Modification on Energy Efficient Design of DVB-T2 Constellation De-mapper

Nourhan M Bahgat et al. Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 5, Issue 6, ( Part -1) June 2015, pp.16-24
www.ijera.com 16 | P a g e
Modification on Energy Efficient Design of DVB-T2 Constellation
De-mapper
Nourhan M Bahgat, Diaa El din S Khalil, Salwa H El-Ramly
Electronics and Communications Engineering Department, Ain Shams University, Cairo, Egypt
Abstract
The second generation of terrestrial digital video broadcasting standard (DVB-T2) offers several advantages for
greater efficiency. Signal Space Diversity (SSD) contains rotated constellation and Q-Delay (RQD), which is
one of advantage that offered to improve the performance over fading channels compared to the non-rotated
modulation. In this journal, the proposed low-power de-mapper design of this work attempts to employ the
introduced SSD to reduce power through replacing LLR calculations by a significantly less complex projection-
based de-mapping whenever possible. It benefits from an algorithm that applies projection-based de-mapping to
significantly reduce LLR computations without deteriorating performance. Two versions are introduced for hard
de-mapping and soft de-mapping. The design uses several techniques simultaneously to be even more energy
efficient without affecting the performance. Prototype results indicate significant reduction of LLR calculations
as Eb/N0 increases with no performance degradation. The idea and energy saving techniques can be easily
applied to any rotated constellation de-mapper.
Keywords - DVB-T2, Energy efficient design, FPGA prototype, Rotated constellation, SSD, Statistical LLR
assignment.
I. INTRODUCTION
SSD is very effective to improve the system
performance in frequency selective terrestrial
channels [1-5]. This concept is implemented through
the use of rotated constellation and the introduction
of Q-delay, where the constellation rotation angle is
optimized by the standard to achieve best
performance [1].
This journal discusses the implementation of a
new DVB-T2 rotated constellation de-mapper which
separated to two main parts; the first part benefits
from a new algorithm [6] that employs projection-
based de-mapping to significantly reduce
computational compared to conventional LLR-based
de-mapping without deteriorating the performance
and the second part benefits from Statistical LLR
assignment [7] that employs standard deviation to
convert the hybrid output that comes from the first
part to soft output to be ready to soft Low Density
Parity Check (LDPC) decoder to get the best
performance. The proposed DVB-T2 rotated
constellation de-mapper output combines between
two kinds of outputs. Hard and, soft outputs; hard due
to using of scheme-C (all bits comparison) and soft
due to normal calculated LLR’s thus, all hard bits
need to convert to soft through proposed LLR
assigning, to be ready for soft LDPC block. Several
implementation techniques are also introduced in the
proposed design to be more energy efficient as [8].
The proposed implementation benefits from a new
algorithm that proposed in [6][7].
The paper is organized as follows: Section II
provides background on the introduced SSD in DVB-
T2. Section III explains the Statistical LLR
assignment. Section IV explains Standard Deviation
Algorithms. Section V highlights prior work in de-
mapper implementation. Section VI presents the
proposed de-mapper implementation. Section VII,
presents the experimental results. And finally, section
VIII summarizes the conclusions.
II. BACKGROUND
Applying Bit Interlever Coding Modulation-SSD
(BICM-SSD) technique with high modulation
diversity and minimum product distance, the
performance over a classical Rayleigh fading channel
becomes equalize to performance over an Additive
White Gaussian Noise (AWGN) channel [1]. The
QAM constellation is rotated by an angle ø based on
channel type, minimum product distance, and
modulation order [1].
Table 1. Optimum angle for different modulation
orders [1]
Modulation QPSK
16-
QAM
64-
QAM
256-QAM
(degrees) 29.0 16.8 8.6 tan-1
(1/16)
RESEARCH ARTICLE OPEN ACCESS

