"Stable" time references

[[email protected]]

I need a relatively stable timebase

Bounce an ultrasonic signal off the pile of horse poop and see how long
it takes to come back. Shorter times = bigger pile. Then you just need
an NIST-traceable horse to calibrate against. And possibly to not care
about RoHS.

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

7 years, wow!

Any other ideas I should entertain?

For a good time, call +1 303 494 4774. Use a landline for lower
latency.

I hear you can get cell phone modules that allow your device to either
make a phone call or send texts over the cell networks. I don't know
if these modules can tell you what time the cell network thinks it is.
You might need a SIM card to make this work.

Matt Roberds

 

[Spehro Pefhany]

On Tue, 03 Dec 2013 06:11:38 -0700, the renowned Don Y

Hi, Don:-

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)

Ah, well... if you pick a common standard frequency like 10MHz, and a
standard footprint like 5 x 7mm, and a standard voltage like 3.3V you
can populate either a $15 TCXO or a $1.00 oscillator on the same PCB.

If you want field options, you could also easily make a 4-pin plug-in
module that would contain just the TCXO (or even an OCXO) and switch
over the clock source (use the 4th pin to ground the tristate output
selection line on the on-board XO). That way the added "cost" per
non-TCXO unit is just a 3-cent 4-pin header eg. SWR201-NRTN-S04-SA-WH.

One thing I've done is to have an oscillator input over an SMA or
whatever and detect the signal & switch over automatically to the
external source if it's present. That takes a few more parts, but
allows the simplest possible connector and is harder (never say
impossible!) for the user to screw up.


Best regards,
Spehro Pefhany

 

[Don Y]

Greetings Joe,

So what you are approaching is a second optional product they can
purchase, your time keeper so to speak.

No! I don't want to be in that business! What I want is a means
by which *they* can "enhance" what I've made TO FIT THEIR NEEDS.
I.e., the guy they hire (in the paragraph above) has no affiliation
with me -- I've just provided a means by which "he" can adjust
the clocks.

[clocks are a flawed analogy because they imply actual time-of-day
is significant. I'm really only interested in ensuring that time
*passes* at the actual rate]

Then you have multiple
variations (GPS, CDMA, RF, etc). Your main product should supply
power to this addon (you dont want another wall wart or such) and
it has to take the pps input. You would be wise to have power
limiting on the supply side and need to handle all the ESD and
EMI issues with the connector and.. and ...

No. Instead, I'm looking at just an "input" for this reference
(DC coupled wrt ground, transformer coupled, optically coupled,

By the time you get done with all that you might as well just put
the TCXO in and be done with it. Cost wise it could quickly be close
to the same.

TCXO (or *any* "fixed design choice") means accuracy is fixed and
can't change after-the-fact as the user's needs change. Annoying
(to the user) to realize you have to buy a different version as
your needs increase. Or, have *your* device retrofitted to
accommodate something this "trivial".

 

[Don Y]

Hi Martin,

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.

Yes. It is the "long term" accuracy that I exploit "on the cheap"
cuz I'm running 24/7/365 (count very long intervals so you are
effectively frequency locking a high frequency "clock" -- MHZ -- to
a subharmonic -- uHz -- derived from the AC mains)

Good point!

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.

Yes. But they rely on that signal (when present) having the
"final accuracy" desired. The LFC approach relies on the *average*
mains frequency having the desired accuracy -- and expects to be
able to watch it over very long periods of time.

For years I've been making "gift timepieces" (the gift aspect being
some unique -- literally "one of a kind" -- and atypical way in which
I "display" the time) that watch the AC mains to discipline the MCU's
XTAL. Cheap solution and works well as most "(wall) clocks" sit in
reasonably well controlled environments.

 

[Don Y]

Hi Jeff,

Relative to what? How about a clue as to your accuracy and stability
requirements? The appropriate device could be anything from a free
running oscillator to a GPSDO.

Exactly! That's the whole point! Let the *user* decide what level
of accuracy he wants/needs. And, whether he is interested in short
term and/or long term goals. Why should he be "stuck" with a
design decision that I've made "for the masses"?

