"Stable" time references

[Don Y]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Thx,
--don

 

[Martin Brown]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Atomic clock. Hydrogen maser.

You don't say how precise you need it or over what sort(s) of timescale
so it is impossible to give a meaningful answer. Either of these
reference standard clocks would be good enough to find the white light
fringe again on separately recorded signals on an VLBI system.

You can buy rubidium frequency standards or modules for example:

http://www.thinksrs.com/products/PRS10.htm
(much cheaper second hand)
http://www.ebay.co.uk/bhp/frequency-standard

These days they almost all VLBI uses H-masers but there were systems in
the 1980's using Rb atomic clocks synchronised to MSF Rugby or DCF77.

 

[Robert Baer]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Thx,
--don

Start with a XTAL oscillator with a constant load; whatever "pulling"
there is, that would tend to make that constant, eh?

 

[Spehro Pefhany]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Thx,
--don

Hi, don:-

Can you quantify "relatively stable"?

You can get used Rubidium 10MHz (sine wave) timebases for about $150.
They're a little bulky, draw significant juice at sometimes
inconvenient voltages, and the sine out probably isn't what you want.
But they're stable in the ppb range. Life is supposed to be 5-10
years, so the used ones are going to die some day not too far away.

http://openhpsdr.org/wiki/images/8/86/Rb-Normal_paper_20.10.09.pdf

I'm using these little guys (below) now.. overkill for my
applications, but they're only US $15 in quantity 1, they use < 20mW,
and are only 5x7mm.

http://www.conwin.com/datasheets/tx/tx355.pdf

If you don't mind OCXO power consumption (a few watts at start-up),
these guys are in the intermediate $55-ish range in singles, and 20ppb

http://www.conwin.com/datasheets/cx/cx207.pdf

Again, used ACROss are available (mostly from China).

Of course you can discipline the oscillator with NTP etc. to reduce
the 1/f noise. The cheaper option above- 200ppb-ish stability, is
going to drift five or ten seconds per year after calibration, and
with their 4.6ppm absolute accuracy spec, two or three minutes per
year before calibration.

Here's an open-source Ham project for a GPS-disciplined oscillator:-

http://ve2zaz.net/GPS_Std/Downloads/VE2ZAZ_GPS_Derived_Std_QEX_09_10_2006.pdf

And, if you happen to have $1K+ to blow, here's a component-size
atomic clock that fits into a relatively compact package and draws
relatively little current.

http://www.symmetricom.com/resources/download-library/documents/datasheets/quantum-sa45s-csac/


Best regards,
Spehro Pefhany

 

[Syd Rumpo]

On 03/12/2013 06:27, Don Y wrote:

To be clear, I am calibrating *time*

Then you will need a sonic screwdriver at the very least.

Cheers

 

[Rune]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Thx,
--don

Hi, I am not quite sure what kind of application or timebase and
accuracy/resolution you are looking for. Considering that you believe a
TCXO to be overkill and you want to keep correct time over a year maybe
an RTC chip would work for you for instance the DS3234 RTC with
integrated MEMS oscilator from Maxim Dallas it has a worst case drift
+-3.5ppm in the temperature range -40 to +85C. I had one installed
outdoors for a year in temperatures going from -25 to +25 over the year.
The 1 year drift was less than 1 minute. You could do fairly well with
this one at your controlled temp, if you are looking for 1 second
resolution at least.


Best Regards,
Rune

 

[Don Y]

Something like this?

[snipped URL due to line length]

Yes, apparently a TCXO.

The problem with a TCXO is EVERYONE pays for that accuracy even if
they don't want it.

I.e., it's not like you can "add" the TCXO after-the-fact to an
existing design (at some price adder) easily. By contrast, a
GPS PPS signal can just be *optionally* connected to a spare
input -- if there's no signal there (in 10 seconds), then the
accuracy associated with the "GPS option" just isn't available.

