Dithering vs. locking all the clocks to the OCXO?
Hi All,
Earlier, I explained that my application require very good relative
stability between various GPSDOs.
A rough estimate of my requirements is:
-Baselines of 100s of meters to 10s of kilometres.
-Sub-nanosecond relative stability (this I forgot to mention earlier -
thanks to TvB for reminding me).
-Time scales of maybe 100s of seconds to 10s of minutes.
-The lower limit on my stability requirement is maybe the 200ps of jitter
that the FPGA will add to the processed data.
My question is this:
At this stage I'm not sure what all the various causes are for the error in
the 1PPS output of a GPS receiver. (I sure I will find the answer to this in
all that references TvB and Magnus gave me).
However, a quick guess would be the delay caused by atmospheric effects (I
don't think thermal noise would play a big role since the antenna is looking
straight up) Also, there will be errors higher up in the food chain, such as
changes in satellite orbits etc. I guess these errors are fairly systematic.
Lower down in the food chain, I presume the M12+T adds further errors to the
signal (viz. the antenna, LNA, TCXO jitter, etc). I presume these errors
would be on faster time scales, smaller and much more stochastic in nature.
If the resolution of my phase comparator is about 100ps, and keeping in mind
that I want relative stability, wouldn't it make sense to lock the M12+T's
on-board TCXO to the OCXO (probably not straight forward to do)? I realise
that one would lose the advantage of any dithering effect which would
quickly average any zero mean effect. I guess this will depend on the nature
of the errors introduced due to clock jitter: Is it Gaussian and zero mean?
I guess one will have to investigate what happens at down-conversion to IF
etc. And, that ultimately it will depend on the size and nature of noise
caused by the TCXO.
If one could closely follow the drifts in atmospheric effects (which would
be the same for short baselines) one will have very good relative stability.
Regards,
Stephan
From: "Stephan Sandenbergh" stephan@rrsg.ee.uct.ac.za
Subject: [time-nuts] Dithering vs. locking all the clocks to the OCXO?
Date: Tue, 4 Jul 2006 14:49:19 +0200
Message-ID: 002101c69f68$44983a00$401c9e89@Stephan
Hi All,
Earlier, I explained that my application require very good relative
stability between various GPSDOs.
A rough estimate of my requirements is:
-Baselines of 100s of meters to 10s of kilometres.
-Sub-nanosecond relative stability (this I forgot to mention earlier -
thanks to TvB for reminding me).
-Time scales of maybe 100s of seconds to 10s of minutes.
-The lower limit on my stability requirement is maybe the 200ps of jitter
that the FPGA will add to the processed data.
My question is this:
At this stage I'm not sure what all the various causes are for the error in
the 1PPS output of a GPS receiver. (I sure I will find the answer to this in
all that references TvB and Magnus gave me).
However, a quick guess would be the delay caused by atmospheric effects (I
don't think thermal noise would play a big role since the antenna is looking
straight up) Also, there will be errors higher up in the food chain, such as
changes in satellite orbits etc. I guess these errors are fairly systematic.
Actually no, not so much as you might expect. Your receivers are close to
each other in GPS terms, you will certainly experience what is referred to as
GPS common view. The receivers will experience almost identical shifts due to
atmospheric effects since they are so close to each other. You will most
definitly have the same satelites overhead, except where local foilage
prohibits the view of a certain satelite.
Lower down in the food chain, I presume the M12+T adds further errors to the
signal (viz. the antenna, LNA, TCXO jitter, etc). I presume these errors
would be on faster time scales, smaller and much more stochastic in nature.
You should be looking at making sure you have a fairly unobstructed view of the
sky and minimal of ground-reflections. A good antenna could reduce dependence
on reflections. BG made some similar comments privately as we met during the
weekend, and I agree.
If the resolution of my phase comparator is about 100ps, and keeping in mind
that I want relative stability, wouldn't it make sense to lock the M12+T's
on-board TCXO to the OCXO (probably not straight forward to do)?
