The windows clock is sometimes correct and sometimes off by 10 or so
minutes.
At 10:11 PM it said the time was 10:01 PM.
I have seen this several times over several days.
but sometimes it is correct.
What is wrong ?
Where do I look ?

On 9/5/2021 10:14 PM, AndyP wrote:
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

Is the 2nd entry (labeled "Answer") of any use?

https://answers.microsoft.com/en-us/windows/forum/all/windows-time-is-always-wrong-in-64bit-win-7/fdfb4472-552b-45d9-bf7d-3173942f1193

On 9/5/2021 10:14 PM, AndyP wrote:
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

Have you looked at the things this article points out?

https://www.4winkey.com/windows-7/windows-7-time-keeps-changing-randomly.html

AndyP wrote:
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

Run this

https://time.gov/

and scroll down to see your total time error.

Check your CMOS battery, as it could be
the reason for the lack of counting properly.
Set multimeter to the 10V scale, place the red
lead on the top of the battery, touch the black
lead to chassis ground (like an I/O screw in the
I/O plate area will work for this purpose). 3.0V
is normal. Below 2.3V or so, time to replace it.

If you leave the PC switched on at the back,
then the ATX PSU provides standby power and then
the CMOS battery is not being used. Your counting
issue likely arises from the PC being completely
devoid of AC power at night (switched off at the back).

Paul

On 9/5/2021 10:14 PM, AndyP wrote:
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

I use an application called Socket Watch. It has a master list of 155
NTP (network time protocol) clocks around the world that are on the
Internet. Some are actual atomic clocks, termed Stratum 1. Others
synchronize to Stratum 1 and are termed Stratum 2. Yet others
synchronize to Stratum 2 and are termed Stratum 3. When Socket Watch is
initialized, it queries all the clocks in its list and scores their
responses. The score is based on how long it takes a clock to respond,
which means a low score is better than a high score.

I set Socket Watch to query the five clocks that have the lowest scores
on the list every hour. The first clock to respond resets my Windows 7
clock. While doing that, Socket Watch again scores those five clocks.
Any of the five that has a higher score than a clock on the master list
is replaced by the clock from the master list with a lower score. In
general, my Windows clock is adjusted by less than a second every hour.

Originally, I paid Can$10 for Socket Watch. Unfortunately, the Canadian
developer is no longer in business. You can get it now as freeware from
CNet at
<https://download.cnet.com/SocketWatch/3001-2350_4-10002491.html>.

Existing NTP clocks get shut down, and new clocks are put online. If
you install Socket Watch, let me know in a reply to this thread. I will
make an updated master list of NTP clocks available to you. I last
updated the list in March 2020.

--
David E. Ross
"A Message to Those Who Are Not Vaccinated"
See my <http://www.rossde.com/index#vaccine>.

On Mon, 06 Sep 2021 01:53:11 -0400, Paul wrote:
> AndyP wrote:
>> The windows clock is sometimes correct and sometimes off by 10 or so
>> minutes.
>> At 10:11 PM it said the time was 10:01 PM.
>> I have seen this several times over several days.
>> but sometimes it is correct.
>> What is wrong ?
>> Where do I look ?
>
> Run this
>
> https://time.gov/
>
> and scroll down to see your total time error.
>
> Check your CMOS battery, as it could be
> the reason for the lack of counting properly.
> Set multimeter to the 10V scale, place the red
> lead on the top of the battery, touch the black
> lead to chassis ground (like an I/O screw in the
> I/O plate area will work for this purpose). 3.0V
> is normal. Below 2.3V or so, time to replace it.
>
> If you leave the PC switched on at the back,
> then the ATX PSU provides standby power and then
> the CMOS battery is not being used. Your counting
> issue likely arises from the PC being completely
> devoid of AC power at night (switched off at the back).
>
> Paul

I could understand why OSes don't use CMOS data to provide time data,
because accessing the CMOS NVRAM is quite slow.

So, why haven't motherboard manufacturers come up with a better solution
ever since the first PC? e.g. use a faster CMOS NVRAM, and/or a more
accurate crystal clock electronic component. Cause I find it disappointing
that PCs' time accuracy is no better than wrist watches.

