Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

Double Block Zero Padding Acquisition Algorithm for GPS Software Receiver

Submitted: 21st October 2018; accepted: 20th December 2018

S.V.S. Prasad

DOI: 10.14313/JAMRIS_4-2018/26 integration of multiple (Nnc≫1) coherent accumu-

lation outputs with a less coherent integration in-
Abstract: terval Tco. [4], [5] In practice, however, both these
Several methods of acquisition have been developed so methods are expensive in computation, the longer
far which aim at accelerating the acquisition process the coherent integration interval the more the Dop-
and detection of weak GPS signals. In many of these pler frequency bins which leads to the smaller step
the search is parallelized in code or frequency space. size in Doppler frequency search. Therefore there
However, sometimes the number of samples in code peri- will be an increase of computational cost in the
ods based on sampling frequency is not equal to required weak GPS signal detection and which is inevitable.
number of Radix-2 FFT algorithm. So it is again comput- The goal or principle of acquisition is to get the
ing the DFT normal method without FFT. The main pur- rough estimates of code phase and carrier frequen-
pose and objective of this project is to implement a fast cy [11] which are initial values or parameters given
and robust for weak signal acquisition algorithm “Double to tracking module.
Block Zero Padding Acquisition” (DBZP) for GPS L1 civilian
signal. Technique is developed even for the signals where
number of samples in the code period taken for corre- 2. Literature Survey
lation is not satisfying the required number for Radix-2 There are many acquisition methods available so
algorithm. It is also suitable for weak GPS signals acquisi- far traditionally. The serial search in time domain is
tion, which require Pre-detection, or integration (correla- the most basic of the search strategies in terms of al-
tion) Time (PIT) to be long. gorithm complexity, but because of this simplicity is
also the slowest of all the acquisition strategies. In the
Keywords: GPS, Acquisition, Weak Signal, Multi-Constel- parallel search all bins of one dimension (frequen-
lation, DBZP, FMDBZP. cy or code delay) are searched at the same time. In
parallel code phase search acquisition scheme if the
input vector to the Fourier Transform is not a power
(radix 2) then DFT is performed [13].
There are several acquisition methods which
1. Introduction exploit the circular correlation, to name a few like
The Global Satellite Navigation technology has coherent integration, half sized circular correlation.
been playing a key role in the positioning and navi- The goal of these techniques is always to reduce the
gation applications. GNSS at present is more useful to computation time and enhance sensitivity. The ac-
the military, safety of life applications and for civilian quisition step requiring a lot of time and operations,
user applications. There is a need at present to design there are many developed techniques, some of them
and development of such system with more accuracy are not based on circular correlation. This GNSS [3]
and reliability than ever before. To use GPS receivers receiver is meant for educational and research pur-
in these challenging conditions and environments, poses, then the software technology is well adopted
high sensitivity algorithms has to be used. Software due to its reconfigurable nature and important flex-
GPS receivers are more flexible and suitable for re- ibility.
search and development and also useful for integra- The efficient and fast acquisition of GPS L1 [2] sig-
tion with sensors. nal is still a challenge. An acquisition technique which
All the conventional acquisition algorithms work seems adapted to these purpose due to its efficiency
well on strong GPS signals with minimum dwell and computational speed is the Double Block Zero
time but for weak signals to detect by these algo- Padding (DBZP) [14], [15]. By comparing 8 of the pre-
rithms [1], [2] the dwell time needs to be increased viously cited acquisition methods, it was concluded
which leads to increase of computational load and that the DBZP seems to be one of the best for the ac-
hence gets slower. For detecting weak GPS signals quisition of weak signals due to the reduced number
(for example, C/N0 < 30 dBHz)[14], a GPS receiver of operations. DBZP is found as a relevant method to
has to perform either a coherent accumulation with be used as a base for developing a new Galileo E1 OS
long integration time T (≫1 ms) or a non-coherent [14] acquisition technique.

