Hi:

I have had excellent results using

http://www.meinberg.de/download/ntp/windows/ntp-4.2.4p8@lennon-o-lpv-win32-setup.exe

on a windows XP platform, but on windows 7 I never get a good sync.
(The NTP server is a high quality statum 2 time server which provides
excellent time to Fedora/Linux clients.)  I was wondering   if anyone
has any experience w/ windows 7 with Meinberg (or any other NTP)
software.

Gautam

Gautam Thaker wrote:
> Hi:
> 
> I have had excellent results using
> 
> http://www.meinberg.de/download/ntp/windows/ntp-4.2.4p8@lennon-o-lpv-win32-setup.exe
> 

The only Meinberg software is the installer.  The rest is a slightly 
dated version of the reference implementation from the University of 
Delaware.

> Hi:
>
> I have had excellent results using
>
> http://www.meinberg.de/download/ntp/windows/ntp-4.2.4p8@lennon-o-lpv-win32-setup.exe
>
> on a windows XP platform, but on windows 7 I never get a good sync.
> (The NTP server is a high quality statum 2 time server which provides
> excellent time to Fedora/Linux clients.)  I was wondering   if anyone
> has any experience w/ windows 7 with Meinberg (or any other NTP)
> software.
>
> Gautam

Gautam, I have also found that Windows Vista and Windows-7 are not such 
good time keepers as Windows XP.  Windows-7 32-bit will work as a 
stratum-1 server with a GPS/PPS serial reference:

  http://www.satsignal.eu/mrtg/stamsund_ntp_2.html

being typically within 50 microseconds, and I have other systems, 
including a Windows-7 64-bit system synced over the LAN:

  http://www.satsignal.eu/mrtg/alta-ntp-b.html

which is within a couple of milliseconds (using 32s poll, 
minpoll=maxpoll=5).  The problem is that the rather excellent 
interpolation scheme developed for Windows cannot be used with the later 
versions from Vista onwards which have a higher clock interrupt rate.

You can see the performance of all my systems here:

  http://www.satsignal.eu/mrtg/performance_ntp.php

If timekeeping on Windows is critical, consider getting a PPS signal sent 
round to the PCs which matter.  You need pay no more that US $35 for a PPS 
source:

  http://www.sureelectronics.net/goods.php?id=99

Be aware that later versions of NTP are available.

Cheers,
David 

David Woolley wrote:
> Gautam Thaker wrote:
>> Hi:
>>
>> I have had excellent results using
>>
>> http://www.meinberg.de/download/ntp/windows/ntp-4.2.4p8@lennon-o-lpv-win32-setup.exe
>>
>>
>
> The only Meinberg software is the installer. The rest is a slightly
> dated version of the reference implementation from the University of
> Delaware.

I'm still compiling (about every two weeks or so) new versions of the 
dev branch code and putting them on my server, i.e.:

http://norloff.org/ntp/dev/

and

http://norloff.org/ntp/

for the regular code which is updated less often.

The way I install ntp on new machines is to use the Meinberg installer, 
then replace the actual binaries with those I've compiled myself.

I've just replaced my running version with code I compiled a few minutes 
ago, it seems to behave well:

C:\install\ntp>ntpq -pnc rv
      remote           refid      st t when poll reach   delay   offset 
  jitter
==============================================================================
+2001:16d8:ee97: .GRMN.           1 u   56   64   17   45.143   -1.955 
  2.284
+2001:16d8:ee97: 192.168.93.123   2 u    -   64   37   44.476   -1.665 
  2.049
+158.38.48.11    139.112.1.21     3 u   42   64    7   17.169   -1.976 
  7.425
+194.63.250.121  192.36.133.17    2 u   42   64    7    8.851    0.347 
  9.408
+192.121.13.5    192.36.134.17    2 u    8   64   17   18.053   -3.110 
  1.971
*79.161.68.172   .GRMN.           1 u   56   64   17   26.805   -0.886 
  1.591
associd=0 status=0618 leap_none, sync_ntp, 1 event, no_sys_peer,
version="ntpd 4.2.7p132@1.2478-o Feb 23 13:55:41 (UTC+01:00) 2011  (1)",
processor="x86", system="Windows", leap=00, stratum=2, precision=-10,
rootdelay=26.805, rootdisp=5.032, refid=79.161.68.172,
reftime=d10f8356.3401a65e  Wed, Feb 23 2011 14:00:06.203,
clock=d10f838e.ab7b3ce8  Wed, Feb 23 2011 14:01:02.669, peer=64862, tc=6,
mintc=3, offset=-0.284, frequency=6.774, sys_jitter=2.574,
clk_jitter=1.605, clk_wander=0.000

