[time-nuts] PC clock comparison software?

[Jack Hudler]

I'll overlook the Windows bashing and try to answer your question.

The architecture of the PC and Windows prohibits anything (except an NMI) from
interrupting time keeping. Time is set in the Windows Kernel by the 2 user level
API's SetSystemTime and SetSystemTimeAdjustment. The later is use to compensate
for the drift of the XO by applying an adjustment on each clock interrupt.
Normally SetSystemTimeAdjustment is only used by the NTP client or Directory
Services in W32Time if it is running otherwise its disabled and the kernel
attempts to calculate and adjustment based on SetSystemTime calls. The code is a
heavy filtered and weird implementation of what may be an Allan deviation (I
only briefly looked at it 5 years ago). 

Windows 9x platform doesn't perform this adjustment however any good NTP client
checking in regularly should alleviate this. 

My point in telling you this is; your problem is with the motherboard; not
Windows. Linux (or your OS of choice) would not do any better.

As I see it there are 2 solutions: (chime in if someone see's others)

1. An NTP time server (sounds like you have this) that's synced to GPS. See
message thread ([time-nuts] Vendors of time sync hardware) which all machines
routinely check in with. 100ms is easy here.
This is what we use at Lone Star Observatory.

2. TAPR Clock-Block http://www.tapr.org/kits_clock-block.html that you can use
with the reference signal, ideally from a disciplined source.

Using the above method (1) our scope can cold start to acquisition of target
without indexing to a known star. Further we divide down the reference signal to
provide an accurate 50ms timer interrupt for the motion base and watchdog.

Option 2 would no doubt provide the most direct and stable clock but you still
need an disciplined source for precision and an NTP source for accuracy
(obviously you'd want this together).  This seems a bit extreme for your use
because it sounds like they are just PC's to you and not part of telescope
control.

Also there is nothing that prevents you from getting your own NTP timer server,
IT need never know and I doubt they care.

Jack Hudler


-----Original Message-----
From: time-nuts-bounces at febo.com [mailto:time-nuts-bounces at febo.com] On Behalf
Of Charles S. Osborne
Sent: Tuesday, December 19, 2006 8:02 PM
To: Time-Nuts
Subject: [time-nuts] PC clock comparison software?

I have a query. Does anyone have a favorite network time sync test software?

Here's my situation. Being at a radio and optical observatory, most
everything we do has varying degrees of time stamp criticality on the stored
data. While I could get obsessive compulsive and try to get everything
sync'd every few seconds with the Z3816A/GPSCon server, much of the data is
quite usable at 100ms or less time accuracy. I'd like better of course.

But for the moment let's just say I need to evaluate where we are, and
improve from that. Most PCs sync to our time server frequently. But PC
clocks being the cheap crap that they are, I need to have some reasonable
way of measuring how quickly they drift. ... particularly if Windows is
experiencing Attention Deficit Syndrome and forgetting to run the core
applications while they obsessively go speak to Father Bill Gates for words
of wisdom or check in with the Mother Ship or whatever it is that they seem
to need an Internet connection to do.

For test purposes I've occassionally set the time server polling routine to
once a day and seen 20 minute errors on delinquent PCs. Other PCs might stay
within a minute. If I try to run the Net-time routine every minute then I
increase the chance that the time server will be busy and not respond. A
goodly handful of these PCs are chugging along quite happily running small
applications on Pentium-166s /Win98se with limited memory. So going to XP or
Linux is not an option in half the PCs. No time, No $$. All the normal
educational, not for profit, zero budget reasons.

Is there a small application that could log the time errors when the system
clock gets set? This would help me decide how often the time needs to update
and what my time error penalty would be if I can't run it that often. I know
temperature will even enter into it as well since several PCs are in
unheated buildings.

It also would be nice if it could log when the time server does not respond.
Some applications respond very ingratiously when the time server won't
answer their query. They hang the program or never resync with the network
even when it comes back. But I can't prove that this is happening. Naturally
our IT folks say the network is running flawlessly. But I'm not that
gulible.

One other idea is to have a central PC poll the outlying PCs for their time
and compare to network time. That reverse net-time sync approach would be
even more useful since the PCs are spread all over many buildings, some a km
of fiber away, and checking/downloading/evaluating their individual logs
would be a chore in itself.

Our network is about 30 PCs on fiber OC12 ( 622Mbps) ATM with 100mbps
ethernet network branches. In the near future some video conferencing and
remote telescope users will add to the chunkiness of the network variable
load. Thoughts on my test software delemma?

Thanks, 73,

Charles S. Osborne, K4CSO
Technical Director

Pisgah Astronomical Research Institute
1 PARI Drive, HC 73, Box 638
Rosman, NC 28772-9614
http://www.pari.edu
828-862-5813 direct
828-862-5554 main
828-862-5877 FAX
cosborne at pari.edu



_______________________________________________
time-nuts mailing list
time-nuts at febo.com
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts