Docstoc

PERFORMANCE ASSESSMENT OF A DVB-T TELEVISION SYSTEM

Document Sample
PERFORMANCE ASSESSMENT OF A DVB-T TELEVISION SYSTEM Powered By Docstoc
					     PERFORMANCE ASSESSMENT OF A DVB-T TELEVISION
                       SYSTEM
           A. Morello, G. Blanchietti, C. Benzi, B. Sacco, M. Tabone
                       RAI Research Centre, Torino, Italy

ABSTRACT

        In the framework of the DVB, RACE dTTb and ACTS VALIDATE European
projects, the RAI Research Centre has directly contributed to the definition and
validation of the DVB system for digital terrestrial television broadcasting (DVB-T).
The system, allowing fixed and (static) portable reception, is based on OFDM
modulation with a variety of modulation/channel coding configurations, and is
characterised by two operational modes, the first one with 2K carriers for
conventional multi-frequency networks (MFN), the second one with 8K carriers to
cover also single frequency networks (SFN).
This contribution presents the results of the VALIDATE laboratory tests carried out
by RAI in winter and spring 1996-1997 on the second dTTb demonstrator,
developed in compliance with the DVB-T specification and including both the 2K and
the 8K options. The system under test included the complete transmission and
reception television delivery chain: picture / sound coding, multiplexing, OFDM
modulation, High Power Amplifier, semi-consumer tuner, coherent OFDM
demodulator, video decoding and display.
The tests investigated the ruggedness of the OFDM DVB system in presence of a
variety of typical channel distortions, such as Gaussian and impulsive noise, tuner
phase noise, echoes with various delays and amplitudes, Doppler frequency shifts,
interference from an other DVB-T signal and from analogue PAL/SECAM TV signals,
transmitter non-linearity.
Even if the equipment being tested at the moment is only a first generation of
equipment, the test results have been largely in line with the performance predicted
by computer simulations, apart from few exceptions, and allow to gain an important
insight on the characteristics of the sophisticated modulation, channel coding and
equalisation techniques of the DVB-T system.
For the sake of conciseness it was not possible to include a description of the DVB-T
system, nor a tutorial on its technical background, but the reader can refer to [2] and
[3] to better understand the measurement results.

1.      INTRODUCTION

Taking into account the rapid progress of digital television technologies in
production, transmission and emission and the new commercial requirements
identified by the broadcasters and the consumer industry, the DVB Project has
started its activity in 1993, aiming at the harmonisation of the strategies for the
introduction of digital television broadcasting in Europe on the various delivery
media. The first significant success of the Technical Module of the DVB project has
been the definition at the end of 1993 of the DVB-S system for direct-to-home
satellite broadcasting of multi-programme TV, followed by the DVB-C system for
cable networks. These two systems were adopted also for TV distribution on satellite
master antenna systems (SMATV) and on microwave point-to-multipoint links
(MMDS). Due to the higher hostility of the terrestrial VHF/UHF propagation
environment and to the longer terms introductory plans, the system for digital
terrestrial broadcasting, named DVB-T, has been agreed only in November 1995,
and approved as an ETSI Standard in February 1997 [1]. The RACE dTTb Project, in
co-operation with other national projects as HD-DIVINE and HDTVT, have studied,
designed and developed the DVB-T system taking into account the service
requirements defined by DVB; among these, the most technically demanding are:
 the possibility to obtain (stationary) portable reception with omnidirectional
   antennas in addition to fixed reception with directive roof-top antennas
 the possibility to operate on single frequency networks (SFN), where the coverage
   of large areas (a region or even a country) is achieved by synchronised
   transmitters operating on the same RF channel, with significant advantages in
   terms of spectrum efficiency.

