Is your clock in sync? (USA)

Brian Gregory wrote:
On 12/07/2020 02:22, lew wrote:
Must also use the same time server.

I generally use at least 4 NTP servers.

I run Meinberg NTP client, the one built in to Windows is just a toy.


And that's the puzzling part.

Why is the Windows one, the way it is ?

Paul

On 2020-07-11 8:50 p.m., Paul wrote:
Rene Lamontagne wrote:
On 2020-07-11 7:26 p.m., Paul wrote:
Rene Lamontagne wrote:
On 2020-07-11 5:22 p.m., Paul wrote:
Big Al wrote:
On 7/11/20 4:07 PM, this is what John Doe wrote:
https://www.time.gov/

My clock is ticking in unison with that clock.

The page says mine is off by 40 seconds, but my clock flips to the
next minute the same time the web page does, so I'm not sure what
they mean other than maybe the rtc on the motherboard.

Al

The RTC is not typically accessed while the OS runs.

It's accessed at some point, to transfer the "software time"
to the RTC for safe keeping. The interface is too slow for any
practical purpose. RTC has 1 second resolution. Hardware
clock (part of software time system) is 100ns or so.

Â*Â*Â*Â* Startup
Â*Â*Â*Â* RTC -- software clockÂ*Â* # prime software clock
Â*Â*Â*Â* ...
Â*Â*Â*Â* software clock -- RTCÂ*Â* # save software clock
Â*Â*Â*Â* shutdown

The RTC clock (32768Hz, ripple divider to 1 second) and
the hardware clock (100MHz BCLK or higher), they drift at
different rates because they use different xtals. Which
means the drift when you shutdown, is different than
the drift rate while the box runs.

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

And time servers have been off by as much as 20 minutes.
If you're getting complaints from your OS that "the clock is off",
it's quite likely to be true. When an official server is "off",
it can just as easily be a "whoa boy" sized error :-) They won't
be off by 3 nanoseconds. Just go whole hog and shoot for 20 minutes.
I have no idea how the thing can go off by 20 minutes. Some
clever software calc of some sort (atomic clock consolidation?).
It'll probably turn out, the whole thing is based on a battery
operated ticky clock on the control room wall :-)

Â*Â*Â* Paul

Damn, my clock drifted since IÂ* checked it earlier at 3:33Â* today it
was out by .321 seconds, now its out by .971 seconds.
What to do, What to do?Â*Â* Sigh. :-)

Rene

Ah, a time fetish ?

GPS discipline.

And either the hardware comes with a downloadable
software to tie the GPS time into the computer,
or you search for applications yourself. I don't
see GPS in any of the screenshots here, but it might
be there.

https://www.meinbergglobal.com/english/sw/ntp.htm

For software, maybe NFS servers in Linux, might
appreciate tight time keeping. I don't know if
Windows shares particularly care.

Way back when, it would have cost a thousand
bucks for this, but it's a lot cheaper now to
discipline a PC.

Hold your GPS near a window, to get some signal.

Â*Â*Â* Paul

Nah, just bored. :-)

Rene

Mine is off by 2.7 seconds. What a slob.

But that's after six days.

Â*Â* Paul

OK, we will let it slide this time !


Rene

lew wrote:

On 2020-07-11, Carlos E.R. wrote:
On 12/07/2020 01.35, Carlos E.R. wrote:
On 12/07/2020 00.22, Paul wrote:

...

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

Wow, one week. I must be spoiled by tools like ntpd in Linux, keeps the
clock accurate continuously within milliseconds using the pool.


I forgot. In a Windows Domain with Active Directory, keeping all the
machines in sync is crucial. A small error and you simply can not login
(Kerberos). It doesn't matter if the clock is correct, but all the
machines need have the same time.

Must also use the same time server. I recall one version of windows
had the time server defaulted to a ms time server; the good thing was
that one can change the time server to any available sever.

Looks like win 10 2004 does not allow changing the server from
default 'time.nist.gov'. AND no longer able to change the time server
to the Fort Collins to be in sync with my 'atomic' wall clocks.

Lew,

Windows 10 ver. 2004 DOES allow changing time servers!

Go to Control Panel--Date and Time--Internet Time--Change Settings The pull down menu has a

few NTP time servers listed or you can manually type in your own.

Just an FYI...

JT


--

Carlos E.R. wrote:
On 12/07/2020 00.22, Paul wrote:

...

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

Wow, one week. I must be spoiled by tools like ntpd in Linux, keeps the
clock accurate continuously within milliseconds using the pool.


But there's an assumption of a first order error there.

It cannot compensate for wander, as wander is random as far
as the observer is concerned.

The drift you're talking about, is the ability to correct if
the clock is consistently 1 second slow every day. Any NTP
implementation worth a damn (Meinberg) will dribble out the
error corrections, compensating for that one.

But on top of the 1 second a day static error, there can be
milliseconds plus or minus at random. And the only way to null
that out, is to up the sync frequency, which is "cheating".
Any oaf can do that.

