This paper proposes a digital pulse blanking algorithm to mitigate pulsed interference from Distance Measurement Equipment (DME) signals on the IRNSS L5 band. The algorithm averages absolute values of analog-to-digital converter (ADC) samples to better detect the envelope of DME interference pulses against noise. It calculates a threshold from the distribution of averaged samples to improve blanking performance compared to using the noise floor. Testing with a hardware simulator and receiver shows the algorithm effectively reduces carrier-to-noise ratio degradation and improves satellite acquisition under moderate to high DME interference levels.
Design of Simulink Model for Constant Envelop OFDM & Analysis of Bit Error Rate
DME Interference Mitigation through Digital Pulse Blanking
1. DME Interference Mitigation through Blanking
Pranav P Ginde, Nikhil M Anand, Harish A H, Ashwitha L S
Accord Software and Systems Private Limited
Bangalore, India
Abstract—The IRNSS Standard Positioning System (SPS)
signals will be transmitted on both L5 (1176.45 MHz) and S1
(2492.028 MHz) band. Like any other GNSS signal present in the
L5 Band, IRNSS L5 signal too is prone to the pulsed interference
from the DME signals, which occupy the frequency band from
960 MHz to 1215 MHz. DME signals are Gaussian shaped pulse
pairs with a certain repetition rate which depends on the air
traffic conditions of the region. Due to tailing of these pulses, it
becomes difficult to clearly distinguish between an L5 signal and
interference pulse in the time domain. This paper describes an
easy to implement time-domain Digital Pulse blanking algorithm
for the mitigation of DME like pulsed RF interference signals.
The basic principle of this algorithm is to average out the noise or
the desired signal samples and determine the envelope of the
interference signal. This averaging process reduces the noise
variance due to which, the interference signal tailing begins to
reveal itself. This revelation can be used to set the threshold
values more accurately, improving the performance of the
blanking algorithm. The performance of the proposed algorithm
is tested using Accord’s IRNSS L5 receiver and Multi-channel
IRNSS Simulator. The analysis and simulation results are
presented along with the results obtained in the actual hardware
setup. The algorithm is found to work very well for a moderate to
high DME interference levels as is evident from the results shown
in the paper.
IndexTerms— Pulse blanking; IRNSS; DME; pulsed
interference.
I. INTRODUCTION
DME (Distance Measurement Equipment) are a
combination of ground based transponders responding to the
air borne interrogators with ranging pulses in the frequency
band 960 MHz-1215 MHz. DME pulse pairs are Gaussian in
nature, with half amplitude pulse width of 3.5 us and with
separation of 12 us between the pulses as shown in the Fig.1. A
number of DME stations in India transmit at frequencies which
cause in-band interference to the IRNSS SPS L5 signal. Since
99.4% of SPS signal power is contained in the 14 MHz
bandwidth, DME interference mitigation is necessary for the
IRNSS receiver operation in these areas.
The mitigation of DME interference basically consists of
pulse detection and subsequent blanking. Both analog and
digital blanking methods are reported in literature [1][2]. The
digital Pulse blanking is easy and convenient to implement in
the hardware. The Blanking algorithm is usually run on the
ADC output samples after comparing against certain threshold.
The assumption made is that the pulses are short and detectable
against the noise floor [3].
Fig.1. DME pulse pair
The detection threshold is a function of the variance of the
thermal noise floor. The quantized ADC output sample is
blanked if it is above this threshold value. The Fig.2 shows 8-
bit ADC output distribution in the presence and absence of the
pulsed interference.
Fig.2. ADC output distributions (a) Without pulsed interference (b) with
pulsed interference
If the Pulsed interference is too strong, the ADC gets
saturated and the peaks are seen at the either ends of the
distribution. Setting the threshold value as that of standard
deviation of thermal noise floor, as shown in the Fig.2(b)
works very well. But when the pulsed interference is weak, the
variance of the noise floor increases, making it difficult to
distinguish between the interference samples from useful signal
samples. In this case, setting the threshold too low will blank a
portion of the useful signal and setting it too high will allow the
interference to creep in into the system. Therefore, the
performance of time domain digital pulse blanking is relatively
poor against weak pulsed interference. Even in the presence of
a strong pulse, due to tailing of the Gaussian shaped pulse,
some of the residual interference signal remains in the blanked
signal.
In the proposed algorithm in this paper, an attempt is made
to alleviate the problem aforementioned, by block averaging
the absolute value of each of the ADC output samples. The
2. threshold for detection is then derived from the distribution of
these transformed output samples.
The algorithm and its implementation aspects are detailed
in the section III. The Hardware setup used for the generation
of DME pulses and collection of data for the analysis is
described in the section II. Section IV presents the test results.
The paper ends with summary and conclusions in the section
V.
II. HARDWARE TEST SETUP
The hardware setup used to validate and test the Digital
Pulse Blanking (DPB) Algorithm is as shown in the Figure 3.