Infact, BG and I is going to engange in such an experiement, we both had been
playing around with that thought. The details vary depending on which
particular receiver one has, both chipset and software plays a factor.
I do infact have a suitable GPS receiver board here in my hand (thanks to BG!).
I realise that one would lose the advantage of any dithering effect which
would quickly average any zero mean effect.
Dithering effect is what you use if you can't do better. It is one of several
methods. If you have better single-shot resolution you will directly lower the
measurement noise-floor and averaging will just work with you.
I guess this will depend on the nature of the errors introduced due to clock
jitter: Is it Gaussian and zero mean?
Yes and no. Yes, there is a large white and Gaussian noise in there, no, it is
not all Gaussian and as you go down in frequency (i.e. up in time - tau) you
will experience non-Gaussian noises too. However, most of that should come from
the GPS system and not the local clocks, really depends on the clocks and
PLL bandwidth, a topic which have been discussed lately.
I guess one will have to investigate what happens at down-conversion to IF
etc. And, that ultimately it will depend on the size and nature of noise
caused by the TCXO.
Indeed. A good TCXO is needed anyway, low noise is crutial since high noise
will effectively lower the selectivity of the receiver.
If one could closely follow the drifts in atmospheric effects (which would
be the same for short baselines) one will have very good relative stability.
Indeed.
Cheers,
Magnus
Are these permanent installations or portable? If portable, how quickly do
they need to lock up to within sub-nanoseconds?
In a permanent (or semi-permanent) installation, it's hard to beat
GPS-steered cesium, with a loop-tau of DAYS to eliminate all the GPS jitter,
ionosphere effects, etc.
-RL
Robert Lutwak, Senior Scientist
Symmetricom - Technology Realization Center
34 Tozer Rd.
Beverly, MA 01915
(978) 232-1461 Voice RLutwak@Symmetricom.com (Business)
(978) 927-4099 FAX Lutwak@Alum.MIT.edu (Personal)
(339) 927-7896 Mobile
----- Original Message -----
From: "Stephan Sandenbergh" stephan@rrsg.ee.uct.ac.za
To: time-nuts@febo.com
Sent: Tuesday, July 04, 2006 8:49 AM
Subject: [time-nuts] Dithering vs. locking all the clocks to the OCXO?
Hi All,
Earlier, I explained that my application require very good relative
stability between various GPSDOs.
A rough estimate of my requirements is:
-Baselines of 100s of meters to 10s of kilometres.
-Sub-nanosecond relative stability (this I forgot to mention earlier -
thanks to TvB for reminding me).
-Time scales of maybe 100s of seconds to 10s of minutes.
-The lower limit on my stability requirement is maybe the 200ps of jitter
that the FPGA will add to the processed data.
My question is this:
At this stage I'm not sure what all the various causes are for the error
in
the 1PPS output of a GPS receiver. (I sure I will find the answer to this
in
all that references TvB and Magnus gave me).
However, a quick guess would be the delay caused by atmospheric effects (I
don't think thermal noise would play a big role since the antenna is
looking
straight up) Also, there will be errors higher up in the food chain, such
as
changes in satellite orbits etc. I guess these errors are fairly
systematic.
Lower down in the food chain, I presume the M12+T adds further errors to
the
signal (viz. the antenna, LNA, TCXO jitter, etc). I presume these errors
would be on faster time scales, smaller and much more stochastic in
nature.
If the resolution of my phase comparator is about 100ps, and keeping in
mind
that I want relative stability, wouldn't it make sense to lock the M12+T's
on-board TCXO to the OCXO (probably not straight forward to do)? I realise
that one would lose the advantage of any dithering effect which would
quickly average any zero mean effect. I guess this will depend on the
nature
of the errors introduced due to clock jitter: Is it Gaussian and zero
mean?
I guess one will have to investigate what happens at down-conversion to IF
etc. And, that ultimately it will depend on the size and nature of noise
caused by the TCXO.
If one could closely follow the drifts in atmospheric effects (which would
be the same for short baselines) one will have very good relative
stability.
