![Quantization Implementation Losses as a Function of AGC Input Signal Level Setting [3]](https://www.researchgate.net/profile/Chris-Hegarty-2/publication/252351232/figure/fig2/AS:655153485066244@1533212171336/Quantization-Implementation-Losses-as-a-Function-of-AGC-Input-Signal-Level-Setting-3_Q320.jpg)
Content uploaded by Chris Hegarty
Author content
All content in this area was uploaded by Chris Hegarty on Jan 06, 2015
Content may be subject to copyright.
Content uploaded by Taehwan Kim
Author content
All content in this area was uploaded by Taehwan Kim on Nov 23, 2014
Content may be subject to copyright.
Suppression of Pulsed Interference through
Blanking
Christopher Hegarty
Federal Aviation Administration, Washington, D.C.
A.J. Van Dierendonck
AJ Systems, Los Altos, CA
Dan Bobyn
Dan Bobyn Engineering Ltd., Calgary, Alberta, Canada
Michael Tran, Taehwan Kim
The MITRE Corporation, McLean, VA
Joe Grabowski
Zeta Associates, Fairfax, VA
BIOGRAPHIES
Dr. Christopher Hegarty received his B.S. and M.S. from
Worcester Polytechnic Institute, and his D. Sc. from The
George Washington University. He has been with The
MITRE Corporation since 1992, most recently as a Project
Team Manager. In August 1999, he began a one-year
assignment as Civil GPS Modernization Project Lead with
the FAA through the Intergovernmental Personnel Act.
He was a recipient of the 1998 ION Early Achievement
Award, and currently serves as Editor of Navigation:
Journal of the Institute of Navigation and as co-chair of
RTCA SC159 Working Group 1 (3
rd
Civil GPS Frequency).
Dr. A.J. Van Dierendonck received a BSEE from South
Dakota State University and MSEE and Ph.D. from Iowa
State University. Currently, he is self-employed under the
name of AJ Systems and is a general partner of GPS
Silicon Valley. In 1993, Dr. Van Dierendonck was awarded
the Johannes Kepler Award by the Institute of Navigation
Satellite Navigation Division for outstanding
contributions to satellite navigation. For 1997, he was
awarded the ION Thurlow award for outstanding
contributions to the science of navigation. A.J. has 25
years of GPS experience and is a Fellow Member of the
IEEE and ION. Recently, he was inducted into the GPS
Joint Program Office’s GPS Hall of Fame. He currently
serves as working group co-chairman of the RTCA SC159
Working Group (WG1) for the 3
rd
Civil GPS Frequency.
Dan Bobyn is an RF engineer, who works full time as an RF
and microwave circuit design consultant and contractor.
He has spent the last 21 years working with and for
companies designing navigation and communications
equipment. He holds a B.Sc. in EE from the University of
Saskatchewan.
Dr. Michael Tran received his BSEE, MSEE, and Ph.D. from
Virginia Tech. Prior to joining The MITRE Corporation in
September 1999, he worked at Texas Instruments for 2 years
and at Stanford Telecom for 4 years in the area of satellite
communications, DSP algorithm coding, and receiver VHDL
algorithm coding.
Taehwan Kim is a lead engineer in MITRE CAASD. Before
joining CAASD in 1997, he worked for STel and Hughes for
10 years supporting NASA satellite communications. He
received his B.S. in mathematics from Seoul National
University, M.S. in Computer Science from the University
of South Carolina, and is pursuing a Ph.D. in EE at the
University of Maryland.
Joe Grabowski received his B.S.EE from Carnegie-Mellon
University and M.S.EE from Purdue University. Since 1990
he has been working at Zeta Associates on various
communications and digital signal processing projects as a
systems engineer. Previously he has worked at ESL Inc.
from 1984 to 1990 also as a systems engineer and at Harris
Corporation from 1978 to 1982 as an analog circuit designer.
ABSTRACT
In November 1999, the Interagency GPS Executive Board
(IGEB) endorsed a set of recommendations on
implementing the third civil GPS frequency (L5) that
included certain measures to be taken within the United
States to ensure that L5 can coexist with government
systems operating at the same or nearby frequencies.
These recommendations were based on analyses
conducted in 1999 that assumed that pulse blanking is
employed by GPS L5 user equipment. This paper
describes the impact of pulsed interference on GPS user
equipment and presents the results of simulation and
hardware tests that were conducted this year to validate
the assumed performance of L5 user equipment with
blanking.
INTRODUCTION
In November 1999, the Interagency GPS Executive Board
(IGEB) endorsed a set of recommendations on
implementing the third civil GPS frequency (L5) [1] that
included certain measures to be taken within the United
States to ensure that L5 can coexist with government
systems operating at the same or nearby frequencies.
