Home made GPS disciplined atomic clock
Hello everyone
Has anyone tried to replace Tbolt's OCXO with Efratom LPRO
Rubidium oscillator?
I just did that it seems to work, but I have no idea is it really better
or worse (or same) as original Thunderbolt. No ideas how to measure it.
Is it possible to measure allan deviation or such things by using the
logs generated by Tboltmon.exe?
So it would be nice to know if anyone has tried that before and done
some measurement against cesium or such standard to get exact results.
I think that it should be a much better (in theory) than OCXO which
comes short therm stability (what I'm actually seeking for). It should
be much more accurate with long holdovers also.
This is very simple modification by the way. Infact my original plan was
to use the 1PPS to synchronize the LPRO C-field with separate control
electronics (based own design with PIC and 24-bit DAC). But then I
noticed with spectrum analyzer measurements that the Tbolt's OCXO has
almost exactly same output level than LPRO: about +7 dBm and of course
they both have 10 MHz output frequency. So it would be possible to
replace OCXO with LPRO without any level matching or so... Also, with
changing the disciplining settings (DAC voltages) even the control
voltage of OCXO can be fed directly to LPRO's C-field without any extra
electronics. This enables to use Tbolt's advanced steering algorithms
for C-field control without any own programming work here.
About the settings: Ko (Hz/V) for LPRO is about 0.006 Hz/V (if
calculated correctly from datasheet values) and minimum value what
Thunderbolt accepts for this is 0.01. However even 0.1 seems to work
(may be even better, causes smaller changes of DAC voltage). Min voltage
must be set to 0.0V, because it's not allowed to feed any negative
voltages to LPRO's C-field. However the LPRO manual says that it will
not break with negative voltages under 8V. Initial DAC Voltage should
be set somewhere about 2.5 volts, I run it some time with GPS lock and
noticed that good initial voltage with my LPRO was 2.56. Loop Dynamics
is big mystery. I'd like to think that very slow loop should be good for
rubidium. So I tried 1000 secs as a test. But if anyone has tried this
before could give some advice about loop dynamics?
There are even places for SMA-connectors of Thunderbolt PCB. One is the
10 MHz and another is the control voltage. After doing the modification
it may be a good idea to start with factory settings to reset any
"learned" data for OCXO and right after that reset change the DAC
voltage settings.
So... it seems to work. But it's quite difficult to tell anything of
it's performance. I just logged it for couple of hours (started after it
has been couple of hours up). Within that very short test time it seems
that at least the PPS offset value stays now inside same ns value at
longer times than with OCXO. With OCXO it changed many ns between
readings, quite randomly - but with rubidium the change is usually
couple of hundred picoseconds. Could this be a sign of better short
therm drift / random walk performance?
Here's a link for the log:
http://www.amigazone.fi/files/gpsdo/tbolt-lpro-test.log
(Log format: TOW, PPS offset, DAC voltage, Disciplining mode & activity)
The log was created with following settings:
Time Constant: 1000
Damping factor: 1.2
Ko: 0.01
Initial DAC volt: 2.558V
--
73's
Esa
OH4KJU
I think that it should be a much better (in theory) than OCXO which
comes short therm stability (what I'm actually seeking for). It should
be much more accurate with long holdovers also.
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse. So do not replace the TBolt
OCXO with a LPRO if short-term stability is your goal. See:
TBolt OCXO plots:
http://www.leapsecond.com/pages/gpsdo/
http://www.leapsecond.com/pages/tbolt-tc/
LPRO plots:
http://www.leapsecond.com/museum/lpro/
However, if long-term, GPS-unlocked, holdover performance
is the goal, then using a Rb would make a good choice.
This is very simple modification by the way. Infact my original plan was
to use the 1PPS to synchronize the LPRO C-field with separate control
...
See John Miles work to replace the Thunderbolt OCXO:
http://www.thegleam.com/ke5fx/tbolt.htm
/tvb
Here's a link for the log:
http://www.amigazone.fi/files/gpsdo/tbolt-lpro-test.log
(Log format: TOW, PPS offset, DAC voltage, Disciplining mode & activity)
I'll have a look at this; but it's not accessible for some reason.
/tvb
The challenge is to detect a failure of the GPS source (LOS) before the DPLL
moves the OCXO.
I used to design Stratum clocks for a large telecom company, and I used
several trick do detect a phase ramp on the digital phase detector; this was
used to declare a probable bad source. At that point, we halted the
movement of the DPLL and observed the phase detector activity. We had two
DPLLs, and if both detected a phase ramp, we declared the source bad.