Here is someone recording the wander of a computer clock.

https://www.satsignal.eu/ntp/2009-03-Feenix_offset.png

Paul

On 12/07/2020 02:52, Paul wrote:
Brian Gregory wrote:
On 12/07/2020 02:22, lew wrote:
Must also use the same time server.

I generally use at least 4 NTP servers.

I run Meinberg NTP client, the one built in to Windows is just a toy.


And that's the puzzling part.

Why is the Windows one, the way it is ?

Â*Â* Paul

1) Not invented here syndrome.
2) Lets make absolutely sure time.windows.com gets so little traffic we
can run it on a raspberry pi.

--
Brian Gregory (in England).

On 12.07.20 14:54, Brian Gregory wrote:
On 12/07/2020 02:52, Paul wrote:
Brian Gregory wrote:
On 12/07/2020 02:22, lew wrote:
Must also use the same time server.

I generally use at least 4 NTP servers.

I run Meinberg NTP client, the one built in to Windows is just a toy.


And that's the puzzling part.

Why is the Windows one, the way it is ?

Paul

1) Not invented here syndrome.
2) Lets make absolutely sure time.windows.com gets so little traffic we
can run it on a raspberry pi.


I use Atomic clock sync , download page:
https://www.worldtimeserver.com/atomic-clock/

On 7/11/2020 1:59 PM, Big Al wrote:
On 7/11/20 4:07 PM, this is what John Doe wrote:
https://www.time.gov/

My clock is ticking in unison with that clock.

The page says mine is off by 40 seconds, but my clock flips to the next minute the same time the web page does, so I'm not sure what they
mean other than maybe the rtc on the motherboard.



My clock is off by .013 seconds. I can live with that.


--
Ken

On 2020-07-12, JT wrote:
lew wrote:

On 2020-07-11, Carlos E.R. wrote:
On 12/07/2020 01.35, Carlos E.R. wrote:
On 12/07/2020 00.22, Paul wrote:

...

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

Wow, one week. I must be spoiled by tools like ntpd in Linux, keeps the
clock accurate continuously within milliseconds using the pool.


I forgot. In a Windows Domain with Active Directory, keeping all the
machines in sync is crucial. A small error and you simply can not login
(Kerberos). It doesn't matter if the clock is correct, but all the
machines need have the same time.

Must also use the same time server. I recall one version of windows
had the time server defaulted to a ms time server; the good thing was
that one can change the time server to any available sever.

Looks like win 10 2004 does not allow changing the server from
default 'time.nist.gov'. AND no longer able to change the time server
to the Fort Collins to be in sync with my 'atomic' wall clocks.

Lew,

Windows 10 ver. 2004 DOES allow changing time servers!

Go to Control Panel--Date and Time--Internet Time--Change Settings The pull down menu has a

few NTP time servers listed or you can manually type in your own.

Just an FYI...

JT


Thanks. I had used the 'settings' to try to navigate to the date/time
setting; unable to find a place. Just tried again & went much deeper
in navigation & the same date/time panel showed up as the one for
the control panel.

A more "useful & friendly" ui? From what I have been using, windows
is becoming more unuseful & unfriendly as more & more stuff are being
hidden.

Brian Gregory wrote:
On 12/07/2020 02:22, lew wrote:
Must also use the same time server.

I generally use at least 4 NTP servers.

I run Meinberg NTP client, the one built in to Windows is just a toy.

Sheesh! Don't you guys have your *own* caesium standard!?

FYI, 'we' (HP) used to make this stuff.

On 2020-07-11 1:07 p.m., John Doe wrote:
https://www.time.gov/

My clock is ticking in unison with that clock.


Mine is only 0.011s off...

....at least, on one particular load of that page.

:-)

On 2020-07-11 4:42 p.m., Carlos E.R. wrote:
On 12/07/2020 01.35, Carlos E.R. wrote:
On 12/07/2020 00.22, Paul wrote:

...

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

Wow, one week. I must be spoiled by tools like ntpd in Linux, keeps
the clock accurate continuously within milliseconds using the pool.


I forgot. In a Windows Domain with Active Directory, keeping all the
machines in sync is crucial. A small error and you simply can not login
(Kerberos). It doesn't matter if the clock is correct, but all the
machines need have the same time.


It depends what you mean by "small error". It's been a while since I had
to work with an AD environment, but I believe you had to be about 5
minutes out for it to be a problem.

Frank Slootweg wrote:

Sheesh! Don't you guys have your *own* caesium standard!?

FYI, 'we' (HP) used to make this stuff.

Curious Mark has had a look at a 5061A ...

Paul wrote:
Rene Lamontagne wrote:
On 2020-07-11 7:26 p.m., Paul wrote:
Rene Lamontagne wrote:
On 2020-07-11 5:22 p.m., Paul wrote:
Big Al wrote:
On 7/11/20 4:07 PM, this is what John Doe wrote:
https://www.time.gov/

My clock is ticking in unison with that clock.

