Hi all,

    I know it is off topic, but I don�t know other place to put this
question.
    What happens is that I want to synchronize my clock just once for day
using ntpd -q (or ntpdate). The bad thing about it is that my internal clock
is a really bad timekeeper (it losses about 3 seconds in a day).
    Anyone know what I can do to make my clock more stable? (I can�t run
ntpd cause I have no permanent connection to internet, just 1 or 2 minutes
per day).

Thanks for any help,

V�tor

On 2004-08-27, vitor wrote:

> What happens is that I want to synchronize my clock just once for day
> using ntpd -q (or ntpdate). The bad thing about it is that my internal
> clock is a really bad timekeeper (it losses about 3 seconds in a day).
> Anyone know what I can do to make my clock more stable? (I can�t run
> ntpd cause I have no permanent connection to internet, just 1 or 2
> minutes per day).

Yes, you can run ntpd.

Use the LocalCLK refclock in addition to some remote time servers and
let ntpd continue to run after your internet connection goes down. Then
just restart ntpd with the '-g' option every time your external
connection is established.

Your ntp.conf should look something like this:

<snip: ntpd files and logging >

# Avoid using restrict lines until you know this works...

# Local CLK
server 127.127.1.1              # LCL, local clock
fudge  127.127.1.1 stratum 12   # increase stratum

# Remote time server(s)
# use iburst to get fast sync when ntpd starts
server ntp.your.isp iburst # for example
server pool.ntp.org iburst # can be repeated to get multiple servers

-- 
Steve Kostecke <kostecke@ntp.org>

Hi all,

    Thanks for your answer Steve, but I�m really more interested in some
other solution (hardware). Something like changing my RTC, the oscillator
(crystal) of my motherboard or any other suggestion. The problem is that I
have no ideas in where/how to begin.

Thanks for any advice,

V�tor

----- Original Message ----- 
From: "Steve Kostecke" <kostecke@ntp.org>
Newsgroups: comp.protocols.time.ntp
To: <questions@ntp.org>
Sent: Friday, August 27, 2004 11:21 AM
Subject: [ntp:questions] Re: Internal Clock


> On 2004-08-27, vitor wrote:
>
> > What happens is that I want to synchronize my clock just once for day
> > using ntpd -q (or ntpdate). The bad thing about it is that my internal
> > clock is a really bad timekeeper (it losses about 3 seconds in a day).
> > Anyone know what I can do to make my clock more stable? (I can�t run
> > ntpd cause I have no permanent connection to internet, just 1 or 2
> > minutes per day).
>
> Yes, you can run ntpd.
>
> Use the LocalCLK refclock in addition to some remote time servers and
> let ntpd continue to run after your internet connection goes down. Then
> just restart ntpd with the '-g' option every time your external
> connection is established.
>
> Your ntp.conf should look something like this:
>
> <snip: ntpd files and logging >
>
> # Avoid using restrict lines until you know this works...
>
> # Local CLK
> server 127.127.1.1              # LCL, local clock
> fudge  127.127.1.1 stratum 12   # increase stratum
>
> # Remote time server(s)
> # use iburst to get fast sync when ntpd starts
> server ntp.your.isp iburst # for example
> server pool.ntp.org iburst # can be repeated to get multiple servers
>
> -- 
> Steve Kostecke <kostecke@ntp.org>
> _______________________________________________
> questions mailing list
> questions@ntp.org
> http://mailman.ntp.org/mailman/listinfo/questions

On Fri, 27 Aug 2004 15:55:03 -0300, V�tor Claudino dos Santos wrote:
>     Thanks for your answer Steve, but I�m really more interested in some
> other solution (hardware). Something like changing my RTC, the oscillator
> (crystal) of my motherboard or any other suggestion. The problem is that I

You have named quite a few possible solutions there. If the budget can
accommodate a hardware solution, buy a new motherboard with a more
accurate hardware clock.

> have no ideas in where/how to begin.

ntpd can indeed provide a workaround in software for bad hardware
clocks. There's also hwclock(8) which can be run in startup scripts
and/or from cron.

On my systems with dead CMOS batteries I do this in startup scripts:
    ntpdate pool.ntp.org. pool.ntp.org. && ntpd
Thus the clock is in synch before trying to start ntpd.
-- 
  /dev/rob0 - preferred_email=i$((28*28+28))@softhome.net
  or put "not-spam" or "/dev/rob0" in Subject header to reply

This is a multi-part message in MIME format.
--------------010607030606020104030103
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

You didn't mention what O/S or what hardware you are using.