Specifically, the IGEB recommendations included that:
1. The new L5 signal structure proposed in [2] be
implemented.
2. The L5 signal be implemented with minimum received
signal power at –154 dBW (6 dB higher than the L1
C/A code).
3. If the Department of Defense (DoD)’s proposal for
updated spectrum certification of Link 16 is approved
through the established National Telecommunications
and Information Administration (NTIA)/
Interdepartment Radio Advisory Committee (IRAC)
process, the DoD will increase the operational
management of Link 16 Time Slot Duty Factor (TSDF).
4. In the United States, the Federal Aviation
Administration (FAA) will regionally reassign
Distance Measuring Equipment (DME)/Tactical Air
Navigation (TACAN) within +/-9 MHz of 1176.45
MHz as necessary.
5. The DoD will include a priced option in the full-rate
production Multifunctional Information Distribution
System (MIDS) frequencies in the 960 – 1215 MHz
band.
Importantly, these recommendations were based entirely
on analytical analyses conducted in 1999. In order to
validate some of the critical underlying assumptions, an ad
hoc Working Group (WG1, Validation of Coexistence) of
the IGEB was established earlier this year. This paper
describes software simulations and hardware tests that
were performed under the auspices of WG1 to validate the
performance of L5 user equipment using blanking to
suppress pulsed interference. The paper first provides
some background information on the effects of pulsed
interference on GPS receivers without blanking, and an
overview of pulse blanking.
EFFECTS OF PULSED INTERFERENCE ON GPS
RECEIVERS
The effects of pulsed interference on GPS receivers can
vary widely depending on the characteristics of the
received interfering signal (peak power, duty cycle, pulse
width) and the exact implementation of the receiver. Very
strong pulsed signals can cause problems even during
their “off” state, since active components in the GPS
receiver may require “recovery” time after a pulse to
resume normal operation. As illustrated in Figure 1, all
electronic amplifiers saturate when the input signal
reaches some level (i.e., amplifiers cannot continue to
provide constant gain with only finite power available).
Typical amplifier output levels at the point of saturation
range from 0 to 20 dBm for commercial GPS receivers
(usually towards the higher end of the range for later gain
stages in the front-end).
Input (dBm)
Output (dBm)
saturation
Figure 1. Amplifier Saturation
Of greater interest is the power level needed at the
antenna output port to saturate the receiver.
Determination of this level, as a function of frequency,
requires careful analysis of the entire receiver front-end.
In-band pulses will typically saturate the last gain stage
first. For one commercial receiver design that was looked
at in detail, this last gain stage is designed to provide 15
dB of headroom above the normal thermal noise level at
that stage. For this design, an in-band pulsed signal with
peak power of –85 dBm at the antenna output port will
saturate the front-end. Interference sources at frequencies
offset from the center of the receiver passband require
progressively larger amplitudes to saturate the receiver
front-end as that frequency offset increases. Near-band
interference will tend to saturate the intermediate gain
stages with medium-level input interference, and out-of-
band interference signals will saturate only the first radio
frequency (RF) stages of the front-end. This characteristic
is due to the interspersing of filters between each RF and
Intermediate Frequency (IF) gain stage, which leads to
increasing frequency selectivity as the interference signal
proceeds through the front-end.
Near-band and out-of-band pulses can saturate radio
frequency (RF) gain stages first, since these stages
precede highly selective intermediate frequency (IF) filters
that protect later IF gain stages.
Figure 2, created through SPICE (Simulation Program with
Integrated Circuit Emphasis) simulations, shows recovery
times required by discrete and monolithic microwave
integrated circuit (MMIC) amplifiers typical of those used
in commercial GPS receivers. The results indicate that
recovery times of approximately 40 ns/dB of input level
beyond the saturation point is typical for active stages in
commercial GPS receivers. It should be noted, however,
that today’s commercial receivers are not built to operate
in a pulsed environment. Prudent design measures for L5
would include raising the saturation levels for some active
stages, and designing these stages to recover more
rapidly in the event of a saturating pulse. With careful
design, it is anticipated that recovery times of a few
hundred nanoseconds can be practically implemented.
Recovery Times vs Drive Level
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0 10 20 30 40 50
Drive level beyond saturation (dB)
Recovery Time (nSec)
INA02186 MMIC
HBFP0405
Bipolar
Figure 2. Recovery Times for Typical Discrete and
MMIC Amplifiers
Strong pulses will also saturate the analog-to-digital (A/D)
converter in GPS receivers. A/D saturation is actually a
beneficial event in that it limits the amount of pulse energy
that enters the correlators. When the A/D is being
saturated by a pulse, the desired signal is completely
suppressed.
