# On the adaptive DVB-S2 physical layer: design and performance

**ABSTRACT** The successful DVB standard has now evolved into the DVB-S2 standard, which promises to bring very significant capacity gains. The main DVB-S2 feature is its adaptive air interface, where coding and modulation techniques are varied flexibly to maximize performance and coverage. This article addresses the design of the entire DVB-S2 communication chain, considering practical algorithms for coding, modulation, predistortion, carrier and SNR estimation, frame synchronization, and data recovery. The design complexity is exacerbated by the fact that DVB-S2 foresees 28 different coding/modulation pairs, demanding specific optimization and variable frame length. The performance achieved considering all possible impairments is compared to the ideal performance achievable in the Gaussian channel in terms of integral degradation, which ranges from 0.4 to 2,5 dB in going from QPSK to 32-APSK.

**1**Bookmark

**·**

**349**Views

- [Show abstract] [Hide abstract]

**ABSTRACT:**In this paper, the SNR estimation error of Decision Directed SNR estimation method in AWGN is investigated, which uses samples received in reference decision region. In communication system receiver, when SNR estimation scheme using error vectors between ideal sample points and received sample points of reference region is adopted, the samples contain incorrectly received samples due to AWGN. Consequently, the mean of estimated reference constellation point is shifted and Decision Directed SNR estimation is inaccurately performed. These effects are explained by modified probability density function and difference between actual SNR and estimated SNR is theoretically derived and quantatively analyzed. It is proved that SNR estimation error obtained through computer simulation is matched up with derived one, and SNR estimation performance is enhanced significantly by adopting suggested correction scheme.The Journal of Korea Navigation Institute. 01/2012; 16(6). - [Show abstract] [Hide abstract]

**ABSTRACT:**The current DVB-S2 standard specifies the use of constant energy level pilots for receiver synchronization and equalization algorithms. However, these are unsuitable for APSK modulations due to the nonlinear response of the satellite power amplifier. In this paper, we investigate the performance of two low-complexity techniques for nonlinear compensation in DVB-S2 systems, i.e. static predistortion and cluster based sequence equalizer (CBSE). We also propose how multilevel pilot structures, matched to each technique, can be used for carrier recovery. Simulation results are presented in terms of total degradation wherein both techniques are shown to achieve a significant improvement over the conventional system.Advanced Satellite Multimedia Systems Conference (ASMS) and 12th Signal Processing for Space Communications Workshop (SPSC), 2012 6th; 01/2012 - SourceAvailable from: ocean.kisti.re.kr[Show abstract] [Hide abstract]

**ABSTRACT:**In the channel-varying environment, it is very important to estimate the signal to noise ratio(SNR) of received signal and to transmit the signal effectively for the modern communication system. The performance of existing non-data-aided (NDA) SNR estimation methods are substantially degraded for high level modulation scheme such as M-ary APSK or QAM. In this paper, we propose a SNR estimation method which uses zero point auto-correlation of received signal per block and auto-/cross- correlation of decision feedback signal in OFDM system. Proposed method can be studied into two Types; Type 1 can estimate SNR by zero point auto-correlation of decision feedback signal based on the second moment property. Type 2 uses both zero point auto-correlation and cross-correlation based on the fourth moment property. In block-by-block reception of OFDM system, these two SNR estimation methods can be possible for the practical implementation due to correlation based the estimation method and they show more stable estimation performance than the previous SNR estimation methods. Also, we mathematically derive the SNR estimation expression according to computational difference of auto-/cross-correlation. Finally, Monte Carlo simulations are used to verify the proposed method.The Journal of Korean Institute of Electromagnetic Engineering and Science. 01/2010; 21(9).

Page 1

IEEE Wireless Communications • December 2005

62

1536-1284/05/$20.00 © 2005 IEEE

SNR

estimation

Freq/phase

tracking

DeMUX

Pilots

Data

^θ0

Buffer

Buffer

Frame

synch

The successful DVB

standard has now

evolved into the

DVB-S2 standard,

which promises to

bring very significant

capacity gains. The

main DVB-S2 feature

is its adaptive air

interface, where

coding and

modulation

techniques are varied

flexibly to maximize

performance and

coverage.

ACCEPTED FROM OPEN CALL

INTRODUCTION AND MOTIVATION

The digital video broadcasting-satellite (DVB-S)

standard is now the most widely adopted trans-

mission protocol for broadcasting by satellite.

This fact has prompted the development of the

evolved standard identified as DVB-S2 [1],

which promises to yield very significant system

capacity gains, as well as the extension of the set

of services and applications that can be delivered

to an enlarged mass market. In particular,

enhanced broadcasting will be paired to interac-

tive communications for broadband access to the

Internet and multimedia content distribution. In

essence, DVB-S2 can be considered a valid alter-

native for solving the last mile dilemma and

bridging the digital divide.

The necessary bandwidth resources for large

throughputs can only be found at extremely high

frequencies, such as Ka-band and above. Here,

atmospheric effects on system capacity can

become truly dramatic, and can only be maxi-

mized by abandoning the conservative worst case

design approach in favor of flexible and adaptive

transmission mode selection to efficiently exploit

the available power and spectrum resources.

This is the underlying rationale of the adaptive

coding and modulation (ACM) air interface

adopted within the DVB-S2 standard. The cost

of adaptivity is the increased complexity repre-

sented by the large number of modulation/cod-

ing sets and the ever more critical parameter

estimation and synchronization procedures.

The design of DVB-S2-compatible data recov-

ery and synchronization procedures is an interest-

ing open problem, calling for smart and effective

solutions accounting for actual impairments, such

as phase noise and nonlinearities, and operating

at extremely low signal-to-noise ratios (SNRs),

close to the Shannon limit. The design complexity

is exacerbated by the fact that DVB-S2 foresees

28 different coding/modulation pairs, demanding

specific optimization and variable frame length,

which calls for optimized design of the frame syn-

chronization subsystem. The latter can benefit

from the use of advanced post-detection integra-

tion techniques described in the following. This

work aims at proposing algorithms that embrace

the overall degrees of freedom provided by adap-

tivity, while keeping complexity at a minimum.

These issues have been separately addressed

for data recovery in [2] and frame synchronization

in [3]. Differently, this article considers the entire

transmission chain to assess overall performance,

addressing at the same time the data decoding

and synchronization issues. Performance is thus

comprehensive of all nonidealities affecting the

actual system, contrasted with the ideal additive

white Gaussian noise (AWGN) case, to evaluate

the so-called integral degradation.

This article introduces the reader to the

essential elements of the DVB-S2 adaptive phys-

ical layer and guides him/her through a descrip-

tion of a possible algorithmic implementation.

More in detail, the transmit-receive chain com-

prises high order modulation schemes, resilient

to nonlinear effects; low density parity check

(LDPC) codes; fractional predistortion tech-

niques, operating after the shaping filter to

GIANNI ALBERTAZZI, STEFANO CIONI, GIOVANNI E. CORAZZA, MASSIMO NERI,

RAFFAELLA PEDONE, PAOLA SALMI, ALESSANDRO VANELLI-CORALLI, AND MARCO VILLANTI,

ARCES, UNIVERSITY OF BOLOGNA

ABSTRACT

The successful DVB standard has now

evolved into the DVB-S2 standard, which

promises to bring very significant capacity gains.

The main DVB-S2 feature is its adaptive air

interface, where coding and modulation tech-

niques are varied flexibly to maximize perfor-

mance and coverage. This article addresses the

design of the entire DVB-S2 communication

chain, considering practical algorithms for cod-

ing, modulation, predistortion, carrier and SNR

estimation, frame synchronization, and data

recovery. The design complexity is exacerbated

by the fact that DVB-S2 foresees 28 different

coding/modulation pairs, demanding specific

optimization and variable frame length. The per-

formance achieved considering all possible

impairments is compared to the ideal perfor-

mance achievable in the Gaussian channel in

terms of integral degradation, which ranges from

0.4 to 2.5 dB in going from QPSK to 32-APSK.

ON THE ADAPTIVE DVB-S2 PHYSICAL LAYER:

DESIGN AND PERFORMANCE

Work supported by the European Space Agency (ESA)

project no. 16532/02/NL/EC.

Page 2

IEEE Wireless Communications • December 2005

63

counteract nonlinear effects and mitigate the

intersymbol interference (ISI); nonlinear and lin-

ear distortions, introduced by the onboard high

power amplifier (HPA) and input/output multi-

plexing filters (I/O muxes); frame synchroniza-

tion and symbol timing estimation; frequency

offset and phase recovery followed by data detec-

tion. Interestingly, the actual integral degrada-

tion of this transmission chain affected by all

possible impairments can be limited by judicious

design to rather small values, ranging from 0.4 to

2.5 dB going from quadrature phase shift keying

(QPSK) to 32-amplitude and PSK (APSK), as

shown in this article. To achieve these satisfacto-

ry results, predistortion techniques are essential

for 16- and 32-APSK modulations. In the follow-

ing, channel coding and modulation schemes as

well as predistortion techniques are described.

We illustrate the proposed DVB-S2 digital

receiver, the performance of which is also dis-

cussed. Finally, overall conclusions are drawn.

CHANNEL CODING AND

MODULATION SCHEMES

The most significant feature introduced in the

DVB-S2 standard [1] is the capability to select the

coding-modulation pair to optimize the system

capacity in terms of achievable system through-

put, under the hypothesis of a finite user popula-

tion [4]. The coding and modulation schemes

selected for DVB-S2 leverage the most recent

and powerful advancements in the field of com-

munication theory. In particular, powerful coding

schemes that essentially achieve channel capacity

are now available with simple implementation

and affordable complexity. Specifically, the DVB-

S2 forward error correction (FEC) encoding

encompasses three components: an outer system-

atic Bose-Chaudhuri-Hocquenghem (BCH) code,

an inner LDPC code, and a block bit interleaver.

The BCH [5] block codes are introduced to

remove the possible error floor produced by LDPC

undetected errors. However, the major error cor-

recting capability definitely comes from the LPDC

element. The LDPC code family was first pro-

posed by Gallager in 1962 [6], and then rediscov-

ered during the 1990s (e.g., [7]). LDPC codes are

(k, n) linear block codes with a sparse parity check

matrix and very large n (e.g., n in the thousands).

