To: Geodetic Stations

From: Ed Himwich (weh@ivcss.gsfc.nasa.gov) and
      Hayo Hase (hase@wettzell.ifag.de)

Re: Clock Offsets

Date: October 9, 2001

Introduction

The clock offsets between the VLBI-formatter and the global GPS-time
must be known on the sub-microsecond level. The offset is used not only
to align the data in the correlator, but it will be also used in VLBI
data analysis to improve the accuracy of the final values of UT1 and UT1
rate observables derived from quasar observations. The IVS is pushing to
achieve higher accuracies in its products, in particular for UT1 (2-3
microseconds) and UT1 rate (0.3-0.5 microseconds). More care in handling
clock offsets at the stations and during data analysis will be required
to achieve this. This memo addresses what stations need to do to support
this.

Please note this memo is also posted on the IVS network coordinator's
web page, http://ivscc.gsfc.nasa.gov/nc/nc.html.

Hardware and Software Requirments

Every geodetic station needs in hardware:
  
- 1 PPS signal from the formatter 
- 1 PPS signal from a GPS receiver
- counter with sub-microsecond resolution and an interface for the Field
  System (serial, IB)
- two commands in their SNAP "midob" procedure: 
  (1) "gps-fmout", to measure the sub-second portion of the clock
      offset uding the signals and counter mentioned above,
  (2) "sy=run setcl &", to check for clock jumps of more than one
      second.
  With these two commands in midob, the correlator can recover the data,
  in most cases, even if the formatter jumps during an experiment.

Important Issues

A few important points need to be emphasized about the use of
"gps-fmout" in the Field Systen (FS):

(1) The command actually comes in two forms: "gps-fmout" and
"fmout-gps". Either form can be used, but it must agree with the wiring!
If the GPS 1 PPS starts the time interval counter, then "gps-fmout" is
used. On the other hand if the formatter 1 PPS starts the counter, then
"fmout-gps" is used. 

(2) We really want to use the formatter, not the Maser. The Maser may
have some unknown offset relative to the formatter. The formatter is the
most reliable choice. Report only the formatter to GPS offset.

(3) We use "fmout", the formatter output 1 PPS and not the formatter
input 1 PPS tick. On a Mark III/IV formatter, use the signal labeled
"CLOCK 1 PPS", not the one labeled "HOUSE 1 PPS". The latter is just a
conditioned feed through of the input signal, probably the Maser 1 PPS.

(4) GPS time and formatter time must not have an offset of integer
seconds when the experiment starts. This must be checked after setting
the formatter time by comparing it to GPS. For Mark IV, VLBA, and S2
systems, you can use "fmset" (in the standard FS display, use
Ctrl-Shift-T) to view the formatter time.

(5) The sub-second portion of the GPS and formatter offset, measured by
"gps-fmout", must not exceed +/- 30 milliseconds at anytime, including
the start of the experiment. This is due to a limitation of some of the
correlators, described in the section below.

How to Handle Clock Jumps
 
In the event of a clock jump in the formatter during an experiment, do
not automatically reset the formatter! It can actually degrade the
results if you do. There are only two conditions under which the
formatter needs a reset after a jump. These are due to the restrictions
in the Haystack, Washington, and Bonn correlators (please note different
restrictions may apply to K4 and S2 correlators). These correlators can
handle integer second offsets of ONLY a few seconds and/or subsecond
jumps ONLY in a small range:
(1) The integer second portion must be within +/- 5 seconds, measured by
    "sy=run setcl &",
(2) the subsecond portion must be within +/- 30 milliseconds, measured
    by "gps-fmout".
Any jump that leaves the offset outside these ranges require a reset of
the formatter by the operator. For example, after a jump, if the offset
(not the size of the jump) is +0.025, -1.019, or +2.972 seconds, no
reset should be made; for offsets of -0.060, -2.658, or +10.0 seconds a
reset is required (these examples are worked through in a separate
section below). Please note that the sub-second portion of a clock jump
is usually a small multiple of 200 nano-seconds, which does not require
a reset.

More Information

For more information on implementing gps-fmout at your station please
read /fs/docs/gpsoff.txt on the FS FTP areas (accessible only if you
have an FS FTP password) of atri.gsfc.nasa.gov and kurp-ftp.hut.fi.

Example Offset Details

The logic for accepting or rejecting the example offsets in the section
on clock jumps is articulated here.

No reset for:
+0.025   offset is smaller than +/- 30ms => ok.
-1.019   offset has an integer jump (not seen by the counter, but caught
         by SETCL) and 19ms < +/- 30ms
+2.972   offset has an integer jump (not seen by the counter but caught
         by SETCL), and (3-2.972=) +28ms < +/- 30ms

Reset needed for:
-0.060   offset is larger than +/- 30ms
-2.658   offset has an integer jump of 2 (ok),
         but (3-2.658=) .342 ms > +/- 30ms (not ok)
+10.00   offset has a too large integer jump