(a) (b)
Figure 1. (a) Classical and (b) rotated constellations
In classical non-rotated 2m QAM, the number of
projections to be computed at the de-mapper on both
I and Q axes is 2×2m/2. For example, in 16-QAM, m
equals 4, then the number of projections are 8, as
shown in Fig. 1(a). In rotated 2m QAM, the number
of projections to be computed at the de-mapper on
both I and Q axes becomes 2m. Due to rotation, each
constellation point has a unique projection on each of
the I and Q exes. For example, in 16-QAM, m equals
4, then the number of projections are 16, as shown in
Fig. 1(b). Such increase in number of projections can
be used by the receiver to have more accurate results.
III. LLR BASED DE-MAPPING
The de-mapping proposed in the DVB-T2
standard [1] computes the Log-Likelihood-Ratios
(LLRs) for all bits based on the received I and Q
components. The exact LLR for a bit bi in the
received cell C is given by:

Where I and Q are the coordinates of the received
cell, Ix and Qx are the coordinates of a certain
constellation point x, σ2 is the noise variance, and
finally Ci0 and Ci1 are the sets of constellation points
having bi equals 0 and 1, respectively. Such complex
calculation is usually simplified by the max-log (ML)
approximation for a simpler form that is generally
more suitable for hardware implementation [9], given
by:

 
It is worth highlighting that even the simplified
LLR (2) is still computationally intensive as it
involved several multiplication operations for each
bit.
IV. PROJECTION BASED DE-
MAPPING
LLR based de-mapping technique does not use
the uniqueness of the I and Q components
(projections) of the rotated constellation points. The
main idea of projection based de-mapping is to
calculate the closest constellation point (X/Y) to the
received cell (I/Q) on each axis separately i.e. the
points (X/Y) having minimum 1D distances from the
received (I/Q) components to the different (Ix/Qx)
projections. The two constellation points
corresponding to the minimum distance on each axis
are used to make a decision based on a selected
scheme. If no decision is possible, the de-mapping
falls back to the conventional LLR technique [6].
There are four schemes proposed in [6] for decision.
The four schemes are: A (bit-by-bit comparison):
Each bit of X is compared with corresponding bit of
Y. If they match, such bit value is final. Else, LLR is
calculated for that bit. B (bit-by-bit comparison
excluding weak points): If either X or Y belongs to
weak points, then all bits are calculated by
conventional LLR. Else, scheme A is applied. A
weak point on an axis is a point having its projection
rather close to its neighbor compared to other points
on that axis. C (all-bits comparison): All bits of X are
compared with corresponding bits of Y. If they
match, all bit values are final. Else, LLR is calculated
for all bits. D (single-axis for weak points): If X (or
Y) belongs to weak points, and then all bits of Y (or
X) are final. Else, scheme A is applied. Such four
schemes actually present different points of trade-off
between reducing computational complexity and
maintaining BER performance. Scheme C (all-bits
comparison) is adopted for the proposed de-mapper
implementation since it demonstrated the best BER
performance that matches the reference performance
of all-bits LLR de-mapping [6]. Fig. 3 presents an
illustrative flowchart for the adopted scheme C.
Figure 2. Sub-regions of rotated QAM-64
constellation [8]

Figure 3. Flowchart for scheme C of [6] (all-bits
comparison)
V. STATISTICAL LLR ASSIGNMENT
High performance communication system use
soft decoding in the receiver. In such case, hard bits
from projection based de-mapping need to be
converted to LLRs before passing to subsequent
blocks as illustrated in Fig 4 for DVB-T2 system. A
simple assignment of maximum and minimum LLR
values to 1 and 0 respectively does not give optimum
BER performance and an elaborate statistical
calculation is required to achieve best BER
performance [7]. Since, the proposed de-mapper
benefits from projection based de-mapping to save
calculations and energy, the output of such de-
mapper will be hybrid combination between hard bits
and soft LLR values. If the de-mapper is followed by
hard LDPC decoder a simple comparator can be used
to make the hard decision for LLR values. However,
if the de-mapper is followed by soft LDPC decoder,
the statistical calculations must extract the LLR
values of the output to perform the calculation and
also extract the hard bits to replace [7] Fig 5
illustrates the process.
The hybrid output of the de-mapper contains; the
soft outputs (LLR’s) represented by the symbols (S1,
S2, S4) and the hard outputs (bits) represented by the
symbols (H3, H5). The separator is used to split the
input stream into two sub-streams. The soft outputs
sub-stream has only the S symbols and the hard
outputs sub-stream has only the H symbols. The
output soft data pass to statistics calculation block to
calculate mean and standard deviation per FEC frame
[7]. These values are used to perform LLR
assignment of the separated hard bits in the next FEC
frame. The merger block combines between the
assigned values representing the hard bits into their
position to form a stream of all soft outputs.
Figure 4. Statistical LLR assignment
Figure 5. Statistical LLR assignment block diagram
[7]
In [7], two techniques are proposed for standard
deviation calculations. One is the double sided
standard deviation (σds), where standard deviation is
calculated for all LLR values. The other is the single
sided standard deviation, where two standard
deviation are calculated for the positive and negative
LLRs (σps,σns). No difference in performance exists
between the two techniques [7]. In our proposed
implementation the single sided technique is adopted
to allow combined calculations to save energy.
The LLR assigned values (LLRest
1
, LLRest
0
)
calculated for the 1’s and 0’s independently. Each
one depends on its corresponding mean and standard
deviation as given in the following equations [7]:
(3)
(4)
(5)