Your "HiFi" lets you decide is you want to buy a crappy cassette
deck or a high end DAT or a CD, etc. Why should the manufacturer
*pick* one of these options and lock you into it -- forever?
(Ah, because the manufacturer wants to sell you another model next
year...)

I also don't like the idea of what sounds like a transfer standard.
With a synthesizer, you can drive whatever clock your driving directly
from your timebase.


One of my early GPSDO contrivances would take about an hour to recover
from every time I opened the outside shop door. That was eventually
fixed by double insulating the OCXO. I also had plenty of fun trying
to keep the 5 and 10 MHz outputs from leaking out of the box, and
trashing my WWV reception.


How long? With all due respect, you're being far too vague here.
Different technologies offer different short and long term accuracies.

Exactly! A frequent technique I use is to monitor the AC mains
to achieve long term accuracies -- but requiring day(s) to lock
to the "average line frequency". If you're running 24/7/365,
this isn't an issue -- except in your "initial accuracy".

That's always the case. More accuracy and stability always costs
more, usually on a non-linear scale. At the high end, a small
increase in accuracy and stability can cost outrageous amounts in
construction costs. However, the worst cost of accuracy and stability
is your time in calibrating the reference OCXO. I generally have to
tinker with the OCXO for several days before it settles down.
<http://www.febo.com/pages/stability/>
Note the differences between accuracy and stability. Also note that
the Allen variance goes out to 10^4 seconds (about 7 days) and that
the next order of magnitude in accuracy requires 10 times that or 70
days.


Who's time? There are subtle differences between UTC, astronomical,
navigational, space, and political clocks. Einstein's time dilation
is involved. Not the table at the top of the page:
<http://www.leapsecond.com/java/gpsclock.htm>
Looks like about 16 seconds difference between UTC and GPS time.

"Time" in the sense of "time passing" -- not in the sense of
"the (current) time is:"

A mechanical wind up clock would work if you don't bother to specify
your accuracy and stability requirements.

Yes, it would! Provided there was a means of "observing" it!

I can't tell what you're trying to accomplish, but if all you want is
accurate time, and not accurate frequency, then all you need is an
accurate clock. The 1pps from most GPS receivers will drive almost
any kind of clock with good accuracy (as long as you don't have any
missing pulses). If you don't mind the clock wandering around a bit
during the day, a WWVH 60 KHz clock might suffice.

On the other foot, if all you want is a stable frequency, you can get
close with an OCXO or possibly a TCXO. An OCXO is probably best,
especially if you add extra insulation. Start at 10 MHz and divide
down to 5, 2.5, and 1 MHz depending on what you need. I have such a
system distributing 5 and 10 MHz to all my test equipment that will
accept an external reference. On a good day in the ionosphere, I get
about 0.01 ppm accuracy against WWV.

Incidentally, I have a pile of about 200 Novatel/CMC Allstar 12 GPS
receiver boards if you want to play. They're old (2001), not the
best, and don't have a frequency output, but are good enough for 1pps
and for building ham-radio APRS trackers. The catch is that you'll
have to build an interface board to make it useful:
<http://802.11junk.com/jeffl/Novatel/Allstar 12 GPS/>

I've got a drawer full of different GPS devices -- which is where
the idea for an "external reference" originated. What I'm looking
for is other options that I may not have considered. Esp if they
affect the sort of "interface" I would have to implement/support

E.g., using an *external* TCXO would mean I would have to be able
to capture a very high frequency signal. Very different approach
than talking to a time server over the network or listening to
a PPS signal.

 

[Don Y]

Hi Syd,

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



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

If I had one, I wouldn't be wasting my time on matters
as *mundane* as these! :>

 

[Don Y]

Greetings Spehro,

On Tue, 03 Dec 2013 06:11:38 -0700, the renowned Don Y


Ah, well... if you pick a common standard frequency like 10MHz, and a
standard footprint like 5 x 7mm, and a standard voltage like 3.3V you
can populate either a $15 TCXO or a $1.00 oscillator on the same PCB.