JJ wrote:
> On Mon, 06 Sep 2021 01:53:11 -0400, Paul wrote:
>> AndyP wrote:
>>> The windows clock is sometimes correct and sometimes off by 10 or so
>>> minutes.
>>> At 10:11 PM it said the time was 10:01 PM.
>>> I have seen this several times over several days.
>>> but sometimes it is correct.
>>> What is wrong ?
>>> Where do I look ?
>> Run this
>>
>> https://time.gov/
>>
>> and scroll down to see your total time error.
>>
>> Check your CMOS battery, as it could be
>> the reason for the lack of counting properly.
>> Set multimeter to the 10V scale, place the red
>> lead on the top of the battery, touch the black
>> lead to chassis ground (like an I/O screw in the
>> I/O plate area will work for this purpose). 3.0V
>> is normal. Below 2.3V or so, time to replace it.
>>
>> If you leave the PC switched on at the back,
>> then the ATX PSU provides standby power and then
>> the CMOS battery is not being used. Your counting
>> issue likely arises from the PC being completely
>> devoid of AC power at night (switched off at the back).
>>
>> Paul
>
> I could understand why OSes don't use CMOS data to provide time data,
> because accessing the CMOS NVRAM is quite slow.
>
> So, why haven't motherboard manufacturers come up with a better solution
> ever since the first PC? e.g. use a faster CMOS NVRAM, and/or a more
> accurate crystal clock electronic component. Cause I find it disappointing
> that PCs' time accuracy is no better than wrist watches.

The drift is a function of the quartz crystal, plus
the sophistication of the associated parts. I think there
is a small bit of room for improvement there, but I couldn't
tell you what the cost/benefit ratio of such changes would be.

The PC implementation draws 10uA from the 3V or 3.3V source.

Your wrist watch draws 2uA from its power source. Not a lot
of drive is required to work the resonant behavior of the
quartz (piezoelectric) crystal.

The 8uA difference might partially be attributed to
the design of the CMOS well (leakage through the
transmission gates). It's strange that behaviorally,
this number seems so consistent (the time for a CR2032
cell to wear out when used constantly). As the PCH technology
level changes, the CMOS well leakage should randomly
change too.

To buttress the timekeeping function with external
facilities, is an unnecessary burden on the user. If
GPS disciplined, GPS only reliably works with
an outdoor antenna. I sorta ran mine sitting in a
window, and in view of the sky, but it also had
outages for periods of time, where the two items
in the constellation it could reliably see, would
have too low a signal amplitude.

There are also LF radio time broadcasting systems, but
their existence is not guaranteed. Those probably have
better wall penetrating power than GPS signals, but the
transmitter can hardly cover whole continents and uses
a fair amount of power. The coverage maps for terrestrial
schemes, does not guarantee 24 hour a day availability.
Maybe an hour a day in the evening, there might be a
good signal. That means you'd sync up once a day with
any regularity, unless you happened to be right next
to the antenna field.

I bet we could come up with a way to include this
function (distributed) on web servers. But then we
would be doing a disservice to machine owners who keep
their client machines airgapped.

A cesium timepiece would not run off 10uA. Just the
operating frequency considerations alone, even the
front end divider of the signal would burn way more
electricity than a quartz timepiece. The quartz on
a motherboard really is a digital watch, because it
uses the same crystal as a watch does (32KHz).

Connecting up a GPS, onto a serial port, is probably
the closest thing to an able assistant. Just don't
use it on OS platforms that have location() services
that will be using the serial stream to their advantage.
I considered GPS to be "reasonably safe" on Windows XP :-)

Paul

On Sun, 5 Sep 2021 22:14:31 -0700, AndyP wrote:
>
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

I can't tell you what's wrong, though some others have made
suggestions that you should definitely follow.

But I do have a workaround: set Windows to synchronize with Internet
time.

https://www.majorgeeks.com/content/page/synchronize_clock_with_an_int
ernet_time_server.html
(The page tries to cover Windows 10, 8, and 7 all together, so look
past the screen shots that don't match Windows 7.)

Here's a specifically Windows 7 video:
https://www.youtube.com/watch?v=-H9pK37EIWM

--
Stan Brown, Tehachapi, California, USA https://BrownMath.com/
https://OakRoadSystems.com/
Shikata ga nai...

AndyP wrote:
> The windows clock is sometimes correct and sometimes off by 10 or so
> minutes.
> At 10:11 PM it said the time was 10:01 PM.
> I have seen this several times over several days.
> but sometimes it is correct.
> What is wrong ?
> Where do I look ?

Is your comp over 5 yrs old?
Check the battery.
To add do Paul NoSpam's excellent post, check the battery with the power
off and comp unplugged from a/c power. With the computer unplugged
from a/c power, press the start button once wait 5 min, then test the
battery.