-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com]On
Behalf Of Tom Van Baak
Sent: Sunday, January 25, 2009 11:03 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Home made GPS disciplined atomic clock
I think that it should be a much better (in theory) than OCXO which
comes short therm stability (what I'm actually seeking for). It should
be much more accurate with long holdovers also.
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse. So do not replace the TBolt
OCXO with a LPRO if short-term stability is your goal. See:
TBolt OCXO plots:
http://www.leapsecond.com/pages/gpsdo/
http://www.leapsecond.com/pages/tbolt-tc/
LPRO plots:
http://www.leapsecond.com/museum/lpro/
However, if long-term, GPS-unlocked, holdover performance
is the goal, then using a Rb would make a good choice.
This is very simple modification by the way. Infact my original plan was
to use the 1PPS to synchronize the LPRO C-field with separate control
...
See John Miles work to replace the Thunderbolt OCXO:
http://www.thegleam.com/ke5fx/tbolt.htm
/tvb
Here's a link for the log:
http://www.amigazone.fi/files/gpsdo/tbolt-lpro-test.log
(Log format: TOW, PPS offset, DAC voltage, Disciplining mode & activity)
I'll have a look at this; but it's not accessible for some reason.
/tvb
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.
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse.
Basicly I'm seeking an accurate frequency standard for RF lab. It should
be always as accurate as possible, regardless the state of GPS receiving
etc.
Before doing this modification I did some test runs Trimble versus LPRO
with phase comparator circuit. I noticed that Trimble is accurate as
long as it gets the GPS signal and phase change between LPRO and Trimble
was changing evenly. It is even accurate after the GPS drops (holdover
mode) but after the signal comes back the things start to go badly
wrong. It starts to roll it's phase / 1 PPS back to alignment woth GPS
time and this caused very badly looking phase activity when compared to
LPRO.
Another bad issue was that if there's a change in satellite receiving
(satellite hopping or some) it causes rapid change the PPS offset and
OCXO frequency starts to roll to get the 1 PPS back to alignment. So it
seems that Trimble's main principle is 10000000 pulses per PPS, with no
exceptions and when the PPS goes off the 10 MHz must also go off to get
the 1 PPS back to aligment. So there's no constant 10 MHz frequency
either. That's not acceptable because in normal use I should be always
aware of GPS receiving states - I'd just like to trust that I'm getting
accurate 10 MHz - any time!
So I become to think that may be very slow loop dynamics will solve that
problem (if the DAC value isn't changed at every little change at
satellite reception). And for that purpose the rubidium sound better
than OCXO.
I also got misunderstanding from this:
http://www.ptsyst.com/GPS10RB-B.pdf
It claims that rubidiums will have good short therm drift.
My problem here is that there's no way to measure the different setups
because my only rb is now part of the experiment. All I can do is the
log them and look the change between PPS timing offset readings.
When doing the GPS vs. LPRO phase comparison told before I noticed that
the changes of PPS offsets are correlated the phase changes between LPRO
and Tbolt output, when observed quite short time. So it seems that the
PPS offset is somehow accurate measurement of oscillator stability as well.
I also done some noise measurements with spectrum analyzer between LRPO
and Trimble outputs. LPRO had lower noise floor around fundamental.
See John Miles work to replace the Thunderbolt OCXO:
http://www.thegleam.com/ke5fx/tbolt.htm
Hmm. May be the OCXO on my tbolt is then somehow bad if the LPRO should
be even worse? It has Trimble label on and the unit is manufactured on
2005, in China.
Is there any logs available with that better OCXO? It would be nice to
see the PPS offsets variance between readings with that oscillator.
I'll have a look at this; but it's not accessible for some reason.
Oops.. Now you should get it.
--
73s!
Esa
OH4KJU
Esa
You can change the Thunderbolt recovery mode from holdover.
One test you can perform that should give an indication of the location
of the Allan intercept is to:
-
connect the receiver to an antenna.
-
let it run for a few days so the Kalman filter learns the drift,
tempco and other parameters. -
manually disable the disciplining leaving the thunderbolt connected
to the antenna. -
Log the Thunderbolt PPS offset (plus time stamp) for a day or more.
-
Analyse the resultant data to determine the relative Allan deviation
between the receiver and the 10MHz source.
Ulrich's Plotter is good for this (use the overlapping ADEV algorithm -
although TOTDEV and Theo1 are even better estimators of the Allan deviation)
All going well, you will see a minimum in the Allan deviation versus tau
plot.
