[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ntp-hackers
Subject:    [ntp:hackers] error estimation on client side
From:       Christopher Hammond <ChrisH () Navsys ! com>
Date:       2004-03-29 19:05:04
Message-ID: 05CD3BFBAA9DB545BC62C48C4540507AC2A078 () mail ! mail
[Download RAW message or body]

Hello, 
I'm needing to estimate error on the client side of an sntp to ntp v4
transaction.
Background:
       I'm attempting to narrow the search space for a GPS receiver through
Network based assistance (NTP)...and chose SNTP because it would shorten my
"Time to First Fix" (which is the main requirement for the system).  It
appears that it takes around 16 seconds for the NTP algorithm to settle and
this is far too long for my TTFF.  SNTP was chosen because it's "one shot"
feature would return a time faster (albeit erroneous) , but we also need an
error estimate to improve the search space determined by the SNTP return.
Do you recommend an alternative approach to get accurate time very fast with
an estimate of error as well?

I see that ntpd/ntp_loopfilter.c will has a RMS jitter called sys_jitter
that is used in loopstats file as the error.
I also see that sys_jitter is gotten from the struct timex ntv.

couple of questions:
	1) is there an open source client that does this (win32 prefered as
project mandates this OS)
	2) is error estimation even possible with SNTP given that it is a
"one-shot" transaction-- I would guess that the jitter is obtained through
averaging the round trip time and determining the standard deviance from the
average
	3) am i out of my mind??

Any advice is greatly appreciated.
regards
chris



C. L. Hammond
Engineer
NAVSYS Corporation
(719)481-4877 x 169
chrish@navsys.com
**************************************************

**************************************************

_______________________________________________
hackers mailing list
hackers@ntp.org
http://mailman.ntp.org/mailman/listinfo/hackers
[prev in list] [next in list] [prev in thread] [next in thread]