On Mon, 06 Sep 2021 07:38:53 -0400, Paul wrote:
>
> To buttress the timekeeping function with external
> facilities, is an unnecessary burden on the user. If
> GPS disciplined, GPS only reliably works with
> an outdoor antenna. I sorta ran mine sitting in a
> window, and in view of the sky, but it also had
> outages for periods of time, where the two items
> in the constellation it could reliably see, would
> have too low a signal amplitude.
>
> There are also LF radio time broadcasting systems, but
> their existence is not guaranteed. Those probably have
> better wall penetrating power than GPS signals, but the
> transmitter can hardly cover whole continents and uses
> a fair amount of power. The coverage maps for terrestrial
> schemes, does not guarantee 24 hour a day availability.
> Maybe an hour a day in the evening, there might be a
> good signal. That means you'd sync up once a day with
> any regularity, unless you happened to be right next
> to the antenna field.
>
> I bet we could come up with a way to include this
> function (distributed) on web servers. But then we
> would be doing a disservice to machine owners who keep
> their client machines airgapped.
>
> A cesium timepiece would not run off 10uA. Just the
> operating frequency considerations alone, even the
> front end divider of the signal would burn way more
> electricity than a quartz timepiece. The quartz on
> a motherboard really is a digital watch, because it
> uses the same crystal as a watch does (32KHz).
>
> Connecting up a GPS, onto a serial port, is probably
> the closest thing to an able assistant. Just don't
> use it on OS platforms that have location() services
> that will be using the serial stream to their advantage.
> I considered GPS to be "reasonably safe" on Windows XP :-)
>
> Paul

How acurate is GPS based time synchronization? Does it take a long time to
initialize (getting ready to constantly receive data)? How frequent can it
receive update of the time (e.g. every 10 seconds)? And how much is a "good
enough" GPS device? Preferrably a standalone one which doesn't require
internet.

FYI...

For a side work, I'm involved with a small non commercial organization which
has a very limited budget. Like, about US$75, or maybe US$120 if I'm pushing
it (don't know if it's still enough). I would need two GPS devices, since I
need two laptops to be time synchronized. The laptops are used outdoor on
the field, mostly in rural places where cellular signal is very weak -
making it unreliable for internet access. Both laptops are far apart from
about 100 meters, up to about 2 kilometers.

The time itself doesn't have to be accurate to the world clock. I just need
the two laptops to have identical time for at least 2 hours with only up to
1/10th of a second time difference.

"JJ" <jj4public@gmail.com> wrote in message
news:rkt0xifyuael$.1h11lye0naoj$.dlg@40tude.net...

> How acurate is GPS based time synchronization? Does it take a long time to
> initialize (getting ready to constantly receive data)? How frequent can it
> receive update of the time (e.g. every 10 seconds)? And how much is a
> "good
> enough" GPS device? Preferrably a standalone one which doesn't require
> internet.

My understanding is that GPS time is very accurate - to far more accuracy
that the number of decimal places that are quoted on GPS clocks. All the
satellites transmit a timestamp that is accurate enough that the differences
in timestamps due to differing distances between satellites and receiver can
be used to calculate the receiver's position.

GPS can take up to 15 minutes (or is it 7.5 minutes) to get a first GPS fix,
because the "almanac" and "ephemeris" data (knowledge of where all the
satellites are in the sky at a particular time) is only transmitted that
often. Having got a GPS fix, subsequent ones can be polled once a second -
maybe more frequently but I've not come across any third-party software that
polls more frequently. I imagine that power consumption increases as you
poll more frequently. It may well be that a timestamp to normal levels of
accuracy (say +/- 1 millisecond) can be obtained immediately, even before
the ephemeris/almanac are updated, though it may need special control of the
GPS to return this partial information before a full
latitude/longitude/altitude fix is known. It is possible for apps to
download the almanac/ephemeris by internet connection even when the version
sent in the GPS signal is "stale": I have an Android app "GPS Status" which
returns all sorts of GPS info, and has the ability to synchronise with the
almanac as long as you have a wifi or mobile internet connection.

Having got one good GPS fix, you are set up for life, until you turn off the
GPS (as opposed to simply not polling it for results). My phone has GPS
turned on 24/7 and it almost always gets a fix within a couple of seconds,
as long as there is sight of several satellites. If I were to turn off GPS
(in the Android pull-down menu) and move to a very different location before
turning back on, I may have to wait for the broadcast almanac to be next
transmitted - or use the downloadable version.

