdf
doesn't work, use instead:
http://tycho.usno.navy.mil/ptti/1987papers/Vol%2019_19.pdf

On Thu, Nov 28, 2013 at 2:03 PM, Azelio Boriani
azelio.boriani@screen.it wrote:

On 11/28/13 1:35 AM, Wolfgang Wallner wrote:

I think it's the right place..  There's plenty of Allan deviation plots
and data here for just about any kind of oscillator you care to name.

One way to generate realistic spectra is to take white noise from a
random number generator and run it through a filter which has the right
shape (e.g. 1/f, etc.).

Essentially this is implementing the Leeson model explicitly.

A year or so ago, I was looking for a similar thing to simulate human
heartbeats (which have a 1/f characteristic, just like other
oscillators).
http://www.febo.com/pipermail/time-nuts/2013-February/074505.html

see also http://paulbourke.net/fractals/noise/

On Thu, 28 Nov 2013 10:35:33 +0100
Wolfgang Wallner wolfgang-wallner@gmx.at wrote:

Hehe. Join the club. But treat carefully, this can become quite addictive ;-)

Can you explain what you are exactly doing? You talk about noise, but
only mention allan deviation. ADEV is the right tool to measure stability,
but not so much phase noise.

What part of the oscillator? Noise of the feedback electronics?
Noise of the output stage? Noise from environmental factors?
Noise intrinsic to the quartz crystal resonator?

How accurate do you want the model to be?

How do you simulate the complete oscillator?
(this is a big topic of its own, and definitly not easy)

Did you read Enrico Runbiolas Book "Phase Noise and Frequency Stability
in Oscillators"? If not, you should start with that. It gives a nice
overview of all the basics you need to understand this topic.

Gangepain did a lot of research on noise sources and stability of
quartz oscillators. You might want to look up papers from him.
(there was somewhere a collection of them)

A couple of weeks ago, i did a literature search on various stuff
around low noise/high stability oscillators. But i didn't have the
time to sort those papers yet, much less read them. But i can search
for things in there, if you tell me what you are looking for.

What are you looking for?

Well, Thieles group does mostly wireless sensor networks (catually most
of the TIK institute does wireless sensor networks in one form or another).
Their use for an "accurate" clock is to minimize the on time of the RF
circuit in order to minimize power consumption. IIRC their goal was to
get down from synchronisation window of 1s to 0.1 on a time scale of
a couple of minutes to a couple of hours. And the whole calculation
had to be simple enough to be done on an 8bit ATMega while not
taking considerable computation time.
These requirements lead to a rather simple model of clock deviation.

I dont know these two papers, but they dont look too bad.

As for the age. Most of the theoretical work on quartz oscillators
was done in the 70s and 80s. Also most of the books on quartz
oscillators are from that time. There are very few books from the 90s
and later.

febo.com (John Ackermann) and leapsecond.com (Tom van Baak) have both some
data on various crystall oscillators.

Yes. Because in the class of cheap AT cut oscillators, you dont worry about
allan variance. The instability due to temperature dependence of your
system is much higher than the temperature-free (in)stability. The ADEV
becomes "relevant" only after you do at least a temperature compensation
or temperature control.

Don't worry, you are at the right place :-)

		Attila Kinali

--
1.) Write everything down.
2.) Reduce to the essential.
3.) Stop and question.
-- The Habits of Highly Boring People, Chris Sauve

Ciao,

On Thu, 28 Nov 2013 14:03:06 +0100
Azelio Boriani azelio.boriani@screen.it wrote:

This host has ceased to exist. Can you tell us the title of the
paper and the names of its authors?

		Attila Kinali

--
1.) Write everything down.
2.) Reduce to the essential.
3.) Stop and question.
-- The Habits of Highly Boring People, Chris Sauve

Hello Wolfgang,

lots of interesting reading about oscillator noise:
http://rubiola.org/index.html

There are also some phase noise related publications from Ulrich L. Rohde:
http://www.tu-cottbus.de/fakultaet3/de/fakultaet/institute/stiftung/prof-dr-ing-habil-dr-hc-mult-ulrich-l-rohde/technical-publications.html

Best regards,
Adrian

Wolfgang Wallner schrieb:

Wolfgang,

A colleague of mine wrote this simulator based on a multirate filterbank:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5461653&queryText%3Dm+brooker+phase+noise

Since you are in the academia I'd assume you'll be able to access it?