Means two different "models" have to be maintained. Or, the user

If you want field options, you could also easily make a 4-pin plug-in
module that would contain just the TCXO (or even an OCXO) and switch
over the clock source (use the 4th pin to ground the tristate output
selection line on the on-board XO). That way the added "cost" per
non-TCXO unit is just a 3-cent 4-pin header eg. SWR201-NRTN-S04-SA-WH.

"Grandma" is probably not going to be comfortable making that sort
of change. Or, will be calling for "Support" when she's bent a
pin (and doesn't yet know it).

One thing I've done is to have an oscillator input over an SMA or
whatever and detect the signal& switch over automatically to the
external source if it's present. That takes a few more parts, but
allows the simplest possible connector and is harder (never say
impossible!) for the user to screw up.

This is the approach I think I am going to settle on. Let the
user connect <something> having <some_characteristics> and
<somehow> figure out how to use that information. (i.e.,
don't *assume* it's 1Hz -- or, perhaps *look* at it and if
it *seems* like ~1PPS, then make that assumption to save the
user having to *tell* me it is 1.0000Hz and not 29.97Hz, etc.)

 

[[email protected]]

Greetings Joe,

So what you are approaching is a second optional product they can
purchase, your time keeper so to speak.

No! I don't want to be in that business! What I want is a means
by which *they* can "enhance" what I've made TO FIT THEIR NEEDS.
I.e., the guy they hire (in the paragraph above) has no affiliation
with me -- I've just provided a means by which "he" can adjust
the clocks.

[clocks are a flawed analogy because they imply actual time-of-day
is significant. I'm really only interested in ensuring that time
*passes* at the actual rate]

If it doesn't, what are you prepared to do about it?

 

[whit3rd]

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

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

Does this relate to time-of-day time? If you have internet connection,
NTP (US naval observatory network time protocol) does a very good
job, even compensates for transmission delays.

Milliseconds over a one-year period makes it a solution with 10**-9
accuracy. It's commonly used in a mode where you only update
once a day, but that's something you can program around.

Many crystal clocks (not all) can get 10**-7 precision over short
periods (but even an ovenized crystal clock isn't as good as the atomic
time standards).

 

[qrk]

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

You can buy a GPS time reference module from Trimble. About $40 or $50
from a hobby shop in England. I've made up two of these systems for
our shop. The module has a 10MHz and a 1 PPS output. You can use these
outputs to train a master clock as they do have a bunch of jitter.

 

[Mark Borgerson]

In article <[email protected]>, [email protected] says...

As someone else mentioned, the stabilized RTC chips from
Maxim/Dallas can be pretty good. I've used them for years
in moored oceanographic instruments that can spend a year
out of touch with the world----including mains power and
GPS signals. We often get instruments back whose clocks are within
30 seconds of each other after a year under water. That's a
difference of about 1PPM. These moored instruments do have
the advantage that the temperature probably doesn't change
more than 5 deg C. over the interval and the instruments are
within about 5 deg. C. of each other.

Another option might be a receiver for the 60KHZ time broadcasts
if the system is going to a place where those can be received.
The receivers must be fairly low power and cheap----there's one
built in to my $35 Casio wristwatch.

Mark Borgerson

 

[Paul E Bennett]

[%X]

[clocks are a flawed analogy because they imply actual time-of-day
is significant. I'm really only interested in ensuring that time
*passes* at the actual rate]

If it doesn't, what are you prepared to do about it?

In the words of Ford Prefect "Time is an illusion; Lunchtime doubly so" ;-)

Following the actual rate of time would rather depend on how the anomilies
in the gravity field affect its apparent passing and the viewpoint you are
observing its passing from.

--
********************************************************************
Paul E. Bennett IEng MIET.....<email://[email protected]>
Forth based HIDECS Consultancy.............<http://www.hidecs.co.uk>
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-510979
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

 

[[email protected]]

[%X]

[clocks are a flawed analogy because they imply actual time-of-day
is significant. I'm really only interested in ensuring that time
*passes* at the actual rate]

If it doesn't, what are you prepared to do about it?