On Sun, 5 Sep 2021 at 23:15:01, David E. Ross <not_me@not_there.invalid>
wrote (my responses usually follow points raised):
[]
>I set Socket Watch to query the five clocks that have the lowest scores
>on the list every hour. The first clock to respond resets my Windows 7
>clock. While doing that, Socket Watch again scores those five clocks.
>Any of the five that has a higher score than a clock on the master list
>is replaced by the clock from the master list with a lower score. In
>general, my Windows clock is adjusted by less than a second every hour.

Seems a competent approach.
>
>Originally, I paid Can$10 for Socket Watch. Unfortunately, the Canadian
>developer is no longer in business. You can get it now as freeware from
>CNet at
> <https://download.cnet.com/SocketWatch/3001-2350_4-10002491.html>.
>
>Existing NTP clocks get shut down, and new clocks are put online. If
>you install Socket Watch, let me know in a reply to this thread. I will

I've installed it.

>make an updated master list of NTP clocks available to you. I last
>updated the list in March 2020.
>
Are you the developer then? I was wondering how it learns about new time
servers.
--
J. P. Gilliver. UMRA: 1960/<1985 MB++G()AL-IS-Ch++(p)Ar@T+H+Sh0!:`)DNAf

The squeamish will squeam a lot.
(Barry Norman on the film "300", in Radio Times 30 March-5 April 2013.)

On 9/7/2021 6:43 AM, J. P. Gilliver (John) wrote:
> On Sun, 5 Sep 2021 at 23:15:01, David E. Ross <not_me@not_there.invalid>
> wrote (my responses usually follow points raised):
> []
>> I set Socket Watch to query the five clocks that have the lowest scores
>> on the list every hour. The first clock to respond resets my Windows 7
>> clock. While doing that, Socket Watch again scores those five clocks.
>> Any of the five that has a higher score than a clock on the master list
>> is replaced by the clock from the master list with a lower score. In
>> general, my Windows clock is adjusted by less than a second every hour.
>
> Seems a competent approach.
>>
>> Originally, I paid Can$10 for Socket Watch. Unfortunately, the Canadian
>> developer is no longer in business. You can get it now as freeware from
>> CNet at
>> <https://download.cnet.com/SocketWatch/3001-2350_4-10002491.html>.
>>
>> Existing NTP clocks get shut down, and new clocks are put online. If
>> you install Socket Watch, let me know in a reply to this thread. I will
>
> I've installed it.
>
>> make an updated master list of NTP clocks available to you. I last
>> updated the list in March 2020.
>>
> Are you the developer then? I was wondering how it learns about new time
> servers.
>

No, I am not the developer. I just updated the list yesterday. I did
it manually by navigating through the Web site at
<https://support.ntp.org/bin/view/Servers/WebHome>. Are you interested
in a copy? If so, reply here.

--
David E. Ross
"A Message to Those Who Are Not Vaccinated"
See my <http://www.rossde.com/index#vaccine>.

On Tue, 7 Sep 2021 at 08:11:21, David E. Ross <not_me@not_there.invalid>
wrote (my responses usually follow points raised):
>On 9/7/2021 6:43 AM, J. P. Gilliver (John) wrote:
>> On Sun, 5 Sep 2021 at 23:15:01, David E. Ross <not_me@not_there.invalid>
>> wrote (my responses usually follow points raised):
>> []
>>> I set Socket Watch to query the five clocks that have the lowest scores
[]
>> Are you the developer then? I was wondering how it learns about new time
>> servers.
>>
>
>No, I am not the developer. I just updated the list yesterday. I did
>it manually by navigating through the Web site at
><https://support.ntp.org/bin/view/Servers/WebHome>. Are you interested
>in a copy? If so, reply here.
>
Couldn't do any harm, thanks! Is it obvious where to put it?
--
J. P. Gilliver. UMRA: 1960/<1985 MB++G()AL-IS-Ch++(p)Ar@T+H+Sh0!:`)DNAf

The squeamish will squeam a lot.
(Barry Norman on the film "300", in Radio Times 30 March-5 April 2013.)

Sorry, this is a laptop.
More screws than I can count.
Always plugged in to A/C.
Battery at 70% (safe max charge).