Some O/S's have a means of correcting the tick rate by changing the 
divisor used to divide the crystal frequency down to a reasonable tick 
rate; e.g. 100 Hz.

You can replace the crystal with one of better quality.

You can vary the frequency slightly by connecting a very small trimmer 
capacitor across the crystal and adjusting it.

You can replace the whole oscillator with one of better quality.  This 
is going to require that you remove components and/or cut traces on your 
motherboard with the risk that it might never work again!

You can purchase, for around US $200, a GPS timing receiver.  If you can 
locate the antenna where it has an unobstructed view of the sky,  you 
can use it as a reference clock for ntpd.   A GPS receiver should allow 
you to synchronize within a few hundred microseconds of UTC.


V�tor Claudino dos Santos wrote:

>Hi all,
>
>    Thanks for your answer Steve, but I�m really more interested in some
>other solution (hardware). Something like changing my RTC, the oscillator
>(crystal) of my motherboard or any other suggestion. The problem is that I
>have no ideas in where/how to begin.
>
>Thanks for any advice,
>
>V�tor
>
>----- Original Message ----- 
>From: "Steve Kostecke" <kostecke@ntp.org>
>Newsgroups: comp.protocols.time.ntp
>To: <questions@ntp.org>
>Sent: Friday, August 27, 2004 11:21 AM
>Subject: [ntp:questions] Re: Internal Clock
>
>
>  
>
>>On 2004-08-27, vitor wrote:
>>
>>    
>>
>>>What happens is that I want to synchronize my clock just once for day
>>>using ntpd -q (or ntpdate). The bad thing about it is that my internal
>>>clock is a really bad timekeeper (it losses about 3 seconds in a day).
>>>Anyone know what I can do to make my clock more stable? (I can�t run
>>>ntpd cause I have no permanent connection to internet, just 1 or 2
>>>minutes per day).
>>>      
>>>
>>Yes, you can run ntpd.
>>
>>Use the LocalCLK refclock in addition to some remote time servers and
>>let ntpd continue to run after your internet connection goes down. Then
>>just restart ntpd with the '-g' option every time your external
>>connection is established.
>>
>>Your ntp.conf should look something like this:
>>
>><snip: ntpd files and logging >
>>
>># Avoid using restrict lines until you know this works...
>>
>># Local CLK
>>server 127.127.1.1              # LCL, local clock
>>fudge  127.127.1.1 stratum 12   # increase stratum
>>
>># Remote time server(s)
>># use iburst to get fast sync when ntpd starts
>>server ntp.your.isp iburst # for example
>>server pool.ntp.org iburst # can be repeated to get multiple servers
>>
>>-- 
>>Steve Kostecke <kostecke@ntp.org>
>>_______________________________________________
>>questions mailing list
>>questions@ntp.org
>>http://mailman.ntp.org/mailman/listinfo/questions
>>    
>>
>
>  
>

--------------010607030606020104030103
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
  <meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
  <title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
You didn't mention what O/S or what hardware you are using.<br>
<br>
Some O/S's have a means of correcting the tick rate by changing the
divisor used to divide the crystal frequency down to a reasonable tick
rate; e.g. 100 Hz. <br>
<br>
You can replace the crystal with one of better quality.<br>
<br>
You can vary the frequency slightly by connecting a very small trimmer
capacitor across the crystal and adjusting it.<br>
<br>
You can replace the whole oscillator with one of better quality.&nbsp; This
is going to require that you remove components and/or cut traces on
your motherboard with the risk that it might never work again!<br>
<br>
You can purchase, for around US $200, a GPS timing receiver.&nbsp; If you
can locate the antenna where it has an unobstructed view of the sky,&nbsp;
you can use it as a reference clock for ntpd.&nbsp;&nbsp; A GPS receiver should
allow you to synchronize within a few hundred microseconds of UTC.<br>
<br>
<br>
V&iacute;tor Claudino dos Santos wrote:<br>
<blockquote type="cite"
 cite="midmailman.4.1093679811.26599.questions@ntp.org">
  <pre wrap="">Hi all,

    Thanks for your answer Steve, but I&acute;m really more interested in some
other solution (hardware). Something like changing my RTC, the oscillator
(crystal) of my motherboard or any other suggestion. The problem is that I
have no ideas in where/how to begin.

Thanks for any advice,

V&iacute;tor

----- Original Message ----- 
From: "Steve Kostecke" <a class="moz-txt-link-rfc2396E" href="mailto:kostecke@ntp.org">&lt;kostecke@ntp.org&gt;</a>
Newsgroups: comp.protocols.time.ntp
To: <a class="moz-txt-link-rfc2396E" href="mailto:questions@ntp.org">&lt;questions@ntp.org&gt;</a>
Sent: Friday, August 27, 2004 11:21 AM
Subject: [ntp:questions] Re: Internal Clock


  </pre>
  <blockquote type="cite">
    <pre wrap="">On 2004-08-27, vitor wrote:

    </pre>
    <blockquote type="cite">
      <pre wrap="">What happens is that I want to synchronize my clock just once for day
using ntpd -q (or ntpdate). The bad thing about it is that my internal
clock is a really bad timekeeper (it losses about 3 seconds in a day).
Anyone know what I can do to make my clock more stable? (I can&acute;t run
ntpd cause I have no permanent connection to internet, just 1 or 2
minutes per day).
      </pre>
    </blockquote>
    <pre wrap="">Yes, you can run ntpd.

Use the LocalCLK refclock in addition to some remote time servers and
let ntpd continue to run after your internet connection goes down. Then
just restart ntpd with the '-g' option every time your external
connection is established.

Your ntp.conf should look something like this:

&lt;snip: ntpd files and logging &gt;

# Avoid using restrict lines until you know this works...

# Local CLK
server 127.127.1.1              # LCL, local clock
fudge  127.127.1.1 stratum 12   # increase stratum

# Remote time server(s)
# use iburst to get fast sync when ntpd starts
server ntp.your.isp iburst # for example
server pool.ntp.org iburst # can be repeated to get multiple servers

-- 
Steve Kostecke <a class="moz-txt-link-rfc2396E" href="mailto:kostecke@ntp.org">&lt;kostecke@ntp.org&gt;</a>
_______________________________________________
questions mailing list
<a class="moz-txt-link-abbreviated" href="mailto:questions@ntp.org">questions@ntp.org</a>
<a class="moz-txt-link-freetext" href="http://mailman.ntp.org/mailman/listinfo/questions">http://mailman.ntp.org/mailman/listinfo/questions</a>
    </pre>
  </blockquote>
  <pre wrap=""><!---->
  </pre>
</blockquote>
</body>
</html>

--------------010607030606020104030103--

In article <mailman.4.1093679811.26599.questions@ntp.org>,
V�tor Claudino dos Santos wrote:

In your original post you said:

>>> my internal clock is a really bad timekeeper (it losses about 3
>>> seconds in a day).

3 in 86400 is about 35 parts per million. If you look at the specifications
for the typical crystal used in computers, you find the _basic_ tolerance
is plus/minus 50 ppm, with a _stability_ of plus/minus 100 ppm over the
temperature range of -10 to +60 C on top of that. What you have may seem
bad as a time standard, but it's fairly good for a computer.

>I<B4>m really more interested in some other solution (hardware). Something
>like changing my RTC, the oscillator (crystal) of my motherboard or any
>other suggestion.

The standard crystal that you would be able to buy isn't even as good as
the results you are getting now.  The replacement could be up to four
times LESS accurate, and still be with specification.  Years ago, some
motherboards had a small variable capacitor to drag the clock frequency
somewhat closer to nominal than the US$0.20 crystal was cut for (0.02
percent [200 ppm] was typical), but the capacitors cost a few cents, and
the setting of the clock took a half minute more - time is money - and
they've been eliminated as not cost effective. 

>The problem is that I have no ideas in where/how to begin.

There are vendors who will sell you a better crystal but the cost is not
going to be pleasent. Buying a GPS clock may be a lot less expensive.

You don't mention what operating system this computer is using. With
some versions of Unix, there is an clock tweaking application that can
be run, but in general this all assumes that the crystal (which is the
primary frequency determining mechanism) is kept at a _relatively_
constant temperature - say within a few degrees C.

I'm slightly curious why you need time more accurate than 3 seconds
a day, but don't feel the need to use NTP.

       Old guy

Thanks for your answer Gilbert, below is some comments...