58
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

3. Double Block Zero Padding 4. Implementation of DBZP

Double Block Zero Padding is a method more suit- In total, there are five steps involved in this algo-
able for acquisition of weak signals. This method was rithm. Before defining and explaining step by step, we
also referred to as Circular Correlation by Partition first define two parameters namely
and Zero Padding. This method performs long coher- – Coherent Integration Time (TC),
ent integration with fewer operations and higher sen- – Maximum and Minimum Doppler Frequency
sitivity than other FFT based techniques. The concept (fD,Max, fD,Min).
of this technique relies on the use of partial correla- The coherent integration time is defined as cor-
tions on duration equivalent to a few tens of chips. relation time. The minimum integration time is 1 ms
The computation time gain of this method is the FFT and is limited by the spreading sequence period. The
processing on small size vectors instead of large size minimum and the maximum Doppler frequency pa-
vectors. rameters represent the minimum and maximum vari-
ations from center carrier frequency as there is a rela-
tive motion between receiver and the satellites.

Step 1: Incoming signal Preprocessing. Figure 2

shows the preprocessing of incoming signal which
is first step of this algorithm. The incoming complex
data from the front end is converted into base band
just by multiplying with the complex exponential
function exp(j*2*p*f*n*T).
The resulting TC ms long baseband samples are split
into equal length of M blocks as shown in figure 2.

Fig. 1. Block diagram of DBZP

Figure 1 shows the block diagram of DBZP

scheme, where the search is parallel in both code
and frequency simultaneously and with partial
correlations, where partial correlation length is
generally taken as a fraction of spreading code period. Fig. 2. Pre-processing of the incoming signal
This property enables this algorithm faster as well as
highly computationally efficient. Step 2: Local PRN spreading code. Figure 3 shows
The Doppler bins in double block-zero padding the generating local PRN spreading code. Second
and their step value are fixed step in this method consists of generating local PRN
spreading code, which is further split into same num-
f D ,Max − f D ,Min ber of blocks as in step 1. After that each block is ap-
Nb
= = 2 f D ,Max × TC (1) pended with same number of zeros to make it 2Nb
1
TC Samples as shown in figure 3. The second box repre-
sents appended zeros of Nb samples, thus by adding
As given below, the number of Doppler bins is those two blocks it makes 2Nb samples per each block.
always equal to the number of code delay blocks

– one block duration tb is:

TC 1
tb
= = (2)
Nb 2 f D. Max

– the number of samples per block Nb is equal to:

=N spb N s /Nb = t b × f s (3)

– the Doppler frequency resolution Df is:

2 f D. Max 1
=
∆f = (4)
Nb TC Fig. 3. Pre-processing of the local code

Articles 59
 Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

Step 3: Split signals and Partial correlations. The Step 5: code block permutations. So far we tested for
third step is to take the correlation output using FFT. only one code delay time slice between [0, tb]. C/A code
The first two blocks of the incoming signal samples blocks are permutated circularly in order to test for all
(2Nb samples) and the first block of local PRN code code delays. On first permutation of local spreading
along with the combined appended zeros block Nb block the last block appears to be the first block and
samples of code and Nb samples of zeros are to be cor- the first block appears to be the second block.
related using FFT method.
Zero-Padding is used in the DBZP acquisition
method; when the padding of zeros is not used, the 5. Experimental Results
autocorrelation which is normalized function has The DBZP and Parallel search in code and frequen-
a peak which suffers a degradation due to attenua- cy algorithm is implemented in MATLAB. A Parallel
tion. When it is padded with zeros we can easily ob- search in code phase (1 ms integration time) + Paral-
serve that there will be no degradation and the peak lel search in Frequency (10 ms integration time) Ac-
is isolated without any attenuation. quisition Algorithm using DFT.
T0 + ( k − 1 )TC + lt b ,T0 + ( k − 1 )Tc + ( l + 1 ) t b  (5) GPS IF Data_1 specifications:

Filename=GPSdata-DiscreteComponents-fs4_774
Furthermore, the phase at if1_1935_2MHz1.bin.
IF (Intermediate Frequency) = 1.19352MHz;
t = To + ( k –1 )TC + lt b (6)
Sampling Frequency = 4.774MHz;
The partial in-phase correlator output is Data Type = int8;