In most cases the value of tau at the minimum will be relatively close
to the value of tau at the Allan intercept.
However to do this successfully your antenna will need a good view of
the sky.
For short tau the GPS receiver noise will dominate.
For long tau the 10MHz source noise and drift will dominate.
Bruce
Esa Heikkinen wrote:
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse.
Basicly I'm seeking an accurate frequency standard for RF lab. It should
be always as accurate as possible, regardless the state of GPS receiving
etc.
Before doing this modification I did some test runs Trimble versus LPRO
with phase comparator circuit. I noticed that Trimble is accurate as
long as it gets the GPS signal and phase change between LPRO and Trimble
was changing evenly. It is even accurate after the GPS drops (holdover
mode) but after the signal comes back the things start to go badly
wrong. It starts to roll it's phase / 1 PPS back to alignment woth GPS
time and this caused very badly looking phase activity when compared to
LPRO.
Another bad issue was that if there's a change in satellite receiving
(satellite hopping or some) it causes rapid change the PPS offset and
OCXO frequency starts to roll to get the 1 PPS back to alignment. So it
seems that Trimble's main principle is 10000000 pulses per PPS, with no
exceptions and when the PPS goes off the 10 MHz must also go off to get
the 1 PPS back to aligment. So there's no constant 10 MHz frequency
either. That's not acceptable because in normal use I should be always
aware of GPS receiving states - I'd just like to trust that I'm getting
accurate 10 MHz - any time!
So I become to think that may be very slow loop dynamics will solve that
problem (if the DAC value isn't changed at every little change at
satellite reception). And for that purpose the rubidium sound better
than OCXO.
I also got misunderstanding from this:
http://www.ptsyst.com/GPS10RB-B.pdf
It claims that rubidiums will have good short therm drift.
My problem here is that there's no way to measure the different setups
because my only rb is now part of the experiment. All I can do is the
log them and look the change between PPS timing offset readings.
When doing the GPS vs. LPRO phase comparison told before I noticed that
the changes of PPS offsets are correlated the phase changes between LPRO
and Tbolt output, when observed quite short time. So it seems that the
PPS offset is somehow accurate measurement of oscillator stability as well.
I also done some noise measurements with spectrum analyzer between LRPO
and Trimble outputs. LPRO had lower noise floor around fundamental.
See John Miles work to replace the Thunderbolt OCXO:
http://www.thegleam.com/ke5fx/tbolt.htm
Hmm. May be the OCXO on my tbolt is then somehow bad if the LPRO should
be even worse? It has Trimble label on and the unit is manufactured on
2005, in China.
Is there any logs available with that better OCXO? It would be nice to
see the PPS offsets variance between readings with that oscillator.
I'll have a look at this; but it's not accessible for some reason.
Oops.. Now you should get it.
One test you can perform that should give an indication of the location
of the Allan intercept is to:
Ok, thanks for your clear instructions!
My test periods have been much too short, if the Kalman filter learning
takes even days! But with these instructions I'lll get better data for
OCXO vs. LPRO comparison and maybe also the OCXO health.
Ulrich's Plotter is good for this
Hmm. Is that software available somewhere?
No luck with quick Google tour...
However to do this successfully your antenna will need a good view of
the sky.
And that's also one of my problems here. Many trees in the yard. No
problems with normal "hand GPS" reception but when it comes to these
timing systems this could be one explanation of these strange timing
changes at satellite "hops" already noted. However the antenna sees most
of the sky clearly but not so close to horizon. Will the Northen
position (Lat 62.33302) also cause inaccuracy to GPS?
--
73s!
Esa
OH4KJU
Google df6jb plotter.
It's great!
73 de Norm n3ykf
Esa Heikkinen wrote:
One test you can perform that should give an indication of the location
of the Allan intercept is to:
Ok, thanks for your clear instructions!
My test periods have been much too short, if the Kalman filter learning
takes even days! But with these instructions I'lll get better data for
OCXO vs. LPRO comparison and maybe also the OCXO health.
Ulrich's Plotter is good for this
Hmm. Is that software available somewhere?
No luck with quick Google tour...
However to do this successfully your antenna will need a good view of
the sky.
And that's also one of my problems here. Many trees in the yard. No
problems with normal "hand GPS" reception but when it comes to these
timing systems this could be one explanation of these strange timing
changes at satellite "hops" already noted. However the antenna sees most
of the sky clearly but not so close to horizon. Will the Northen
position (Lat 62.33302) also cause inaccuracy to GPS?