The DVB-T system [1, 2] can transmit digital multi-programme, conventional-
definition television signals in MPEG-2 MP@ML format, but it is open to evolve
towards HDTV by using higher MPEG-2 levels and profiles. The transmission
scheme is based on a multi-carrier coded modulation named C-OFDM [3],
characterised by two operational modes, the first one with 2K carriers for
conventional multi-frequency networks (MFN), the second one with 8K carriers to
cover also single frequency networks (SFN). This modulation is particularly suitable
to operate on the terrestrial multipath propagation channel because of the narrow-
band characteristic of each data carrier and of the presence of a "guard interval"
(duration Tg) which separates adjacent symbols and avoids inter-symbol
interference in the presence of echoes. The DVB-T system offers a bit-rate capacity
ranging from 5 Mbit/s to 31.5 Mbit/s depending on the chosen level of m-QAM
modulation (m=4, 16, 64), the inner code rate (1/2, 2/3, 3/4, 5/6 or 7/8) and the guard
interval duration (Tg/Tu==1/4, 1/8, 1/16 or 1/32; Tu=useful symbol duration). It is
optimised for 8 MHz channels (European UHF channellisation), but it can be easily
adapted to 7 MHz and 6 MHz channels by adjusting the receiver sampling frequency.
The system bit-rate capacity and the computer simulation performance over various
multipath channels are given in the Annex, which repeats the information of
Appendix A of the DVB-T specification.

The ACTS VALIDATE Project [4] has been set up at the end of 1995 by the
European Commission in order to validate the DVB-T specification in practical
situations, by means of interworking tests between different pieces of equipment,
extensive laboratory tests, real over-air transmissions using single frequency
networks or interleaved channels. In addition the project is investigating the
transmitter and domestic repeater technology and developing planning methods
suitable for digital terrestrial television broadcasting.

In the framework of the VALIDATE activity, a complete DVB-T chain was set-up at
the RAI Research Centre of Turin, including real time MPEG-2 encoders / decoders
(for three programmes), a Transport Stream multiplexer/demultiplexer, the dTTb C-
OFDM modem, an RF transmitter on UHF channel 28 (or 43), a “quasi-consumer”
tuner.
Among the DVB-T compliant pieces of equipment up to now available to the
members of the VALIDATE Project, the dTTb demonstrator is the first one
implementing not only the 2K, but also the 8K FFT mode, and therefore it can give
important confirmations on the specification performance in SFNs.
Table 1 lists the DVB-T system configurations which have been tested.

                             TABLE 1 - Tested Modes
        Modulation        Code Rate        Carriers            Ru [Mbps]     Guard Interval 
          QPSK               1/2          2K & 8K                5.53                1/8
         16QAM               3/4          2K & 8K                18.10              1/32
         64QAM               2/3          2K & 8K                19.91          1/4, (1/8 **)
                   Note (*): Tg/Tu, Tg=guard interval duration, Tu=useful symbol duration
                   Note (**): =1/4 unless otherwise explicitly indicated

The results presented in this contribution cover the system performance with
Additive White Gaussian Noise (AWGN), with impulsive noise, with static and time-
varying echoes and with non-linear distortions. These figures are important to verify
some of the key elements debated during the DVB-T standard definition: the system
sensitivity to phase noise in the 8K mode, the implementation margin, the possibility
to operate with high-level, long-delay echoes (e.g. 0 dB amplitude and 200 s delay,
for SFNs), the steepness of the failure characteristic with echoes outside the guard
interval, the possibility to serve static or even mobile portable receivers.


2.      LABORATORY SET-UP AND BASIC TEST PROCEDURE

The laboratory performance evaluation of the DVB-T system is carried out by
changing the parameters of the channel impairments (noise, interference, multipath
echoes, phase noise, frequency off-sets, non-linearity) and measuring the BER at
the receiver side, after Viterbi error correction. Additional observations on the
decoded picture are carried out to verify if, as indicated by theory, BER=2.10-4 after
Viterbi decoding allows to deliver good service quality (i.e., less than an error event
per hour), thanks to the powerful outer error correction by Reed-Solomon (188,204)
coding.
Figure 1 shows a simplified scheme of the RAI laboratory set-up, simulating the
channel impairments at the intermediate frequency (IF, 35.5 MHz), in order to exploit
the flexibility of the laboratory instruments and the availability of SAW filters at these
frequencies. Up and down frequency converters (quasi-transparent in terms of
system performance degradation), from IF to radio frequency (RF, UHF channels
28 or 43) and viceversa, allows to test the RF devices (transmitter and receiver).