Regards,
Stephan
time-nuts mailing list
time-nuts@febo.com
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
From: "Robert Lutwak" Lutwak@Alum.mit.edu
Subject: Re: [time-nuts] Dithering vs. locking all the clocks to the OCXO?
Date: Tue, 4 Jul 2006 10:34:01 -0400
Message-ID: 000701c69f76$e5219b20$6600a8c0@lutwakhome
Are these permanent installations or portable? If portable, how quickly do
they need to lock up to within sub-nanoseconds?
In a permanent (or semi-permanent) installation, it's hard to beat
GPS-steered cesium, with a loop-tau of DAYS to eliminate all the GPS jitter,
ionosphere effects, etc.
Actually, for these distances, pulling some fibre and do two-way time transfer
should not be too hard. Acheiving sub-nanosecond relative timing should not at
all be unfeasable but should rather be consider fairly easy.
Cheers,
Magnus
Hello,
Please excuse the fact that this is not what we all consider precise
frequency. I am selecting crystals to use for a medium to high volume
application.
The basic criterion are:
- used with a VLSI inverter based oscillator with
3.3 V supply, but somewhat high impedance output. - frequency: 27.000 MHz fundamental within 40 (or so) ppm over
temperature including 7 to 10 years of aging. - HC49S case
- no production line trimming
- low cost
After looking at the design issues, I wonder if some of you
haven't faced similar designs and have some suggestions
regarding the following issues:
-
Is there a free (or inexpensive) spice that works well
simulating crystal oscillators?
(I did see the submissions to this email list last September on LTSpice.)
(I have tried using the SIMetrix Intro package,
and it generally seems to produce good results,
but I found it's awkward or impossible to measure
frequency to lots of precision (taking an FFT helps),
and some results regarding oscillator frequency
shifts as a result of small delay changes in the
inverter seemed questionable to me.) -
Are there some good papers dealing with the practical aspects of:
- transforming the Q limiting element of the design from
a series R to a parallel R - Rs vs. ESR
- exceeding the correlation drive level of a crystal
- measuring the drive level & calculating the maximum drive level
- negative R and oscillator startup margin
- consumer video color subcarrier tolerance
- transforming the Q limiting element of the design from
-
Do you have any suggested suppliers for crystals with the
following specifications:- HC49S
- 27.000 MHz
- 10 ppm tolerance
- 10 ppm stability
- 18 pF load
- trim sensitivity less than 15 ppm/pF
- 7 (or better yet 10) year aging less than 10 ppm
- drive level up to 1 mW (may not need to be this high if
the ESR is < 30 ohms or I better understand our maximum
inverter output voltage)
If someone is interested in providing comments on the app note
I'm working on, that would also be appreciated.
Thanks
Stan Searing
However, a quick guess would be the delay caused by atmospheric effects (I
don't think thermal noise would play a big role since the antenna is
looking
straight up).
Seems to me that thermal noise depends only on the resistive
impedance of the antenna and input circuit. See Johnson
noise. Orientation of the antenna would affect reception of
external noise radiation. Don't think cosmic background
radiation is a problem for the usual GPS ground antenna.
What causes the "atmospheric effects"? I'd expect radiation to
slow down a bit as it passes through water. Are there heat
effects as well? Does the density of the atmosphere change
enough to make a 10E-10 second delay possible? If so, it
seems there would be non-negligible changes due to turbulence
over a kilometer or so.
I have two each HP 58532A antenna, 58535A distribution amplifier,
and Z3801A receiver connected to a laptop running a common (not
HP) program that provides receiver status reports. The antennae
are mounted 4 feet apart on a sturdy pole (6 inch plastic pipe).
There are nearby trees above the pole. One has to live with one's
mate when it comes to unnatural things that are visible from the
street.
I ran it for a while and got diverted to other matters, but I
remember seeing relative variations in altitude and position.
A Racal-Dana 1992 counter set for phase angle showed a constantly
increasing phase difference between the receivers, with readings
taken hours apart. I have not automated data collection yet.