where σps is the positive LLRs distribution standard
deviation, σns is the negative LLRs distribution
standard deviation, μps is the mean of the positive
side (the ones) of the LLR histogram distribution. μns
is the mean of the negative side (the zeros) of the
LLR histogram distribution and c is the Empirical
scaling factor. The empirical equation obtained by
analyzing performance for different SNR values. The
optimum c equation is divided into two parts
depending on the SNR range. For SNR values from 0
to 5 dB, the optimum c is constant while for SNR >
5dB, the optimum c is dependent on SNR as given in
(3) [7].
VI. PRIOR WORK IN IMPLEMENTATION
The most relevant work has been recently
published by [8]. It proposes a new de-mapper design
that reduces the computational effort of LLR
calculations by reducing the set of distance
calculations based on the signs of the received I and
Q. For 64-QAM and 256-QAM, the constellation
space is divided into four sub-regions as shown in
Fig. 4. Based on the signs of the received I and Q,
one sub-region is selected for LLR calculations rather
than the whole constellation space. Hence, the
number of distances in LLR calculations is reduced.
In addition, it proposes a simple approximate formula
for Euclidean distance calculation that gets rid of
multiplication operations and further reduces the
computational effort of LLR calculations, given by:
(6)
VII. PROPOSED IMPLEMENTATION
In this work, a new energy efficient design of the
DVB-T2 constellation de-mapper is presented and
prototyped. It leverages the benefit of significantly
lower computational effort of projection-based de-
mapping to avoid the rather large computation effort
of LLR calculation whenever possible, allowing all
LLR calculation blocks to be clock-gated. Also, it
reduces the computational effort of LLR calculation
by using the simplified Euclidean distance of [8] and
by a proposed efficient minimum search technique
that combines distance calculation with minimum
search to get rid of all intermediate distance storage
and by a pipelined use of multipliers. The design
supports all modulation types in DVB-T2 standard,
namely QPSK, 16-QAM, 64-QAM, and 256-QAM
through a 2-bit selector. Also, it is reconfigurable
where projection-based de-mapping can be activated
or not through an enable. In case projection-based de-
mapping is deactivated, the related blocks are clock-
gated. Received I and Q are 9 bits each for high
enough resolution. Each LLR value is 11 bits to
accommodate all values without truncation.
Fig. 6 presents a block diagram that illustrates
the proposed de-mapper design. The de-mapper gets I
and Q bits from the cell de-interleaver. The (control
block) passes them to the two (minimum projection
distance calculators) and clock-gate all LLR related
blocks. Once the X-bits and Y-bits, corresponding to
the nearest constellation point projections on the I
and Q axes respectively, are calculated, a
(comparator) is used to check for matching. If they
match, the bits are final and no need for LLR
calculations, so the LLR related blocks remain clock-
gated. If they are different, a flag is asserted to
(control block) to enable the LLR related block and
pass the stored I and Q bits to them. The two
(minimum Euclidean distance calculators) compute
the minimum distances for the Ci0 and Ci1 sets of
constellation points using the simplified formula of
(3). Then the (LLR calculator) computes the LLR
values for each bit according to (2). And finally, a
(hard decision) circuit converts LLR values to bits.
After the LLR calculations are done the (control
block) clock-gates the related blocks again. A
pipelined minimum search technique is implemented
in all the four minimum distance blocks. It combines
the distance calculation steps with on-the-fly
minimum search algorithm that updates the minimum
value once a new distance is ready. Such design
eliminates the need for distance storage, hence
reduces hardware significantly. Also, it achieves the
best latency reaching the minimum distances as it just
takes one extra clock cycle after last distance
calculation.
Fig 7 presents a block diagram that illustrates the
proposed statistical LLR assignment design. The
design separates the hybrid input stream to three sub-
streams one for soft (LLRs), other sub-streams for
hard bits, one of them for zero hard bit and another
for one hard bits. The soft (LLRs) passes to LLR
separator block to separate it to two passes, one for
LLR which represent one (LLR > 0) and another for
LLR which represent zero (LLR ≤ 0) to calculate
mean and standard deviation for positive and
negative sides to calculate LLRest
1
and LLRest
0
with
(1), (2). Empirical scaling factor block, out its value
depends on SNR value. Finally, the mux block
combines LLRs from three paths (original soft,
LLRest
1
and LLRest
0
) and merges them correctly in
order to be ready to soft LDPC block.

Figure 6. Block diagram of the proposed de-mapper
design
Figure 7. Block diagram of the proposed statistical
LLR assignment
The Empirical factor (c) shown in (5) is complex
to implement because it contains real numbers. To
avoid any use of real numbers and
multiplications/divisions, (5) is converted to (7)
through approximations and introducing a factor of
16. This factor preserves the accuracy of having
decimal fractions of original equation and is reversed
after multiplication by the standard deviation in LLR
calculations simply using shift operation.
(7)
While [7] calculates the mean and standard
deviation for each FEC frame (16200 bits), this work
proposes to do the calculations on small sub-frames
each is 1024 bits that is a power of 2 for binary
numbers. The mean divided by shift operation. The
sub-frame is more accurate than FEC frame (16200
bits) because the mean on few values give more
accuracy. The basic operations like addition and
subtraction are easy to implement in an FPGA
because synthesis tools have optimized
addition/subtraction units based on FPGA
architecture. Multiplication, division and square root
are complex operations; square root in particular, is
computationally intensive as it involves convergence
and approximation techniques. Many algorithms have
been developed to implement it on FPGAs. But there
is a need of an algorithm which should be more
efficient in terms of time, speed and area. Thus, to
calculate the square root of standard deviation in this
energy efficient design an approximation is needed to
reduce it to simple formulas that only uses addition,
subtraction, and shift. After plotting the square root
function, it could be simplified as follows: The
function is split into three ranges as represented in
equation (8). The first range from 0 to 2048, is
difficult to approximate. However to avoid complex
calculations, it is saved as 12-bit LUTs. The second
range from 2048 to 8192 and the third range from
8192 to 16383 use linear approximation and hence
reduce hardware and guarantee the speed.
(8)
The proposed de-mapper is implemented as
VHDL then synthesized and prototyped on VertixII-
pro FPGA XC2VP30 (same as [8] for comparison).
Table 2 presents the FPGA hardware resources
utilization in the proposed design as well as the
reference design [8] for comparison. Hardware
overhead of projection based de-mapping is only
12% compared to LLR blocks. Hence, energy
overhead is roughly estimated as 12% when
projection-based de-mapping is inconclusive and
LLR de-mapping has to be used. On the other hand,
energy saving is roughly estimated as 88% when
projection-based de-mapping is conclusive and LLR
de-mapping is not used. The actual energy saving is
actually less considering the leakage power
dissipated whether clock-gating is applied or not.
Compared to [8], the hard de-mapper proposed
design, with the projection-based de-mapping
overhead and with Euclidean distance calculation for
all constellation points, uses 40% less hardware. This
is mainly attributed to the efficient minimum search
technique and the pipelined use of multipliers. Such
hardware reduction directly translates to energy
saving since the same clock speed is used in both
designs and both meet the latency constraint imposed
by cell duration from standard.

In this work also, a design of statistical LLR
assignment is proposed to utilize the concept of
single-sided standard deviation algorithm not double-
sided standard deviation algorithm to combine
calculation of standard deviation (σ). The proposed
Statistical LLR assignment is implemented as VHDL
then synthesized and prototyped on VertixII-pro
FPGA XC2VP30. Table 3 presents the FPGA
hardware resources utilization in the proposed design.
Table 2. The FPGA hardware resources without
statistical LLR assignment
Table 3 The FPGA hardware resources with
statistical LLR assignment
It is worth highlighting that no memory blocks
are used in soft de-mapper proposed design since all
standard deviation probabilities values are stored in
registers to improve speed and simplify design and
further from the root as possible. Adding the
proposed Statistical LLR assignment to propose
design of DVB-T2 rotated constellation hard de-
mapper is implemented as VHDL then tested by
Modelsim simulation then synthesized and
prototyped on VertixII-pro FPGA XC2VP30. Table 3
presents the FPGA hardware resources utilization in
the proposed design. It is worth highlighting that
proposed full design (soft de-mapper) or (the
proposed statistical LLR assignment with proposed
DVB-T2 rotated constellation de-mapper) still less
than [8]. Thus, it is still more efficient energy. The
full system (the proposed Statistical LLR assignment
with proposed design of DVB-T2 rotated
constellation de-mapper) saves 22% more than [8] in
hardware resources. The proposed Statistical LLR
assignment with proposed design of DVB-T2 rotated
constellation de-mapper (soft de-mapper) saves
almost 36 % when projection-based de-mapping is
conclusive and LLR de-mapping is not used.
Figure 8. The experimental measurement setup
VIII. EXPERIMENTAL RESULTS
An experimental measurement setup is used to
simulate the hardware performance of the proposed
design. While the DVB-T2 system obviously has
complex bit-chain processing steps, the steps
performed at transmitter are sequentially reversed at
the receiver. Hence, it is acceptable to only consider
the processing steps related to the block under
inspection. As a result, in the measurement setup, a
mapper is used with random bits, followed by
channel model and additive white Gaussian noise
(AWGN), then quantization representing the receiver
ADCs, before entering the de-mapper. Fig. 8
illustrates the experimental measurement setup.
Matlab is used to generate I and Q bits going into the
de-mapper hardware and perform BER calculations.
A random data bits stream is generated then passed
through an ideal mapper to generate I and Q values of
corresponding cells to be transmitted. The cells pass
through a classical Rayleigh channel with no erasure.
Then, AWGN is added with controlled SNR level.
The resultant I and Q values are then quantized into 9
bits each (same as hardware design). The data are
then passed to Modelsim to simulate the de-mapper
hardware. Finally, the de-mapped bits are passed
back to Matlab for BER calculation.
Experimental results for 64-QAM are presented
in this section for illustration. Fig. 9 plots the BER
vs. Eb/N0(dB) for the proposed prototype simulation
with projection-based de-mapping activated and
disabled, along with ideal Matlab simulation of LLR
based de-mapping. In addition, it presents the
prototype simulation and the ideal Matlab simulation
of [8] for comparison.
Hardware
Resources
Hard De-mapper
Proposed Design
Design
of [8]
Scheme-C
and LLR
blocks
LLR
blocks
LUTs 2,928 2,559 4,667
Flip-Flops 1,741 1,328 791
DSP blocks
(multipliers)
8 8 20
Memory
blocks
0 0 0
Hardware
Resources
Soft De-mapper
Proposed Design
Design
of [8]Full
Statistical
LLR
assignment
block
LUTs 3,280 658 4,667
Flip-Flops 1,814 281 791
DSP blocks
(multipliers)
13 5 20
Memory blocks 0 0 0