The page says mine is off by 40 seconds, but my clock flips to the
next minute the same time the web page does, so I'm not sure what
they mean other than maybe the rtc on the motherboard.

Al

The RTC is not typically accessed while the OS runs.

It's accessed at some point, to transfer the "software time"
to the RTC for safe keeping. The interface is too slow for any
practical purpose. RTC has 1 second resolution. Hardware
clock (part of software time system) is 100ns or so.

Startup
RTC -- software clock # prime software clock
...
software clock -- RTC # save software clock
shutdown

The RTC clock (32768Hz, ripple divider to 1 second) and
the hardware clock (100MHz BCLK or higher), they drift at
different rates because they use different xtals. Which
means the drift when you shutdown, is different than
the drift rate while the box runs.

Windows time synchronizes infrequently, and so an error
can accumulate between sync points. A period of a week
is selected, to avoid overloading timeservers. Timeservers
(should) have anti-hammering provisions - if you set the
sync interval to one minute, a good server will cut you off.
The pool.org server might not, because it's a pool and
has some form of load balancer. I don't personally
encourage the usage of pool.org . Use one of the government
ones, leave it set to a week etc.

And time servers have been off by as much as 20 minutes.
If you're getting complaints from your OS that "the clock is off",
it's quite likely to be true. When an official server is "off",
it can just as easily be a "whoa boy" sized error :-) They won't
be off by 3 nanoseconds. Just go whole hog and shoot for 20 minutes.
I have no idea how the thing can go off by 20 minutes. Some
clever software calc of some sort (atomic clock consolidation?).
It'll probably turn out, the whole thing is based on a battery
operated ticky clock on the control room wall :-)

Paul

Damn, my clock drifted since I checked it earlier at 3:33 today it
was out by .321 seconds, now its out by .971 seconds.
What to do, What to do? Sigh. :-)

Rene

Ah, a time fetish ?

GPS discipline.

And either the hardware comes with a downloadable
software to tie the GPS time into the computer,
or you search for applications yourself. I don't
see GPS in any of the screenshots here, but it might
be there.

https://www.meinbergglobal.com/english/sw/ntp.htm

For software, maybe NFS servers in Linux, might
appreciate tight time keeping. I don't know if
Windows shares particularly care.

Way back when, it would have cost a thousand
bucks for this, but it's a lot cheaper now to
discipline a PC.

Hold your GPS near a window, to get some signal.

Paul

Nah, just bored. :-)

Rene

Mine is off by 2.7 seconds. What a slob.

But that's after six days.

Paul

This is my latest result. I'm missing a connection
from the 1PPS active high signal, to the DCD pin on
my TTL serial dongle. And unfortunately, I bought the
wrong one, not knowing that I needed to hook up such
crap to use the thing :-/ I need one with a full
breakout board, rather than the four wire cable one.

https://i.postimg.cc/hvyYYF4R/flakey...experiment.gif

1PPS +--+ +--+
| | | |
---------+ +----------------------+ +-------- (tied to DCD carrier detect
^ to generate interrupt and
| strobe in the time)
| XXXXXXX $GPGGA...
| ^
+-- this edge ties |
to this GPGGA msg +

Without the PPS available as a interrupt on the serial port,
the clocking in off the $GPGGA message is flakey pastry. Timing
depends on message length.

Paul

Andy Burns wrote:
Frank Slootweg wrote:

Sheesh! Don't you guys have your *own* caesium standard!?

FYI, 'we' (HP) used to make this stuff.

Curious Mark has had a look at a 5061A ...

I had no idea which Mark you were talking about, until I did a Google
search on 'HP 5061A' and then found CuriousMarc's (no space, 'c' instead
of 'k') 'HP 5061A Cesium Clock' YouTube video (as recent as March 20!).

https://www.youtube.com/watch?v=eOti3kKWX-c

The leader says '1968 ...' which is when I started at HP.

I was not directly involved in caesium standards, but we could win
over a matron (? female head of a hospital) to our medical equipment
with them.

She was unimpressed with the state-of-the-art electronics [1] in our
medical equipment and the sales rep was about to give up, when he
referred to our caesium standards: "They're not really all that good!
They're off by a second every couple of million years!" *That* impressed
her! If a company can pull that off, they must be something special.
They were a loyal customer ever since.

[1] Such as protection against microAmp currents in heart-monitoring
equipment.

On 12/07/2020 03.52, Paul wrote:
Brian Gregory wrote:
On 12/07/2020 02:22, lew wrote:
Must also use the same time server.

I generally use at least 4 NTP servers.

I run Meinberg NTP client, the one built in to Windows is just a toy.


And that's the puzzling part.

Why is the Windows one, the way it is ?

On the ntpd site I seem to recall there were comments about this. Simply
Windows could not do it (discipline the local clock).


--
Cheers, Carlos.