1999 Joint Meeting EPTF - IEEE IFCS

GROUP-DELAY ERRORS DUE TO COHERENTINTERFERENCE

F.G. ASCARRLJNZ'. T.E. PARKERt and S.R. JEFFERTSt

'University of Colorado. BoulderCO. USA

Time and Frequency Division
325 Broadway. Boulder CO. USA

ABSTRACT dispersive
propagation
medium.Consider
using a
sinusoidal signal describedby
Timeandfrequencytransfertechniquesthat
rely on spread spectrum groupdelay measurements are J2 P sin (ut)
prone to potentiallylargetimingerrorscaused by
coherentinterferencethat is timedelayed.Reducing where the power level is P. the frequency is a and f is
thelevelof thecoherentinterference.increasingthe
time. The phase. 4 of this signal after a propagation
pseudo-random code
rate.
using
or a fractional
correlator can minimize these timing errors. time of r, is given as

1. INTRODUCTION

The
time-and-frequencycommunity
uses where T isthephasedelay. The groupdelay, is
spread spectrum conununication links for the transfer of defined by
timeandfrequencyinformationfromlaboratoryto
laboratoryormorecommonly,thedisseminationof
time and frequency to users across the world The two
systems that are widely used for the transfer of time-
and-frequencyinformation are theglobalpositioning
system (GPS) and two-waysatellite time-and-frequency Since the medium is assumed to be non-dlspersive, the
transfer (TWSTFI'). In both of these systems a timing groupdelay is equal to the phase delay when there no is
signalisencodedon a carriersignalusingbi-phase interfering signal. Consider the case where a signal of
spread spectrum modulation Data taken over the past power P,,,, with the same frequency q delayed by time
ten years indicate that although T W S m is a two-way rd is summed with theon@ signal. The phase of the
system where there should be common mode rejection interfering signal with respect to the main signal is 9
of pathdelays,thedelaystability of this methodof described by
transfer has generallynotbeenmuchbetter than
common view GPS.
However,there
recent
is
a e=m,,
TWSTFT link that shows a factor of 5 improvement in
stability over otherTWSTFT links [l]. In this paper we where rd is thetimedelaybetweenthe main and
discuss limitations on time and frequency transfer due interfering signal. Figure [l] shows that the combined
to hdamental limitations
in groupdelay
measurements. Both GPS and TWSTFT relyon a
'

groupdelay measurement transfer

to time and
frequencyinformation.
Understanding groupdelay
measurementsandtheirlimitations are essentialto
establishingastablemethod of timeandfrequency
transfer. The simplegroupdelay theory in the case of a
propagating sine waveis presented first and is followed
by the groupdelay theory of a spread spectrum signal
with cornlators as the detecting elements.

2. SIMPLE GROUP-DELAY THEORY

Let us assume a groupdelay measurement in

which we are trying to measure the propagation timeof Figure 1. Vector diagram showing the relation
a signal from point A to point B through some non- between the main signal with power P, the delayed
signal with power Pm, andthe resultant vectors u m
with powerPr.

U.S. Government work not protected by U. S. copyright

signal has an angular deviation a and a new level P,. To
keep the math simple we will make the assumption tllat
the delayed signal is small compared to the main signal.
so that we can use the s m a l l angle approximation
sin(a) = a . (5)
Starting with
period of dr, is equal totheperiod of the canier
frequency m. The conclusion is that a group-delay
measurement may not be a good estimate of the time
delay in the presence of delayed coherent interference.
For large rd thepeak groupdelay diverges from the
phase delay without bound even for small values of P,,,.
Figure 2 is a plotof AT,. for increasing delay rd.
between the main and interfering signal.

3. EARLYLATE CODE CORRELATOR

Using a spread spectrum modulated signal to

weget. after using some trigonometric identities and
the small-angle approximationane'qression for a
described by
l k s elpession for a is used to obtain the groupdelay
variation.
Arg is the Werence between the phase delay and the
group-delay when there is time delayed coherent
interference. For e.uample. if the groupdelay were
being measured to try to determine the electrical delay
of a cable. Arg would be the time error due to VSWR in
thecable being measured. Thegroupdelay variation
has a peak value that varies linearly with ?d, and goes
through a full cycle when the phase between the main
measurethe group-delay will limit the group-delay
variations due to coherent interference. The group-
delay deviation from the phase delay goes to zero as the
timedelay rd. between the main signal and delayed
signal becomes larger and the delayed signal is no
longer correlated with the main signal. An example of
a system that uses a spread spectrum signal to measure
time delays isGPS. For GPS. the delayed coherent
interference could be due to multipath effects.
Multipath signals can belarge.in some cases the
multipath signals may be only 10 dB below the direct
signal. Multipath effects are well documented and the
following derivations can be found in many
publications [2.3]. The results can be generalized to
cover delayed coherent interference from any source
such as: multipath, reflections due to impedance
mismatch in long antenna cables, multiple reflections
through filters. near-field reflections in antennas. and
reflections in waveguides. A simplified modelof a
noncoherent earlyAate (EL)correlator such as is used
in both GPS and T W S m systems is described below.
The complete derivations of these equations are
n

t
4PN(t-T-Tc/Z)

PN(t-T+ Tc/2)

I I I
l Figure 3. Code loop detector model.
50 100 150 200 presentedin [2,3]. In Figure 3. the E L codeloop
rd (m) detector is displayedin its simplest form. The input
signal Y(t) is split and is the input to the early and the
Figure 2. The group delay error A 5 as a function of late comelator channels. The reference pseudo-random
the delaybetween the main and interfering signal rd code sequence (PN) to the early and late correlators is
for the an interfering signal level -20 dB from the spaced by a chip. The time variable is t, the reference
main signal. sequence time estimate is T, and the chpping period is
T,. In both the early and late correlation channels, the
signal and thedelayed signal goes through 2n. The output of the correlator is band-pass filtered and

U. S. Government work notprotected by U. S. copyright.

199
squared. The code detector error signal S(E) isthe code detector error signal for the case of no
difference of the early and late correlation clmnel interference. When the delay-locked loop is locked the
oui,. its. The error signal S(E) is dnven to 0 by the function S(&) is equal to 0
delay locked loop @LL) in normal tracking operation.
The tracking-loop error (t-T) is given by E.
P.RjN(rg-r-+)=PS&(rg -r++>, (13)
4. MEASURING GROUP-DELAY WITH A
CORRELATOR rB = r , (14)
First we consider a groupdelay measurement and rg measures the desired quantity r.
without an interfering signal. The spread-spectrum Now consider the case where there is a
signal is propagating from A to B in a nondispersive coherent interfering signal that is delayed by rd. The
medium. The spread spectrum signal is input signal Y(t) is the sum of the direct signal and one
delayed interfering signal:

where P is the power level of the signal. PN(t) is the
pseudo-randomcode sequence and W is the carrier
frequency. After a propagation time 5; the signal
is received and decorrelated by the E L correlator. The
error signal S ( & ) is gven by
y(t)=J2P.PN(t-r).sin(w.t+wt) (15)
+~.PN(t-r-r,).sh(o.t+or+8),
where T is the propagation time.w is the dxect signal
carrier frequency and rd is the t h e delayof the
reflected interfering signal with respect to the dlrect
signal. B is the phase of the interfering signal with
respect to the main signal and is related to rd by
equation (4). The error signal from the codeloop
detector S(&) is

S ( C ) = P . R ; ~ ( ( E - % ) + P ~ * R -$)
~ ~ ( E(16)
+~~

where Rm(t) is the autocorrelation function of PN(t)

+ 2p. R , (6 - %) P m R ,
* (6 + 5, - 4) COS(WT,)
and E is defined as thecode timing error. The code - P . R ~ N ( ( s + ~ ) - P , . R ~ ~+%)
(~+rl
timing error is
- 2 P . R , ( € + ~ ) . P , R p N ( & + t , +%).cos(oT,).
The error signal is a fimction of r-& the timedelay
between the direct and reflected signal and P,,,/P the
where r- is the measured groupdelayand r is the ratio of reflected signal power to direct signal power.
propagation delay as defined above. Figure 4 shows the Figure 5 shows the code detector error signal for the

-1
-1Tc
’
-2Tc 0 1 Tc 2Tc
-1Tc
-2Tc 0 1 TC ZTC
E Loop Timing Error E Loop Timing Error

Figure 4. S(&) without interference. Figure 5. S(€)with interference.

U. S. Government work not protected by U. S. copyright.

200
case where interference is present.
During tracking mode S(&) is 0 and E is not 0
as in the case where there is no interference. The group-
delay error drg is obtained by solving for E with the
condition S(&)=O. drg is a complicated function of P,
P,,,, u1 and ?d. For f d < 0.5 T,. the eqxession for d7, is
manageable and is given by
For P,,,<<P. and when ?d 0.5 T,. Equation (17)
reduces to
which is identical to the expression for the sine wave
signal (see Equation 9). The function for the group-
delay error changes when ?d becomes larger than T,.
Witha spread spectrum signal the interfering signal
begins tolose correlation with the mahi signal as r,,
becomes larger than the c h p spacing of the correlator
(in this example T, = 400 ns). A plot of the groupdelay
error as a rd varies from zero to 1.5 T,, is shownin
Figure 6 for the case of an interfering signal that is -20
dB lower in power than the direct signal. The carrier
20 ' . I I I , ' ' ' ' ' ' ' 1
Figure 6. Group delay error as a function of Td.
The chip period is 400x1s and the detector is a
1/2chipE L correlator.
frequency is 70 MHz and the chip rate is 2.5 MHz. The
error becomes smaller for ~4 > 200 11s. Understanding
the groupdelay error function can help us improve the
spread spectrum time and frequency transfer systems.
The following section discusses some possibilities.
U. S. Government work not protected by U. S. copyright.
S. MINIMIZING GROUP-DELAY ERRORS IN
SPREAD SPECTRUM SYSTEMS
The most obvious method of reducing the
groupdelay error is by reducing the reflected signal
power. P,,. Antennas can be positioned to reduce the
effects of multipath. All impedances should be
matchedto minimize signal reflections in cables and
system components such as filters and amplifiers.
Phase stable cables should be used
to
minimize
mapping small phase changes into large delay errors
[41.
The other method is to eqdoit the correlation
properties of the spread spectrum system. Recall that as
f d becomes larger tlm l d f the clup spacing of the
correlator the groupdelay error begins to decrease.
One methodof decreasing the groupdelay error is
therefore to decrease the correlator spacing [3].
Decreasing the correlator spacing comes at an e'xpense.
The linear tracking region and the acquisition
performance are degraded [3]. It may not be possible to
operate under these conditions in cases where the
signal-to-noise ratio is poor. Figure 7 shows the group-
-201 , , , , , , , , I , ,
0 loo 200 300 m 500 600
ry (ns)
Figure 7. Group delay error as a f i c t i o n of zd.
The chip period is 400 ns and the detector is a 1/10
chip En correlator.
delay error as a function of ?d for the same signal
conditions as in Figure 6 but with a one-tenth chip UL
correlator. In other words the reference codeto the
early and late correlators is s p a d by one-tenth of a
chip.
Another method of reducing the group-delay
error is to use a higher chip rate. By increasing the chip
rate you decrease the correlation time window during
which the interfering signal affects the groupdelay
measurement. Figure 8 displays the group-delay error
for the same signal levels as in Figure 6 but the spread
spectrum signal is generated by a 20 MHz chip rate.
The E/L correlator spacing is one-half of a chip.
201
 In the TWSTFT system one of the sources for
tlle groupdelay error is the VSWR in the 30 m to 60 m
fimctionon the phase of the carrier. The VSWR in
ground station components isclosely coupled to the
environment and may be the major cause of instabilih
in TWSTFT. New technology that isnow available
may be used to increase the stability of current methods
of time and frequency transfer. Some of this
technology includes optical fiber RF links. modems tllat
are capable of chip rates in excess of 20 MHz. phase
stable coasial cables. and better antenna designs.

-l0 t REFERENCES

Larson,K.. “Comparison of GPS Carrier Phase and

0 100 200 300 400 500 600 Two-way Satellite Time and Frequency Transfer.”
4r (m)
to be published in Roc. 30” Precision Time and
Time Interval Systems and Applications Meetings.
Figure 7. Group delay error as a function of Td. 1998.
The chip period is 50 ns and the detector is a one-
half chip En correlator. Holmes. J. K..Coherent Spread Spectrum S’stenrs,
John Wiley and Sons. Inc.. NewYork. 1982.
cables that go from the laboratory to the earth station Van Dierendonck A. J., “GPS Receivers,” Gfobal
[4]. The cables lengths are such that reflected signals Positioning Systems: Theory and Applications.
due to VSWR are delayed from the main signal by 300
Vol. 1,American Institute of Aeronautics and
to 400 ns. As has been illustrated in Figure 6, the Astronautics, Inc., Washington. 1996.
groupdelayerror has the potential of being large.
Making thereflected signal delay larger than 600 11s
Ascamnz, F. G., Parker, T. E. and Jefferts. S. R.
(greater than 1.5 Tc)could make the groupdelay error “Pseudo-random code correlator timing errors due
smaller without changing the chip rate of the modems. to multiple reflections in transmission lines,” to be
Th~scould be accomplished by using an optic fiber RF published in hoc. 30* Precision Time and Time
link with a propagation delay greater than 600 ns. The
Interval Systems and Applications Meetings, 1998.
delayed coherent interference caused by reflections at
the ends of the fiber would be delayed by greater than
600 ns and wouldno longer be correlated withthe
direct signal. To implement this method of mitigating
the groupdelay error, optical fiber should be used and
not a longer RF cable. Reflections in the RF cable can
occur at random places along the cable due to dielectric
imperfections, kinks, or temperature gradients.
Reflections in optical fibers are, for the mostpart,
constrained to occur only at the ends.

CONCLUSIONS

All groupdelay measurements are sensitive to

delayed coherent interference. Groupdelay
measurements using a swept sine wave have a group-
delay error function that is proportional to the time
delay of the interfering signal. The spread spectrum
groupdelay measurement systems have the advantage
that the groupdelayerror function is 0 outside the
correlator bandwidth. In addition, using the time
correlation properties of spread spectrum systems can
minimize the groupdelay error function. Systems that
suffer delay errors due to VSWR problems cannot be
calibrated due to the sensitivity of the groupdelay error

U. S. Government work notprotected by U. S. copyright.

202