From jimr@maia.usno.navy.mil Tue Jul 13 10:44:36 EDT 1999
Received: (from jimr@localhost)
	by maia.usno.navy.mil (8.8.6 (PHNE_17135)/8.8.6) id KAA25789
	for gpst@maia; Tue, 13 Jul 1999 10:39:31 -0400 (EDT)
From: Jim Ray (USNO 202-762-1444) 
Message-Id: <199907131439.KAA25789@maia.usno.navy.mil>
Subject: Evaluation report for Javad receiver
To: gpst@maia.usno.navy.mil
Date: Tue, 13 Jul 1999 10:39:31 EDT
X-Mailer: Elm [revision: 212.4]
Status: RO

EVALUATION OF TIMING PERFORMANCE OF A JAVAD LEGACY RECEIVER
===========================================================
   J.R. Ray, J.R. Rohde, and M.S. Carter
   (U.S. Naval Observatory)


SUMMARY
-------
We have evaluated a Javad Legacy receiver compared with two other GPS
receivers, all deployed at USNO using H-maser frequency standards.  The
Javad system, which was equipped with a JPS Regant single-depth antenna,
appears to have been subject to a high level of multipath variation at
low elevation angles, particularly below about 20 degrees.  There is no
way to establish whether this is intrinsic to the Javad system itself or
caused by the particular siting of its antenna.  While this would be
expected to degrade troposphere and station height estimates, the observed
clock performance, based on data analysis for the week 21-27 March 1999,
was nonetheless outstanding.

DATA COLLECTION
---------------
The Geosciences Research Division at NOAA/NGS was loaned a Javad Legacy
receiver for testing and evaluation during Spring, 1999.  During a portion
of the loan period, the receiver was deployed at USNO to test the system
connected to a H-maser external frequency standard.  Also running at USNO
during the same period was an AOA TurboRogue SNR-12 receiver and an Ashtech
Z-XII3 receiver (modified for time transfer applications), both connected to
H-maser frequency standards.  Deployment information for all three receivers
is tabulated below.

     Table 1.  Receivers evaluated during 18 March - 02 April 1999
     ====================================================================
     name  receiver        antenna       freq. std.   location at USNO
     ====  ==============  ============  ===========  ===================
     USNJ  JPS Legacy      REGANT_SD_E   MC3 H-maser  Bldg 52 roof, W end
     USNO  AOA TR SNR-12   AOAD/M_T      MC3 H-maser  Bldg 52 roof, E end
     USNB  Ashtech Z-XII3  ASH D/M       MC2 H-maser  Bldg 78 roof
     ====================================================================

The USNO receiver is a permanent installation which reports data to the
IGS.  AOA TurboRogues use a P1-P2 cross-correlation technique to measure the
pseudorange at the L2 frequency and they report C/A pseudorange for the L1
frequency.

The USNJ receiver was equipped with a single-depth antenna.  It was also
deployed for a period using a Javad dual-depth antenna but unfortunately
no useful data were collected during that time.

The USNB receiver is a transportable GeTT system developed by the CODE-
OFMET group for time transfer applications (F. Overney, L. Prost, G. Dudle,
Th. Schildknecht, G. Beutler, J. Davis, J. Furlong, and P. Hetzel, Proc.
12th European Frequency and Time Forum, 94-99, 1998).  It is deployed at
USNO for an extended demonstration period.

Both the Javad and Ashtech receivers employ a non-cross-correlation,
Y-codeless method for tracking the pseudoranges at both the L1 and L2
frequencies, which is distinct from TurboRogues.  The non-cross-correlation
method gives P2 observables with higher SNR values, especially at lower
elevation angles.  (See IGS Mail No. 2320, which is available at
ftp://igscb.jpl.nasa.gov/igscb/mail/igsmail/igsmess.2320.)

The two master clocks, MC2 and MC3, are both steered Sigma Tau H-masers
operating in separate environments in different buildings.  MC3 is
tightly steered to MC2 as a back-up clock.  The differences between these
two clocks are continuously monitored and are avaiable at
ftp://tycho.usno.navy.mil/pub/usnodc-amc.gpscp/.

In most respects, USNJ and USNO shared more common elements than USNB,
particularly the frequency standards and local antenna conditions.  The
Bldg 78 rooftop is much more cluttered with other antennas and equipment
than Bldg 52, and greater problems with multipath might be expected.

The comparison test was not ideal in that each receiver used a separate
antenna each at a different location.  This was done for expediency and to
minimize disruption of the operational systems.  As we will see, this
limitation was not serious in establishing the basic performance of the
Javad receiver for timing purposes.  For a detailed comparison of the
relative merits of each receiver type, a common antenna would be required.
Such a test would not necessarily be representative of actual user experience,
however, since each receiver would normally use an antenna from the same
vendor.

The table below compares several measures of data quantity and quality
for GPS week 1002, 21-27 March 1999, as reported by the UNAVCO utility
"teqc."  Of the possible observations above 10 degrees elevation, the
Javad and AOA receivers recorded "complete" usable data similar fractions
of the time, 92% and 94% respectively.  The Ashtech was more successful
at 99%.  Using teqc's estimates of pseudorange multipath variation at L1
(MP1) and at L2 (MP2), the differences among the receivers are much more
striking.  The overall multipath variation (for all observations above
10 degrees) is similar for the AOA and Ashtech receivers at L1 but
distinctly better for the Ashtech at L2.  This is probably a direct 
consequence of the different code tracking techniques used, which favors
better P2 data for the Ashtech, especially at lower elevation angles.
Overall, the multipath variations seen by the Javad are much larger than
for either of the other receivers.  However, examining the elevation
dependence it is seen that this is caused by observations at low elevations,
below 40 degrees and most dramatically below about 20 degrees.  At high
elevation angles, the Javad actually shows the smallest multipath effects.