To+( k −1)Tc +( l +1)tb

Data contains strong satellite signals with PRNs
i ( k ) = ∫ Blc+1 ( t ) * Bls+1dt (7) 15, 18, 21, 22 along with weak signals 3, 6, 26.
To+( k −1)Tc +ltb

Finally Table 1. PRNs that are having a threshold 2.5 and above
Peak Code
Il ( k ) = PRN
Metric phase
Doppler
A
d ( k ) Rc1 ( ετ ( k , l ) ) cos (π f Dt b + ε ∅o ( k , l ) ) sinc (π f Dt b ) + 15 7.20 4606 1937.72
2
η Il ( k ) (8) 22 7.09 852 1710.08
21 5.26 1741 -557.24
No No N b 18 4.35 2656 253.17
σ
= 2
n = (9)
4t b 4TC

In Table 1, the peak metric is the ratio of first peak
The partial correlator outputs for all permutations
to second peak ratio. The figure 4 gives the peak for
of the C/A code blocks are placed in a matrix form
PRN 15.
(M × Nb).
The peaks for the PRNs 22, 21 and 18 are in the
Figure 4, Figure 5 and Figure 6 respectively. The
Step 4: FFT application. FFT is applied to the par- above same results may not be acquired for the weak
tial correlation matrix output of size M × Nb. An M signals. For the weak satellite vehicles like PRN 3, 6,
point FFT is applied to each column corresponding 26 double block zero padding method is used, which
to each delay and stored in another matrix column suffers degradation due to attenuation. When it is
wise. The mathematical expression (A/2) Rc1 (𝜀𝜏) padded with zeros we can easily observe that there
sinc(�fDtb) is constant for all 𝑙 and can be approximat- will be no degradation and the peak is isolated with-
ed by (A/2) Rc1 (𝜀𝜏) sinc(�fDtb) in the neighborhood of out any attenuation
𝜀𝜏 = 0. Thus, the Fast Fourier Transform (FFT) of the
partial correlator outputs will give the DBZP outputs.
L ( k , m) =
A
d ( k ) Rc1 (∈r ) sinc (π f Dt b ) F ( cos (π f Dt b + 2π f D lt b +
2
)
+ ∈∅0 ( k ,0 ) ) + ηl ( m ) ) (10)

where:

( k ) π f Dt b + ( ( Nb − 1) / Nb ) (π f Dt c − m ) + ∈∅ ( k ,0)
0

– m = 0…M – are the points FFT is taken which cor-
responds to a Doppler frequency bin. Fig. 4. Acquisition Result for PRN 15

60 Articles
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

The Figure 8 shows the graph between acquisition

metric and PRN number which in turn gives the in-
formation between acquired and not acquired signals.

Double block zero padding (DBZP) acquisition Al-

gorithm for 10 ms Integration time
The DBZP method for 10 ms Integration time re-
sults the phase of the code and Doppler frequency as
shown in Table 2.

Table 2. PRNs that are having a threshold 2.5 and

above (DBZP)
Peak
PRN Code phase Doppler
Metric
Fig. 5. Acquisition Result for PRN 22
22 6.25 852 1800

18 5.27 2656 400

21 5.17 1741 -600
15 4.70 4606 2000
26 4.22 3419 -3000
6 3.56 3591 -3700
3 2.83 4342 2000

The acquisition results for weak signals by using

DBZP acquisition algorithm is shown in the Figure 9,
Fig. 6. Acquisition Result for PRN 21 Figure 10 and Figure 11 respectively. The peak gives
the PRNs 22, 18, 21 respectively.

Fig. 7. Acquisition Result for PRN 18

Fig. 9. Acquisition Result for PRN 22 (DBZP)

Fig. 8. Acquisition Results PRN vs Acquisition Metric Fig. 10. Acquisition Result for PRN 18 (DBZP)

Articles 61
 Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

Table 3. Comparison results (DBZP vs DFT method)

using DFT (10 ms) Doppler

Parallel Code Phase Search

(100-% absolute error)

% absolute error
DBZP Doppler

|(X-Y)/X*100|
(10 ms PIT)