In the words of Ford Prefect "Time is an illusion; Lunchtime doubly so" ;-)

Following the actual rate of time would rather depend on how the anomilies
in the gravity field affect its apparent passing and the viewpoint you are
observing its passing from.

I just thought the choice of the word "ensuring" was irresistible. ;-)

 

[Maynard A. Philbrook Jr.]

Bounce an ultrasonic signal off the pile of horse poop and see how long
it takes to come back. Shorter times = bigger pile. Then you just need
an NIST-traceable horse to calibrate against. And possibly to not care
about RoHS.

Ok, Which horse do you use for a NIST?

Example:

Years ago, a friend of mine saw this nice 2 HP bench grinder, made
in CHINA of course. Now since this deal was irresistible, he purchased
this grinder to replace a 0.5 HP grinder he had with a large stone to
sharpen things like axes, knifes etc..

Upon setting it up and trying it out, he found out quickly that just
about anything would slow it down to a crawl at a point to which it
would trigger the centrifugal switch.. After playing around, he came to
the final conclusion that the horses in CHINA must be a lot smaller than
those in the US!

Hence, which horse do you use for a NIST calibration ?

Jamie

 

[Don Y]

Hi Jim,

I compared my ancient HP 5328A with the
010 high stability time base option to
a Sulzer Laboratories frequency standard
that I rescued from the recycling bin.

The two agreed within something like 30
parts per billion over a week's time.

The 5328A is nice because it has a 10mhz
bnc on the back and standby mode that keeps
the crystal warm. Plus you have a counter
timer.

"10MHz" has been referenced several times in this
thread. Of course, any frequency can be a "reference"
(to the stated accuracy *of* that reference). But, is
"10MHz" a "standard" -- even if informally so -- in
much the same way that 1PPS is a defacto "standard"
output from a GPS Rx?

I.e., if I design an input that expects a digital
signal and, if I see "more than several" transitions
in a given "nominal, uncalibrated second", *assume*
it is a 10MHz reference? And, if I see *some*
transitions but not "enough" in that same second,
assume it is 1PPS? Otherwise, assume there is no
reference connected?

Does this cover the majority of cases? Or, do you
need to be able to *specify* that you have a 23.5KHz
"reference" attached? (remember, I'm not just looking
for a "frequency" but a "standard")

Thx,
--don

 

[Don Y]

Hi Mark,

As someone else mentioned, the stabilized RTC chips from
Maxim/Dallas can be pretty good. I've used them for years
in moored oceanographic instruments that can spend a year
out of touch with the world----including mains power and
GPS signals. We often get instruments back whose clocks are within
30 seconds of each other after a year under water. That's a
difference of about 1PPM. These moored instruments do have
the advantage that the temperature probably doesn't change
more than 5 deg C. over the interval and the instruments are
within about 5 deg. C. of each other.

I think I can guarantee the 5C range -- though I'll be up around
room temperature (whereas I suspect you were closer to 0C?)
Not sure if that lessens or exacerbates the problem...

But, that seems lie a "design *in*" solution vs. a "bolt on"
*addition*. I.e., I assume you need a bus (even if it's a
"one wire") to talk to the device -- you don't just "listen".

Another option might be a receiver for the 60KHZ time broadcasts
if the system is going to a place where those can be received.
The receivers must be fairly low power and cheap----there's one
built in to my $35 Casio wristwatch.

Accuracy? Do you have any way of knowing how often your
watch *loses* access to that reference? (it's actually more
than I was asking for as I don't really care what "time"
it is -- though I can obviously work with that)

Thx,
--don

 

[Don Y]

Does this relate to time-of-day time? If you have internet connection,
NTP (US naval observatory network time protocol) does a very good
job, even compensates for transmission delays.

It's to drive the basic timebase in the system. ToD ends up being
one aspect of that -- but that's not an essential function.

I'm trying to avoid unnecessary reliance on external references
THAT THE USER CAN'T CONTROL as they represent potential attack
vectors.