(yes, I am aware that you can do differential stuffing, have
different product/model numbers, etc. -- that's not what I'm
looking for.)

Simple divider to get 1 PPS

I'm not worried about getting a particular frequency/pulse rate
out of it. Rather, trying to ensure that "X units of time"
is, in fact, "X units of time" (to some cost related level of
performance)

 

[Don Y]

Hi Mike,

On 03/12/2013 06:27, Don Y wrote:

Why is a TCXO overkill,

It's overkill in the context of the XTAL option presented
immediately before it. TCXO has to go in at time of manufacture.
So, everyone pays for it -- even if they don't want or *need*
it (or, you start dealing with different versions of a product...
one with TCXO, one without).

XTAL is dirt cheap -- "everyone" can afford one (but the
accuracy sucks -- esp without tweaks)

or even a VCTCXO, when it will do the job and is
quite an affordable solution <$5 ?

I'm looking for something that folks can *opt* to use -- if they
want to bear the cost. (I consider GPS to be pretty pricey way
of doing this but someone else may not...). And, that doesn't
require shipping a product back to factory for an "upgrade"
(suddenly the $5 TCXO costs $100 -- shipping, parts, labor, test)

 

[Don Y]

Hi Robert,

By LFC, do you mean picking up radio time broadcasts? There are

Sorry, "Line Frequency Clock" (AC Mains). But, the radio approach
is one I hadn't thought of!

others RF broadcasts beyond the nominal "atomic clock" radio
broadcasts. HDTV broadcasts all have a time signal, if any FM radio
stations in your area support RDS (most should), that has a time
signal, the digital radio broadcasts all(?) do (although those are not
very popular in the US).

Thanks, I'll look into it. I've not been keen on relying on an
external reference simply because it presents another attack vector
(why expose something if you don't *have* to?)

Other than that, the only thing that comes to mind is picking the
phases off the mains (that won't get you an actual time setting, but
it'll keep your clock synchronized).

Yes -- my LFC reference.

While there's been some talk of
relaxing the standards, over longer periods (~one day) the grids are
adjusted to run at exactly a net 50 or 60Hz.

To some extent, I can deal with *really* long sampling intervals
(to get finer precision/accuracy).

However, there *has* been talk of easing this (in US -- I think Europe
expects EXACTLY 50Hz*60sec*60min*24hr cycles in a day!). I am not sure
what the benefits would be (why dick with something this fundamental
if there isn't some *big* payoff?) but am leary that all these
alternative power sources (e.g., solar) coming on line may cause
something like this to be sacrificed on the altar of expediency...

While there is short
term drift (at high loads the generators tend to drag a bit), the
operators then bump the frequency up (or down, if necessary) to
compensate once enough error has accumulated (usually no more than
about 10 seconds, absolute). The change is usually on the order of
+/-.01 or .02Hz. You do need a lot of filtering to read that though.
Many actual line powered clocks depend on that for their timekeeping.

Yes. I've looked at some power line monitors and rarely see any
"events" outside the 0.05Hz band.

Given your comments one other possibility is probably out of budget,
but Chip Scale Atomic Clocks can make the problem mostly go away - set
it and forget it. Several small devices are available, unfortunately
they're not cheap. One from Symmetricom is about $1500. ~ppb drift
per month. Lower power consumption than a TCXO too (with several
orders of magnitude better performance). Some of the competitors
appear to be cheaper.

Ow, ow, ow... I'm really trying to be cost-conscious, here.
Getting as much as I can for as few dollars -- even if it requires
more "smarts"/planning.

THE LFC option is one I have typically used in the past. Let it
act as a reference over a *long* integration period for the finer
grained local clock. Takes a while to settle but for things
running 24/7 that's not an issue (as long as your "nominal" is
pretty close and you open up the loop a bit to start with).