Cheers,

Stephan.

On 28 November 2013 11:35, Wolfgang Wallner wolfgang-wallner@gmx.at wrote:

Unfortunately that was a contribution from Magnus in 2010

(see  www.febo.com/pipermail/time-nuts/2010-April/046932.html )

that I have simply reported without verifying the link and found that
link unusable after sending the message. My best guess is this:

http://www.crya.unam.mx/radiolab/recursos/Allan/Kasdin-Walter.pdf

based on a search on FLFM (flicker of frequency).

On Thu, Nov 28, 2013 at 8:22 PM, Attila Kinali attila@kinali.ch wrote:

On 11/29/13 5:56 AM, Azelio Boriani wrote:

one limitation of the Kasdin-Walter method is that it is "batch mode",
and doesn't lend itself to an implementation which is continuous.

The paper does have a nice discussion of why the "white noise into a
filter" technique doesn't work very well if the slopes you need aren't
integer powers of frequency. Integer powers in frequency correspond to
rational functions in filter characteristics, which are straightforward,
but how do you make a 1.5th order filter section or half a pole or zero?

The fractal literature, though, may provide mechanisms that might be
useful.

On 11/28/2013 08:18 PM, Attila Kinali wrote:

On the contrary, ADEV (and MDEV even more so) is designed to separate
noise-types such that they can be estimated separately. It is only of
lately that you can use both phase-noise plots and timer-based ADEV/MDEV
plots to achieve the same thing, not that the phase-noise
characteristics have been hard to estimate from, but rather that it has
been hard to achieve qualitative measures for precision sources such
that all noise-forms can be used for estimation. This have been easier
in the TIC-driven measurement.

Strongly recommended reading. His view is phase-noise driven.

I think one should study "both sides of the coin" since they have
different benefits and different usages.

If you want more modern papers on simulation of noise, then there is a
few from JPL that applies, but the site where unfortunatly dropped in
one of the server-roundups.

The specification for temperature variations is a poor excuse too. Some
vendors have learned that the hard way.

I agree. :)

Cheers,
Magnus

On 11/29/2013 04:11 PM, Jim Lux wrote:

Actually, NIST (or actually this was in it's NBS days) did a few good
articles, comparing the Mandelbrot simulation method with their filter
method. Turns out that you need to dimension the filter to the
simulation length, as the number of lead-lag sections needs to cover the
range where 1/f slope is needed and then the density of them (lead-lag
pole/zeros per decade) will control how close it will approximate, that
is, how little "pass-band" ripple there is from the ideal. Also, you
need to apply the corrections to start the filter up in the correct state.

It's non-trivial to do well.

There are many many methods to do this. Everyone has a favorite.

Cheers,
Magnus

On 11/29/13 8:50 AM, Magnus Danielson wrote:

That's essentially what the Kasdin-Walter paper talks about.  The number
of taps/sections is adjusted to approximate whatever curve you want
"well enough".

Then, they sort of shunt all that with an FFT based method.. Generate
white noise, filter it with a FFT convolution scheme where you've loaded
the bins of the FFT with the desired power spectrum.

And, I suspect, non-trivial to do with low computational complexity.

No doubt about it.

Jim,

On 11/29/2013 07:27 PM, Jim Lux wrote:

... which fails to reference the right papers:

NBS Report 9284 "The generation and recognition of flicker noise" by Jim
Barnes.
http://tf.boulder.nist.gov/general/pdf/190.pdf‎

NBS Technical Note 604 "Efficient Numerical and Analog Modeling of
Flicker Noise Processes" by Jim Barnes.
http://tf.nist.gov/timefreq/general/pdf/29.pdf‎

Jim Barnes and Chuck Greenhall "Large sample simulation of flicker noise"
http://tycho.usno.navy.mil/ptti/1987papers/Vol 19_19.pdf
This one has nice plots about different amount of stages, however you
really want the follow-up correction and addenda
http://tycho.usno.navy.mil/ptti/1992papers/Vol 24_44.pdf

This is the W. Riley list of references:
http://www.wriley.com/Refs.htm

The paper which is filename is FlfmSimPtti.pdf has the propper title
"FFT-Based Methods for Simulating Flicker FM" by Charles A. Greenhall of
JPL. Should have remembered Chuck's name in the previous post, but I was
tired. The Kasdin-Walter paper was proposed as a replacement, there are
similarities, but do read Chuck's fine paper!