  Table 2.  Data comparisons for GPS week 1002 (21-27 March 1999)
  =======================================================================
  statistical measure        Javad/USNJ      AOA TR/USNO     Ashtech/USNB
  =======================    ============    ============    ============
  % total observations            92%             94%             99%
    complete (> 10 deg)

  total MP1 RMS (> 10 deg)      1.02 m          0.57 m          0.55 m
    MP1 RMS, 85-90 deg          0.11 m          0.15 m          0.18 m
    MP1 RMS, 15-20 deg          1.76 m          0.93 m          0.81 m

  total MP2 RMS (> 10 deg)      1.14 m          0.80 m          0.56 m
    MP1 RMS, 85-90 deg          0.11 m          0.15 m          0.16 m
    MP1 RMS, 15-20 deg          2.12 m          1.34 m          0.78 m
  =======================================================================

With the data available, it is not possible to determine whether the
high low-elevation multipath variation experienced by the Javad
receiver (USNJ) is due to the intrinsic performance of the receiver,
the instrinsic performance of its single-depth antenna, the particular
siting of its antenna, or other effects.  Test data using the dual-depth
antenna might have been helpful in this regard, but those test data were
not successful.

CLOCK ANALYSIS
--------------
To evaluate the timing performances of the receivers, the Javad USNJ data
were added to the USNO "Final" clock solutions for GPS week 1002.  Results
are posted at http://maia.usno.navy.mil/gpsclocks/index.html in the
sub-directory "1002" for all solution clocks relative to USNO.  The basic
analysis procedures are described at the web site.  Only observations
above 15 degrees elevation are used.  Differences in the pseudorange
observables due to cross correlation versus non-cross correlation receiver
types was compensated for (see IGS Mail No. 2320).  Antenna phase pattern
corrections were applied based on results by Gerry Mader (see
http://www.grdl.noaa.gov/GRD/GPS/Projects/ANTCAL/).

Over the entire week, most of the clock differences observed were due to
discontinuties at day boundaries.  This can be seen in the two following
tables, which show clock differences first within individual days, then
over the full week.  For the comparisons involving USNB, which used a
separate frequency standard (MC2) from the other two receivers (MC3),
the MC2-MC3 H-maser differences can be significant.  However, on 23-25
March, the clock differences involving USNB were much larger than the
differences in H-masers.  These could possibly reflect slight differences
in temperature sensitivity within the Ashtech system, particularly in
its antenna and RF cable.  On the other hand, the USNJ (Javad) and USNO
(AOA TR) clocks behaved very similarly within each one-day period.

  Table 3.  RMS clock differences within observation days, GPS week 1002
  =========================================================================
  WRMS clock diff.   USNJ - USNO    USNB - USNO    USNB - USNJ    MC2 - MC3
  ================   ===========    ===========    ===========    =========
  21 March 1999         49 ps          58 ps          92 ps         47 ps
  22 March 1999         24 ps          30 ps          29 ps         34 ps
  23 March 1999         25 ps         127 ps         129 ps         32 ps
  24 March 1999         20 ps         112 ps         109 ps         26 ps
  25 March 1999         26 ps         107 ps         115 ps         55 ps
  26 March 1999         32 ps          85 ps          68 ps         28 ps
  27 March 1999         22 ps          27 ps          29 ps         54 ps
  =========================================================================

  Table 4.  RMS clock differences for full GPS week 1002
  =========================================================================
                     USNJ - USNO    USNB - USNO    USNB - USNJ    MC2 - MC3
  ================   ===========    ===========    ===========    =========
  WRMS clock diff.      666 ps         857 ps         619 ps        89 ps
  adjusted WRMS *       666 ps         852 ps         613 ps
  =========================================================================
  * after quadratic subtraction of MC2-MC3 variation for differences 
    involving USNB

All the clock differences are much larger when considered over the full
seven-day span due to day-boundary discontinuities.  With three pairs of
clock differences, we can infer the RMS magnitudes of the clock differences
due to individual receiver systems assuming that their variations are
independent (after correcting the USNB differences caused by MC2-MC3
variations).  The results are given in the following table.

           Table 5.  Inferred RMS day-boundary clock
	   variations attributable to individual receivers
           ===============================================
                     USNJ (Javad)         216 ps
                     USNO (AOA TR)        630 ps
                     USNB (Ashtech)       574 ps
           ===============================================

DISCUSSION
----------
The Javad receiver is inferred to produce the smallest clock discontinuities
at day boundaries, at least over the one week analyzed.  Its variation
is only slightly larger than expected based on the formal errors of the
clock estimates.  The day-boundary discontinuities of the other two 
receivers are considerably larger.  This result is suprising in view of
the larger multipath variation of the Javad data (Table 2) and could be
interpretted as excluding multipath as a cause of day-boundary clock
discontinuities.  Such an interpretation would probably be simplistic,
however, partly because of the small sample size of just 6 day boundaries.
Moreover, it should be noted that low-elevation data contribute geodetic
information primarily for tropospheric and station position (especially
height) parameters, whereas high-elevation data are more important for
clock estimates.  So the Javad's large multipath variation at low elevation
angles may not be controlling.  In this view, the very low multipath of
the Javad at high elevation angles would be more important.  More data
will be needed to test this hypothesis.

Even if the Javad low-elevation multipath does not degrade the clock
estimates, it is expected to be undesirable for high-quality tropospheric
and height determinations.  As noted previously, we cannot isolate the
cause of the USNJ high multipath variation as being internal to the Javad
system or related to the environment of the antenna.

Regardless of the uncertainties of the multipath cause and its possible
relationship to day-boundary clock discontinuities, this test does
demonstrate that the Javad Legacy receiver can provide excellent
timing performance.