Esa
You'll need a good view of the sky to the south where the GPS SVs will
be located.
You'll also need to set the elevation mask appropriately.
Multipath will be more problematic with low elevation SVs.
It would also be helpful if you plot the SV tracks across the sky (as
seen by a GPS receiver) as this will show if and where obstructions are
significant.
There's a lot of software out there for doing this.
Bruce
Hello again...
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse. So do not replace the TBolt
OCXO with a LPRO if short-term stability is your goal. See:
I'm wondering what could be the cause of this. According to operating
manual LPRO's output should be crystal oscillator (VCXO) generated
signal, which is synchronized to rubidium. So why it is so much worse
than any other crystal oscillator (or other Rd oscillators). Are there
any schematics for LPRO available anywhere?
I cannot see the any phase noise difference between Trimble's OCXO and
LPRO with spectrum analyser. Measured with different spans from 500 kHz
to 200 Hz, using resolution bandwiths 300 Hz to 6 Hz. So the noise which
is causing bad short term drift must be very close to fundamental.
It seems that only way to see this noise is to use phase detector
circuit but my problem with that is that I haven't got any good
reference for it and this kind of equipments are quite hard to find here
in Finland. It would be nice to see what kind of noise there are, to
design the filter bandwith for external OCXO lock circuit.
Other idea to bring that noise visible could be multiplying it with some
kind of comb generator circuit (might be hard to build one). Then it
would be possible to measure it's harmonics. Not sure if there's enough
level present anymore at GHz frequencies...
What kind of test setup did you use when getting this result:
LPRO plots:
http://www.leapsecond.com/museum/lpro/
--
73s!
Esa
OH4KJU
Esa
Wenzel has a introduction to low cost phase noise measurement at:
http://www.wenzel.com/documents/measuringphasenoise.htm
Its relatively easy to assemble such a system.
A PC sound card can be used as a spectrum analyser for measuring phase
noise to within a few Hz of the carrier.
There are low noise amplifiers with lower flicker noise than Wenzel's
low noise FET input audio amplifier
http://www.wenzel.com/pdffiles1/pdfs/lowamp.pdf.
It is also possible to build a variant of Wenzel's amplifier that doesnt
require selection of JFETs.
If one wishes to measure the Allan variance of an oscillator, the 3
cornered hat technique can be used with 3 oscillators having similar
performance.
For longer tau (1000sec or more depending on the OCXO being
characterised) the PPS output of a good gps timing receiver can be used
as a reference.
However a clear view of the southern sky is required,
Bruce
Esa Heikkinen wrote:
Hello again...
Right, it all depends on what stability you're after. The OCXO
will have much better short-term stability than the LPRO -- the
LPRO is close to ten times worse. So do not replace the TBolt
OCXO with a LPRO if short-term stability is your goal. See:
I'm wondering what could be the cause of this. According to operating
manual LPRO's output should be crystal oscillator (VCXO) generated
signal, which is synchronized to rubidium. So why it is so much worse
than any other crystal oscillator (or other Rd oscillators). Are there
any schematics for LPRO available anywhere?
I cannot see the any phase noise difference between Trimble's OCXO and
LPRO with spectrum analyser. Measured with different spans from 500 kHz
to 200 Hz, using resolution bandwiths 300 Hz to 6 Hz. So the noise which
is causing bad short term drift must be very close to fundamental.
It seems that only way to see this noise is to use phase detector
circuit but my problem with that is that I haven't got any good
reference for it and this kind of equipments are quite hard to find here
in Finland. It would be nice to see what kind of noise there are, to
design the filter bandwith for external OCXO lock circuit.
Other idea to bring that noise visible could be multiplying it with some
kind of comb generator circuit (might be hard to build one). Then it
would be possible to measure it's harmonics. Not sure if there's enough
level present anymore at GHz frequencies...
What kind of test setup did you use when getting this result:
LPRO plots:
http://www.leapsecond.com/museum/lpro/
Its relatively easy to assemble such a system.
A PC sound card can be used as a spectrum analyser for measuring phase
noise to within a few Hz of the carrier.
I still need some high quality reference oscillator. Do you have any
clue how much those Wenzel oscillators cost? There wasn't any prices on
website, may be the only way is to ask?
There seems to be interesting alternatives for the output oscillator too
(to clean the LPRO signal). One of those was even named "timekeeper".