Sorry about the wrapping, the key point is that all my servers agree 
that the local clock is just a ms or two off, this is pretty good, 
particularly since it is less than 5 minutes since I restarted ntp.

Terje



-- 
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"

On 2011-02-23, Terje Mathisen <"terje.mathisen at tmsw.no"> wrote:

> I'm still compiling (about every two weeks or so) new versions of the 
> dev branch code and putting them on my server, i.e.:
>
> http://norloff.org/ntp/dev/
>
> and
>
> http://norloff.org/ntp/
>
> for the regular code which is updated less often.

Which is listed under "Ports" at http://support.ntp.org/links

-- 
Steve Kostecke <kostecke@ntp.org>
NTP Public Services Project - http://support.ntp.org/

On 2011-02-23, David J Taylor <david-taylor@blueyonder.co.uk.invalid> wrote:
....
>
> If timekeeping on Windows is critical, consider getting a PPS signal sent 
> round to the PCs which matter.  You need pay no more that US $35 for a PPS 
> source:
>
>   http://www.sureelectronics.net/goods.php?id=99

Except there is no indication that this thing puts out a PPS signal. 


>
> Be aware that later versions of NTP are available.
>
> Cheers,
> David 
>

Once you replace the binaries in

C:\Program Files\NTP\bin

do you need to do anything extra (such as use Instsrv.exe) to register
the new binary. I get a popup saying

"Erorr 5: access is denied".

Any hints welcome.

Gautam

Once you replace the binaries in

C:\Program Files\NTP\bin

do you need to do anything extra (such as use Instsrv.exe) to register
the new binary. I get a popup saying

"Erorr 5: access is denied".

Any hints welcome.

Gautam

sorry about the double post - not sure how that happened.
Anyway, i have uninstalled and reinstalled meinberg. Meinberg is
running now, according to window services status, but I am unable to
see how well it is doing as running the meinberg
NTP_time_server_monitor appears to start up but immediately exists w/
no error messages in any event logs or anywhere tha I can find. And
running C:\Program Files\NTP\bin\ntpq.exe -p    command simply returns
w/

"C:\Program Files\NTP\bin\ntpq.exe : connect : No error"

(I am run both Meinbert NTP time server monitor and ntpq.exe  often
before without problems.)

Any suggestions? Once I get this to work I will update the binaries
w/  those from  file

"ntp-4.2.7p67-win32.zip"

Gautam

On Wed, Feb 23, 2011 at 10:56 AM, unruh <unruh@wormhole.physics.ubc.ca> wro=
te:
> On 2011-02-23, David J Taylor <david-taylor@blueyonder.co.uk.invalid> wro=
te:
> ...
>>
>> If timekeeping on Windows is critical, consider getting a PPS signal sent
>> round to the PCs which matter. =A0You need pay no more that US $35 for a=
 PPS
>> source:
>>
>> =A0 http://www.sureelectronics.net/goods.php?id=3D99
>
> Except there is no indication that this thing puts out a PPS signal.

You need to download and read the user manual.  The PPS output is on pin "T=
P9".

But really, I think the Sure Electronics unit was only used as an
example to show that there are GPS receivers that would work and don't
cost much.  There are many others also under $50 and almost all are
from 10X to 1000X more accurate than needed for NTP.