The DME interference is generated at a frequency of 1176.42
MHz, which coincides with the IRNSS SPS L5 center
frequency. The number of pulse pairs is limited to 3000 pp/s
and the duty cycle is varied from 10% to 90%. The DME
Interference signal is combined with the output of Accord’s
multi-channel IRNSS Simulator and fed to an IRNSS L5
Receiver.
A. Generation of DME Interference Signal
The DME – Gaussian pulses is generated in the Matlab
using (1) and the sample values fed to an Arbitrary Waveform
Generator (AWG).
( ) ( )2 22 2
6
11 2
( )
12
4.5
t t t t
DMEP t e e
t e
e s
(1)
The output of AWG was earlier up-converted to 1176.42
MHz by mixing it with the signal of same frequency from a
signal generator as shown in the Fig.4 (a). During setup
validation, it was found that there was a small DC offset in the
AWG output. Due to which the carrier was found to leak in the
output creating a DME source with both pulsed and continuous
interference. This setup is rectified by two stageup conversion
as shown in the Fig.4(b), where the AWG is used to generate a
modulated DME signal at 20 MHz. This AWG output is then
up converted to 1176.42 MHz by mixing it with a frequency of
1156.42 MHz and then filtering with a Cavity filter at
1176.42MHz of bandwidth 30 MHz. Thus creating a DME,
pulsed interference only source.
Figure 3: Block diagram of Hardware setup for validating DPB algorithm.
Fig.4. DME Pulse Generation (a) Architecture 1- Direct Modulation
(b) Architecture 2- two-stage up-conversion.
B. Accord’s IRNSS simulator and Receiver
Accord’s GNSS Simulator is a configurable, multichannel,
multi-frequency, and multi constellation GNSS RF signal
Simulator. This Simulator was configured and used to simulate
7-channels of IRNSS SPS signal. A scenario for the User at a
static position was used for testing.
The IRNSS Receiver used is again an Accord’s L5, 7-
channel receiver. The data used for analysis, development and
testing was collected using this receiver. The DPB algorithm
was implemented in the Correlator-FPGA part. The details of
the algorithm and implementation specific details are covered
in the section that follows.
III. ALGORITHM AND IMPLEMENTATION
As explained earlier in the traditional time-domain Digital
Pulse blanking algorithm, it is difficult to clearly distinguish
the desired signal samples from the interference signal samples
due to the tailing of the Gaussian shaped pulse [4]. Because of
this tailing, a part of the interference signal gets buried in the
noise. Therefore some residual interference signal always
remains in the signal after blanking. This problem is alleviated
by the algorithm presented in the Fig.6.
The basic principle of this algorithm is to average out the
noise or the desired signal samples and determine the envelope
of the interference signal. This averaging process reduces the
noise variance and the interference signal tailing begins to
reveal itself as shown in the Fig.5. This revelation can be used
to set the threshold values more accurately, thus improving the
performance of the blanking algorithm. The Fig.5 shows a
possible blanking threshold TH with traditional time domain
blanking algorithm and threshold set after taking a moving
average of the |x(n)|with the mean value subtracted. There is a
marked improvement in the threshold value calculation, which
is close to the start of the Gaussian interference pulse. Moving
average is used here for the illustration purpose only, whereas,
in the proposed algorithm Block averaging is used in order to
detect the envelope of the interference signal more accurately.
The Algorithm is described here with reference to the
receiver hardware. As shown in the Fig.6, the output of ADC is
split into two signal paths inside the FPGA (Spartan 3 -
3. XCS35000). In one path, 2-bits are derived out of 8-bit ADC
output and sent to the correlator for further processing. In the
parallel path, the DPB algorithm is implemented of which the
Block averaging filter, Distribution generation, Threshold
estimation and the Comparison and Blanking are the main
constituent blocks.
Fig.5. Comparison of thresholds for traditional DPB and proposed algorithm
Fig.6. Block Diagram of the proposed DPB Algorithm
A. Block averaging filter
The absolute value of each of the quantized 8-bit ADC
output samples is taken first and then passed through a block-
averaging filter of N samples. Block averaging is achieved by
an N-sample moving average filter and then down sampling the
output of the filter by N samples. The Fig.7shows input x(n) to
the FPGA and corresponding signal at the output of the Block
averaging filter z(n)for a strong DME pulsed interference for
N=8.
1-NNp
Npn
|x(n)|
N
1
z(n) Where, ..2,1,0p (2)
The samples were collected over 4 ms at sampling
frequency of 30.1 MHz. As seen in theFig.7, the envelope of
the DME pulse becomes clearer against the averaged noise
floor. The samples of this envelope are compared against the
estimated threshold value in real time.
B. Distribution Generation
The threshold value is calculated from the distribution of
the detected envelope samples. This distribution is again
generated inside the FPGA from z(n) by the Distribution
generation block and periodically read by the DSP for user
display and calculation of the threshold. The assumption made
in the calculation of the threshold is that the DME interference
environment does not change drastically over the period of
calculation of threshold.