Weaker pulses (those that do not saturate active stages or
the A/D) may be viewed as adding to the receiver’s noise
floor. As will be discussed in more detail later, the
degradations to receiver functions depend on the power
spectral density of the pulsed signal. Data demodulation
is very sensitive to pulse width as well. This paper
focuses exclusively on short-pulsed interference, expected
at or near L5, which is characterized by many pulses within
each data symbol.
Automatic Gain Control (AGC) effects are also very
important with pulsed interference. An AGC is needed
with multibit A/Ds to properly set the A/D input level to
minimize signal-to-noise losses due to quantization. As
shown in Figure 3 [3], A/D quantization losses for
common 2-bit and 3-bit implementations can increase
significantly if the signal level at the A/D input is not
properly maintained.
The presence of pulsed interference will degrade the
performance of slow AGCs (typical in commercial
receivers) unless preventative measures are taken. Slow
AGCs set the A/D input levels based on average signal
power averaged over pulse “on” and pulse “off”
conditions and thus the quantizer levels are not properly
set for either condition. Consider the effect of strong
pulsed interference. The slow AGC will correctly
determine the increase in the root mean square (rms) signal
voltage. The AGC will then decrease the signal level
entering the A/D (or equivalently increase the quantizer
input levels), which increases quantization losses for
those periods of time between pulses (the only periods of
time when the desired signal is not suppressed). With
weak interference present a similar degradation occurs
when the pulses are not present. In addition, when the
pulses are present, the AGC sets the A/D input level too
high. A fast AGC would solve these problems. Increasing
the number of A/D bits would alleviate the problems, since
performance is no longer as sensitive to threshold settings
(see Figure 3). However, adding bits decreases benefits
from quantizer limiting of pulse energy.
Figure 3. Quantization Implementation Losses as a
Function of AGC Input Signal Level Setting [3]
PULSE BLANKING
Pulse blanking is a simple technique to suppress pulsed
interference by having the A/D output zeroes when a
pulse is detected. Pulse detection, for strong pulses, is
not difficult since the received GPS signals are buried
beneath thermal noise. Pulse detection may be
accomplished using analog power measurements or,
nearly-equivalently, implemented digitally by analyzing
histograms of the A/D output levels in real-time. With
either method, careful AGC implementation is necessary to
minimize quantization losses. For instance, when using
the histogram method, quantizer output samples that have
been determined to correspond to the reception of strong
pulses should be discarded from use in determining A/D
input levels since these samples would result in improper
quantizer level thresholds. Careful design is also required
to ensure that the receiver will blank any input signals that
are strong enough to saturate the front-end.
It should be noted that blanking is not the optimal
technique for pulse suppression. Given a general received
signal of the form:
)t(n)t(I)t(s)t(r ++= (1)
where s(t) is the desired signal, I(t) is interference (pulsed
or continuous), and n(t) is additive white gaussian noise,
it is straightforward to show that if I(t) is perfectly known,
optimal estimation of certain parameters of the desired
signal (e.g., time of arrival) involves interference excision
(i.e., computation of r(t) – I(t)). If some parameters of I(t)
are unknown (e.g., time-of-arrival, amplitude), but can be
well-estimated as in the case of a very strong pulsed
signal, optimal estimation of some parameters of the
desired signal again involves interference excision. This
approach was not pursued as a minimum requirement for
L5 user equipment, since it requires a very linear front-end
(i.e., many A/D output bits), which was not deemed
practical. In contrast, blanking is very simple to implement
and does not even require a multi-bit receiver. It can be
viewed as near-optimal in the sense that it completely
excises strong pulsed interference, at the expense of
complete suppression of the desired signal during the
strong pulses.
PERFORMANCE OF PULSE BLANKING
With one strong-pulsed signal present, perfect blanking
results in a signal-to-noise ratio (SNR) degradation of
10log(1-PDC
B
) where PDC
B
is the duty cycle of the
blanking signal. This SNR degradation follows from the
fact that when the strong pulses are present, blanking
completely suppresses the desired signal (a 20log(1-PDC
B
)
SNR degradation), but also completely suppresses the
thermal noise (a 10log(1-PDC
B
) SNR gain).
A more detailed equation [4] that was used for computing
the S/N
0
degradation to an L5 receiver in the analyses that
led to the IGEB recommendations in November 1999 is:
)10PDC1log(10
)PDC1log(205.39N/S
N
1i
10
R
B
Beff,0
i
∑
=
+−−
−+=
(2)
where
iidBi
dclog10dBm 97P)R( ++= , (3)
PDC
B
(pulse duty cycle – blanker) is the total duty cycle of
all pulses strong enough to activate the blanker, N is the
total number of low-level undesired received signals (i.e.,
those not strong enough to trip the blanker), P
i
is the peak
received power of the i-th undesired signal, and dc
i
is the
duty cycle of the i-th low-level signal. Importantly, this
equation is only valid when the pulses are very short
relative to the minimum predetection integration time used
by the receiver (1 – 10 ms, depending on the mode of
operation). This constraint is satisfied for the sources of
emission of concern near L5.