-- =

=3D=3D=3D=3D=3D
Chris Albertson
Redondo Beach, California

Steve Kostecke wrote:
> On 2011-02-23, Terje Mathisen<"terje.mathisen at tmsw.no">  wrote:
>
>> I'm still compiling (about every two weeks or so) new versions of the
>> dev branch code and putting them on my server, i.e.:
>>
>> http://norloff.org/ntp/dev/
>>
>> and
>>
>> http://norloff.org/ntp/
>>
>> for the regular code which is updated less often.
>
> Which is listed under "Ports" at http://support.ntp.org/links
>
Yeah, I know. :-)

Terje

-- 
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"

Gautam Thaker wrote:
> Once you replace the binaries in
>
> C:\Program Files\NTP\bin
>
> do you need to do anything extra (such as use Instsrv.exe) to register
> the new binary. I get a popup saying
>
> "Erorr 5: access is denied".
>
> Any hints welcome.

Are you on Vista or W7?

If so, you need to be Administrator to touch these files...

You can start ntpd.exe from the command line and see what happens.

Terje

-- 
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"

Gautam Thaker wrote:
> Once you replace the binaries in
> C:\Program Files\NTP\bin
> do you need to do anything extra (such as use
>  Instsrv.exe) to register the new binary.
>  I get a popup saying
> "Erorr 5: access is denied".
> Any hints welcome.


Stop the NTP service,
 before coping a newer ntpd.exe into the
  "C:\Program Files\NTP\bin\" directory,
Start the NTP service.

-- 
E-Mail Sent to this address <BlackList@Anitech-Systems.com>
  will be added to the BlackLists.

Update:

ntp-4.2.7p98-win-x86-bin.zip

is running for me on my Win7 box. I will see how well it syncs time
and report.

My problem appear to be at least in part that

ntp-4.2.7p67-win32.zip

binaries did not work when I used them to over write what Meinberg
installer puts down (4.2.4 based.) But I did a clean install (again),
and used 4.2.7p98 binaries from

http://davehart.net/ntp/win/x86/

and things are looking better and hopefully I get a good clock sync.
After about 15 minutes the status line is reporting:

State	Remote		                 Refid	  Stratum	Type		When	Poll	Reach
Delay	Offset	Jitter
*	        192.168.214.3		192.168.35.253	2	Unicast server	44	64	377
0.003	16.141	3.174

Delay of "0.003" looks a bit suspicious, that is 3 usec delay, too low
though I have direct connection to the Remote over a gig network.
Gautam
PS thanks to everyone for their help.

Gautam Thaker wrote:
> ntp-4.2.7p98-win-x86-bin.zip
> and things are looking better and hopefully I get a good clock sync.
> After about 15 minutes the status line is reporting:
> State	   Remote          Refid St  Type When Poll Reach Delay Offset Jitter
> * 192.168.214.3 192.168.35.253 2 Unicast 44   64   377  0.003 16.141  3.174
> Delay of "0.003" looks a bit suspicious, that is 3 usec
>  delay, too low though I have direct connection to the
>  Remote over a gig network.

If reach goes away,
 try ntpq -n -c "las" -c "rv &1"
  (or which ever association is the server you are trying to sync to)
  check the flash status word for hints of why it is not being used.
 e.g. peer distance threshold exceeded


FYI, sometimes a giga-bit adapter can cause extra jitter.
 Some say they solve this by locking the adapter down to 100Mbit
  & disabling interrupt coalescing.

-- 
E-Mail Sent to this address <BlackList@Anitech-Systems.com>
  will be added to the BlackLists.

> Once you replace the binaries in
>
> C:\Program Files\NTP\bin
>
> do you need to do anything extra (such as use Instsrv.exe) to register
> the new binary. I get a popup saying
>
> "Erorr 5: access is denied".
>
> Any hints welcome.
>
> Gautam

Gautam,

If you are using Vista or Window-7, I advise against installing in the 
\Program Files\ tree, and using a directory such as C:;\Tools\NTP\ 
instead:

  http://www.satsignal.eu/ntp/setup.html

The Meinberg installation sets up the NTP service for you.

