GPS-based reference clock

[Andrey]

Hello All,

How can I make reference oscillator (clock) out of generic consumer GPS
Receiver?
I hope to get stability 10-8 seconds over 8 hours.

There is off-the-shelf equipment doing just that, but prices are way over my
budget.

Appreciate your advices.

Sincerely,

Andrey Gleener

 

[Tim Shoppa]

Hello All,

How can I make reference oscillator (clock) out of generic consumer GPS
Receiver?
I hope to get stability 10-8 seconds over 8 hours.

The standard reference on this kind of project is "A GPS based
Frequency Standard", Brooks Shera, QST July 1998, pp 37-44. I think the
article is available on the net somewhere, and if you Google you will
find many many examples of implementations.

Most variations start with a GPS with PPS output, a OCXO with
electronic tweaking (e.g. HP 10811), and a microprocessor with
accompanying circuitry to measure the phase difference and tweak the
oscillator. For even better stability rubidium standards are available
in the low k$ range.

There is off-the-shelf equipment doing just that, but prices are way over my
budget.

Z3801A's were commonly available surplus in the $300 range a few years
ago. They aren't as common now but I thijnk you can find them for just
a little more. Certainly there are non-Z3801A units coming off of
cellphone sites regularly.

Tim.

 

[budgie]

The standard reference on this kind of project is "A GPS based
Frequency Standard", Brooks Shera, QST July 1998, pp 37-44. I think the
article is available on the net somewhere, and if you Google you will
find many many examples of implementations.

(snip)

That article desribes a GPS-conditioned VCXO using the 1pps output from a GPS
receiver ("GPSR)". This requires long conditioning times, in the order of
hours, before real long-term performance is achieved.

There are numerous GPSR that provide a 10kHz output which facilitates a far
shorter settling time in a hardware-only device, resulting in a much simpler
configuration.

 

[artie]

Hello All,

How can I make reference oscillator (clock) out of generic consumer GPS
Receiver?
I hope to get stability 10-8 seconds over 8 hours.

There is off-the-shelf equipment doing just that, but prices are way over my
budget.

Appreciate your advices.

Sincerely,

Andrey Gleener

Tim pointed to the Z8301A -- still available on eBay, for $300 to $400,
and still very accurate. They have a GPS receiver and a double-oven
VCXO.

 

[Allan Herriman]

(snip)

That article desribes a GPS-conditioned VCXO using the 1pps output from a GPS
receiver ("GPSR)". This requires long conditioning times, in the order of
hours, before real long-term performance is achieved.

There are numerous GPSR that provide a 10kHz output which facilitates a far
shorter settling time in a hardware-only device, resulting in a much simpler
configuration.

.... and worse phase noise.

However, the OP is only interested in "stability 10-8 seconds over 8
hours" so this may not matter.

Regards,
Allan

 

[budgie]

... and worse phase noise.

I have no intention of entering a debate on that, but my jury is still out.

However, the OP is only interested in "stability 10-8 seconds over 8
hours" so this may not matter.

Indeed.

 

[Andrey]

Thank you all very much for the advices.

I checked for Z8301A on EBay did not find it so far. I will keep looking.
far
shorter settling time in a hardware-only device, resulting in a much simpler
configuration.

This is also interesting statement unfortunately not supported by specific
references. For my purposes 10 MHz is preferable output.

Sincerely,

Andrey

 

[budgie]

Thank you all very much for the advices.

I checked for Z8301A on EBay did not find it so far. I will keep looking.

far
shorter settling time in a hardware-only device, resulting in a much simpler
configuration.

This is also interesting statement unfortunately not supported by specific
references. For my purposes 10 MHz is preferable output.

I'm sure it is, as it often is for others including me. I originally built a
10MHz TV-derived frequency reference, but was never entirely happy with that.
With the impending demise of analog TV here, I looked at various alternatives.
Eventually I modified the TV-derived scheme to use the GPSR 10kHz output. and
built my own 10MHz GPS-derived frequency reference.

I'm not aware of any GPSR that have a 10MHz output, but there are numerous
projects very similar to mine (which I only discovered after reinventing the
wheel). One by James Miller G3RUH is fully documented at:

http://www.jrmiller.demon.co.uk/projects/ministd/frqstd.htm

 

[Slurp]

Hello All,

How can I make reference oscillator (clock) out of generic consumer GPS
Receiver?
I hope to get stability 10-8 seconds over 8 hours.

There is off-the-shelf equipment doing just that, but prices are way over
my
budget.