Paul wrote:
> AndyP wrote:
>> The windows clock is sometimes correct and sometimes off by 10 or so
>> minutes.
>> At 10:11 PM it said the time was 10:01 PM.
>> I have seen this several times over several days.
>> but sometimes it is correct.
>> What is wrong ?
>> Where do I look ?
>
> Run this
>
> https://time.gov/
>
> and scroll down to see your total time error.
>
> Check your CMOS battery, as it could be
> the reason for the lack of counting properly.
> Set multimeter to the 10V scale, place the red
> lead on the top of the battery, touch the black
> lead to chassis ground (like an I/O screw in the
> I/O plate area will work for this purpose). 3.0V
> is normal. Below 2.3V or so, time to replace it.
>
> If you leave the PC switched on at the back,
> then the ATX PSU provides standby power and then
> the CMOS battery is not being used. Your counting
> issue likely arises from the PC being completely
> devoid of AC power at night (switched off at the back).
>
>    Paul

JJ wrote:

> How acurate is GPS based time synchronization? Does it take a long time to
> initialize (getting ready to constantly receive data)? How frequent can it
> receive update of the time (e.g. every 10 seconds)? And how much is a "good
> enough" GPS device? Preferrably a standalone one which doesn't require
> internet.
>
> FYI...
>
> For a side work, I'm involved with a small non commercial organization which
> has a very limited budget. Like, about US$75, or maybe US$120 if I'm pushing
> it (don't know if it's still enough). I would need two GPS devices, since I
> need two laptops to be time synchronized. The laptops are used outdoor on
> the field, mostly in rural places where cellular signal is very weak -
> making it unreliable for internet access. Both laptops are far apart from
> about 100 meters, up to about 2 kilometers.
>
> The time itself doesn't have to be accurate to the world clock. I just need
> the two laptops to have identical time for at least 2 hours with only up to
> 1/10th of a second time difference.

GPS has very good temporal characteristics.

The problem is interfacing it with the PC. Not
all the implementations make this particularly easy.

This is the one I got. You don't have to buy this one,
and instead you could look for one better suited to
your circumstances.

https://www.adafruit.com/product/746

That one has TTL level serial port and a separate PPS signal.

*******

Part of what GPS does, is the satellites send "messages".

They come out of the receiver, as a serial stream at
9600 baud or so. The receiver listens to multiple
satellites and delivers messages from them. Where
I'm sitting, the receiver can see seven birds, but
only two have "excellent" (green) signal level.

<message...> Contains time information
<message...> Contains latitude and longitude information
<message...> (Sufficient info to record waypoints)
<message...>
....quiet...
<message...>
<message...>
<message...>
<message...>

Now, the PPS is a Pulse Per Second signal, which
has a very accurately timed edge. The signal can
be tied to DCD or RI on a full serial port, as a
means to "get an interrupt event in on the serial port".
The more direct that interrupt path (such as inband
ints on PCI Express full serial ports), the lower
the latency for interrupt servicing and registering
"the accurate second".

The trick then, is what driver you use, and what
the driver is looking for. Like, if the driver is
looking for a Data Carrier Detect interrupt, then
you would tie the PPS signal (TTL level) to DCD
(TTL level).

This is probably why I bought the module with the
raw signals on it... expecting trouble.

If you don't use PPS, and just receive messages, there
will be some additional error.

The edge of PPS could be as accurate as 10nsec. It
is a digital signal clocked from something on the
processor inside the receiver. I don't think the
signal is clocked with the GPS carrier frequency
(which is higher), so the clocking source must
be something intermediate.

On a desktop PC, the ideal interface card would be
a PCI Express serial port with TTL level interface.
When I bought my PCI Express card years ago (before
even thinking about doing such a project), I didn't
get the "full featured" card, and now the company
that made the chips, was bought out and crushed
(that was OxSemi). I could have had excellent results,
if I hadn't been a cheapskate when the cards were
available.

There is some company with a fab in India making
silicon for serial ports, and that would be the
next thing for me to investigate.

You on the other hand, need something a bit more
neat and tidy. A USB interface device with the
PPS tied already to a usable signal, and then
the resolution is the USB2 polling interval
(125usec or more of jitter perhaps).

So yes, on the face of it, it has great potential.
The "message" content says what time it is, in
advance of the rising edge of PPS. The rising edge
of PPS "clocks in" the message, for anything listening.
The NTPD driver absorbs the message and stores it.
Then, when a PPS rising edge causes an interrupt,
the driver passes the timestamp up to the OS level
subsystem.