The formulation of this equation included several
simplifying assumptions. First, the equation
conservatively assumes that pulses never collide, thus
strong pulses never suppress weak ones, and duty cycles
are directly summed. Second, a highly simplified
correlation model is used that assumes that weak pulse
energy is spread evenly across the 20 MHz L5 signal null-
to-null bandwidth by correlation of the interference with
the replica L5 codes. More refined correlation models are
described later in this paper.
Because analyses based on this equation (with an
assumed 1 µs pulse detection delay) directly led to the
recommendations in [1], IGEB WG1 efforts have focused
on validating its accuracy. The next two sections present
software simulation and hardware tests that were
performed for this purpose.
SOFTWARE SIMULATIONS
As one means to verify L5 receiver performance in the
presence of pulsed interference, a high fidelity software
simulation of a GPS/Wide Area Augmentation System
(WAAS) single-channel receiver was developed. The
simulation emulated a GPS/WAAS L5 receiver beginning
with modeled analog inphase (I) and quadraphase (Q) IF
received signals sampled at 40 Msamples/s. A 20 MHz
two-sided receiver front-end bandwidth was emulated
using a four-pole low-pass filter. The I and Q samples
were then quantized into 3 bits (8 levels), and correlated
with early, late, and prompt samples of replica L5 I (I5) and
Q (Q5) codes (including the L5 Neuman-Hoffman codes
[5]). The correlator outputs, integrated over 10 ms for
GPS, or 2 ms for WAAS, were subsequently processed to
form carrier phase, code phase, and bit estimates.
A third-order carrier phase loop was implemented using
the loop filter formulation from [6]. A phase locked loop
(PLL) with a 10 Hz loop bandwidth was implemented using
the Q5 channel for GPS. A Costas loop with a 10 Hz loop
bandwidth was implemented for WAAS signal phase
tracking (note that the WAAS L5 signal was assumed to
only include an I5 code with no quadrature channel).
The code phase tracking loop used a first-order early-
minus-late power delay locked loop (DLL) with carrier
aiding (operating on the Q5 channel for GPS). Early-late
spacing was set at 0.5115 L5 chips (2 samples at the 40
MHz sampling rate). The code tracking loop bandwidth
was set at 5 Hz. This unusually large code loop
bandwidth was selected to allow collection of closed loop
error statistics with reasonable run times (on a 450 MHz
computer, one simulated receiver second took about 3 min
to run). A limited number of extremely long runs were
conducted using a 0.05 Hz loop bandwidth to verify the
expected linear proportionality between loop bandwidth
and tracking error variance.
Data demodulation was accomplished with 3-bit soft-
decision Viterbi decoding of the 50 bps GPS data (250 bps
WAAS data).
Analog pulse detection was emulated by adding the
squares of the IF I and Q samples to form an
instantaneous power estimate, and then smoothing this
estimate using a first-order digital filter with a 0.25 µs time
constant. The smoothed power estimate was compared
with a threshold consistent with a peak received power
level of –86.5 dBm at the antenna output port. This level
is more than 10 dB above the ambient noise power level of
–97 dBm. The threshold level was selected to be high
enough to preclude undesired blanking in the event of a
continuous interference source [4].
Calibration
To ensure the simulation was working properly, its
outputs were carefully calibrated against theory prior to
adding any pulsed interference. The predicted variance
(in radians squared) for a PLL tracking the Q5 code in a
GPS L5 receiver is [7]:
0
2
N/S
B2
φ
φ
=σ (4)
where B
φ
is the loop bandwidth.
The predicted code tracking variance (in code chips
squared) for a DLL tracking the Q5 code is:
2
c
2
cc0
2
c
2
2
cLL
2
df)fdTsin()f(H)f(fSTN/S)2(
df)fdT(sin)f(H)f(S)TB5.01(B2
ππ
π−
=σ
∫
∫
∞
∞−
∞
∞−
τ
(5)
where B
L
is the code tracking loop bandwidth, H(f) is the
frequency response of the bandlimited front-end, S
c
(f) is
the power spectral density of the spreading code, d is the
chip spacing (in L5 code chips), and T is the predetection
integration time.
Equation (5) is based on [8] with modifications to account
for the power split between the I5 and Q5 channels, and
also to account for the implemented bandlimiting of the
front-end (perfect filtering is assumed in [8]).