Fig.7. (a) Input signal x(n) and its zoomed version (b) Output z(n)and its
zoomed version
The distribution of z(n) can also be directly calculated from
the distribution of x(n). This will avoid the |x(n)| operation and
Block averaging filter block in the FPGA. Assuming that the
PDF of x(n) is Gaussian with mean zero and standard deviation
, the distribution of y(n) in the Fig.6, is a half normal
distribution given by (3).
0,
2
exp
2
)( 2
2
y
y
yfY
where,
2
(3)
From (2) where and assuming each sample
value y(n) to be independent and identically distributed with
the PDF of the PDF of z(n), ) is (N-1) times
convolution of with itself, scaled by N. The calculated
and theoretical PDFs of z(n) are as shown in theFigure 8(a).
Figure 8(b) shows the distribution of z(n) in the presence of a
strong DME interference.
Figure 8: (a) PDF of z(n) , (b) PDF of z(n) with pulsed interference
C. Threshold calculation
The threshold calculation involves finding the tail end of
the distribution of z(n) as shown in the Figure 8(b), and then
subtracting the mean of this distribution from this value. This
task is performed in the DSP. Since the |x(n)|operation is
performed and then block averaging is carried out, the
envelope gets shifted by an amount equal to the mean of the
distribution. Therefore the mean needs to be subtracted to set
the threshold value accurately.
4. (4)
D. Comparison block:
The samples of the detected Interference envelope z(n) are
compared against this thresholdThin real time. If z(n) exceeds
this value a detection pulse is generated which is used to stall
the Accumulator in the Correlator section of the FPGA. In the
Fig.6, an alternate path is shown for the ADC outputs from
which 2-bits are generated and sent to the Correlator section for
further processing. The delay between the two paths is matched
so that the detection pulse stalls the Accumulator exactly at the
instant the DME interference sample is at the input of it. Thus
the blanking part is achieved by stalling the Accumulator
(Fig.6).
IV. TEST RESULTS
Using the setup as described in the section II and with the
proposed algorithm (described in section III) implemented in
the FPGA and DSP of the IRNSS receiver, tests were
conducted to record the CNR degradation at different duty
cycles of the Pulsed interference and at different power levels.
The results are as plotted in the Fig.9. The results are compared
against the theoretical degradation curve equation given in [1].
It is seen that for a strong to moderate DME signal interference
levels, the algorithm performs really well. The Acquisition
plots in Fig.10(a) and Fig.10(b) further validate this and
TABLE 1 quantifies the acquisition performance of the
proposed algorithm.
Fig.9. CNR degradation with varying duty cycles and Interference power
levels
Fig.10. (a) Acquisition plot before Pulse Blanking, (b) Acquisition plot after
Pulse Blanking.
TABLE 1. ACQUISITION PERFORMANCE
Sl No Signal Condition Peak to Average
Correlation value
1.
IRNSS signal without Pulsed
interference
5.3797
2.
IRNSS signal with Pulsed
Interference
2.7243
3.
IRNSS signal + Pulsed Interference
+Pulse Blanking enabled
5.0778
The Fig.11 shows a snapshot of the DME signal captured at
IF in the receiver triggered with respect to the Pulse detection
output of Comparison block in the FPGA. The detection pulse
is in excellent coherence with the DME pulse. The slight delay
observed is attributed to the propagation delay of the signal
from input of ADC to the comparison block output in the
FPGA.
Fig.11. Output Observed in digital Oscilloscope.
V. SUMMARY AND CONCLUSIONS
A time domain digital pulse blanking was presented along
with the implementation details. The results presented were
obtained on an actual hardware, which show that the proposed
algorithm works very well for a moderate to high DME
interference signal. The tests were conducted with the AGC
disabled. The authors are presently working on a method to
make use of the DME pulse detection signal in order to control
the AGC gain. They also plan to explore the wavelet based
methods and feasibility of implementing the methods in
hardware for weak pulsed interference mitigation.
REFERENCES
[1] Hegarty, C., A.J. Van Dierendonck, D. Bobyn, M. Tran, T. Kim, J.
Grabowski, “Suppressing of Pulsed Interference through Blanking.”
Proceedings of Proceedings of the IAIN World Congress, San Diego,
CA, June 2000
[2] Grabowski, J, Hegarty, C, “Characterization of L5 Receiver
Performance Using Digital Pulse blanking”, Proceeding of The Institute
of Navigation GPS Meeting, Portland, OR, Sept 2002
[3] Bastide, F., Akos, D., Macabiau, C., Roturier, B., "Automatic Gain
Control (AGC) as an Interference Assessment Tool," (ION GPS/GNSS
2003), Portland, OR, September 2003, pp. 2042-2053.
[4] Grace XingxinGao, “DME/TACAN Interference and its Mitigation in
L5/E5 Bands”, ION Global Navigation Satellite Systems Conference
2007, Fort Worth, Texas, September 2007