*******

For example, this one has a USB port for interface,
and PPS is tied to RI (presumably at PCB level, not
a hidden signal).

https://www.adafruit.com/product/4279

You can see PPS is tied to RI via a copper track.
This is probably a two layer PCB, and the signal
in question changes layers (part of the signal
is red, part is blue).

https://learn.adafruit.com/assets/76956

https://cdn-learn.adafruit.com/assets/assets/000/079/165/large1024/adafruit_products_Ultimate_GPS_USB_Fab_Print.png?1565208353

Yet some of the software, assumes DCD has PPS tied to it.

https://kb.meinbergglobal.com/kb/time_sync/ntp/ntp_for_windows/using_pps_signals_on_windows

It would help if everyone played by the same rules,
or at least in the case of the Adafruit product,
if they placed gold stake pins so the RI and DCD
could be wired to PPS with a jumper plug. (If you
arrange the gold pins at angles to the source pin,
the jumper plug can be rotated 90 degrees to pick
up the gold pin with the signal you want. That's how
we'd do it in "the jumpering era".)

That's to give you some idea what a lack of
standards causes... Almost there... except for
a tiny, nagging, detail.

*******

This is to give you some idea what my ntp.conf looks like.

ntp.conf

restrict default noquery nopeer nomodify notrap
restrict -6 default noquery nopeer nomodify notrap

# allow status queries and everything else from localhost
restrict 127.0.0.1
# IPV6 stuff... restrict -6 ::1

# Use drift file
driftfile "C:\Program Files\NTP\etc\ntp.drift" <=== need some place writeable!

# your local system clock, should be used as a backup
# (this is only useful if you need to distribute time no matter how good or bad it is)
#server 127.127.1.0
#server 127.127.20.5 minpoll 4 prefer # NMEA serial port
#server 127.127.22.1 minpoll 4 # PPS - serialpps.sys

server 127.127.20.5 minpoll 4 mode 18 prefer # NMEA serial port, 16 = 9600 baud, 2 = $GPGGA

# but it operates at a high stratum level to let the clients know and force them to
# use any other timesource they may have.
#fudge 127.127.1.0 stratum 1

fudge 127.127.20.5 flag1 0 time2 0.447

# Use a NTP server from the ntp pool project (see http://www.pool.ntp.org)
# Please note that you need at least four different servers to be at least protected against
# one falseticker. If you only rely on internet time, it is highly recommended to add
# additional servers here.
# The 'iburst' keyword speeds up initial synchronization, please check the documentation for more details!
# server 0.us.pool.ntp.org iburst minpoll 6 maxpoll 7
# server 1.us.pool.ntp.org iburst minpoll 6 maxpoll 7
# server 2.us.pool.ntp.org iburst minpoll 6 maxpoll 7

# End of generated ntp.conf --- Please edit this to suite your needs

*******

"What is the fudge line for in NTP ? "
https://www.tek-tips.com/viewthread.cfm?qid=336666

The choices:

PCI Express serial port - lowest interrupt latency and jitter
USB2 serial port - larger latency and jitter
No DCD/RI and PPS used - good to large fraction of a second

HTH,
Paul

On 9/7/2021 9:46 AM, J. P. Gilliver (John) wrote:
> On Tue, 7 Sep 2021 at 08:11:21, David E. Ross <not_me@not_there.invalid>
> wrote (my responses usually follow points raised):
>> On 9/7/2021 6:43 AM, J. P. Gilliver (John) wrote:
>>> On Sun, 5 Sep 2021 at 23:15:01, David E. Ross <not_me@not_there.invalid>
>>> wrote (my responses usually follow points raised):
>>> []
>>>> I set Socket Watch to query the five clocks that have the lowest scores
> []
>>> Are you the developer then? I was wondering how it learns about new time
>>> servers.
>>>
>>
>> No, I am not the developer. I just updated the list yesterday. I did
>> it manually by navigating through the Web site at
>> <https://support.ntp.org/bin/view/Servers/WebHome>. Are you interested
>> in a copy? If so, reply here.
>>
> Couldn't do any harm, thanks! Is it obvious where to put it?
>

I will compose complete instructions, effectively describing what I did
yesterday. Stand by.

--
David E. Ross
"A Message to Those Who Are Not Vaccinated"
See my <http://www.rossde.com/index#vaccine>.

On Tue, 7 Sep 2021 10:57:53 -0700, AndyP wrote:
>
> Sorry, this is a laptop.
> More screws than I can count.
> Always plugged in to A/C.
> Battery at 70% (safe max charge).

If it's always plugged in, how can it stay at 70% charge?

--
Stan Brown, Tehachapi, California, USA https://BrownMath.com/
https://OakRoadSystems.com/
Shikata ga nai...