Cheers,
David 

>>   http://www.sureelectronics.net/goods.php?id=99
>
> Except there is no indication that this thing puts out a PPS signal.

Bill, do you need to add a couple of wires to get an RS-232 level signal, 
as I describe here:

  http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm#sure

For just over US $30 (and delivered to the UK for a few dollars more) it's 
not bad, considering that the GPS 18x LVC comes to approaching US $100 
over here!

Cheers,
David 

<dwmalone@maths.tcd.ie> wrote in message 
news:201102241413.aa58488@walton.maths.tcd.ie...
> In comp.protocols.time.ntp you write:
>
>>Bill, do you need to add a couple of wires to get an RS-232 level 
>>signal,
>>as I describe here:
>
>>  http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm#sure
>
>>For just over US $30 (and delivered to the UK for a few dollars more) 
>>it's
>>not bad, considering that the GPS 18x LVC comes to approaching US $100
>>over here!
>
> Cute - I might get one - even my soldering might be up to doing that ;-)
>
> David.

It's surface mount stuff, David, which I don't normally do, but attaching 
a couple of leads is OK if you are careful.  I happen to have a very 
narrow bit for my soldering iron, which IIRC is 800F, so hot enough to 
melt what they call solder these days!  But perhaps more modern irons 
would be OK in any case.  As you can tell, it's enamelled copper wire I 
use, and I would suggest trying to layout and scrape and tin the ends 
before soldering.  Usual anti-static precautions apply.  I've ordered a 
second one as a spare - the unit seems very sensitive, but it would be 
most interesting to see how it compares with a more expensive box.

Do let me know how you get on, and don't worry if the combined postal time 
is 2-3 weeks!

Cheers,
David 

David J Taylor wrote:
>>>   http://www.sureelectronics.net/goods.php?id=99
>>
>>
>> Except there is no indication that this thing puts out a PPS signal.
> 
> 
> Bill, do you need to add a couple of wires to get an RS-232 level 
> signal, as I describe here:
> 
>  http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm#sure
> 
> For just over US $30 (and delivered to the UK for a few dollars more) 
> it's not bad, considering that the GPS 18x LVC comes to approaching US 
> $100 over here!

Thats nothing. Garmin DE wants 180�
.... and a piece of your immortal soul ;-(


uwe

On 2011-02-24, David J Taylor <david-taylor@blueyonder.co.uk.invalid> wrote:
>>>   http://www.sureelectronics.net/goods.php?id=99
>>
>> Except there is no indication that this thing puts out a PPS signal.
>
> Bill, do you need to add a couple of wires to get an RS-232 level signal, 
> as I describe here:
>
>   http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm#sure
>
> For just over US $30 (and delivered to the UK for a few dollars more) it's 
> not bad, considering that the GPS 18x LVC comes to approaching US $100 
> over here!

There are no specs in the writeup as to what the timing is on that pin
T9. Since it seems to be being used purely to drive an LED, its timing
requirements are maybe 1/10 sec. But for timing purposes we want sub
usec. Does the pulse out of T9 meet some sort of "The upgoing edge of
the pulse occurs within 500ns of the seconds rollover" or something?
On the GPS18x there specs say it is withing that. What is it on this
board? Has anyone tested it? (It would be great if it really were good
enough).
Note that the cmos inverter and the RS232 drive will both put in delays.
Anyone got any idea of how much?

>
> Cheers,
> David 
>

> There are no specs in the writeup as to what the timing is on that pin
> T9. Since it seems to be being used purely to drive an LED, its timing
> requirements are maybe 1/10 sec. But for timing purposes we want sub
> usec.

No.  You may want sub-microsecond, a few microseconds may do for others.

> Does the pulse out of T9 meet some sort of "The upgoing edge of
> the pulse occurs within 500ns of the seconds rollover" or something?
> On the GPS18x there specs say it is withing that. What is it on this
> board? Has anyone tested it? (It would be great if it really were good
> enough).

