Experimental Physics and
Industrial Control System

If there's a genuine drift I'd expect that to be something like
the SCAN value + the processing time.

Maren

Michael Davidsaver wrote:

I mean, the difference between two
consecutive samples is always higher than the SCAN value, accumulating an
error.

Just to make certain I understand you.  Do you mean that the difference
increases with each iteration?  Or that the scan start time is moving
(precessing) relative to start of the system clock's second?

In case you have not seen it, the code in question is periodicTask().

http://bazaar.launchpad.net/~epics-core/epics-base/3.15/view/head:/src/ioc/db/dbScan.c#L561





On 1/3/2013 4:36 PM, Mark Rivers wrote:

What is the magnitude of the error?  How long does it take before the
error is 1 second, for example?

-----Original Message-----
From: [email protected]
[mailto:tech-talk-bounces@aps.anl.gov] On Behalf Of
[email protected]
Sent: Thursday, January 03, 2013 3:09 PM
To: [email protected]
Subject: time drift in camonitor timestamps

Hi,

We have realized that when monitoring periodic PVs, provided by soft
IOCs in
standard machines under Windows/Linux/MacOS, it is observed a drift in
the
timestamps (server time) . I mean, the difference between two
consecutive samples is always higher than the SCAN value, accumulating an
error.

It seems that  this issue does not occur in RT systems such as VxWorks
(as
expected).

We are wondering if this is a known issue. In our opinion, this fact
could
lead to
malfunction problems, especially if it is not taken into account.

Does anyone in the list experienced this problem?

thank you