PSN-L Email List Message

Subject: Re[2]: from Angel in Panama
From: "a.rodriguez" stuff@.................
Date: Mon, 5 May 2003 07:02:23 -0500

Hello Keith,

Thank you for you convincing e-mail.  I have done a bit more work and
even gone so far as to re-read the TAC32 documentation. I think I see
what the main display is telling me.  I also got out the
scope and found the PPS signal at the right places and now think I
understand things a bit better.

In the end I will have to caulk up some of my confusion to WinOncore.
While in the MotBin mode a selftest with WinOncore reported nothing
was working.  Still does.

I am back no to TAC32 with a 1pps and again feel that it is all the
accuracy that I need.

Saturday, May 3, 2003, 8:44:23 PM, you wrote:

k> NTP offsets of 100's of milliseconds are not unusual.  It depends on how far
k> away the server is.

The server wherever it is is far away, believe me, sometimes I think live at the ends
of the earth. Can't even get a resistor sometimes.

k> Can you tell me more about how you have Tardis
k> configured?

Tardis is just set up to get the time from the Naval server and set
the pc's clock, nothing fancy.  I have several station that do not have
GPS's and need to continue to use Tardis and NPT.  I now can see that
I may have to figure out how to make some corrections to the station.

Seislog, the data capture software I use, will allow me to give the
station a time correction.  I just need to figure out what I want that
correction to be.  While I was comparing the TAC32/GPS to TARDIS/NPT
the difference was quite stable and consistent at -.0820 seconds.

k> Just telling it to set the time every second may not do what
k> you want.  Most NTP client software tries to keep the PC clock within a
k> second of UTC.

k> GPS by itself can deliver time well below a microsecond without any trouble,
k> better if you have a good receiver.  I used to work for TrueTime (now a
k> division of Symmetricom) and we and our competitors claim accuracy in the
k> area of 25 nanoseconds RMS, 100 ns peak.  Many of them use Motorola OnCore
k> receivers to get there.

k> The real trick is getting the time from the receiver to the PC.  If you just
k> use the serial message, all of the precision from the GPS has gone out the
k> window.

Yes, I came to that realization a while back.  GPS without 1pps is
worse than NTP.

k> Some people use the 1 Pulse Per Second (PPS) output from the
k> receiver into one of the RS-232 control lines on the PC serial port to
k> improve the timing.  Then they dig deep into the PC to use the interrupts
k> for the best response.  Even then, you can't get the PC's clock set closer
k> than a few milliseconds.  The people who wrote TAC32 claim 25 milliseconds,
k> which is pretty reasonable.

k> Based on your setup, I would question Tardis and NTP before TAC32 and GPS.
k> At these levels, you can ignore cable delay from the antenna.  Even 300
k> meters of cable will only shift the time about a microsecond or so.

Thanks for for getting me back on track.




Public Seismic Network Mailing List (PSN-L)

[ Top ] [ Back ] [ Home Page ]