Figure 9. BER performance comparison for 64-QAM
It is worth highlighting that the simulations of
[8] are not repeated. The curves data are exactly
copied from the relevant figure in [8]. No impact on
BER performance is observed when using projection
based de-mapping vs. LLR based de-mapping for all
cells. This agrees with reference [6]. This can be
attributed to the conservative projection-based de-
mapping scheme used. Projection-based de-mapping
is only used when noise is rather low that both
received I and Q projections are closest to correct
constellation point. So, the probability of incorrect
decision is very low. Both curves of the proposed
prototype simulation are close to LLR ideal Matlab
simulation with observed increase in BER resulting
from quantization, which is not represented in such
ideal Matlab simulation. BER performance of the
proposed prototype is better compared to that of [8].
This can be attributed to the reduced subset of
constellation points used in [8] to reduce LLR
calculations as explained in section VI.
Figure 10. LLR usage vs. Eb/N0 values
The proposed Statistical LLR assignment is
implemented as VHDL then tested by Modelsim
simulator. Fig. 11 illustrates simulation results of the
statistical LLR assignment, while the design is
verified to work correctly in simulation. In simulation
the reset is used for initial state, when (reset=”1”), the
system reset; but when (reset=”0”), the system waits
the rising edge of clk to work. The snr (signal-to-
noise ratio); is the system input which is function in
empirical factor (c). llr signal is the system input (soft
input which it comes from separator block),
Count_one and count_zero signals count positive and
negative input LLRs. The sum_llrs_one, sum_llrs_
zero, sum_squred_llr_one, sum_ squred_ llr_ zero
signals are intermediate signals that used to calculate
mean and standard deviation for the 0 and 1
independently. The est_llr_one and est_llr_zero
signals; are the system output. Thus, the design
implement LLRest
1
and LLRest
0
which shown in (3),
(4). It is also important to evaluate the computational
effort reduction and corresponding energy saving, as
well as, to highlight their dependence on Eb/N0.
With the conservative projection-based de-mapping
scheme used, LLR calculations are skipped only
when noise affecting the cell is low enough that both
projections lead to same constellation point. The
probability of such case certainly increases as Eb/N0
increases.
Fig. 10 presents the normalized LLR usage vs.
Eb/N0(dB). When Eb/N0 is small, the received data
is rather noisy and projection based de-mapping will
only succeed in very few cells while LLR
calculations will be needed for almost all cells. As
Eb/N0 increases, the data becomes less noisy and
projection based de-mapping will succeed with more
cells, saving more LLR calculations. Such saving
continues to improve significantly as Eb/N0
continues to increase as indicated in Fig. 8 thus it
normalized the usage of LLR; hence benefits from
projection based de-mapping and clock-gate all LLR
calculation blocks. For example, at Eb/N0 of 27dB,
the proposed design saves almost half of the LLR
calculations. This translates to roughly 44% of
energy saving. (As estimated in section VII, 88%
energy saving is achieved when LLR calculation is
skipped).

Figure 11. The proposed Statistical LLR assignment
IX. HARDWARE IMPLEMENTATION
FPGA’s are equipped with a lot of resources that
allow them to hold large digital systems on a single
chip. FPGA vendors provide tools that allow the
designer to build embedded systems on FPGAs. One
of the new concepts that discuss in this experiment is
the System-on-Chip approach (SoC). Xilinx provides
a tool for building an embedded SoC on its FPGAs,
and we are going to use this tool in this experiment.
This tool is called Xilinx Embedded Development
Kit (EDK). The EDK allows the designer to build the
processor system based on an embedded processor in
Xilinx called MicroBlaze. The tool provides a C/C++
compiler for the processor and VHDL/Verilog for
coprocessor.
In this journal, we changed C program to adopt
proposed de-mapper. Make the serial port transmit
two input numbers and receive one output then test it,
I/O display on “Hyper Terminal” or “Matlab”. While
the C program is verified to work correctly in
simulation, start to change VHDL code to proposed
de-mapper. The VHDL code in MicroBlaze writes as
finite state machine. The problem in the VHDL code
template is that the result is immediately written after
the last input is read. However, proposed de-mapper
depends on clock cycle. To overcome this problem,
add a new state to the state machine, and modify the
state transitions to introduce this new state between
the "Read_Inputs" and "Write_Outputs" states. The
new state in this project called "processing" state that
depends on FSL (Fast Simplex Link) bus clock
(FSL_CLK). The FSL bus provides a point to point
communication between any two peripherals modules
on the FPGA. To write output, the flag up to transit to
"Write_Outputs" state as illustrate in Fig.12.
Figure 12 Proposed de-mapper finite state machine
In the beginning, Generating bit stream for proposed
de-mapper hardware implementation. The hardware
synthesized on Xilinx VertixII-pro FPGA XC2VP30.
Table 4 presents the FPGA hardware resources
utilization in the hardware proposed design.