                           Down                              Up
     Transmitter          Converter                       Converter           Tuner
                     RF(tx)                                        RF(rx)

      OFDM                Multipath                                           OFDM
     modulator                                                               receiver
                              Channel                              IF(rx)

                     IF(tx)                                                             BER
                                                                        Spectrum
     MPEG MUX                                       interference        Analyzer        Meter
                                        AWGN
      & Coders
                    Figure 1 - Block diagram of the RAI laboratory set-up

The following basic test procedure allows to evaluate the DVB-T system
performance in the presence of a variety of different channel impairments, by
measuring the noise margin loss ###C/N [dB] (see Figures 1 and 2) at the
reference BER*=2.10-4 after Viterbi decoding:
 the channel impairment under consideration is switched off. The additive noise
   level is increased until BER* is achieved. The corresponding carrier-to-noise ratio
   at the receiver input (C/N*) is noted (measured in the receiver noise bandwidth of
   7.6 MHz)
 the channel impairment (e.g. the interfering signal) is switched on. The additive
   noise level N is reduced by ###C/N dB, corresponding to the noise margin loss.
   The impairment level (e.g. the interference power I) is increased until the
   reference bit-error ratio BER* (after Viterbi decoding) is obtained again. The
   impairment level (e.g. I) is noted.
The procedure can set the noise margin loss in pre-determined steps (e.g., ###C/N
= 1, 2, ... , ### dB), and identify the corresponding channel impairment level (e.g., I),
or viceversa. It should be noted that ###C/N = ### corresponds to the case without
additive noise, where the channel impairment produces itself the reference BER*.
                     BER



                                                    C/N



                          -4
                   2 10                                              BER*


                                             without         with
                                           impairment      impairment

                                                                            C/N
                                                C/N*


  Figure 2- Measurement of the noise margin loss ###C/N produced by a channel
                                  impairment

3.       SYSTEM IMPLEMENTATION LOSSES OVER AWGN CHANNEL
In order to identify the implementation losses introduced by the various elements of
the DVB-T chain, the BER versus C/N curves after Viterbi decoding have been
measured, and the noise margin losses have been identified as summarised in
Table 2. The laboratory set-up connections, indicated for example as IF(tx)- RF(rx),
make reference to Figure 1.
The basic modem implementation losses and the total chain degradation are
referred to the computer simulation figures (see Table A1 of the Annex, AWGN,
ideal receiver). Two figures are reported when different results were obtained on
UHF channel 28 (530 MHz) and on UHF channel 43 (650 MHz).

                    TABLE 2 - System implementation losses C/N
                                                       C/N [dB]
                                  IF(tx)-IF(rx)         IF(tx)-RF(rx)             RF(tx)-RF(rx) (*)
     Modulation     FFT         Basic modem         Tuner degradation              Total chain
     and coding              implementation losses including phase noise            degradation
                                                    ch. 28         ch.43          ch. 28    ch.43
     QPSK 1/2     1/8 2K               1.7                   0.0                   1.7        1.8
                       8K              1.6                   0.0                   1.8        1.8
  16QAM 3/4       1/32 2K              1.6                   0.0                   1.7        1.7
                       8K              1.5                   0.1                   1.7        1.7
  64QAM 2/3       1/4 2K               2.4            0.2          0.5             2.8        3.1
                       8K              2.3            0.3          0.7             2.9        3.3
Note (*): HPA producing spectrum shoulders of the order of - 38 dB