On 9/7/2021 12:47 PM, David E. Ross wrote:
> On 9/7/2021 9:46 AM, J. P. Gilliver (John) wrote:
>> On Tue, 7 Sep 2021 at 08:11:21, David E. Ross <not_me@not_there.invalid>
>> wrote (my responses usually follow points raised):
>>> On 9/7/2021 6:43 AM, J. P. Gilliver (John) wrote:
>>>> On Sun, 5 Sep 2021 at 23:15:01, David E. Ross <not_me@not_there.invalid>
>>>> wrote (my responses usually follow points raised):
>>>> []
>>>>> I set Socket Watch to query the five clocks that have the lowest scores
>> []
>>>> Are you the developer then? I was wondering how it learns about new time
>>>> servers.
>>>>
>>>
>>> No, I am not the developer. I just updated the list yesterday. I did
>>> it manually by navigating through the Web site at
>>> <https://support.ntp.org/bin/view/Servers/WebHome>. Are you interested
>>> in a copy? If so, reply here.
>>>
>> Couldn't do any harm, thanks! Is it obvious where to put it?
>>
>
> I will compose complete instructions, effectively describing what I did
> yesterday. Stand by.
>

Okay. On the Web, go to <http://www.rossde.com/test/export%20servers/>.
Download both ReadMe.txt and servers.lst. Read the ReadMe.txt file.
If you have questions, my E-mail address is near the end of that file.

--
David E. Ross
"A Message to Those Who Are Not Vaccinated"
See my <http://www.rossde.com/index#vaccine>.

Toshiba conserves batter life by not charging all the way unless I
override it.

Old laptop and battery is still good !

Stan Brown wrote:
> On Tue, 7 Sep 2021 10:57:53 -0700, AndyP wrote:
>>
>> Sorry, this is a laptop.
>> More screws than I can count.
>> Always plugged in to A/C.
>> Battery at 70% (safe max charge).
>
> If it's always plugged in, how can it stay at 70% charge?
>
>

On Tue, 07 Sep 2021 14:41:25 -0400, Paul wrote:
>
> GPS has very good temporal characteristics.
>
> The problem is interfacing it with the PC. Not
> all the implementations make this particularly easy.
>
> This is the one I got. You don't have to buy this one,
> and instead you could look for one better suited to
> your circumstances.
>
> https://www.adafruit.com/product/746
>
[snip]
>
> For example, this one has a USB port for interface,
> and PPS is tied to RI (presumably at PCB level, not
> a hidden signal).
>
> https://www.adafruit.com/product/4279

I'll try to find one. Thank you.

> Yet some of the software, assumes DCD has PPS tied to it.
>
> https://kb.meinbergglobal.com/kb/time_sync/ntp/ntp_for_windows/using_pps_signals_on_windows
>
[snip]
> This is to give you some idea what my ntp.conf looks like.
>
> ntp.conf
>
> restrict default noquery nopeer nomodify notrap
> restrict -6 default noquery nopeer nomodify notrap
>
> # allow status queries and everything else from localhost
> restrict 127.0.0.1
> # IPV6 stuff... restrict -6 ::1
>
> # Use drift file
> driftfile "C:\Program Files\NTP\etc\ntp.drift" <=== need some place writeable!
>
> # your local system clock, should be used as a backup
> # (this is only useful if you need to distribute time no matter how good or bad it is)
> #server 127.127.1.0
> #server 127.127.20.5 minpoll 4 prefer # NMEA serial port
> #server 127.127.22.1 minpoll 4 # PPS - serialpps.sys
>
> server 127.127.20.5 minpoll 4 mode 18 prefer # NMEA serial port, 16 = 9600 baud, 2 = $GPGGA

I actually use Meinberg NTP software. Though I'll have to synchronize the
time while the cellular signal is still reliable for internet access.

I didn't know that it also support GPS. But it's good that it does.

On Tue, 7 Sep 2021 15:47:32 -0700, AndyP wrote:
>
> Stan Brown wrote:
> > On Tue, 7 Sep 2021 10:57:53 -0700, AndyP wrote:
> >> Sorry, this is a laptop.
> >> More screws than I can count.
> >> Always plugged in to A/C.
> >> Battery at 70% (safe max charge).
> >
> > If it's always plugged in, how can it stay at 70% charge?
>
> Toshiba conserves batter life by not charging all the way unless I
> override it.
>
> Old laptop and battery is still good !

Cool -- I didn't know there were any laptops that did that.

--
Stan Brown, Tehachapi, California, USA https://BrownMath.com/
https://OakRoadSystems.com/
Shikata ga nai...