Yes, I would also be delighted to see measurements, but already it's 
apparent from my own tests that it's far better than no PPS at all.

> Note that the cmos inverter and the RS232 drive will both put in delays.
> Anyone got any idea of how much?

74HC04 ~ 15ns, RS-232 driver ~1.5us

This is a $35 device, not a $100-$200 one.

Cheers,
David 

On Thu, Feb 24, 2011 at 9:47 AM, unruh <unruh@wormhole.physics.ubc.ca> wrote:

> There are no specs in the writeup as to what the timing is on that pin
> T9. Since it seems to be being used purely to drive an LED, its timing
> requirements are maybe 1/10 sec. But for timing purposes we want sub
> usec. Does the pulse out of T9 meet some sort of "The upgoing edge of
> the pulse occurs within 500ns of the seconds rollover" or something?
> On the GPS18x there specs say it is withing that. What is it on this
> board? Has anyone tested it? (It would be great if it really were good
> enough).
> Note that the cmos inverter and the RS232 drive will both put in delays.
> Anyone got any idea of how much?

I agree, this is the problem with the Sure Electriconic GPS, basicaly
Chinese/eBay level of documentation and customer suport.  You are left
to measure the unit yourself  One could say "That's what you get for
$35" but no.  I bought a Motorola GPS for $15 and it is very well
documented and I know that the PPS is within 50ns (and I verify this
by observation.)  To be fair the Motorola unit is used, removed from
service and sold at ~10X less then new price.  Many of these are
available.  These are GPSes from Rockwell and others too, all of these
documented to "written in USA standards"   If you are loking for under
$40 GPS with good PPS specs you have many options

You do need to do some study before you buy.  I tried and older hand
held GPS I just happened to own that I used for sailboat racing and
it's timing performance was so poor as to be unusable.  But I just
happened to own it already, nothing to loose by trying.

One of the very best affordable GPS is the "Thunderbolt"  It uses a
GPSDXO and then drives the PPS from that.  Typical price is $150 but
you can have one for $100 if you hunt and wait.

Now about the question about delay.  Even with a Thunderbolt of the
Motorola MT+ Where the pules is spec'd to a few nanoseconds yu still
don't know the delay because there are cables and interfaces and then
more of the same inside the computer.   Your only option if yu need
precision is to have more than one, well really more then two sources
of time and compare and measure them.  You need a bit of gear for
that.   Lacking that you just use the data sheets and add up all the
delays and then tell NTP what the estimated delay is.   You will
likely be "close"

-- 
=====
Chris Albertson
Redondo Beach, California

On 2011-02-24, Chris Albertson <albertson.chris@gmail.com> wrote:
> On Thu, Feb 24, 2011 at 9:47 AM, unruh <unruh@wormhole.physics.ubc.ca> wrote:
>
>> There are no specs in the writeup as to what the timing is on that pin
>> T9. Since it seems to be being used purely to drive an LED, its timing
>> requirements are maybe 1/10 sec. But for timing purposes we want sub
>> usec. Does the pulse out of T9 meet some sort of "The upgoing edge of
>> the pulse occurs within 500ns of the seconds rollover" or something?
>> On the GPS18x there specs say it is withing that. What is it on this
>> board? Has anyone tested it? (It would be great if it really were good
>> enough).
>> Note that the cmos inverter and the RS232 drive will both put in delays.
>> Anyone got any idea of how much?
>
> I agree, this is the problem with the Sure Electriconic GPS, basicaly
> Chinese/eBay level of documentation and customer suport.  You are left
> to measure the unit yourself  One could say "That's what you get for
> $35" but no.  I bought a Motorola GPS for $15 and it is very well
> documented and I know that the PPS is within 50ns (and I verify this
> by observation.)  To be fair the Motorola unit is used, removed from
> service and sold at ~10X less then new price.  Many of these are
> available.  These are GPSes from Rockwell and others too, all of these
> documented to "written in USA standards"   If you are loking for under
> $40 GPS with good PPS specs you have many options
>
> You do need to do some study before you buy.  I tried and older hand
> held GPS I just happened to own that I used for sailboat racing and
> it's timing performance was so poor as to be unusable.  But I just
> happened to own it already, nothing to loose by trying.
>
> One of the very best affordable GPS is the "Thunderbolt"  It uses a
> GPSDXO and then drives the PPS from that.  Typical price is $150 but
> you can have one for $100 if you hunt and wait.
>
> Now about the question about delay.  Even with a Thunderbolt of the
> Motorola MT+ Where the pules is spec'd to a few nanoseconds yu still
> don't know the delay because there are cables and interfaces and then
> more of the same inside the computer.   Your only option if yu need
> precision is to have more than one, well really more then two sources
> of time and compare and measure them.  You need a bit of gear for
> that.   Lacking that you just use the data sheets and add up all the
> delays and then tell NTP what the estimated delay is.   You will
> likely be "close"