It should be remarked that, to estimate the channel frequency response and equalise
the received constellations, the system under test makes use of time-domain and
frequency-domain interpolation (named 2-D algorithm) applied to the scattered pilots.
For =1/4 the residual noise on the estimated channel response produces a C/N
degradation of about 1.6 dB (with respect to the computer simulation figures of the
Annex), while for < 1/4 the degradation can be reduced. This can explain the
superior performance achievable with the smaller guard intervals. To verify this
directly, a reduction of  from 1/4 to 1/8 was introduced with 64QAM rate 2/3, and a
0.4 dB C/N improvement was achieved, in IF(tx)-IF(rx) configuration. Other
algorithms can be implemented in the receiver, averaging in time the estimated
channel response in order to filter-out the residual noise. These solutions can reduce
the implementation margin, but at the cost of a reduction of the channel tracking
speed of the system, in the presence of time varying channels.
From the test results, the following conclusions can be drawn:
 the 2K and 8K systems show similar implementation margins
 IF(tx)-IF(rx): for QPSK 1/2 and 16QAM 3/4 and for <1/4, modem implementation
   loss is of the order of 1.6 dB; for 64QAM 2/3 the loss is of about 2.3 dB for 
   =1/4, and of about 1.9 dB for  =1/8;
 IF(tx)-RF(rx): the “quasi-consumer” tuner introduces a moderate degradation on
   the system performance (about 0.2-0.3 dB on channel 28, and 0.5-0.7 dB on
   channel 41). The 8K mode is not penalised by the tuner phase noise compared
   with the 2K mode;
 RF(tx)-RF(rx): the additional degradation introduced by the transmitter depends
   on the modulation/coding scheme and on the level of HPA non-linear distortion
   (see Section 7). In the tested configuration (about -38 dB spectrum shoulders) it
   is in the range 0.2 to 0.3 dB. The total system implementation margin is of the
   order of 2 dB for QPSK 1/2 and 16QAM 3/4 (which could increase to about 2.5 dB
   for  =1/4) and of 3 dB for 64QAM 2/3 ;
 during the tests, the demonstrator operated reliably without synchronisation
   problems even at high BER levels after Viterbi decoding (up to BER=10-2); at
   BER=2.10-4 the decoded picture quality was unimpaired by errors, and at
             -3
   BER=10 was at the service continuity threshold.

3.1.     Tuner Noise Figures
Noise figure measurements have been made by IRT with the dTTb demonstrator, by
CCETT with the STERNE 4 modem and by DTAG with the DMV modem. In all
cases the manufacturers have indicated that their tuners are quasi-consumer type
tuners. The tuner noise figures measured for the three modems were of the order of
8 dB. The manufacturer that developed the tuner of the dTTb demonstrator
indicated that “The devices used in dTTb (and HDTV-T) demonstrators should lie in
the range 6.2 to 7 dB, valid for 500 and 700 MHz, respectively. New production type
tuners come up with about 7 dB in average (max.10 dB). Therefore the front-end in
the demonstrator represents practically a standard device.”
An information request on this subject has been sent from CEPT-FM PT24 to the
DVB consumer manufacturers. An important European manufacturer has responded
that “Our believe is that the noise figure of 5 dB used so far in the planning of DVB-T
services is not practical. It would mean a significant cost increase in the tuner. At the
moment we see that a reasonable figure would be 8 dB. This is achievable with
careful design and does not require any special measures. Unfortunately we don’t
see any major improvements in the near future.” Another major European
manufacturer indicated that ”the currently achievable noise figure for DVB-T tuners is
8 dB, this value should be used both for smooth introduction of DVB-T systems and
for planning purposes. Although it is not completely impossible to go down to 5 dB,
this is at significant cost increase, coupled with large signal handling penalties”.
At present, a noise figure of 7 dB is being considered within CEPT for planning
purposes.
4.      PERFORMANCE WITH IMPULSE NOISE

The dTTb demonstrator has been tested in the presence of impulsive noise
interference, simulated by means of a 5 ns base-band pulse generator filtered with
an 8 MHz SAW filter centred at IF. The pulse repetition frequency (PRF) was varied
between 50 Hz and 20 kHz. The measurements identified the required carrier-to-
interference ratio C/Io corresponding to a system noise margin loss C/N= 1, 3 and
 dB for BER=2.10-4 after Viterbi decoding. Io is the average power (1) of the filtered
impulsive noise normalised with respect to RPF, so that keeping I o constant the
impulse amplitude becomes independent from RPF. Figure 3 shows the results
relevant to 64QAM rate 2/3, 8K mode.




          Figure 3 - 64QAM 2/3 performance with impulse noise (8K mode)

The shape of the curves can be interpreted as follows:
 for PRF lower than the OFDM symbol rate (about 890Hz in 8k mode), only one
   out of N OFDM symbols is corrupted by an interfering pulse, where N=symbol-
   rate / PRF. Therefore the average BER is proportional to PRF. Since the C/I
   versus BER curves are very flat, of the order of 1 dB per BER decade, the
   resulting slope is negligible.
 for PRF larger than the OFDM symbol rate, M interfering pulses corrupt each
   OFDM symbol, and the interference power is proportional to PRF. Therefore in
   Figure 3 the slope of the curves at the right side of the symbol-rate is of about
   10 dB per decade.