----- Original Message ----- 
From: "Richard B. Gilbert" <rgilbert88@comcast.net>
Newsgroups: comp.protocols.time.ntp
To: <questions@ntp.org>
Sent: Saturday, August 28, 2004 2:59 PM
Subject: Re: [ntp:questions] Re: Internal Clock


> You didn't mention what O/S or what hardware you are using.

I�m using a ROCKY-4784EVG motherboard with RTC being controlled by an INTEL
82801BA (ICH2)
I�ve tested with 2 versions of Linux - Debian 3.0 R2 and Red Hat 7.3
Interprise

> Some O/S's have a means of correcting the tick rate by changing the
> divisor used to divide the crystal frequency down to a reasonable tick
> rate; e.g. 100 Hz.

Do you know if I can do that with the O/S I have? Some time ago in this list
I�ve heard about 'nanokernel'... Can it help me?

> You can replace the crystal with one of better quality.

Ok, here I got a question. The crystal of the motherboard is used for
timekeeping if the computer is always turned on? Some time ago I�ve read
some NIST papers and (if I�m not mistaken) I found that this clock(hardware)
is only used when the computer is turned off. So when the computer is turned
on again, this time is read an put in the software clock. By this papers
looks to me that changing this oscillator is not necessary.

> You can vary the frequency slightly by connecting a very small trimmer
> capacitor across the crystal and adjusting it.
>
> You can replace the whole oscillator with one of better quality.  This
> is going to require that you remove components and/or cut traces on your
> motherboard with the risk that it might never work again!
>
> You can purchase, for around US $200, a GPS timing receiver.  If you can
> locate the antenna where it has an unobstructed view of the sky,  you
> can use it as a reference clock for ntpd.   A GPS receiver should allow
> you to synchronize within a few hundred microseconds of UTC.
>
>
> V�tor Claudino dos Santos wrote:
>
> >Hi all,
> >
> >    Thanks for your answer Steve, but I�m really more interested in some
> >other solution (hardware). Something like changing my RTC, the oscillator
> >(crystal) of my motherboard or any other suggestion. The problem is that
I
> >have no ideas in where/how to begin.
> >
> >Thanks for any advice,
> >
> >V�tor
> >
> >----- Original Message ----- 
> >From: "Steve Kostecke" <kostecke@ntp.org>
> >Newsgroups: comp.protocols.time.ntp
> >To: <questions@ntp.org>
> >Sent: Friday, August 27, 2004 11:21 AM
> >Subject: [ntp:questions] Re: Internal Clock
> >
> >
> >
> >
> >>On 2004-08-27, vitor wrote:
> >>
> >>
> >>
> >>>What happens is that I want to synchronize my clock just once for day
> >>>using ntpd -q (or ntpdate). The bad thing about it is that my internal
> >>>clock is a really bad timekeeper (it losses about 3 seconds in a day).
> >>>Anyone know what I can do to make my clock more stable? (I can�t run
> >>>ntpd cause I have no permanent connection to internet, just 1 or 2
> >>>minutes per day).
> >>>
> >>>
> >>Yes, you can run ntpd.
> >>
> >>Use the LocalCLK refclock in addition to some remote time servers and
> >>let ntpd continue to run after your internet connection goes down. Then
> >>just restart ntpd with the '-g' option every time your external
> >>connection is established.
> >>
> >>Your ntp.conf should look something like this:
> >>
> >><snip: ntpd files and logging >
> >>
> >># Avoid using restrict lines until you know this works...
> >>
> >># Local CLK
> >>server 127.127.1.1              # LCL, local clock
> >>fudge  127.127.1.1 stratum 12   # increase stratum
> >>
> >># Remote time server(s)
> >># use iburst to get fast sync when ntpd starts
> >>server ntp.your.isp iburst # for example
> >>server pool.ntp.org iburst # can be repeated to get multiple servers
> >>
> >>-- 
> >>Steve Kostecke <kostecke@ntp.org>
> >>_______________________________________________
> >>questions mailing list
> >>questions@ntp.org
> >>http://mailman.ntp.org/mailman/listinfo/questions
> >>
> >>
> >
> >
> >
> _______________________________________________
> questions mailing list
> questions@ntp.org
> http://mailman.ntp.org/mailman/listinfo/questions

Thanks for your answer Moe, below is some comments/questions...