Figure 4 plots the standard deviation of the code and
carrier phase tracking errors (as circles) for the simulated
GPS L5 receiver based on simulated receiver operating
period of 33 s. Also shown on the figure (as solid lines)
are the theoretical results, calculated from equations (4)
and (5) after applying input SNR corrections to account
for quantization losses (0.4 dB) and signal losses due to
the bandlimiting (0.5 dB). The 0.5 dB signal loss
correction was only applied to carrier tracking, since
equation (5) implicitly accounts for this effect.
30 32 34 36 38 40 42 44 46 48 50
0
2
4
6
8
10
C/N0 (dB-Hz)
RMS Carrier Phase Tracking Error (deg)
30 32 34 36 38 40 42 44 46 48 50
0
0.5
1
1.5
2
C/N0 (dB-Hz)
RMS Code Phase Tracking Error (m)
Figure 4. Simulation Calibration Results - GPS
Strong Pulse Results
To validate equation (2) in the presence of strong pulses,
the simulation input signal was corrupted by pulsed
interference with peak power level of –74 dBm (above the
blanker threshold). The pulse repetition interval was held
fixed at 12 µs and the duty cycle was increased from 0 to
90 percent (see Figure 5) to produce pulsewidths
consistent with those expected near L5 [4]. The input
SNR, without interference, was set to 41.5 dB-Hz [4].
Figure 5. Square Pulse Inputs (vertical bias added to
facilitate viewing)
The standard deviation of the simulated carrier phase and
code tracking errors are shown in Figure 6 (as circles) as a
function of duty cycle. Also shown on the figure (as solid
lines) are the predicted errors, based on (4) and (5) with
input SNR adjusted by the implementation loss
corrections (described above) and 10log(1-PDC
B
) (which is
what the SNR reduction predicted by equation (2) reduces
to when only strong pulses are present). Note the
excellent agreement. No data bit errors occurred over any
of the 33 s runs.
0 10 20 30 40 50 60 70 80 90
0.4
0.6
0.8
1
1.2
1.4
Pulse Duty Cycle (%)
RMS Code Tracking Error (m)
0 10 20 30 40 50 60 70 80 90
2
3
4
5
6
7
8
9
Pulse Duty Cycle (%)
RMS Carrier Phase Tracking Error (deg)
Figure 6. Strong Square Pulse Simulation Results -
GPS
Weak Pulse Results
To validate equation (2) in the presence of weak pulses,
the simulation input signal was corrupted by pulsed
interference with peak power level of –88 dBm (below the
blanker threshold). The pulse repetition interval was again
held fixed at 12 µs and the duty cycle was increased from 0
to 90 percent. The input SNR, without interference, was
set to 41.5 dB-Hz [4].
No data bit errors were observed. The standard deviation
of the simulated carrier phase and code tracking errors are
shown in Figure 7 (as circles) as a function of duty cycle.
Also shown on the figure (as solid lines) are the predicted
errors, based on (4) and (5) with input SNR adjusted by -
10log(1+10
(Pi+97)/10
dc
i
) (as predicted by equation (2)). In
this case, equation (2) predicted much greater
degradations to carrier phase and code tracking than were
observed. It was speculated that this discrepancy was
mostly due to the correlation model used. As stated
earlier, equation (2) assumes that correlation with the
reference spreading waveform spreads the interference
signal uniformly over the reference waveform’s two-sided
bandwidth.
0 10 20 30 40 50 60 70 80 90
0.4
0.6
0.8
1
1.2
1.4
Pulse Duty Cycle (%)
RMS Code Tracking Error (m)
0 10 20 30 40 50 60 70 80 90
2
3
4
5
6
7
Pulse Duty Cycle (%)
RMS Carrier Phase Tracking Error (deg)
Figure 7. Weak Square Pulse Simulation Results - GPS
Two improved correlation models are shown in Figure 8.
The model illustrated in Figure 8a is well-known and an
equation for the effective interference power spectral
density (psd) derived from this model [3]:
∫
∞
∞−
= df)f(S)f(SI
ICeff,0
, (6)
where S
C
(f) is the normalized psd of the reference
spreading waveform and S
I
(f) is the psd of the interfering
signal, accurately predicts degradations to carrier tracking,
data demodulation, and acquisition from the presence of
non-white interference.
The second correlation model, shown in Figure 8b is not
so well-known. The model in Figure 8a is often misapplied
to predict code tracking degradations due to interference.
Proper analysis of interference effects on code tracking
must take into account the correlation between early- and
late-correlator outputs [8,9]. An equation for the effective
interference psd for Figure 8b is:
∫
∞
∞−
π= df)f(S)fdT(sin)f(S
d
2
I
Ic
2
Ceff,0
, (7)
where d is the spacing between the early and late
correlator outputs (in code chips) and T
c
is the code chip
period.