Is this setup able to quantify atmospheric effects or are there
too many system errors?
Regards,
Bill Hawkins
On Tue, July 4, 2006 19:18, Bill Hawkins said:
What causes the "atmospheric effects"? I'd expect radiation to
slow down a bit as it passes through water. Are there heat
effects as well? Does the density of the atmosphere change
enough to make a 10E-10 second delay possible? If so, it
seems there would be non-negligible changes due to turbulence
over a kilometer or so.
Largest error source for single freq receivers are ionospheric errors,
because of time varying TEC. (The very same effect that makes some
radiowaves sometimes circle the globe.) The next important error source is
the trophosphere, where water vapor content delays the signal somewhat
unpredictable.
--
Björn
The basic criterion are:
- used with a VLSI inverter based oscillator with
3.3 V supply, but somewhat high impedance output.
An inverter is not specified for oscillator duty.
It cannot be analyzed for this application on SPICE.
It will never be a high precision oscillator circuit.
(10 ppm is "high precision" when it comes to using inverters).
If you care about performance, use a transistor to make
the oscillator and use the inverter to convert the
sine wave to logic levels. If you "cannot" use a transistor,
relax your specs about an order of magnitude.
Rick Karlquist
On 7/4/06, Magnus Danielson cfmd@bredband.net wrote:
Actually, for these distances, pulling some fibre and do two-way time transfer
should not be too hard. Acheiving sub-nanosecond relative timing should not at
all be unfeasable but should rather be consider fairly easy.
Do you mean easy for fibre or for a GPSDO?
Regards,
Stephan.
From: "Bill Hawkins" bill@iaxs.net
Subject: Re: [time-nuts] Dithering vs. locking all the clocks to the OCXO?
Date: Tue, 4 Jul 2006 12:18:34 -0500
Message-ID: 006901c69f8d$e1616760$0500a8c0@darius.domain.actdsltmp
However, a quick guess would be the delay caused by atmospheric effects (I
don't think thermal noise would play a big role since the antenna is
looking
straight up).
Seems to me that thermal noise depends only on the resistive
impedance of the antenna and input circuit. See Johnson
noise. Orientation of the antenna would affect reception of
external noise radiation. Don't think cosmic background
radiation is a problem for the usual GPS ground antenna.
There will indeed be a difference due to orientation, as will the weather.
If you have unobstructive clear sky condition by night, you will have the
background temperature of the universe, about 2,7 K. If you have clouds, they
are way hotter and thus will contribute more noise as a result. This is the
difference between direct parabola antennas and offset parabolas, the edges
contribute less noise into the antenna in the offset antenna since it sees
clear sky - if you mounted it properly.
What causes the "atmospheric effects"? I'd expect radiation to
slow down a bit as it passes through water.
Not much. One of the interested parties was the US marine, and their vessels
see a fair amount of water. The selection of the L-band was done with this in
mind.
Are there heat
effects as well? Does the density of the atmosphere change
enough to make a 10E-10 second delay possible? If so, it
seems there would be non-negligible changes due to turbulence
over a kilometer or so.
The really big error is the ionosphere, where the amount of ionised atoms will
cause a shift in phase.
I have two each HP 58532A antenna, 58535A distribution amplifier,
and Z3801A receiver connected to a laptop running a common (not
HP) program that provides receiver status reports. The antennae
are mounted 4 feet apart on a sturdy pole (6 inch plastic pipe).
There are nearby trees above the pole. One has to live with one's
mate when it comes to unnatural things that are visible from the
street.
I ran it for a while and got diverted to other matters, but I
remember seeing relative variations in altitude and position.
A Racal-Dana 1992 counter set for phase angle showed a constantly
increasing phase difference between the receivers, with readings
taken hours apart. I have not automated data collection yet.
Is this setup able to quantify atmospheric effects or are there
too many system errors?
For me it sounds like you are measuring difference in local reflections and
obstructions.
Cheers,
Magnus