LDPC codes can be represented by a bipartite

Tanner graph [7], in which the ith check node is

connected to the jth bit node if and only if the

matrix element at position (i, j) is non-null. The

sparsity of the graph is key to efficient algorithmic

encoding and iterative decoding of these long

codes. LDPC codes are defined as regular if the

number of ones in every row, the so-called check

node degree, and the number of ones in every col-

umn, identified as the bit node degree, are constant;

they are identified as irregular otherwise. The

DVB-S2 standard codes are irregular LDPC codes

with two block length options, n = 64,800 and n =

16,200. The specified parity check matrices have a

well defined structure in order to enable low-com-

plexity systematic encoding and parallel decoding.

In order to cover an exhaustive range of spec-

tral efficiencies, 11 different code rates (1/4, 1/3,

2/5, 1/2, 3/5, 2/3, 3/4, 4/5, 5/6, 8/9, 9/10) are speci-

fied in conjunction with four different modulation

schemes for a total of 28 allowed combinations, as

reported in [1, Table 12]. Besides classic QPSK

and 8-PSK modulation schemes, a hybrid APSK

modulation format is considered in order to have

high bandwidth efficiency and cope with nonlinear

satellite transponders. In essence, APSK is formed

by concentric circles of constellation points. In par-

ticular, 16- and 32-APSK modulations are specified

with two and three rings of points, respectively.

The characteristic parameters of APSK are opti-

mized as a function of the LDPC code rate [1].

Finally, the bit interleaving for 8-PSK, 16-

APSK, and 32-APSK is performed through a

block interleaver that acts after the LDPC

encoder in order to increase code diversity.

NONLINEAR SATELLITE CHANNEL AND

PREDISTORTION TECHNIQUES

A main source of degradation is given by the

presence of saturation-driven HPAs, which are

normally employed for satellite links to maxi-

mize output power and DC/RF conversion effi-

ciency. The HPA is typically described by

nonlinear AM/AM and AM/PM characteristics,

and increases adjacent channel interference

(ACI) due to spectrum sidelobes regrowth, as

well as warping and clustering of the signal con-

stellation. The complexity is enhanced by the

adoption of variable high order modulations,

while frame-by-frame adaptivity results in non-

deterministic waveforms, which increase the sys-

tem distortion sensitivity.

Several techniques have been proposed in the

literature in order to mitigate the impact of satu-

ration-driven HPAs. A straightforward solution is

to introduce input backoff (IBO) to step back

from saturation, with a consequent output backoff

(OBO). For ground terminal amplifiers this may

be acceptable (up to a reasonable level) because

DC/RF conversion efficiency is not at a prime.

For onboard amplifiers, the larger the OBO, the

smaller the coverage and conversion efficiency;

therefore, this approach must be used very spar-

ingly, and other techniques are necessary.

Compensation techniques can be applied at

both transmitter and receiver. Typical receiver

techniques involve the use of equalizers, with the

related drawbacks of increased receiver com-

plexity and impossibility of eliminating the unde-

sirable spectrum spillover. It is therefore

preferable to use transmitter countermeasures to

precompensate for the HPA input and achieve

quasi-linear amplification of the signal. This

approach is commonly referred to as predistor-

tion [8], and can be classified into two main

classes, analog and digital. Analog predistortion

is performed at RF (or possibly at intermediate

frequency [IF]), while digital predistortion is car-

ried out at baseband, either before or after the

pulse shaping filter, identified as data and frac-

tional approaches, respectively. The main advan-

tages of digital predistortion are flexibility,

adaptivity, and ease of implementation.

The proposed predistortion method for DVB-

S2 involves the use of a digital predistorter locat-

ed after the pulse shaping filter, operating on an

oversampled signal. The signal is processed by

The HPA is typically

described by AM/AM

and AM/PM

characteristics, and

induces increased

adjacent channel

interference (ACI)

due to spectrum

sidelobes regrowth,

as well as

warping and

clustering of the

signal constellation.

Page 3

IEEE Wireless Communications • December 2005

64

means of a lookup table (LUT), which stores the

inverted HPA coefficients computed offline

through analytic inversion of a proper HPA

model. The steps needed to obtain LUT coeffi-

cients are:

• HPA model selection

• Parameter extrapolation

• Analytic model inversion

• LUT definition and construction

Regarding the HPA model, a simple yet

robust empirical model is the classic Saleh model

[9]. Here, a more sophisticated and accurate

model is adopted, which involves the use of poly-

nomial ratios with 10 parameters (five for the

amplitude and five for the phase). Given the

measured HPA characteristics, the model param-

eters can be extracted by minimizing the energy

of the difference between the modeled and

experimental HPA curves (minimum mean

square error [MMSE] criterion). These parame-

ters are then applied to the analytically inverted

characteristics to obtain the analytical predistor-

tion transfer function. The last step is the dis-

cretization of the analytical curve in order to

store it into the LUT. The adopted strategy is

linear in power indexing (i.e., table entries are

uniformly spaced along the input signal power

range), yielding denser table entries for larger

amplitudes where nonlinear effects reside.

DIGITAL RECEIVER IMPLEMENTATION

Having described the main baseband transmitter

blocks, attention is now focused on the receiver.