This Greenhall paper is found here:
http://trs-new.jpl.nasa.gov/dspace/handle/2014/11024
http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/11024/1/02-2912.pdf

It is. Hence it is important to read-up.

Not sure which is my favorite just yet.

Cheers,
Magnus

On 11/28/2013 03:28 PM, Jim Lux wrote:

Make sure that you use different noise sources as you add different
distributions up, since they should not correlate with each other for
things to match up as you want.

Cheers,
Magnus

Hello,

thanks a lot for all your feedback (also in the other threads)!
It will take some time to read reed through all your recommendations :)

On 11/28/2013 08:18 PM, Attila Kinali wrote:

The interest comes from IEEE 1588, which is an interesting standard, but
the equipment is quite expensive (a single switch with transparent clock
support for >1000$).
This means it is hard to estimated what could be achieved with 1588, as
it is rather impossible to just buy some devices and try it out.
Questions which would be interesting are things like:

*) what synchronization interval is need to achieve a specific
precision, and where are the limits
*) how does a long daisy chain behave (with/without syntonization in
each hop)
*) what happens when a link breaks and comes up some time later (how far
will the parts drift away)

At my institute (TU Vienna, Computer Engineering) there has been a
bachelor thesis which dealt with simulation of IEEE 1588 in OMNeT++ (a
discrete event simulator).
But the assumptions where rather simple (both of the clock model and the
implemented version of IEEE 1588).

For my master thesis I would like to enhance both aspects.
I would like to do a full implementation of IEEE 1588 and to use a more
realistic clock model.

Implementing IEEE 1588 is rather straight forward. The standard is easy
to read, and it is not rocket science to implement this it in C++.

But as already stated in my first message, I underestimated how hard it
would be to get a realistic clock model.

Hi Wolfgang,

On 11/30/2013 10:32 AM, Wolfgang Wallner wrote:

It's a difficult topic because:

1. There is no defined clock model defined in IEEE 1588, so there is no
   way to generally specify what a 1588 device will do. I've checked this
   with one of the core 1588 guys and he agrees.

2. There is no defined network jitter model. You need to consider if
   equipment is 1588 aware or not, and it may be traffic load dependent,
   and there is a wide range of behaviors which does not fit "normal"
   noises. For 1588 aware switches and routers which works well, a first
   degree damping can be expected, but it is not perfect. The systematic
   behavior to low-frequency "noise" will leak through.

The best you can do is build a few fair models and test with a few
different noise-characteristics and see what they will do.

Remember that it is not just the noise sources of your oscillator, you
have systematic effects to care about, such as aging, initial frequency
offset, initial phase offset, temperature dependence, PLL parameters
etc. etc.

Cheers,
Magnus

Hi Magnus,

On Fri, 29 Nov 2013 17:42:25 +0100
Magnus Danielson magnus@rubidium.dyndns.org wrote:

Could you explain a little bit what you mean here? I don't think
i get exactly what you are hinting at.

		Attila Kinali

--
1.) Write everything down.
2.) Reduce to the essential.
3.) Stop and question.
-- The Habits of Highly Boring People, Chris Sauve

Servus Wolfgang,

On Sat, 30 Nov 2013 10:32:42 +0100
Wolfgang Wallner wolfgang-wallner@gmx.at wrote:

I think you could easily do a PhD on this topic alone.
Thus I would highly recommend to focus on one aspect only
for your master thesis.

For simplicity, i'd first use some numbers on a good OCXO. These are
much better specified and measured than the cheaper ones. E.g. you can
use the Oscilloquartz 8607 as reference. If what data is freely available
online is not enough for you, try contacting the manufacturer. They always
have better data available, but do not publish it (don't ask me why).
But still, they are usually quite generous with handing this data out
for specific projects.

Using a good oscillator will also give you a chance to verify your
model. It should still be close to what the simulation with an ideal
oscillator. Check for any deviation and try to explain it from the model.
If you cannot explain it, it might be a simulation artefact.

it matches those of the oscillators you actually want to study.

		Attila Kinali

--
1.) Write everything down.
2.) Reduce to the essential.
3.) Stop and question.
-- The Habits of Highly Boring People, Chris Sauve