In the tests, BER=2.10-4 corresponded to good picture quality, denoting a good
performance of the synchronisation algorithms. The obtained curves are comparable

1 measured with a HP8136 spectrum analyser (7.6 MHz window, automatic video and resolution
bandwidth).
with the performance obtained with first generation OFDM modems tested in the
same conditions.

5.            PERFORMANCE WITH ECHOES

Digital terrestrial broadcasting is affected by the presence                          of echoes, produced by
the obstacles on the propagation environment (indicated                               as natural echoes). In
addition, the use of synchronised transmitters, in SFN                                configuration, produces
additional echoes (indicated as artificial echoes), usually                           characterised by strong
levels and long delays.

5.1.          Static echoes

A basic service requirement defined by DVB is the reception with roof-top directive
antennas as well as with built-in omnidirectional antennas (static portable receivers).
Therefore the DVB-T system can operate with high level static echoes, produced by
reflections or by SFN transmitters.

        5.1.1. Single echo
The dTTb demonstrator has been tested on UHF channel 28, IF(tx)-RF(rx), in the
presence of echoes inside and outside the guard interval. Figure 4 shows the noise
margin loss (compared to AWGN channel) of 64QAM 2/3 versus the echo delay and
amplitude (C/I=0 dB means that the main path and the echo has equal power). The
guard interval is =1/4, corresponding to 56 s in the 2K mode. Figure 5 shows
similar curves for the 8K mode, with guard interval of 224 s=1/4), suitable for
SFN operation.

                                                               Single Echo
                                                          2k, 64QAM 2/3, =1/4
                                                         Guard Interval = 56 sec
                                  Guard Interval
                                      Limit

     10
      9
                C/I=0 & 0.5 dB
      8
      7
      6
                          C/I=2
      5                                                  C/I=22                 C/I=22
      4                   C/I=4
      3                                                           C/I=24
                          C/I=6
      2                                                                      C/I=26
                                  C/I=8
      1
                                                                                            C/I=28
      0
          0     10   20    30     40      50   60   70     80     90   100    110     120   130   140   150   160   170   180   190   200

                                                             Echo Delay (sec)
  Figure 5 2K mode: noise margin loss with a single echo (versus echo delay and
                                  amplitude)
                                               Single Echo
                                         8k - 64QAM 2/3 - =1/4
                                        Guard Interval = 224 sec

                                                                              Guard Interval Limit

  10
   9
   8
   7                                                                      C/I=0 & 0.5 dB      8                       20
   6
                                                                             C/I=2                          16
   5                                                                                              10

   4                                                                                               12
                                                                              C/I=4
                                                                                                                             22
   3
                                                                                                       14        18
                                                                                C/I=6
   2                                                                                                                         24

   1                                                                                                                         26

   0
       0   20   40     60    80   100     120      140    160       180       200       220   240            260       280        300
                                                Echo Delay (sec)
  Figure 4 8K mode: noise margin loss with a single echo (versus echo delay and
                                  amplitude)

Table 3 summarises the results achieved with the three modulation and coding
schemes under test, for echoes within the guard interval, and searching for the worst
echo phase. The figures reported in the Table give the best and the worst case
achieved with various echo delays within the guard interval.