In Fig. 1 a possible DVB-S2 digital receiver archi-

tecture is depicted. Several subsystems are

involved in the recovery of information. Following

a sequential functional order from startup, the

first operation after matched filtering is clock

recovery, followed by frame synchronization,

which exploits the physical layer frame

(PLFRAME) header. Both operations must be

performed in the presence of large carrier fre-

quency errors. After this initial synchronization

procedure, a demultiplexer is used to separate

pilots from data symbols in a PLFRAME. The

pilot symbol stream is used by the following three

subsystems: the coarse frequency acquisition loop,

the signal-to-noise level estimator, and finally a

phase-locked loop (PLL) for tracking the residual

frequency offset and carrier phase. Once all auxil-

iary parameters are recovered, the data symbols

can be detected by the hard/soft demodulator.

The hard decisions are fed back to the PLL, while

the soft initial a posteriori probabilities (APPs) on

the received information bits are passed on to the

LDPC-BCH decoder. In the following sections

each subsystem is described in more detail.

SYMBOL TIMING RECOVERY

A suitable algorithm for timing recovery in

DVB-S2 is the Gardner estimator [10]. In fact,

this is a non-data-aided (NDA) circuit that is vir-

tually insensitive to modulation format (from

QPSK to 32-APSK) and performs efficiently

even in the presence of quite large carrier fre-

quency errors over the range of Es/N0of inter-

est. After the Gardner timing adjustment, the

sampled symbol is indicated as

rk= dkejφk + nk,

where dkis the transmitted symbol, φkcontains

the carrier frequency and phase offsets, and nk

represents the AWGN sample. In terms of

achievable performance, with an oversample fac-

tor equal to 4, the overall acquisition transient

can be completed in around 105symbols (2–3

PLFRAMEs) independent from the modulation

format, while at steady state the normalized

residual timing jitter is less than 10–2, which is

more than satisfactory.

FRAME SYNCHRONIZATION

Frame synchronization for DVB-S2 can be divid-

ed into acquisition and tracking, as in best prac-

tice. Frame acquisition (i.e., the initial frame

(1)

nFigure 1. Digital receiver block diagram.

Frequency

acquisition

SNR

estimation

Hard/soft

demodulator

LDPC/BCH

decoder

Freq/phase

tracking

Timing

recovery

Matched

filter

DeMUX

Pilots

Data

^θ0

^θk

Buffer

Buffer

Symbol

sampling

Frame

synch

Received

signal

Following a

sequential functional

order from startup,

the first operation

after matched

filtering is clock

recovery, followed

by frame

synchronization,

which exploits the

physical layer frame

header. Both

operations must be

performed in the

presence of large

carrier frequency

errors.

Page 4

IEEE Wireless Communications • December 2005

65

epoch detection) is the most critical phase and

thus is investigated hereafter. DVB-S2 frame

acquisition is performed by exploiting the auto-

correlation properties of the PLFRAME header,

which is a sequence of 26 π/2-binary PSK

(BPSK) symbols, identified as start of frame

(SOF). The received signal is correlated with

locally generated SOF replicas shifted by dis-

crete offsets. Therefore, the frame epoch estima-

tion problem is translated into a detection

problem that has to discriminate between

hypotheses or cells in a discretized uncertainty

region. In particular, we indicate as H1the

hypothesis corresponding to the achieved syn-

chronism, while H0marks all misaligned cells. A

possible detection approach is the threshold

crossing (TC) criterion [11], which amounts to

declaring H1(H0) if the correlation output does

(does not) cross a properly designed threshold.

Frame acquisition in DVB-S2 suffers from two

major impairments: the extremely low SNR, which

can indeed assume negative values in dB, and the

unknown carrier frequency offset and phase. In

addition, at terminal startup the uncertainty region

equals the entire frame length, TF, which in the

worst case is as large as 33,282 QPSK symbols. At

first glance, one could hope to improve perfor-

mance by exploiting multidwell procedures (i.e.,

collecting information from multiple frames before

making a final decision). Unfortunately, this

approach cannot be used in DVB-S2, where the

frame length depends on the selected coding/mod-

ulation pair. In particular, the frame format is sig-

naled to the receiver by a 64-symbol physical layer

signaling (PLS) field. Since the PLS field cannot

be decoded accurately prior to frame synchroniza-

tion, the latter must be performed in single-dwell

fashion. In any case, a genie-aided multidwell case

can be considered as a performance benchmark.

The single-dwell TC procedure is here identified

as 1TC, while a benchmark three-dwell strategy is

indicated as 3TC.

The detection of the SOF auto-correlation

peak is heavily affected by the presence of the

frequency offset, ∆f, which bans the possibility of

correlating coherently over the entire SOF. To

cope with this problem, the classic noncoherent

post-detection integration (PDI) approach could

be adopted, but this is outperformed by the dif-

ferential PDI (DPDI) scheme [12], sketched in

Fig. 2, which performs coherent correlation over

L SOF-segments of length M, followed by differ-

ential detection and combining. M and L should

be optimized considering worst case ∆f. A quick

set for M is given by the coherent integration

length dimensioning (CHILD) rule [13], accord-

ing to which the value of M should approach

3

8∆

(2)

where Tsis the symbol period, to maximize the

SNR at the output of coherent correlation.