What I *don't* want is the "ceramic resonator" sort of "it kinda
works" approach. Yet, don't want to count electrons flying around
a nucleus... :-(

 

[John Devereux]

Hi Mike,



It's overkill in the context of the XTAL option presented
immediately before it. TCXO has to go in at time of manufacture.
So, everyone pays for it -- even if they don't want or *need*
it (or, you start dealing with different versions of a product...
one with TCXO, one without).

XTAL is dirt cheap -- "everyone" can afford one (but the
accuracy sucks -- esp without tweaks)


I'm looking for something that folks can *opt* to use -- if they
want to bear the cost. (I consider GPS to be pretty pricey way
of doing this but someone else may not...). And, that doesn't
require shipping a product back to factory for an "upgrade"
(suddenly the $5 TCXO costs $100 -- shipping, parts, labor, test)

Is it possible to get the effect of a TCXO by thermally characterizing a
normal XO? Then the "value added" would be making the tempco
measurements and setting the correction factors, for those that pay. You
would still need to measure temperature but perhaps you do that already
or in any case it can be done very cheaply if needed.

 

[George Herold]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Is the 60Hz AC line still "correct" once a day?
What about the broadcast time stamp they use for the cheap "atomic clocks" we hang on the wall. (I can't recall the name.)
Just look up on the web and get the time.

Try asking at time-nuts.
http://leapsecond.com/time-nuts.htm

George H.

 

[Don Y]

Hi John,

Is it possible to get the effect of a TCXO by thermally characterizing a
normal XO? Then the "value added" would be making the tempco
measurements and setting the correction factors, for those that pay. You
would still need to measure temperature but perhaps you do that already
or in any case it can be done very cheaply if needed.

That's, in a sense, what I am trying to do "at runtime" -- *if*
I have access to some other reference. I want every device to
be built the same way -- no "separate processing" for different
accuracies, etc.

What I want is user to be able to "add" something to the system
after-the-sale (i.e., I don't even want to be in the business
of *making* that "something") that magically unlocks extra
capabilities (that were present all along just not usable
because the system didn't have this "external reference" to
consult).

E.g., the example, elsewhere, that I posted re: monitoring a
digital input for a PPS signal and, if present (and *appears*
to be "about 1Hz") using it *as* the "1 Hz reference". OTOH,
if it is *absent*, the associated software disables itself
simply because there's nothing it can work off of!

Imagine a timepiece in your home. You can buy a really *good*
timepiece that keeps really accurate time. Or, settle for
"whatever is cheapest". *OR*, hire someone to keep reseting them
on a daily basis!

I.e., folks who don't care about accuracy buy whatever is cheapest
(basic system). The "really good" option isn't available (cuz I
don't want to be in that business of making different models, etc.).
So, those who want better hire someone to keep their clocks
up to date -- an external GPS, LFC, <whatever>. And, no one else
"pays extra" (software is free :> ) for their needs.

 

[Martin Brown]

Is the 60Hz AC line still "correct" once a day?

It is held in theory so that the long term average is right so that any
older synchronous motor based mains kit keeps good time. However, it
runs slow during the day and fast at night which used to annoy
astronomers back in the days when they used mains synchronous motors for
sidereal drives to track the rotation of the Earth.

What about the broadcast time stamp they use for the cheap "atomic clocks" we hang on the wall. (I can't recall the name.)
Just look up on the web and get the time.

Try asking at time-nuts.
http://leapsecond.com/time-nuts.htm

George H.

The usual way is to control a local oscillator by phase locking it to
one or other broadcast reference time signal. The wall clocks use a
standard watch crystal and lock it to the reference second either
analogue with a tunable capacitor or by tweaking the divisor slightly.

 

[Phil Hobbs]

Is it possible to get the effect of a TCXO by thermally characterizing a
normal XO? Then the "value added" would be making the tempco
measurements and setting the correction factors, for those that pay. You
would still need to measure temperature but perhaps you do that already
or in any case it can be done very cheaply if needed.