Stan Brown <the_stan_brown@fastmail.fm> wrote:
> On Tue, 7 Sep 2021 15:47:32 -0700, AndyP wrote:
> >
> > Stan Brown wrote:
> > > On Tue, 7 Sep 2021 10:57:53 -0700, AndyP wrote:
> > >> Sorry, this is a laptop.
> > >> More screws than I can count.
> > >> Always plugged in to A/C.
> > >> Battery at 70% (safe max charge).
> > >
> > > If it's always plugged in, how can it stay at 70% charge?
> >
> > Toshiba conserves batter life by not charging all the way unless I
> > override it.
> >
> > Old laptop and battery is still good !
>
> Cool -- I didn't know there were any laptops that did that.

Yes, sadly Toshiba is one of the few or the only one which has this
feature.

I've ruined my fair share of laptop batteries, so now our laptops are
on an AC timer, which switches the AC/DC adapters on and off for several
periods during the day.

I don't understand why other laptop manufacturers don't have this
feature. It's only SMOP - Small Matter Of Programming -, because the
needed hardware - ACPI (Advanced Configuration and Power Interface) -
and low-level software is already there.

For example, for HP laptops, HP has an optional utility ('HP AC Power
Control Utility') which disables charging during certain (settable)
hours. If such a utility can disable charging during certain hours, a
similar utility should be able to stop (read: disable) charging at XX%
full.

On Thu, 9 Sep 2021 at 14:36:52, Frank Slootweg <this@ddress.is.invalid>
wrote (my responses usually follow points raised):
>Stan Brown <the_stan_brown@fastmail.fm> wrote:
>> On Tue, 7 Sep 2021 15:47:32 -0700, AndyP wrote:
>> >
>> > Stan Brown wrote:
>> > > On Tue, 7 Sep 2021 10:57:53 -0700, AndyP wrote:
>> > >> Sorry, this is a laptop.
>> > >> More screws than I can count.
>> > >> Always plugged in to A/C.
>> > >> Battery at 70% (safe max charge).
>> > >
>> > > If it's always plugged in, how can it stay at 70% charge?
>> >
>> > Toshiba conserves batter life by not charging all the way unless I
>> > override it.

How _do_ you override it? (Not that I'd want to, just curious to know
whether it's a complex process.)
>> >
>> > Old laptop and battery is still good !
>>
>> Cool -- I didn't know there were any laptops that did that.
>
> Yes, sadly Toshiba is one of the few or the only one which has this
>feature.

It was, IIRR, a KB update. There was some discussion of it at the time
(I think it was in the last couple of years of 7). Sorry, don't remember
the KB number - anyone?
>
> I've ruined my fair share of laptop batteries, so now our laptops are

Three here )-:

>on an AC timer, which switches the AC/DC adapters on and off for several
>periods during the day.
>
> I don't understand why other laptop manufacturers don't have this
>feature. It's only SMOP - Small Matter Of Programming -, because the
>needed hardware - ACPI (Advanced Configuration and Power Interface) -
>and low-level software is already there.

No incentive, I think. In the short term, they can sell replacement
batteries; in the longer term, when people can't find them, they can
sell more laptops.
>
> For example, for HP laptops, HP has an optional utility ('HP AC Power
>Control Utility') which disables charging during certain (settable)
>hours. If such a utility can disable charging during certain hours, a
>similar utility should be able to stop (read: disable) charging at XX%
>full.

Indeed. (Though why stop at 70% - wouldn't 95 or 90 be better?)
--
J. P. Gilliver. UMRA: 1960/<1985 MB++G()AL-IS-Ch++(p)Ar@T+H+Sh0!:`)DNAf

"Bother," said Pooh, as Windows crashed into piglet.

"J. P. Gilliver (John)" <G6JPG@255soft.uk> wrote:

[Lots deleted]

> > For example, for HP laptops, HP has an optional utility ('HP AC Power
> >Control Utility') which disables charging during certain (settable)
> >hours. If such a utility can disable charging during certain hours, a
> >similar utility should be able to stop (read: disable) charging at XX%
> >full.
>
> Indeed. (Though why stop at 70% - wouldn't 95 or 90 be better?)

No, the recommended upper limit is a bit lower, 80% IIRC. If you want
the nitty-gritty details, then head over to BATTERY UNIVERSITY,
specifically (as a starter):

'BU-808: How to Prolong Lithium-based Batteries'
<https://batteryuniversity.com/article/bu-808-how-to-prolong-lithium-based-batteries>

Especially see the table on Depth of discharge versus the number of
Discharge cycles.

Bottom line: Not fully charging and not fully discharging prolongs
life and preserves capacity.