Appreciate your advices.

Sincerely,

Andrey Gleener

A simple 50ppm tolerance crystal oscillator (less than a dollar) will meet
your requirements.

At 50ppm error it will drift about 1.5 seconds in 8 hours

Slurp

 

[Ancient_Hacker]

I suspect he meant "ten to the minus eighth power", not "ten minus
two".

 

[Slurp]

I suspect he meant "ten to the minus eighth power", not "ten minus
two".

Ahh OK, I thought he said 10 or 8 seconds, not 10^-8!!!!!

Major problem even with GPS as you will only get an accuracy of ~10^-7
seconds absolute anyway without very expensive kit.

Only option is a phase tracked Rubidium - or if you have megabucks a
caesium!

 

[Tim Shoppa]

Thank you all very much for the advices.

I checked for Z8301A on EBay did not find it so far. I will keep looking.

There are other similar OCXO and rubidium clock devices out there on
the surplus market. If you look at http://www.leapsecond.com/, the
Z3801A-related stuff at http://www.realhamradio.com/, and check out the
time-nuts mailing list (@febo.com) then you will see what's out there
for the picking.

The surplus-timing-source market isn't quite as plentiful as in the
year or two after the telecom crash, but there is at least a trickle.

Your 10^-8 second stability over 8 hours actually does require some
clarification: Do you need 10ns stability after 8 hours of unlock? At
that level OCXO stuff definitely isn't good enough and you'll have to
go to rubidium or cesium, and while rubidium stuff does show up at the
low k$ level new and less than that new, lucking across a cesium
reference for that little money will take extremely good fortune.

If you only meant 1 part out of 10^8, that would be 36 microseconds
over 8 hours, and that is easily done with an OCXO.

10ns precision over any number of hours is going to be pushing the edge
with a GPS-locked device, because these typically do not provide that
sort of precision at the PPS output. Typical $30-$50 range GPS modules
only guarantee under a microsecond of jitter (and typically achieve
100ns) on their PPS output. There are ones that go to the under-100ns
jitter guaranteed level but I don't know of any that promise to the
10ns level. (Some may deliver circa 10ns level under kid-glove
conditions although it's not guaranteed.)

If you need absolute time at the 10ns level from GPS then you need a
truly excellent survey, even then I think it's pushing the edge.

far
shorter settling time in a hardware-only device, resulting in a much simpler
configuration.

This is also interesting statement unfortunately not supported by specific references. For my purposes 10 MHz is preferable output.

I suspect that he misinterpreted your stability requirements, which I
think you could spend a little time clarifying. There are
GPS-locked-oscillator applications which do require the fast lock time
that he was addressing but I don't think your application falls in that
range.

Tim.

 

[Ken Taylor]

looking.

There are other similar OCXO and rubidium clock devices out there on
the surplus market. If you look at http://www.leapsecond.com/, the
Z3801A-related stuff at http://www.realhamradio.com/, and check out the
time-nuts mailing list (@febo.com) then you will see what's out there
for the picking.

The surplus-timing-source market isn't quite as plentiful as in the
year or two after the telecom crash, but there is at least a trickle.

Your 10^-8 second stability over 8 hours actually does require some
clarification: Do you need 10ns stability after 8 hours of unlock? At
that level OCXO stuff definitely isn't good enough and you'll have to
go to rubidium or cesium, and while rubidium stuff does show up at the
low k$ level new and less than that new, lucking across a cesium
reference for that little money will take extremely good fortune.

If you only meant 1 part out of 10^8, that would be 36 microseconds
over 8 hours, and that is easily done with an OCXO.

10ns precision over any number of hours is going to be pushing the edge
with a GPS-locked device, because these typically do not provide that
sort of precision at the PPS output. Typical $30-$50 range GPS modules
only guarantee under a microsecond of jitter (and typically achieve
100ns) on their PPS output. There are ones that go to the under-100ns
jitter guaranteed level but I don't know of any that promise to the
10ns level. (Some may deliver circa 10ns level under kid-glove
conditions although it's not guaranteed.)

If you need absolute time at the 10ns level from GPS then you need a
truly excellent survey, even then I think it's pushing the edge.

specific references. For my purposes 10 MHz is preferable output.

I suspect that he misinterpreted your stability requirements, which I
think you could spend a little time clarifying. There are
GPS-locked-oscillator applications which do require the fast lock time
that he was addressing but I don't think your application falls in that
range.

Tim.