% accuracy
PRN

Y
15 1937.72 1900 1.946 98.053

22 1710.08 1700 0.589 99.410

21 -557.24 -600 7.67 92.326

Fig. 11. Acquisition Result for PRN 21 (DBZP) 18 253.17 300 18.49 81.50

Table 4. Comparison of Weak satellite peak metrics

Parallel Code
Phase Search DBZP Increase
PRN using DFT (10ms PIT) in Peak
(10ms) Peak Metric metric
Peak Metric
26 2.48 4.22 1.701
6 2.12 3.56 1.67
3 1.82 2.83 1.55

Table 5. PRNs that are having a threshold 2.5 and

above
PRN Peak Metric Code phase Doppler
15 10.69 4606 1937.72
Fig. 12. Acquisition Result for PRN 15 (DBZP)
22 6.88 852 1710.08
21 4.66 1741 -557.24
Figure 12 gives the acquisition result for PRN=15
2 3.08 3714 81412.12
using double block zero padding acquisition algo-
10 2.95 203 5243.08
rithm.
6 2.84 3592 -3671.38
3 2.81 4342 1919.51
Comparison of Two Acquisition Schemes
26 2.67 3419 -2970.24
Comparison of the above two methods is present-
ed below and calculated absolute error and accuracy Table 6. Comparison results (DBZP vs FFT method)
taking A Parallel search code phase (1 ms integration Parallel
time) + Parallel search in Frequency (10 ms integra- Code
DBZP
tion time) as reference acquisition scheme. Phase
Doppler
%
PRN Search absolute % accuracy
(10 ms
In DBZP scheme using FFT error
PIT)
PRNs 3, 6, 26 which are low sensitive satellite ve- (10 ms)
Doppler
hicles present in the data are detected by DBZP meth-
15 1937.72 1900 1.946 98.053
od of acquisition. DBZP acquisition method enhanc-
22 1710.08 1700 0.589 99.410
es peak metric of these PRNs approximately by 1.6
21 -557.24 -600 7.67 92.326
times. The table gives the comparison of peak metric
2 81412.12 9900 87.83 12.160
with both the methods and corresponding increase in
10 5243.08 7900 50.67 49.32
the peak metric value in decibels.
6 3671.38 -3700 0.77 99.23
The Table 4 gives the comparison between peak
3 1919.51 1900 1.01 98.98
metrics of weak satellites for 10 ms PIT by imple-
26 2970.24 -3000 1.00 99
menting double block zero padding acquisition algo-
rithm.
The PRNs with the threshold value of 2.5 and By using double block zero padding acquisition
above with code phase and Doppler is given in Ta- method, the PRNs 3, 6, 26 being low sensitive sat-
ble 5. ellite vehicles present in the data enhances peak
The absolute error and accuracy for the PRNs 15, metric by 1.6 times when compared with other tra-
22, 21, 2, 10, 6, 3, 26 by taking parallel code phase ditional acquisition methods by using double block
search using FTT as 10 ms Doppler and 10 ms PIT is zero padding acquisition algorithm which is shown
shown in Table 6. in Table 6.