Integrate
& dump
r(t)
c(t)
Integrate
& dump
r(t)
c(t+d/2)-c(t-d/2)
(a) (b)
Figure 8. Weak Pulse Correlation Models for (a)
Carrier Tracking, Data Demodulation, and Acquisition,
and (b) Code Tracking
To evaluate these improved correlation models,
quantization was disabled in the simulations (to isolate
correlation effects from A/D effects) and the code and
carrier phase error statistics from simulation was compared
to equations (4) and (5) with the input SNR in those
equations replaced with:
eff,00eff,0
IN
S
N
S
+
= (8)
The comparison yielded very close agreement between
simulation and theory (see Figure 9, simulation results are
shown as circles, theory as solid lines). Note that the lack
of degradation in the code tracking performance as the
duty cycle of the pulsed signal was increased is due to the
fact that the simulated early-late DLL with d = 0.5115 chips
is insensitive to interference with power concentrated
tightly about the carrier frequency (as predicted by
equation (7)). The simulated interference waveforms fall
into this category.
0 10 20 30 40 50 60 70 80 90
0
2
4
6
8
Pulse Duty Cycle (%)
RMS Carrier Phase Tracking Error (deg)
0 10 20 30 40 50 60 70 80 90
0
0.5
1
1.5
2
Pulse Duty Cycle (%)
RMS Code Tracking Error (m)
Figure 9. Comparison of Weak Square Pulse Simulation
Results (with A/D Removed) to Improved Theory
Emulated L5 Environment Results
To evaluate the spectral environment near L5, a
measurement campaign was conducted by participants in
IGEB WG1. Signature waveforms, in the time domain, of
emitters that operate at or near L5 have been digitized and
recorded. Data (100 Msample/s, 10 bits/sample) has thus
far been collected from DMEs, Joint Tactical Information
Distribution System (JTIDS), and various radars (AN/TPS-
59, AN/TPS-124, Millstone Hill) using a data collection
system with a passband of 30 MHz centered at L5.
To date, a limited number of simulations have been
conducted using the recorded signature waveforms as
building blocks to reconstruct the current spectral
environment for L5. As one example, Figure 10 plots the
emulated time-domain signal that would be seen by an L5
receiver (DME/TACANs only) at 40,000 ft above
Harrisburg, Pennsylvania, an area in the United States
with a particularly challenging environment for L5 [4].
This signal was provided as the input to the L5 receiver
simulation. The input SNR, before the interference was
included, was set to 41.5 dB-Hz. The results with the
interference included was a degradation in receiver SNR
varying from 3 to 7 dB (as estimated from the code and
carrier phase tracking statistics) over time. The time
variation in SNR was due to random arrival times of the
pulses received from the dozens of visible DME/TACANs,
which produced a variation in duty cycle of the aggregate
signal seen by the receiver. For comparison, equation (2)
as it was applied for IGEB deliberations predicted a 13 dB
SNR degradation. The reason for the discrepancy is that
in the IGEB deliberations, pulses were conservatively
assumed to never collide. This assumption was made
largely to avoid the difficult question of how much credit
to take for overlapping pulses. The question could be
avoided within the United States, since reassigning the
frequencies of DMEs near L5 is not difficult to plan for,
given current U.S. plans to decommission a large number
of these systems in the future. Other nations that would
like to use L5 at high altitudes while retaining a large
number of DMEs may wish to entertain this question.
0 50 100 150 200 250
-140
-130
-120
-110
-100
-90
-80
-70
-60
-50
-40
Time (us)
Received Power (dBm)
Figure 10. Emulated DME/TACAN Environment at
40,000 ft Above Harrisburg, Pennsylvania
HARDWARE TESTS
As a second validation activity, a hardware prototype
blanking receiver was built [10]. The prototype (see
Figure 11) was designed to receive standard L1 GPS
signals in the presence of an externally generated test
interference signal. The platform uses a custom RF front-
end, a modified off-the-shelf GPS receiver and an external
power supply module.
To facilitate system level testing, the custom RF front end
incorporates:
• A GPS L1 Low Noise Amplifier (LNA), as normally
found in an external active antenna,
• The losses of the antenna to receiver interconnecting
cable, and,
• The receiver functions of RF signal down-conversion,
AGC and amplification.
Relative to standard L1 receiver circuits, this front-end
(see Figure 12) provides enhanced RF filter selectivity and
incorporates pulse interference blanking circuits.