----- Original Message ----- 
From: "Moe Trin" <ibuprofin@painkiller.example.tld>
Newsgroups: comp.protocols.time.ntp
To: <questions@ntp.org>
Sent: Saturday, August 28, 2004 9:10 PM
Subject: Re: [ntp:questions] Re: Internal Clock


> In article <mailman.4.1093679811.26599.questions@ntp.org>,
> V�tor Claudino dos Santos wrote:
>
> In your original post you said:
>
> >>> my internal clock is a really bad timekeeper (it losses about 3
> >>> seconds in a day).
>
> 3 in 86400 is about 35 parts per million. If you look at the
specifications
> for the typical crystal used in computers, you find the _basic_ tolerance
> is plus/minus 50 ppm, with a _stability_ of plus/minus 100 ppm over the
> temperature range of -10 to +60 C on top of that. What you have may seem
> bad as a time standard, but it's fairly good for a computer.
>
> >I<B4>m really more interested in some other solution (hardware).
Something
> >like changing my RTC, the oscillator (crystal) of my motherboard or any
> >other suggestion.
>
> The standard crystal that you would be able to buy isn't even as good as
> the results you are getting now.  The replacement could be up to four
> times LESS accurate, and still be with specification.  Years ago, some
> motherboards had a small variable capacitor to drag the clock frequency
> somewhat closer to nominal than the US$0.20 crystal was cut for (0.02
> percent [200 ppm] was typical), but the capacitors cost a few cents, and
> the setting of the clock took a half minute more - time is money - and
> they've been eliminated as not cost effective.
>
> >The problem is that I have no ideas in where/how to begin.
>
> There are vendors who will sell you a better crystal but the cost is not
> going to be pleasent. Buying a GPS clock may be a lot less expensive.

Can you tell me about what cost we�re talking about? I�ve never
worked/studied this so I�ve no idea in how much it would cost to change the
crystal of my motherboard.

> You don't mention what operating system this computer is using. With
> some versions of Unix, there is an clock tweaking application that can
> be run, but in general this all assumes that the crystal (which is the
> primary frequency determining mechanism) is kept at a _relatively_
> constant temperature - say within a few degrees C.

Can you give me some hints here? If it would cost too much to change
components I�ll have to go by this way.

> I'm slightly curious why you need time more accurate than 3 seconds
> a day, but don't feel the need to use NTP.

I�ve some secure server (certificate authoruty) that is kept offline by the
most of time, so I�m using ntp just to make some calibrations, but the ideal
is not to connect this server. I�ve studied about a GPS solution, but
board+antenna+acessories is costing something around U$1000,00 and that is
too much for me :).

>        Old guy
> _______________________________________________
> questions mailing list
> questions@ntp.org
> http://mailman.ntp.org/mailman/listinfo/questions

This is a multi-part message in MIME format.
--------------010507040309020200070604
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

V�tor Claudino dos Santos wrote:

>Thanks for your answer Gilbert, below is some comments...
>
>----- Original Message ----- 
>From: "Richard B. Gilbert" <rgilbert88@comcast.net>
>Newsgroups: comp.protocols.time.ntp
>To: <questions@ntp.org>
>Sent: Saturday, August 28, 2004 2:59 PM
>Subject: Re: [ntp:questions] Re: Internal Clock
>
>
>  
>
>>You didn't mention what O/S or what hardware you are using.
>>    
>>
>
>I�m using a ROCKY-4784EVG motherboard with RTC being controlled by an INTEL
>82801BA (ICH2)
>I�ve tested with 2 versions of Linux - Debian 3.0 R2 and Red Hat 7.3
>Interprise
>
>  
>
>>Some O/S's have a means of correcting the tick rate by changing the
>>divisor used to divide the crystal frequency down to a reasonable tick
>>rate; e.g. 100 Hz.
>>    
>>
>
>Do you know if I can do that with the O/S I have? Some time ago in this list
>I�ve heard about 'nanokernel'... Can it help me?
>
>  
>
>>You can replace the crystal with one of better quality.
>>    
>>
>
>Ok, here I got a question. The crystal of the motherboard is used for
>timekeeping if the computer is always turned on? Some time ago I�ve read
>some NIST papers and (if I�m not mistaken) I found that this clock(hardware)
>is only used when the computer is turned off. So when the computer is turned
>on again, this time is read an put in the software clock. By this papers
>looks to me that changing this oscillator is not necessary.
>  
>
The software clock also has a hardware component; a crystal oscillator.  
Typically, the oscillator frequency is divided  by a constant so as to 
produce a low frequency "tick".  100 Hz or ten millisecond ticks are 
frequently used.   The crystal oscillator my be the same one that drives 
the hardware clock or a different one.  I'm not familiar with your 
motherboard and  would not attempt to guess exactly how the hardware and 
software clocks are implemented.