62 Articles
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 12, N° 4 2018

ning system software receiver (gpSrx) imple-

mentation in low cost/power programmable
processors.” In: Proc. of the 14th International
Technical Meeting of the Satellite Division of
The Institute of Navigation (ION GPS 2001), Salt
Lake City, UT, USA, September 2001, 2851–
2858.
[3] R.E. Best, “Phase Locked Loops: Design, Simu-
lation, and Applications,” 4th ed., Mc-Graw Hill,
1999.
[4] J.B. Bullock, M. Foss, G.J. Geier, M. King, “Inte-
gration of GPS with Other Sensors and Network
Assistance.” In: E. Kaplan, Ch. Hegarty (eds.)
Understanding GPS: principles and applications,
Artech House, London, 2005.
[5] V.M. Jovanovic, “Analysis of strategies for serial-
-search spread-spectrum code acquisition-di-
rect approach.” In: IEEE Transactions on Com-
Fig. 13. Acquisition Results PRN vs Acquisition Metric munications, vol. 36, no. 11, 1988, 1208–1220,
DOI: 10.1109/26.8927.
The graph between acquisition results and acqui- [6] K. Krumvieda, P. Madhani, Ch. Cloman, E. Ol-
son, J. Thomas, P. Axelrad, W. Kober, “A Comple-
sition metric of PRNs using double block zero padding
te IF Software GPS Receiver: A Tutorial about
acquisition method by showing acquired signals in the Details.” In: Proc. of the 14th International
green color and non acquired signals in blue color is Technical Meeting of the Satellite Division of The
shown in Figure 13. Institute of Navigation (ION GPS 2001), 2001,
789–829.
[7] M. S. Braasch, A. J. van Dierendonck, “GPS rece-
6. Conclusion iver architectures and measurements.” In: Proc.
DBZP method is used in this work, where both of the IEEE, vol. 87, no. 1, 1999, 48–64,
code phase and Doppler search take place in paral- DOI: 10.1109 /5.736341.
lel and the FFT vector is small compared to Parallel [8] P. Misra, P. Enge, “Global Positioning System: Si-
code phase search. This makes the algorithm execu- gnals, Measurements and Performance,” Ganga-
-Jamuna Press, 2001.
tion very fast compared to 1+1 ms acquisition meth-
[9] B.W. Parkinson, J.J. Spilker, “Global Positioning
od which is mentioned in the literature. The DBZP System: Theory and Applications. Vol. I.,” Wa-
scheme is implemented and tested in Matlab. It de- shington, DC: American Institute of Aeronautics
tects weak signals that are present in the IF data. The and Astronautics, Inc., 1996.
comparison of results both in absolute error and ac- [10] J.B.-Y. Tsui, “Fundamentals of Global Positio-
curacy is presented. The results are shown in section ning System Receivers: A Software Approach,”
5 which shows that it can detect weak signals as com- John Wiley & Sons, Inc., 2000.
pared to the other existing schemes of acquisition and [11] D.M. Lin, J.B.-Y. Tsui, “Comparison of acquisi-
the execution time is also faster due to the partial cor- tion methods for software gps receiver,” Proc.
relations. Since the FFT length is smaller, the correla- of the 13th International Technical Meeting of
tions time is also lesser. So we can conclude that by the Satellite Division of The Institute of Navi-
gation (ION GPS 2000), Salt Lake City, UT, USA,
replacing the traditional acquisition schemes in the
September 2000, 2385–2390.
software receiver with this new method makes the re- [12] D.M. Lin, J.B.-Y. Tsui, D. Howell, “Direct p(y)-co-
ceiver work effectively even for weak signals. Hence it de acquisition algorithm for software gps rece-
may enable the receiver to give the services indoor or ivers,” Proc. of the 12th International Technical
in heavy canopy or in a high vegetation environments. Meeting of the Satellite Division of The Institute
of Navigation (ION GPS 1999), Nashville, TN,
USA, September 1999, 363–368.
AUTHOR [13] N.I. Ziedan, “GNSS receivers for weak signals,”
S.V.S Prasad – Department of Electronics and Com- Artech House, London, 2006.
munication Engineering, MLR Institute of Technology, [14] M. Foucras, O. Julien, Ch. Macabiau, B. Ekambi,
Hyderabad, India, [email protected]. “A  novel computationally efficient Galileo E 1
OS acquisition method for GNSS software rece-
iver.” In: Proc. of the 25th International Techni-
REFERENCES cal Meeting of the Satellite Division of The Insti-
tute of Navigation (ION GNSS 2012), Nashville,
[1] K. Borre, D.M. Akos, N. Bertelsen, P. Rinder, TN, USA, September 2012, 365–383.
S.H. Jensen, “A software-defined GPS and Galileo [15] K. Mollaiyan, R. Santerre, R. Landry, “Acquisi-
receiver. A single frequency approach,” Birkhäu- tion of Weak Signals in Multi-Constellation Fre-
ser Basel, 2007, quency Domain Receivers,” Positioning, vol. 4,
DOI: 10.1007/978-0-8176-4540-3. no. 2, 2013, 144–152,
[2] D. Akos, P-L. Normark, A. Hansson, A. Rosen- DOI: 10.4236/pos.2013.42014.
lind, Ch. Stahlberg, F. Svesson, “Global positio-

Articles 63