Milliseconds over a one-year period makes it a solution with 10**-9
accuracy. It's commonly used in a mode where you only update
once a day, but that's something you can program around.

Yes, though that doesn't ensure that the device remained that
closely synchronized throughout the entire period -- just that
it was able to get to that precision *at* the end of the year.

Many crystal clocks (not all) can get 10**-7 precision over short
periods (but even an ovenized crystal clock isn't as good as the atomic
time standards).

Yes. What I want is a "cheap" solution for the masses that
others can "upgrade" (without my direct involvement) to fit
their needs.

 

[Mark Borgerson]

Hi Mark,



I think I can guarantee the 5C range -- though I'll be up around
room temperature (whereas I suspect you were closer to 0C?)
Not sure if that lessens or exacerbates the problem...

But, that seems lie a "design *in*" solution vs. a "bolt on"
*addition*. I.e., I assume you need a bus (even if it's a
"one wire") to talk to the device -- you don't just "listen".

I do use an I2C bus (2 wires plus ground) to communicate with the
DS3232 as I set and read the clock and calendar. However, you could
simplifiy things by just using either the 8192Hz clock output or
the 1Hz interrupt output from the DS3232. In some systems, I use
the 1Hz output as an interrupt input in a sofware phase locked
loop to correct the clock on a microprocessor with a less-stable
crystal oscillator.

Accuracy? Do you have any way of knowing how often your
watch *loses* access to that reference? (it's actually more
than I was asking for as I don't really care what "time"
it is -- though I can obviously work with that)

My watch checks that reference about once a day. Sometimes it
misses a day if it spends the night in a place with too much
surrounding metal. I guess the maximum error your would expect
would be the difference between your system clock and the
time standard over the maximum time you might expect the
system to miss the 60KHz signal.


Mark

 

[[email protected]]

WWVB and other time broadcasts are fairly inconvenient sources of
accurate time. There are two big problems: unknown propagation delay and
slow modulation rate. Propagation delay could be as large as ~tens of
milliseconds; you have to know your location to correct. Slow modulation
makes difficult to find exact location of time marks and relate them to
carrier period; it could take hours of averaging.
There is no problem if the accuracy of 1s is sufficient. Getting to the
accuracy of 1us isn't trivial.

Vladimir Vassilevsky
DSP and Mixed Signal Designs
www.abvolt.com

Is the Loran-C navigation system at 100 kHz still working in the USA ?

It should solve the location and propagation delay issues, however,
you would need some means of telling when (ToD) the transmission code
sequence for each transmitter started.

 

[Don Y]

Hi Mark,

I do use an I2C bus (2 wires plus ground) to communicate with the
DS3232 as I set and read the clock and calendar. However, you could
simplifiy things by just using either the 8192Hz clock output or
the 1Hz interrupt output from the DS3232. In some systems, I use
the 1Hz output as an interrupt input in a sofware phase locked
loop to correct the clock on a microprocessor with a less-stable
crystal oscillator.

I'd need someone to put it in a box and agree to sell it as a
product. Hard to see much of a market for it other than as
a 1PPS frequency reference (in that configuration).

I'll keep it in mind and see if I can cajole my buddy to
supporting it as well...

My watch checks that reference about once a day. Sometimes it

Oh, so there is a *time* at which it goes looking for the
signal? (I assumed it was monitoring it constantly and
just freewheeling when it lost signal)

misses a day if it spends the night in a place with too much
surrounding metal. I guess the maximum error your would expect
would be the difference between your system clock and the
time standard over the maximum time you might expect the
system to miss the 60KHz signal.

Yes. Watches also have the benefit of reasonably constant
temperature operation (at least while they are being worn).

I've got an "atomic clock" here. Maybe I should open it up and
see what's inside.

Again, I'd need to find a COTS "module" that just gave me
some convenient signal that I could "listen to" instead of
"interact with". This approach has the downside that if
it can't lock onto the signal *when* it's needed, it's
ineffective (whereas the Dallas device would be pushing out
"valid signal" regardless).

Thx!
--don