--------------010507040309020200070604
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
  <meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
  <title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
V&iacute;tor Claudino dos Santos wrote:<br>
<blockquote type="cite"
 cite="midmailman.6.1093901229.26599.questions@ntp.org">
  <pre wrap="">Thanks for your answer Gilbert, below is some comments...

----- Original Message ----- 
From: "Richard B. Gilbert" <a class="moz-txt-link-rfc2396E" href="mailto:rgilbert88@comcast.net">&lt;rgilbert88@comcast.net&gt;</a>
Newsgroups: comp.protocols.time.ntp
To: <a class="moz-txt-link-rfc2396E" href="mailto:questions@ntp.org">&lt;questions@ntp.org&gt;</a>
Sent: Saturday, August 28, 2004 2:59 PM
Subject: Re: [ntp:questions] Re: Internal Clock


  </pre>
  <blockquote type="cite">
    <pre wrap="">You didn't mention what O/S or what hardware you are using.
    </pre>
  </blockquote>
  <pre wrap=""><!---->
I&acute;m using a ROCKY-4784EVG motherboard with RTC being controlled by an INTEL
82801BA (ICH2)
I&acute;ve tested with 2 versions of Linux - Debian 3.0 R2 and Red Hat 7.3
Interprise

  </pre>
  <blockquote type="cite">
    <pre wrap="">Some O/S's have a means of correcting the tick rate by changing the
divisor used to divide the crystal frequency down to a reasonable tick
rate; e.g. 100 Hz.
    </pre>
  </blockquote>
  <pre wrap=""><!---->
Do you know if I can do that with the O/S I have? Some time ago in this list
I&acute;ve heard about 'nanokernel'... Can it help me?

  </pre>
  <blockquote type="cite">
    <pre wrap="">You can replace the crystal with one of better quality.
    </pre>
  </blockquote>
  <pre wrap=""><!---->
Ok, here I got a question. The crystal of the motherboard is used for
timekeeping if the computer is always turned on? Some time ago I&acute;ve read
some NIST papers and (if I&acute;m not mistaken) I found that this clock(hardware)
is only used when the computer is turned off. So when the computer is turned
on again, this time is read an put in the software clock. By this papers
looks to me that changing this oscillator is not necessary.
  </pre>
</blockquote>
The software clock also has a hardware component; a crystal
oscillator.&nbsp; Typically, the oscillator frequency is divided&nbsp; by a
constant so as to produce a low frequency "tick".&nbsp; 100 Hz or ten
millisecond ticks are frequently used.&nbsp;&nbsp; The crystal oscillator my be
the same one that drives the hardware clock or a different one.&nbsp; I'm
not familiar with your motherboard and&nbsp; would not attempt to guess
exactly how the hardware and software clocks are implemented.<br>
</body>
</html>

--------------010507040309020200070604--

V�tor Claudino dos Santos <vitor@bry.com.br> writes:

> is not to connect this server. I�ve studied about a GPS solution, but
> board+antenna+acessories is costing something around U$1000,00 and that is
> too much for me :).

Then find a cheaper kit. I have bought several kits around $100 or
less. GPS-board, adaptor board and antenna is possible to find at
$30-$40 each.

In article <mailman.7.1093901233.26599.questions@ntp.org>,
V�tor Claudino dos Santos wrote:
>Can you tell me about what cost we<B4>re talking about? I<B4>ve never
>worked/studied this so I<B4>ve no idea in how much it would cost to change the
>crystal of my motherboard.