A modified NovAtel MiLLennium OEM3 series receiver
tracks the received L1 signals after they are down-
converted in the custom RF front end. The signal
interface between the OEM3 and the custom RF front end
is at IF.
A special software load is contained in the OEM3 receiver.
This software allows the receiver to perform P code signal
handoff and straight P code tracking. Thus, the effect of
pulsed interference can be evaluated for wide-band
pseudonoise (PN) codes, similar to L5.
Figure 11. Hardware Prototype Blanking Receiver
Figure 12. Functional Overview of Prototype Receiver
Strong Pulse Tests
To test equation (2) in the presence of strong pulses, the
hardware prototype’s input signal (from a GPS satellite
signal generator) was corrupted using a Rohde & Schwarz
SMT-03 synthesizer with a pulse modulation capability.
Data was logged from the NovAtel OEM-3 receiver board
related to carrier to noise level, lock status and the AGC
level the receiver operated at. The typical test scenario
would collect data without any interference for a period
from 3 to 5 min followed by some kind of pulsed
interference over the same time period. The interval
without interference established a baseline for assessing
loss in carrier to noise and also confirmed that the receiver
was operating properly. Pulse widths were varied from 100
µs to as little as 1 µs while duty cycles were varied from
2.5 percent to 80 percent and pulsed interference power
levels were varied from as little as –100 dBm to as high as
–50 dBm.
Initial test results (see Figure 13) showed significant
discrepancy between the theory of equation (2) and the
prototype measurements. The discrepancy increased as
the pulse widths were decreased. Close examination of the
input signals to the A/D (see Figure 14) revealed that the
analog blanker exhibited turn-on and turn-off lags, and
also that the switching in and out of the attenuators that
blanked the A/D input signal resulted in undesired signal
transients. Figure 14(a) shows the A/D input signal with
blanking disabled. A pulse is clearly present over the time
interval from 0 to 5 µs. An ideal blanker would zero the
input signal over that time period. The prototype blanker
produced the A/D input signal shown in Figure 14(b). A
turn-on delay of around 1.5 µs is evident, followed by
another 0.5 µs of ringing by the attenuator being switched
in. At 5 µs, when the pulsewidth ends, a 1.5 µs blanking
turn-off delay can be seen, followed by another 0.5 µs of
attenuator ringing as the attenuator is switched out.
These reaction lags and ringing clearly significantly
reduced the effectiveness of blanking in this particular
analog realization.
Losses with Pulse Blanking Enabled
Fixed AGC (Modified S/W)
-10.00
-9.00
-8.00
-7.00
-6.00
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
Duty Cycle (%)
C/N Loss (dB)
10 Log 100 usec 50 usec 15 usec 7 usec 4 usec 1 usec
Figure 13. Initial Strong Pulse Hardware Test Results
To evaluate equation (2) with a nearly perfect blanker, a
second strong pulse test was performed. In this test, the
blanking circuitry of the prototype receiver was disabled.
A mixer was inserted between the IF output of the
prototype front-end and the A/D input on the NovAtel
receiver. The SMT-03 pulse generator was used, with
some auxiliary circuitry, to use the mixer as a fast analog
switch that turned off during the synthesized pulses. The
A/D input signal and pulsed switch-control waveform are
shown in Figure 15. Note that the blanking of the input
signal appears nearly instantaneous on the time scale of
the plot. The recorded SNR degradations during this
induced blanking test very closely matched equation (2)
(see Figure 16), indicating that a fast-reacting analog
blanking mechanism could in fact be as effective as
predicted by theory.
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
-1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
Time (uSec)
Voltage
(a)
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
-1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
Time (uSec)
Voltage
Turn-on
delay
Attenuator
switch-in
transients
Turn-off
delay
Attenuator
switch-out
transients
(b)
Figure 14. IF Output (a) without and (b) with Blanking
Enabled
Pulse Blanking Waveforms
External Function Generator Induced Blanking after Pulse Blanker
733 Video Amp driving Mixer at A/D Input
1 microsecond Forced Blanking every 2 microseconds
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
-0.5 0 0.5 1 1.5 2 2.5 3 3.5
Time (microseconds)
Amplitude
Video Output A/D Input
Figure 15. A/D Input with Induced Blanking
Losses for External Forced Pulse Blanking
Pulse Detector Bypassed using 733 Video Amp
ZLW-2 Mixer used at A/D Input, Fixed AGC 5000
-10.00
-9.00
-8.00
-7.00
-6.00
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
Duty Cycle (%)
C/N Loss (dB)
10 Log 100 uSec Bypass 50 uSec Bypass 15 uSec Bypass 7 uSec Bypass 4 uSec Bypass 1 uSec Bypass
Figure 16. Strong Pulse Hardware Test Results with
Induced Blanking
Weak Pulse Tests
To test equation (2) in the presence of weak pulses, the
hardware prototype’s blanking circuitry was disabled and
the input was again was corrupted using the SMT-03
pulse generator. The results for pulsed interference with
varied peak power and duty cycle are shown in Figure 17.