That's a lot of work.

How about putting a normal VCXO in an analogue PLL with an XOR phase
detector--it'll servo right at half scale unless you put in an external
reference. A little 4-pin connector and a baby daughtercard would let
you sell the reference as an add-on.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

 

[[email protected]]

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

If you are making a time base (not just a frequency standard), how are
you going to handle leap seconds ?

 

[George Herold]

Is it possible to get the effect of a TCXO by thermally characterizing a

normal XO? Then the "value added" would be making the tempco

measurements and setting the correction factors, for those that pay. You

would still need to measure temperature but perhaps you do that already

or in any case it can be done very cheaply if needed.

Yes. It's commonly done in land mobile radios that require high
stability but cannot tolerate the power drain of an oven. Also used
for stabilizing oscillators over very wide temperature ranges.

Officially, it's a DTCXO (digital temp compensated crystal osc)[1].

The crystal and oscillator circuitry is characterized over the desired

temperature range, and the required frequency compensation is saved in
a lookup table. It's usually controlled by a PIC. Here's an
example:
<http://www.standardcrystalcorp.com/pdf\o-12.pdf>
0.3 to 1 ppm stability over various temperature ranges (depending on
model). Here's another:
<http://www.ecmelectronics.co.uk/pdf/ecdt25.PDF>

with varying specs.
The big problem is crystal ageing, which can sometime be as large as
the temperature drift. I've forgotten everything I had learned about
accelerated crystal aging, but suffice to say that it's closer to
magic than science.

Several years ago, I interviewed at a company that does time stuff. Xtals and atomic clocks. I had this idea of cutting the ovenized X-tals for operation near 25C (rather than the 70C) and then using TEC's to keep the temperature stable, but near room temp. I was told that aging was bigger problem. Basically due to atoms moving onto or off of the xtal.. the mass loading changes the frequency.
There may be other aging effects too.

George H.

[1] More oscillator acronyms:

<http://en.wikipedia.org/wiki/Crystal_oscillator#Circuit_notations_and_abbreviations>



--

Jeff Liebermann [email protected]

150 Felker St #D http://www.LearnByDestroying.com

Santa Cruz CA 95060 http://802.11junk.com

Skype: JeffLiebermann AE6KS 831-336-2558

 

[Artemus]

Hi,

I need a relatively stable timebase against which to calibrate
my own *local* timebase.

Assume *my* environment is reasonably well controlled (temp, etc.).

Also, assume I can make measurements over VERY long periods of
time (24/7/365 duty).

Ideally, I would like the choice of reference to be something
that user could trade cost/convenience/performance.


To that end, I've got a few basic ideas:
- local XTAL (too easy to "pull", no real advantage over TCXO)
- local TCXO (overkill -- hard to make this "optional")
- external PPS input from GPS (requires view of sky, expensive)
- network time service (hole in firewall, latency issues)
- LFC (low accuracy but dirt cheap, LONG integration periods)

To be clear, I am calibrating *time* -- beyond that is my own
concern.

Any other ideas I should entertain?

Thx,
--don

NIST offers a variety of broadcast time references.
http://www.nist.gov/pml/div688/grp40/wwv.cfm
There is also one on the WWW.
Art

 

[miso]

And, if you happen to have $1K+ to blow, here's a component-size
atomic clock that fits into a relatively compact package and draws
relatively little current.

http://www.symmetricom.com/resources/download-library/documents/datasheets/quantum-sa45s-csac/


Best regards,
Spehro Pefhany

I have two Symetricom references. Just look for them used. Mine were $30
a pop.

The cellular companies dumped the crystal references for Rb year ago for
long operating time without GPS discipline. The old crystal based
references are everywhere, though in the US most were probably crushed.
On ebay they seem to all be out of China. Mine were new old stock from
"I can't say", but these time references are everywhere in high tech
areas. Data centers, research labs, company wide clock references in the
better companies, etc.