Maybe it could be wise to buy one and use it for phase noise measurement
first and then put it in permament use as the output oscillator of the
reference, when the desired loop bandwith is known.
But of course if those units has price tag with four figures or so then
this is only daydreaming... :-)
--
73s!
Esa
OH4KJU
-----Original Message-----
From: time-nuts-bounces@febo.com
[mailto:time-nuts-bounces@febo.com] On Behalf Of Esa Heikkinen
Sent: Tuesday, January 27, 2009 2:46 PM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Home made GPS disciplined atomic clock
Its relatively easy to assemble such a system.
A PC sound card can be used as a spectrum analyser for
measuring phase
noise to within a few Hz of the carrier.
I still need some high quality reference oscillator. Do you
have any clue how much those Wenzel oscillators cost? There
wasn't any prices on website, may be the only way is to ask?
The run of the mill Wenzel "Streamline" series runs about $250 each.
http://www.wenzel.com/pdffiles1/Oscillators/STR_4_to_30.pdf
There seems to be interesting alternatives for the output
oscillator too (to clean the LPRO signal). One of those was
even named "timekeeper".
Wenzel sells "cleanup loops" as a packaged device for a variety of frequencies.
Maybe it could be wise to buy one and use it for phase noise
measurement first and then put it in permament use as the
output oscillator of the reference, when the desired loop
bandwith is known.
But of course if those units has price tag with four figures
or so then this is only daydreaming... :-)
I would imagine a packaged cleanup loop (oscillator and PLL) is right around $1K. I can't recall what we paid for our 10MHz widgets a few years back, but that seems about right.
Clearly, one can build it yourself for less (e.g. you could use a $250 oscillator and use a PLL eval board of some sort to drive the EFC input, for instance) , but that's only if your time is free.
Send Wenzel an email and ask..
Esa,
I'm wondering what could be the cause of this. According to operating
manual LPRO's output should be crystal oscillator (VCXO) generated
signal, which is synchronized to rubidium.
So it is! But the detection process of the "atomic lock" is "noisy"
itself. That makes it necessary to have low pass filters in this pll
that make the overall noise dependend from the xtal oscillator's noise
at short observation times. If you consider that a typical "normal" xtal
oscillator (no oven, no temperature compensation) has an Allan deviation
of 5E-8 @ 1 s then you see that 1E-11 @ 1 s IS ALREADY a BIG improvement
above the "normal" case. 1E-12 @ 1 s is near the best that amateur money
can buy. A lot of high grade OCXOs will be >1E-12 and <1E-11 @ 1 s and
VERY FEW will be <1E-12 @ 1 s
73s Ulrich, DF6JB
-----Ursprungliche Nachricht-----
Von: time-nuts-bounces@febo.com
[mailto:time-nuts-bounces@febo.com] Im Auftrag von Esa Heikkinen
Gesendet: Dienstag, 27. Januar 2009 17:34
An: Discussion of precise time and frequency measurement
Betreff: Re: [time-nuts] Home made GPS disciplined atomic clock
Hello again...
Right, it all depends on what stability you're after. The OCXO will
have much better short-term stability than the LPRO -- the LPRO is
close to ten times worse. So do not replace the TBolt OCXO
with a LPRO
if short-term stability is your goal. See:
I'm wondering what could be the cause of this. According to operating
manual LPRO's output should be crystal oscillator (VCXO) generated
signal, which is synchronized to rubidium. So why it is so much worse
than any other crystal oscillator (or other Rd oscillators).
Are there
any schematics for LPRO available anywhere?
I cannot see the any phase noise difference between Trimble's
OCXO and
LPRO with spectrum analyser. Measured with different spans
from 500 kHz
to 200 Hz, using resolution bandwiths 300 Hz to 6 Hz. So the
noise which
is causing bad short term drift must be very close to fundamental.
It seems that only way to see this noise is to use phase detector
circuit but my problem with that is that I haven't got any good
reference for it and this kind of equipments are quite hard
to find here
in Finland. It would be nice to see what kind of noise there are, to
design the filter bandwith for external OCXO lock circuit.
Other idea to bring that noise visible could be multiplying
it with some
kind of comb generator circuit (might be hard to build one). Then it
would be possible to measure it's harmonics. Not sure if
there's enough
level present anymore at GHz frequencies...
What kind of test setup did you use when getting this result:
LPRO plots:
http://www.leapsecond.com/museum/lpro/
--
73s!
Esa
OH4KJU
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.