The frame acquisition performance is mea-

sured in terms of mean acquisition time, for

which a practical performance specification of 2 s

can be assumed at terminal startup. In the pro-

posed design, a serial search procedure is consid-

ered where possible false alarms are recovered by

a tracking circuit that restarts the procedure after

a penalty time of Tps. Passive correlation (SOF

matched filtering) is used for coherent integra-

tion to reduce delay at the price of increased

complexity. Assume that Es/N0is in the typical

DVB-S2 range [–2.3, 0.7] dB, the frequency off-

set is ∆f = 5 MHz, and the baud rate is 27.5

Mbaud, and let the SOF length of 26 symbols be

subdivided in M = 2 and L = 13. Thresholds are

optimized in order to minimize the mean acquisi-

tion time at the worst SNR, –2.3 dB. Observing

the performance results summarized in Table 1

for Tpin the range [0–2TF], it can be noted that

1TC is indeed the optimal choice and is largely in

spec. This is due to the fact that DPDI yields low

false alarm and missed detection probabilities,

avoiding the need for multiple dwells. As a fur-

ther advantage, 1TC entails at the same time the

smallest complexity, and does not require PLS

decoding. Hence, DPDI is perfectly suited to

frame synchronization for the DVB-S2 standard.

An alternative frame acquisition design

approach is reported in [3]. There, SOF differ-

ential combining is performed without coherent

accumulation, which can lead to inferior perfor-

mance, especially for smaller ∆f. This requires

the exploitation of the Reed-Muller encoded

PLS field to aid frame acquisition, introducing

considerably increased complexity in the detec-

tor. In summary, the adoption of the DPDI

scheme with optimized coherent correlation

length allows the requirements to be satisfied

with reduced complexity.

COARSE CARRIER FREQUENCY RECOVERY

Coarse frequency recovery is also a critical step

due to the fact that DVB-S2 mass market ter-

minals will typically incorporate low-cost oscil-

lators, which introduce large initial frequency

offsets (e.g., 5 MHz at 27.5 Mbaud). Fortu-

M

fTs

≈

,

nFigure 2. Differential post detection integration block diagram.

ΣM

r(t)rMF(t)rm

cm*

xi

(m+∆)Ts+δ

MF

ΣL-1

Ts

(.)*

I I

n n n nTable 1. Frame synchronization performance results.

Ndwells

Es/N0(dB)

Penalty

time

Normalized

threshold

Mean acquisition

time (ms)

3 –2.32 TF8, 1, 1578

1–2.32 TF8244

10.72 TF842

1 –2.31 TF8 162

1 0.71 TF8 32

1–2.3043.2

1 0.7041.3

Page 5

IEEE Wireless Communications • December 2005

66

nately, after frame synchronization is achieved

it is possible to exploit the pilot fields intro-

duced in the DVB-S2 PLFRAME. In particu-

lar, the specified length for the pilot field is Np

= 36. In these conditions, a pilot-aided Men-

gali and Morelli (M&M) algorithm [14]

appears to be a valid candidate, because it

allows to have low estimation error variance

(approaching the Cramer-Rao bound) and a

sufficiently large pull-in range, according to the

requirements. The M&M frequency estimate is

computed as

(3)

where N is a design parameter, and

(4)

Npbeing the observation length in symbols. The

optimum choice for N is Np/2. The auto-correla-

tion function, R(m), is defined as

−

∑

(5)

The frequency estimate can be further improved

by averaging the auto-correlation function over

MP consecutive pilot fields over several

PLFRAMEs. In this case the auto-correlation

function is replaced by its average

(6)

Ri(m) being the auto-correlation contribution of

the ith pilot field defined in Eq. 5. As a practical

example, consider MP= 660, which corresponds

to 30 QPSK PLFRAMEs, or equivalently to an

estimation window of 36 ms. The normalized

residual frequency root mean square (RMS)

error for the averaged pilot-aided M&M estima-

tor is less than 10–4for Es/N0larger than 1 dB.

Interestingly, this result shows performance

improvements over the scheme proposed in [2].

SNR ESTIMATION

Adaptive physical layer receivers require accurate

estimation of the received SNR, for two basic rea-

sons: first, this estimate can be fed back to the

network to determine the most suitable

coding/modulation pair according to the channel

conditions experienced by the specific user; sec-

ond, this information is needed by the soft demod-

ulator to compute the APPs on the received

symbols. A suitable algorithm to this purpose is

the SNR Estimator (SNORE) [15], which is here

performed in a data-aided fashion over the pilot

field after frequency and timing corrections have

taken place. Considering a possible DVB-S2 sys-

tem at Ka-band, it can be assumed that channel

propagation impairment (atmospheric attenua-

tion) is a very slowly varying process compared to

SNR estimation time; thus, quasi-stationary chan-

nel conditions can be considered. Briefly, the use-

ful power, PS, can be estimated as follows:

=

1

(7)

where Npis the number of known symbols for

coherent accumulation, and dkis a QPSK pilot

symbol. On the other hand, the estimate of the

total received power, PR, is given by

(8)

so that the noise (plus possible interference)

power estimate is P

SNR estimate is computed as the ratio P

P

tions, the SNR estimate can further be averaged

over the last W values to reduce the estimation

error standard deviation. Assuming Np= 36 and

W = 50, the estimated SNR is unbiased and the

RMS error is on the order of 3 × 10–2, which is

perfectly adequate for both uses outlined above.