You can pick up a Ball-Efratom rubidium standard on Ebay easily enough; I
think even their older GPS units are available from time to time. Failing
that an HP via the same source - they all pop up from time to time.

Cheers.

Ken

 

[Allan Herriman]

I have no intention of entering a debate on that, but my jury is still out.

Oops, I believe I misinterpreted your post.

You are suggesting that locking an external VCXO to the 10kHz output
instead of the 1pps output will give better results. That is a good
suggestion, both in terms of implementation ease and phase noise
performance.

Regards,
Allan

 

[budgie]

Oops, I believe I misinterpreted your post.

You are suggesting that locking an external VCXO to the 10kHz output
instead of the 1pps output will give better results. That is a good
suggestion, both in terms of implementation ease and phase noise
performance.

It certainly provides implementation ease, as shown in the G3RUH project I
referenced elsewhere in this thread. It also provides for incredibly good cold
start capability, with final stability within three minutes of power application
(even using 1998 ROM data).

When your *reference* frequency is from a division chain , then the higher up
that chain you take the feed to the phase comparator the better the loop
parameters can be and the better the phase noise also. In the case of the GPSR
I employ (Conexant Jupiter) the 1pps rising edge is cited as being coincident
with the edge of the 10k stream. But whether the 10k stream is simply a local
loop locked to the 1pps or a higher point in a division chain which produces the
1pps is not explained by the data sheet, and Navman (who acquired the
Rockwell-Conexant product line) decline to provide any support for the product.

Regarding phase noise, I started a thread under the heading "Jitter measurement"
about three weeks ago in this group because I wanted to quantify that area of my
unit.

In the case of my GPS-derived 10MHz unit using the 10khz stream, comparisons
against the 10MHz OCXO in my frequency counter have been very promising,
although I am unable to quantify the performance.

 

[Joseph2k]

It certainly provides implementation ease, as shown in the G3RUH project I
referenced elsewhere in this thread. It also provides for incredibly good
cold start capability, with final stability within three minutes of power
application (even using 1998 ROM data).

When your *reference* frequency is from a division chain , then the higher
up that chain you take the feed to the phase comparator the better the
loop
parameters can be and the better the phase noise also. In the case of
the GPSR I employ (Conexant Jupiter) the 1pps rising edge is cited as
being coincident
with the edge of the 10k stream. But whether the 10k stream is simply a
local loop locked to the 1pps or a higher point in a division chain which
produces the 1pps is not explained by the data sheet, and Navman (who
acquired the Rockwell-Conexant product line) decline to provide any
support for the product.

Regarding phase noise, I started a thread under the heading "Jitter
measurement" about three weeks ago in this group because I wanted to
quantify that area of my unit.

In the case of my GPS-derived 10MHz unit using the 10khz stream,
comparisons against the 10MHz OCXO in my frequency counter have been very
promising, although I am unable to quantify the performance.

Actually, if you really get into the GPS signal structure all of the various
clocks are all plesiosynchronous. All of the signals from 1 pps to 10.23
MHz chipping rates and even the L1 and L2 carrier frequencies all are
synchronous (at each transmitter). This was done explicitly to allow post
processing phase accumulation error corrections. And resultant accuracy
improvements. The real trick is in compensating for Doppler and
relativistic effects.

 

[nzstranger]

Hello All,

How can I make reference oscillator (clock) out of generic consumer GPS
Receiver?
I hope to get stability 10-8 seconds over 8 hours.

There is off-the-shelf equipment doing just that, but prices are way over my
budget.

Appreciate your advices.

Sincerely,

Andrey Gleener

In the New Zealand Power industry, we use a Tektron Clock to synch our
relays, this is more accurate than you described, but that cant be bad.


I dont know if its comercialy available, nor the schematics, but it
could be a starting point?

Tim Strange.

 

[Don Bowey]

Actually, if you really get into the GPS signal structure all of the various
clocks are all plesiosynchronous.

The correct term is plesiochronous.

 

[Joseph2k]

The correct term is plesiochronous.

That's what i get for not verifying every term a telco uses.

 

[[email protected]]

If anyone's still on this thread, I'd appreciate any tips or links to
help me get whatever I need to generate a hardware clock that stays
within, say, 1ms, over 1 or more days. I can build/program any
required additional electronics, as this clock system will be part of
a larger production-line calibration facility anyway, including some
PCs. I imagine GPS would be a good way to go, but what
hardware/software should I look for, and where? I also imagine it
might be a PC peripheral, and therefore possibly also supply a time
reference to our network for other things. Cheap would be good, but
time is money.

Thanks for reading.