If the delay is constant (eg through a RS232 chip) you can compensate.
If it is variable (eg in the GPS chipset if it issues a PPS when it
feels like it) it is useless.

The Sure unit seems to have no data sheets. 
It states that the "timing" is 20ns, but I certainly do not believe that
the output of the PPS TP9 is to withing 20ns. 


>

unruh wrote:
> If the delay is constant (eg through a RS232 chip)
>  you can compensate.
> If it is variable (eg in the GPS chipset if it issues
>  a PPS when it feels like it) it is useless.
>
> The Sure unit seems to have no data sheets.
> It states that the "timing" is 20ns, but I certainly
>  do not believe that the output of the PPS TP9 is to
>  withing 20ns.

All of the Skylab 16 series data sheets appear to say:
 "Accuracy Timing (PPS) 60ns RMS"

some also say
 "Time pulse Signal (Default 200ms pulse/sec)"

others also say
 "Time pulse Signal (100ms)"


FYI, I spotted some end user test data on the PPS
<http://www.leapsecond.com/pages/MG1613S/>

-- 
E-Mail Sent to this address <BlackList@Anitech-Systems.com>
  will be added to the BlackLists.

On 2011-02-25, E-Mail Sent to this address will be added to the BlackLists <Null@BlackList.Anitech-Systems.invalid> wrote:
> unruh wrote:
>> If the delay is constant (eg through a RS232 chip)
>>  you can compensate.
>> If it is variable (eg in the GPS chipset if it issues
>>  a PPS when it feels like it) it is useless.
>>
>> The Sure unit seems to have no data sheets.
>> It states that the "timing" is 20ns, but I certainly
>>  do not believe that the output of the PPS TP9 is to
>>  withing 20ns.
>
> All of the Skylab 16 series data sheets appear to say:
>  "Accuracy Timing (PPS) 60ns RMS"
>
> some also say
>  "Time pulse Signal (Default 200ms pulse/sec)"
>
> others also say
>  "Time pulse Signal (100ms)"
>
>
> FYI, I spotted some end user test data on the PPS
><http://www.leapsecond.com/pages/MG1613S/>

Very nice -- Gives me more faith that it is accurate.

E-Mail Sent to this address will be added to the BlackLists wrote:
> All of the Skylab 16 series data sheets appear to say:
>   "Accuracy Timing (PPS) 60ns RMS"
>
> some also say
>   "Time pulse Signal (Default 200ms pulse/sec)"
>
> others also say
>   "Time pulse Signal (100ms)"
>
>
> FYI, I spotted some end user test data on the PPS
> <http://www.leapsecond.com/pages/MG1613S/>
>

Thanks!

That was a very nice writeup of the board that I'm awaiting, it 
definitely looks to be at least as good as my Garmin 18 (which is 
specified to give out PPS with 1-2 us accuracy).

If the GPS chip used had an output format like the Oncore which includes 
the sawtooth offset it seems like it could be close to UT+ in timing 
performance.

Terje

-- 
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"