FINE CARRIER FREQUENCY AND PHASE TRACKING

The residual frequency offset after coarse acqui-

sition and phase estimation (penalized by strong

phase noise) is performed through a second-

order PLL that exploits all possible aiding mech-

anisms, using a hybrid data-aided and

decision-directed approach. The initial phase

estimate, θ

hood (ML) feedforward estimator over a pilot

sequence, and is employed by a decision-direct-

ed second-order PLL.

Within a PLFRAME, whenever a new pilot

field occurs, the PLL operates according to a

data-aided approach exploiting the known pilot

symbols (Fig. 1).

LDPC DECODING

LDPC decoding is performed through the iterative

sum-product algorithm, which updates the APP

values after each iteration, exploiting the Tanner

graph. This is a message passing algorithm, where-

^N= P

^R– P

^S. Finally, the

^Sover

^N. Exploiting the slowly varying channel condi-

^0, is obtained by a maximum likeli-

ˆP

N

r

R

P

k

k

NP

∑

=

=

1

2

1

ˆ

Re,

*

P

N

r d

k kS

P

k

NP

∑

=

{}

1

2

R m

( )

R m

i

( ),

i

MP

∑

=

=

1

R m

( )

Nm

r d

k k

rd

p

k m k m

−

k m

=

Np

( )( ).

**

≡

−

−

1

1

w

N m N

)(

m N N

2

N

NNN N

N

m

ppp

pp

=

−−+−−

−+−

3

1

4631

2

()(

() )

,

ˆ

f

arg{ ( ) *(

R m R

)},1

T

wm

m

m

N

∑

0

1

1

π

2

=−

=

n n n nFigure 3. PER comparison between AWGN channel (IC) and non linear dis-

torted channel, IBO = 2 dB, with channel parameter estimation.

Eb/N0 (dB)

12

1.E-04

PER

1.E-05

1.E-03

1.E-02

1.E-01

1.E+00

1311 109876543210

1/2 - QPSK (IC)

1/2 - QPSK

3/5 - QPSK (IC)

3/5 - QPSK

2/3 - QPSK (IC)

2/3 - QPSK

3/4 - QPSK (IC)

3/4 - QPSK

2/3 - 8PSK (IC)

2/3 - 8PSK

3/4 - 8PSK (IC)

3/4 - 8PSK

5/6 - 8PSK (IC)

5/6 - 8PSK

3/4 - 16APSK (IC)

3/4 - 16APSK

5/6 - 16APSK (IC)

5/6 - 16APSK

3/4 - 32APSK (IC)

3/4 - 32APSK (IBO=3.5dB)

Page 6

IEEE Wireless Communications • December 2005

67

by at the jth iteration the message sent from a bit

node v to a check node c is denoted Λj,v→c, and the

message in the opposite direction Λj,c→v. As shown

in [7], these messages are computed as

(9)

(10)

where Λk

soft demodulator related to rk. This simple

decoding algorithm proceeds iteratively until the

code parity check constraints are all verified, or

a maximum number of iterations is reached. As

shown in Eq. 9, the check node updating rule

takes into account the tanh(⋅) function, which

introduces significant computation complexity.

For this reason, proper approximations of this

function are recommended for use in actual

hardware implementations.

PERFORMANCE RESULTS

0is the initial APP value given by the

The entire DVB-S2 transmission chain has been

simulated using a purposefully developed C++

software tool. The final results are reported in

terms of packet error rate (PER) in steady state

conditions, after successful frame synchroniza-

tion and coarse carrier frequency recovery,

where the residual offsets and uncertainties are

continuously tracked by the receiver. The user

packet length is assumed to be equal to 1504 bits

(MPEG format), while the long LDPC coded

block (64,800 symbols) is selected. A reference

PER target of 10–4is specified here, which serves

as a benchmark for all Eb/N0performance losses.

The performance results reported in Fig. 3

show the comparison between DVB-S2 FEC

code performance in an AWGN channel (labeled

ideal channel, IC), and a channel with nonlinear

distortion and parameter estimation inaccura-

cies. In the latter case, the onboard HPA works

with an IBO = 2 dB that corresponds to OBO =

0.43 dB for both QPSK and 8-PSK modulations,

and 0.63 dB for the 16-APSK case. Furthermore,

a carrier phase noise mask compliant with the

DVB-S2 standard [1] has been adopted. With

the adoption of the estimation blocks described

above, actual values for the estimation inaccura-

cies have been considered. In particular, the nor-

malized symbol timing offset has a standard

deviation of 10–2, and the residual normalized

frequency error has a standard deviation of 10–4,

5 × 10–5, and 3 × 10–5for QPSK, 8-PSK, and 16-

APSK, respectively. The degradation with

respect to the IC case ends up being on the

order of 0.25 dB for QPSK, 0.4 dB for 8-PSK,

1.0 dB for 16-APSK, and finally 1.45 dB for 32-

APSK. These very small losses with respect to

IC are achievable thanks to the efficient predis-

tortion technique that has been purposefully

designed. This fact is confirmed by the results of

Fig. 4, where the performance with and without

predistortion is compared for IBO = 2 dB. Apart

from QPSK and 8-PSK schemes, in which the

degradation is limited, predistortion introduces a

gain of about 2 dB for 16-APSK, while 32-APSK

is not reported as it is completely unreliable

without compensation techniques for IBO values

lower than 4 dB.