Nearly identical curves resulted over a range of pulse
widths from 1 to 100 µs. Also shown in Figure 17 are the
predicted degradation curves based on equation (2) with
one slight adjustment (the noise floor of the prototype
equipment was lower than the –97 dBm level assumed in
equation (3)). In general, the agreement was very good.
The discrepancies, especially for the –85 dBm peak power
case, can be explained by the fact that the prototype
equipment operates with a slow AGC, and quantizes the
signal with 6 digital output levels (see earlier discussion).
-10.00
-9.00
-8.00
-7.00
-6.00
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
Duty Cycle (%)
C/N Loss (dB)
Predicted -100 dBm Predicted -95 dBm Predicted -90 dBm Predicted -85 dBm -100 dBm
-95 dBm -90 dBm -85 dBm
Figure 17. Weak Pulse Hardware Test Results
SUMMARY AND CONCLUSIONS
Blanking has been shown, through high-fidelity simulation
and hardware testing, to be an effective means to combat
pulsed interference. A simple equation that was used by
the IGEB for long-range spectrum planning has been
shown to very accurately predict the performance of
blanking for strong pulses. The equation is not nearly so
accurate for weak pulses, but is adequate for its intended
purpose of providing a computationally simple means of
evaluating the spectral environment for L5 for the purpose
of long-range systems planning. Improved correlation
models were described that more accurately predict weak
pulse effects. However, these models require additional
input information on both the interfering signal (power
spectrum) and L5 receiver (early-late correlator spacing).
Application of the simple equation with the assumption
that pulses do not collide has been shown to be very
conservative. This conservatism has been acceptable
within the United States, since a resulting decision to
reassign the frequencies of DMEs near L5 is not difficult
to plan for. Other nations that would like to use L5 at high
altitudes while retaining a large number of DMEs may wish
to revisit this assumption.
ACKNOWLEDGMENTS
The authors would like to thank the participants of IGEB
Working Group 1 for their contributions to the simulations
and hardware tests.
REFERENCES
[1] Anon., Implementation of a Third Civil GPS Signal:
Final Report, Interagency GPS Executive Board,
November 2, 1999.
[2] Van Dierendonck, A.J., and J.J.Spilker, “Proposed
Third Civil GPS Signal at 1176.45 MHz: In-
Phase/Quadrature Codes at 10.23 MHz Chip Rate,”
Proceedings of The Institute of Navigation Annual
Meeting, Cambridge, Massachusetts, June 1999.
[3] Van Dierendonck, A. J., “GPS Receivers,” Global
Positioning System: Theory and Applications, B.
Parkinson and J. J. Spilker, Jr., Ed., Washington, D.C.:
AIAA, Inc., 1996.
[4] Hegarty, C., T. Kim, S. Ericson, P. Reddan, T.
Morrissey, and A.J. Van Dierendonck,
“Methodology for Determining Compatibility of GPS
L5 with Existing Systems and Preliminary Results,”
Proceedings of The Institute of Navigation Annual
Meeting, Cambridge, MA, June 1999.
[5] Anon., NAVSTAR GPS L5 Civil Signal
Specification, Draft 8, RTCA, Inc., Washington,
D.C., June 13, 2000.
[6] Stephens, S.A., and J.C. Thomas, "Controlled-Root
Formulation for Digital Phase-Locked Loops," IEEE
Transactions on Aerospace and Electronic Systems,
January 1995.
[7] Hegarty, C., Evaluation of the Proposed Signal
Structure for the New Civil GPS Signal at 1176.45
MHz, The MITRE Corporation, Working Note
WN99W34, McLean, Virginia, June 1999.
[8] Kolodziejski, K., and J. Betz, Effect of Non-White
Gaussian Interference on GPS Code Tracking
Accuracy, The MITRE Corporation, MITRE
Technical Report MTR99B21R1, Bedford,
Massachusetts, June 1999.
[9] Lopez-Almansa, J.-M., and P. Pablos, “Derivation of
an Analytical Model for Estimation of Measurement
Errors Induced by an Arbitrary Disturbance on a
Coherent Delay Locked Loop,” Proceedings of The
Institute of Navigation Annual Meeting, June 1998.
[10] Bobyn, D., AJ Systems L5 Receiver Prototype
(Hardware Part No. 9911002) Documentation
Package, Revision 0, Dan Bobyn Engineering Ltd.,
Calgary, Alberta, Canada, October 18, 1999.
