From jimr@maia.usno.navy.mil Thu Dec  3 09:30:20 EST 1998
Received: (from jimr@localhost)
	by maia.usno.navy.mil (8.8.6 (PHNE_14041)/8.8.6) id JAA25879
	for gpst@maia; Thu, 3 Dec 1998 09:24:50 -0500 (EST)
From: Jim Ray (USNO 202-762-1444) 
Message-Id: <199812031424.JAA25879@maia.usno.navy.mil>
Subject: 1 pps monitor problems at AMC2 & USNO
To: gpst@maia.usno.navy.mil
Date: Thu, 03 Dec 1998 9:24:49 EST
X-Mailer: Elm [revision: 212.4]
Status: RO

[It is not really my place to send this notice.  However, since the
problem described here has led to considerable misunderstandings among
users of the 1 pps monitor data from the AMC2 station in Colorado
Springs, which was evident yesterday at the PTTI meeting, this report
is being sent even though the information may be incomplete.  --Jim]


Since the replacement of the TurboRogue SNR-12 receiver in Colorado
Springs (AMCT -> AMC2) on 20 October 1998, a persistent problem has
been observed in the monitor 1 pps output data.  (The 1 pps data are
available at the site ftp://tycho.usno.navy.mil/pub/usnodc-amc.gpscp/
in the file "amc1-amct.1pps".  Please see the "readme" file there for
explanations.)  At irregular intervals, several times a week, spikes
are recorded that deviate by almost 3 ns from neighboring values.

After conferring with Larry Young (JPL), Paul Wheeler (Time Service
Dept., USNO) did some checking and found that the problem was caused
by the odd shape of the 1 pps output pulse from the TR.  He showed me
an oscilloscope trace that looked something like:


                           |
                           |
                           |                   +++++++++++++++
                           |                 +
                           |                +
                           |              +
                           |  +          +
                           | + +       +
                           | + +      +
                           |+   +   +
                           |+   +  +
                     1V    +     +
                          +|
                          +|
                         + |
                         + |
                        +  |
                        +  |
                       +   |
                       +   |
         -------------+----------|------------------------
                               ~3ns

The rise is initially steep but it reaches a peak (> 1V) then dips back
down to close to the 1V level before rising again more slowly to the pulse
plateau.  With the trigger level set for 1V there are occasional instances
when the mid-rise dip comes down to 1V causing a secondary trigger about
3 ns later.  According to Paul, all 3 TRs that we own have this 1 pps pulse
shape.  Only the receiver newly installed in Colorado Springs seems to have
a dip that is sometimes low enough to cause a secondary trigger at 1 V.
The other 2 receivers have not been observed to have 1 pps glitches (or at
least not often enough to cause concern).

Larry suggests that the poor pulse shape appears to be caused by a bad
impedance match which produces a reflected signal in a cable about 1.5 ns
long.  Paul has argued that the TRs have "no proper driving circuit" and
a variety of different impendances have not improved the pulse shape.
Larry does not observe similar performance in the TRs he has tested.  (If
anybody has a resolution to this discrepancy, I have not heard about it.)

Regardless of the cause, a simple solution is to lower the trigger level
to 0.5V, which was done on 12 November 1998 for the receiver here at USNO
(after the disruption caused by having to remove and reinstall the antenna).
This is just a precaution in this case since we have not seen any problems
here.  Note that this action will introduce a small bias in the time series
of 1 pps measurements for (Master Clock 2 - USNO) compared to before and
after the change so we would not be able to calibrate for the clock reset
due to the antenna work on 12 November as accurately as before.  But this
is a one-time problem only.

At Colorado Springs, the situation is complicated by the measurement system
there which is used to monitor a number of other things.  Apparently it is
not so easy to change the trigger level for the TR only without adversely
affecting other unrelated measurements.  Paul is continuing to look into
other possible solutions for that receiver.  In the meantime, the 1 pps
data from AMC2 must be filtered to remove the spikes before being used to
interpret the TR receiver performance.