The most significant overall achievements are

summarized in Fig. 5, which reports the integral

degradation for all modulation formats as a

function of the HPA OBO. The integral degra-

dation is a novel figure of merit that includes the

overall Eb/N0loss with respect to the ideal chan-

nel case due to both linear/nonlinear distortion

and channel parameter recovery imperfections

(time, phase/frequency, and SNR). In the figure

it can be noted that the minimum integral degra-

dation is in the order of 0.47 dB for QPSK, 0.78

dB for 8-PSK, 1.61 dB for 16-APSK. and 2.59 dB

for 32-APSK, respectively. These results confirm

that the proposed design solutions are able to

ΛΛΛ

k

j v

,

c

k

i

j

cv

i

i k

≠

dv

,

,

→

−→

=

=+∑

0

1

1

Λ

Λ

k

j c

,

v

i

j c

,

v

i

i k

≠

dc

tanh

tanh

→

−

→

=

= ⋅

2

1

2

1

1

∏ ∏

nFigure 4. PER comparison in presence of nonlinear distorted channel, IBO =

2 dB, with and without predistortion technique.

Eb/N0 (dB)

8

1.E-04

PER

1.E-05

1.E-03

1.E-02

1.E-01

1.E+00

8.5 97.57 6.565.554.54 3.532.521.510.50

1/2 QPSK

1/2 QPSK - no Pred

2/3 8PSK

2/3 8PSK - no Pred

3/4 16APSK

3/4 16APSK - no Pred

nFigure 5. Integral degradation vs. OBO for several modulation formats.

OBO (dB)

1.8

0.4

Integral degradation (dB)

0.0

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

4.0

4.4

4.8

21.61.4 1.210.80.60.40.20

QPSK

8PSK

16APSK

32APSK

Linear

Page 7

IEEE Wireless Communications • December 2005

68

provide performance compliant with the DVB-

S2 specifications.

CONCLUSIONS

This article provides general guidelines for the

implementation of an adaptive physical layer

compliant with the DVB-S2 standard. These

guidelines can be summarized as follows. Frame

synchronization can be performed through a sin-

gle-dwell TC strategy coupled to a robust post

detection integration scheme, such as DPDI,

which exploits the SOF only, with no introduc-

tion of further complexity. Timing acquisition

can be performed through the Gardner estima-

tor, while coarse frequency estimation can be

achieved through the Mengali and Morelli algo-

rithm. Adaptivity is enabled by a SNORE algo-

rithm to estimate the experienced SNR

conditions. Fine carrier phase and frequency

tracking is maintained through a hybrid PLL. By

coupling the LDPC capabilities with the

advanced predistortion techniques based on

LUTs, data recovery performance shows surpris-

ingly small integral degradation for all

coding/modulation pairs. The interesting result is

that by considering an actual Ka-band satellite

link severely affected by strong linear and non-

linear distortion, and taking into account non-

ideal parameter estimation, it is still possible to

fully exploit the adaptive physical layer potential,

and to effectively achieve the capacity gains fore-

seen by the DVB-S2 standard.

REFERENCES

[1] ETSI EN 302 307, “Second Generation Framing Struc-

ture, Channel Coding and Modulation Systems for

Broadcasting, Interactive Services, News Gathering and

Other Broadband Satellite Applications,” Jan. 2004.

[2] E. Casini, R. De Gaudenzi, and A. Ginesi, “DVB-S2

Modem Algorithms Design and Performance over Typi-

cal Satellite Channels,” Int’l. J. Satell. Commun. and

Net., vol. 22, June 2004, pp. 281–318.

[3] F. W. Sun, Y. Jiang, and L. N. Lee, “Frame Synchroniza-

tion and Pilot Structure for Second Generation DVB via

Satellites,” Int’l. J. Satell. Commun. and Net., vol. 22,

June 2004, pp. 319–39.

[4] R. Rinaldo and R. De Gaudenzi, “Capacity Analysis and

System Optimization for the Forward Link of Multi-

beam Satellite Broadband Systems Exploiting Adaptive

Coding and Modulation,” Int’l. J. Satell. Commun. and

Net., vol. 22, June 2004, pp. 401–23.

[5] F. J. MacWilliams and N. J. A. Sloane, The Theory of

Error-Correcting Codes, North-Holland Mathematical

Library, 1978.

[6] R. G. Gallager, “Low Density Parity Check Codes,” IRE

Trans. Info. Theory, 1962, vol. 8, pp. 21–28.

[7] D. J. Mackay, Information Theory, Inference and Learn-

ing Algorithms, Cambridge Univ. Press, 2003.

[8] Y. Nagata, “Linear Amplification Technique for Digital

Mobile Communications,” Proc. VTC, 1989, pp. 159–64.

[9] P. Salmi, M. Neri, and G. E. Corazza, “Design and Per-

formance of Predistortion Techniques in Ka-band Satel-

lite Networks,” Proc. 22nd AIAA ICSSC 2004, Monterey,

CA, May 2004.

[10] F. M. Gardner, “A BPSK/QPSK Timing-Error Detector for

Sampled Receivers,” IEEE Trans. Commun., vol. 34, May

1986, pp. 399–406.

[11] G. E. Corazza, “On the MAX/TC Criterion for Code Acquisi-

tion and Its Application to DS-SSMA Systems,” IEEE Trans.