You're referring to changing the crytal - I don't think I'd recommend this
even if you could get a compatible crystal (International Crustal is one
of the more well know vendors, with Raltron and Unites States Crystal
being less well known - Vectron is well known for crystal oscillators,
with Dale and Raltron less well known.)  The part you could use is very
much motherobard dependent. This would also void any warrenty you might
have with the board. In the case of every motherboard I can look at at
the moment, the crystal is soldered directly to the motherboard, except
where the vendor is using an oscillator chip - only one old board I have
has this as a plug in componant - all of the rest have the oscillator
soldered in place.  Some of my other motherboards use a "Dallas
Semiconductor" Real Time Clock", These have a Lithium battery, oscillator
and the crystal all in one sealed package. The "good news" about that is
that these chips are _usually_ in sockets and are easy to replace. The
bad news is that you are not likely to find one that keeps better time
than what you have now.

In your other reply, you mention using an Intel 82801BA.   According
to a google search, this _appears_ to have an external crystal, but
I don't see the specifications for it.

>> You don't mention what operating system this computer is using. With
>> some versions of Unix, there is an clock tweaking application that can
>> be run,

>Can you give me some hints here?

In your other reply, you mention using Debian 3.0R2 and RH7.3.

[compton ~]$ whatis adjtimex
adjtimex             (2)  - tune kernel clock
adjtimex             (8)  - display or set the kernel time variables
[compton ~]$

I don't recognize the motherboard or chipset you mention, but Red Hat 7.3
was (as I recall) x86 only. In that hardware family, time is based on the
real time clock (also known as the CMOS clock), which often runs from a
low frequency clock (32768 Hz, or 1.8432 MHz are common). When the system
boots, it reads the CMOS clock time to get an rough idea of what time
and date is. The boot process programs a timer on the motherboard such
that the low frequency oscillator is divided down to 100 ticks per second,
and this is fed as IRQ0 tp the CPU. If you look at the contents of
/proc/interrupts, the count for IRQ 0 is increasing at that 100/second.
The kernel then uses that interrupt to keep track of RELATIVE time.

Now, if your system is running continuously and kept as a relatively
constant temperature (such that the crystal for this oscillator is at
a constant temperature), then you can use the adjtimex to fine tune
the kernel's idea of relative time.   Turning the system off and on
costs you in two ways - first the temperatures inside the case will
vary widely, and this will have a significant effect on the accuracy
of the relative time. Second, when the kernel reads the CMOS time at
boot, that time has a resolution of 1 second.

Most of the crystals used in this type of service are sensitive to
the temperature of the crystal - in some cases, very sensitive.
That is the rational for keeping the system running all the time.
One thing that I did to improve the stability of a system that used
a crystal you could touch was to enclose the _crystal_ in a small
chunck of styrofoam which then acted as an insulator to slow the
rate of change of the temperature of the crystal. 

>I<B4>ve some secure server (certificate authoruty) that is kept offline
>by the most of time, so I<B4>m using ntp just to make some calibrations,
>but the ideal is not to connect this server.

Well, the server has to be connected in order to serve certificates,
but I guess you mean that you don't want to have it connected to the
internet. That would be understandable.

>I<B4>ve studied about a GPS solution, but board+antenna+acessories is
>costing something around U$1000,00 and that is too much for me :).

That cost is VERY high.  Is this because of customs duties or import
taxes, or are you just looking at Mercedes and Cadillacs instead of
Volkswagens, Chevies and Toyotas?

        Old guy

Here you can find cheap Motorola Oncore receivers, and antennas.

        http://www.synergy-gps.com/Excess_Inventory.pdf

Here you used to be able to buy Interface boards for above

        http://www.tapr.org/tapr/html/Fvpib.html

Tapr still have Garmin interface boards

        http://www.tapr.org/tapr/html/Fgarminib.html

Jupiter OEM boards and extras might be found here

        http://gpskit.nl/prices-en.htm

Here are GPS boards from an electronics supplier in Sweden. You should
be able to find similar boards from suppliers in Brazil.

        http://www.elfa.se/elfa/produkter/en/6924.htm

Often you need some kind of interface board. That takes care of giving
the GPS board a clean and stable 5V (or maybe 3.3V). Most GPS boards
have TTL levels for serial signals and PPS. To connect these you need
a MAX232 chip or similar. Othervise your computer serial port might
damage the GPS board. Google abound and you will find schematics if
you cannot find a interface board that fits your receiver.

Another source for a receiver is ebay and other online
auctions. You might search for 'Motorola Oncore'.

Good luck!

/Bj�rn 

V�tor Claudino dos Santos <vitor@bry.com.br> writes:

> Great! Can you point some directions?
> 
> V�tor