In the old days, the HP labs in Santa Clara had a cesium clock in the
lobby. Supposedly it was the reference for the facility, but who knows.
It could have been a busted cesium clock that some engineer just hacked
to display time from another source.

 

[Maynard A. Philbrook Jr.]

Is it possible to get the effect of a TCXO by thermally characterizing a

normal XO? Then the "value added" would be making the tempco

measurements and setting the correction factors, for those that pay. You

would still need to measure temperature but perhaps you do that already

or in any case it can be done very cheaply if needed.

Yes. It's commonly done in land mobile radios that require high
stability but cannot tolerate the power drain of an oven. Also used
for stabilizing oscillators over very wide temperature ranges.

Officially, it's a DTCXO (digital temp compensated crystal osc)[1].

The crystal and oscillator circuitry is characterized over the desired

temperature range, and the required frequency compensation is saved in
a lookup table. It's usually controlled by a PIC. Here's an
example:
<http://www.standardcrystalcorp.com/pdf\o-12.pdf>
0.3 to 1 ppm stability over various temperature ranges (depending on
model). Here's another:
<http://www.ecmelectronics.co.uk/pdf/ecdt25.PDF>

with varying specs.
The big problem is crystal ageing, which can sometime be as large as
the temperature drift. I've forgotten everything I had learned about
accelerated crystal aging, but suffice to say that it's closer to
magic than science.

Several years ago, I interviewed at a company that does time stuff. Xtals and atomic clocks. I had this idea of cutting the ovenized X-tals for operation near 25C (rather than the 70C) and then using TEC's to keep the temperature stable, but near room temp. I was told that aging was bigger problem. Basically due to atoms moving onto or off of the xtal.. the mass loading changes the frequency.
There may be other aging effects too.

George H.

[1] More oscillator acronyms:

<http://en.wikipedia.org/wiki/Crystal_oscillator#Circuit_notations_and_abbreviations>

Get some caesium and make your own clock ref!

Jamie

 

[Tom Miller]

On Tue, 03 Dec 2013 13:35:21 +0000, John Devereux




Is it possible to get the effect of a TCXO by thermally characterizing
a

normal XO? Then the "value added" would be making the tempco

measurements and setting the correction factors, for those that pay.
You

would still need to measure temperature but perhaps you do that
already

or in any case it can be done very cheaply if needed.
Yes. It's commonly done in land mobile radios that require high
stability but cannot tolerate the power drain of an oven. Also used
for stabilizing oscillators over very wide temperature ranges.

Officially, it's a DTCXO (digital temp compensated crystal osc)[1].

The crystal and oscillator circuitry is characterized over the desired

temperature range, and the required frequency compensation is saved in
a lookup table. It's usually controlled by a PIC. Here's an
example:
<http://www.standardcrystalcorp.com/pdf\o-12.pdf>
0.3 to 1 ppm stability over various temperature ranges (depending on
model). Here's another:
<http://www.ecmelectronics.co.uk/pdf/ecdt25.PDF>

with varying specs.
The big problem is crystal ageing, which can sometime be as large as
the temperature drift. I've forgotten everything I had learned about
accelerated crystal aging, but suffice to say that it's closer to
magic than science.

Several years ago, I interviewed at a company that does time stuff.
Xtals and atomic clocks. I had this idea of cutting the ovenized X-tals
for operation near 25C (rather than the 70C) and then using TEC's to keep
the temperature stable, but near room temp. I was told that aging was
bigger problem. Basically due to atoms moving onto or off of the xtal..
the mass loading changes the frequency.
There may be other aging effects too.

George H.

[1] More oscillator acronyms:

<http://en.wikipedia.org/wiki/Crystal_oscillator#Circuit_notations_and_abbreviations>

Get some caesium and make your own clock ref!

Jamie

Cesium is really nasty stuff.

And making a Cs beam tube is not for chickens.