Commun., vol. 44, no. 9, Sept. 1996, pp. 1173–82.

[12] G. E. Corazza et al., “Differential and Non-Coherent

Post Detection Integration Techniques for the Return

Link of Satellite W-CDMA Systems,” IEEE PIMRC, Lisbon,

Portugal, 2002.

[13] G. E. Corazza, R. Pedone, and M. Villanti, “Frame Acquisi-

tion for Continuous and Discontinuous Transmission in the

Forward Link of Ka-Band Satellite Systems,” 6th European

Wksp. Mobile/Pers. Satcoms and 2nd Adv. Sat. Mobile Sys.

Conf., vol. 1, ESA-ESTEC, Noordwijk, The Netherlands,

21–22 Sept. 2004, pp. 211–18.

[14] U. Mengali and M. Morelli, “Data-Aided Frequency

Estimation for Burst Digital Transmission,” IEEE Trans.

Commun., vol. 45, Jan. 1997, pp. 23–25.

[15] D. R. Pauluzzi and N.C. Beaulieu “A Comparison of

SNR Estimation Techniques for the AWGN Channel”

IEEE Trans. Commun., vol. 48, Oct. 2000, pp. 1681–91.

BIOGRAPHIES

GIANNI ALBERTAZZI (galbertazzi@arces.unibo.it) received a

Dr.Ing. degree in telecommunications engineering in 2002

from the University of Bologna, Italy, defending a thesis on

turbo coding techniques. In January 2002 he joined the

Advanced Research Center for Electronic Systems (ARCES)

of the same university as a Ph.D. student. His research

activity is concerned with digital transmission, mobile and

satellite CDMA systems, and especially channel coding

techniques, with particular interests in turbo and LDPC

code construction.

STEFANO CIONI (scioni@arces.unibo.it) received a Dr.Ing.

degree in telecommunication engineering and a Ph.D. from

the University of Bologna in 1998 and 2002, respectively.

From March 2002 to October 2002 he was a visiting

researcher at the European Space Agency (ESA) on ACM

techniques for future broadband satellite networks. His

research activities and interests include synchronization

techniques, medium access control resource allocation

algorithms, OFDM, and iterative decoding techniques joint

with channel parameter estimation.

GIOVANNI E. CORAZZA (gecorazza@deis.unibo.it) received a

Dr.Ing. degree (summa cum laude) in electronic engineer-

ing in 1988 from the University of Bologna and a Ph.D. in

1995 from the University of Rome “Tor Vergata,” Italy. He

is a full professor in the Department of Electronics, Com-

puter Science, and Systems (DEIS) of the University of

Bologna. He is Chairman of the Advanced Satellite Mobile

Systems Task Force (ASMS-TF).

MASSIMO NERI (mneri@deis.unibo.it) received a Dr.Ing.

degree in telecommunications engineering from the Uni-

versity of Bologna in 2002. Since October 2002 he has

been with the Department of Electronics, Computer Sci-

ence and Systems (D.E.I.S.) at the University of Bologna,

where he started his Ph.D. study in 2003. His research

activities and interests include nonlinearity mitigation tech-

niques, multi-user detection, OFDM, and coding theory.

PAOLA SALMI (psalmi@ieee.org) received a Dr.Ing. degree in

telecommunications engineering and a Ph.D. in electronics

and computer science from the University of Bologna in

1996 and 2001, respectively. From 2001 to 2004 she was

at the ARCES of the University of Bologna. During 2001

she was a summer manager at AT&T Shannon Laboratory,

New Jersey. She currently works with the Optical Research

Team of Andrew Wireless Systems. Her research activities

include digital communication theory, 3G systems, synchro-

nization techniques, and nonlinear effects on optical links.

ALESSANDRO VANELLI-CORALLI (avanelli@deis.unibo.it) received

a Dr.Ing. Degree (cum laude) and a Ph.D. in electronics

engineering from the University of Bologna in 1991 and

1996, respectively. In 1996 he joined the University of

Bologna, where he is currently an assistant professor. Dur-

ing 2003 and 2005 he has been a visiting scientist at Qual-

comm Inc., San Diego, California. He is responsible for the

R&D group of the ASMS-TF.

MARCO VILLANTI (mvillanti@deis.unibo.it) received a degree

in telecommunications engineering (summa cum laude) in

2001 and a Ph.D. in telecommunications in 2005 from the

University of Bologna, where presently he is a post-doctor-

al researcher with DEIS and ARCES. He was a visiting schol-

ar at the University of California, San Diego, in 2003 and

2005. His research interests are in communications theory,

acquisition and synchronization for CDMA, TDMA, and

UWB systems.

RAFFAELLA PEDONE (rpedone@deis.unibo.it) received a degree

in electronic engineering (summa cum laude) from the Uni-

versity of Bologna in 2002, where she is presently a Ph.D.

student in telecommunications with DEIS. Her interests are

in digital communications, with particular focus on acquisi-

tion and synchronization for spread spectrum and TDMA

systems.

By coupling the LDPC

capabilities with

the advanced

predistortion

techniques based on

LUTs, data recovery

performance show

surprisingly small

integral degradation

for all coding/

modulation pairs.

#### View other sources

#### Hide other sources

- Available from Stefano Cioni · May 19, 2014
- Available from etu.edu.tr