Table 4 The FPGA hardware resources for EDK tool
X. CONCLUSION
Two versions are introduced for DVB-T2 rotated
constellation de-mapper, hard de-mapping and soft
de-mapping. Proposed design of hard de-mapper is
presented and prototyped. The design benefits from
projection based de-mapping to significantly reduce
power by saving LLR calculations. It also used
several techniques to be even more energy efficient.
Experimental results clearly indicate there is no
performance degradation when projection based de-
mapping is used. Experimental results also indicate
significant reduction of LLR calculations and
corresponding energy saving as Eb/N0 increases.
Proposed design of soft de-mapper is presented and
prototyped. [7] presents two algorithms to transform
from the hybrid output stream from proposed hard
de-mapper to all-soft stream in order to guarantee the
best performance of the soft LDPC decoder. These
two algorithms were tested and gave almost the same
performance at low SNR values and a slightly better
performance at a moderate SNR range so, the
proposed design of statistical LLR assignment utilize
the concept of single-sided standard deviation,
prototyped by VHDL and simulated by Modelsim.
The proposed full system (soft de-mapper) saves
almost 36 % from energy when LLR calculation is
skipped.
REFERENCES
[1] Digital Video Broadcasting (DVB); Frame
structure channel coding and modulation for
a second generation digital terrestrial
television broadcasting system (DVBT2),
ETSI Standard EN 302-755, v1.2.1, Feb.
2011.
[2] J. Boutros, E. Viterbo, “Signal space
diversity: A power and bandwidth-efficient
diversity technique for the Rayleigh fading
channel,” Transactions on Information
Theory, vol. 44, no. 4, pp. 1453-1467, Jul.
1998.
[3] X. Giraud, E. Boutillon, and J. C. Belfiore,
“Algebraic tools to build modulation
schemes for fading channels,” Transactions
on Communications, vol. 43, no. 3, pp.938–
952, May 1997.
[4] C. Abdel Nour and C. Douillard, “Rotated
QAM Constellations to Improve BICM
Performance for DVBT2,” International
Symposium on Spread Spectrum Techniques
and Application, pp. 354-359, Aug. 2008.
[5] C. Abdel Nour and C. Douillard, “Improving
BICM performance of QAM constellations
for broadcasting applications,” International
Symposium on Turbo Codes and Related
Topics, pp. 55-60, Sep. 2008.
[6] M.Z. Saleh, M. Sadek, S.H. Elramly, “Novel
hybrid soft hard demodulation schemes for
RQD constellation,” International
Conference on Telecommunications and
Signal Processing, pp. 684-688, Jul. 2012.
[7] M.Z. Saleh, M. Sadek, S.H. Elramly,“A
Modified Hybrid RQD Demapper based on
Novel Hard to Soft Assigning Algorithm, "In
Application of Information and
Communication Technologies (AICT), 7th
International Conference on (pp. 1-6). IEEE,
October 2013.
[8] M. Li, C. Abdel Nour, C. Jego, C. Douillard,
“Design of rotated QAM mapper/demapper
for the DVB-T2 standard,” Workshop on
Signal Processing Systems, pp. 18-23, Oct.
2009.
[9] Digital Video Broadcasting (DVB);
Implementation guidelines for a second
generation digital terrestrial television
broadcasting system (DVB-T2), DVB
Document A133, Feb. 2012.
[10] M. Li, C. Abdel Nour, C. Jego, J. Yang, C.
Douillard, “Design and FPGA Prototyping
of a bit-interleaved Coded Modulation
Receiver for the DVB-T2 Standard”,
Workshop on Signal Processing Systems,
Nov. 2010.
[11] M. Li, C. Abdel Nour, C. Jego, J. Yang, C.
Douillard, “FPGA implementation of a
shuffled iterative Bit-Interleaved Coded
Modulation receiver,” Workshop on Signal
Processing Systems, Oct. 2011.
Hardware Resources
64-QAM with serial
port
LUTs 2,303
Flip-Flops 1,849
DSP blocks
(multipliers)
3
Memory blocks 4

Modification on Energy Efficient Design of DVB-T2 Constellation De-mapper

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to Modification on Energy Efficient Design of DVB-T2 Constellation De-mapper (20)

Recently uploaded (20)

Modification on Energy Efficient Design of DVB-T2 Constellation De-mapper