   TABLE 3 - Noise margin losses for a single echo within the guard interval
          Echo level                        Noise margin losses C/N [dB]
                                  Reference: AWGN channel performance, IF(tx), RF(rx)
            C/I [dB]          QPSK 1/2 (=1/8)   16QAM 3/4 (=1/32) 64QAM 2/3 (=1/4)
                               2K         8K        2K           8K         2K         8K
               0            4.3###5.0 4.1###4.6     8.7       8.1###      5.4###    6###7
                                                                8.7         7.9
               1            4.1###4.9 3.9###4.4     8.5       6.2###      5.2###    5.8###
                                                                8.4         7.8       6.8
               2            3.2###3.6 3.0###3.7 5.3###        5.0###      4.1###    4.6###
                                                    5.7         5.9         5.8       5.3
               4            2.1###2.4 1.9###2.3 3.4###        3.1###      2.5###    2.8###
                                                    3.9         3.8         4.1       3.1
               6            1.3###1.5 1.2###1.4 2.2###        2.0###      1.5###    1.7###
                                                    2.3         2.4         2.6       1.9
               8            0.8###1.0 0.8###1.0 1.4###        1.3###      0.9###    1.2###
                                                    1.5         1.5         1.7       1.4
              10            6.0###0.7 0.5###0.7 0.9###        0.8###      0.5###    0.7###
                                                    1.0         1.0         1.1       0.8
The following conclusions can be drawn:
 the noise margin losses depend on the coding rate: lower coding rates (e.g. 1/2)
   allow to reconstruct more efficiently the OFDM carriers suppressed by the
     frequency-domain notches, making the performance less sensitive to the
     multipath echo configuration. Therefore high coding rates (5/6 and 7/8) should be
     avoided when SFN and portable reception are addressed;
   the system can satisfactorily operate even with a 0 dB echo (within the guard
     interval), without synchronisation or instability problems. The measured noise
     margin losses are lower than 5 dB, 8 dB and 9 dB, with coding rates 1/2, 2/3 and
     3/4, respectively (it should be noted that these figures refer to different m-QAM
     constellations). For 64QAM rate 2/3, the figures obtained by computer simulations
     give C/N=4 ### 4.5 dB with perfect channel estimation;
   the 8K and 2K modes shows comparable performance (to be noted that for the
     same, the 8K mode shows a guard interval 4 times longer than the 2K mode);
    the delay Tu/6 (within the guard interval for =1/4) produces periodical notches
     affecting one scattered pilot out of two (assuming a worst case phase), and
     degrades the channel estimation performance. This explains the degradation
     peaks within the guard interval in Figures 4 and 5, relevant to 64QAM 2/3. These
     additional degradations not reported in Table 3;
   as expected from theory, for 2K and 8K modes and =1/4, the performance
     degradation is steep for echoes outside the guard interval (see Figures 4 and 5).
     For example in the 8K mode a strong echo with C/I=10 dB can be accepted only
     for delays shorter than  =245 s, corresponding to /Tg=1.09; a weak echo at
     C/I=20 dB can drive the system out of service for  =275 s, corresponding to
     /Tg=1.23. The peak of degradation outside the guard interval for delays around
     Tu/3 is due to an aliasing effect in the channel estimation algorithm based on
     scattered pilots.     An alternative channel estimation method [5] is being
     implemented at RAI in order to achieve a more gradual performance degradation
     for echoes outside Tg;
   both for 2K and 8K, the particular echo delay of 1 s forced the system in a
     high BER status (even without noise). This effect has still to be investigated in
     order to introduce the required modifications in the receiver.

        5.1.2. Multiple echoes
Additional tests have been carried out with six echoes, which is the maximum
allowed by the present set-up at RAI. In the following the tests are indicated as
“fixed”, “portable”, “dense SFN” and “regional SFN”, with reference to the reception
conditions in which similar echo configurations could be found in the service area.
The first two cases correspond to the six strongest echoes of Annex B of the DVB-T
specification. For the dense and regional SFNs, the echo configuration is reported in
the footnotes 2 , 3 (=delay in s, =C/I in dB).
Table 4 compares the test results with the simulation results (when available) given
in the Annex A of the DVB-T specification.

     TABLE 4 - Noise margin losses C/N with multiple echoes within the guard
                                     interval
                                                                 Noise margin losses C/N [dB]
                                                            (reference: AWGN channel performance)
           Reception condition                   QPSK 1/2 =1/8)     16QAM 3/4 =1/32)     64QAM 2/3 (=1/8)
                                                  2K         8K         2K         8K         2K        8K
                  Fixed                           0.4      0.4          0.8       0.6        0.7       0.5
             (Ricean channel)                    (0.5*)   (0.5*)      (0.5*)     (0.5*)     (0.6*)    (0.6*)

2 ray 1:=0, =0; ray 2:=7.8, =9.3; ray 3:=11.6, =5.5; ray 4:= 17.5, = 16.1; ray 5:= 20, = 14.5 ; ray
6:=23.4, =23.4;
3 ray 1:=0, =0; ray 2:=/3, =9.3; ray 3:=/2, =5.5; ray 4:=7/10, =16.1; ray 5:=/5, =14.5 ; ray
6:=/10, =23.4;
              Portable                 2.5       2.6        4.3        4.4        3.4       3.5
         (Rayleigh channel)           (2.3*)    (2.3*)     (4.2*)     (4.2*)     (2.8*)    (2.8*)
   Dense SFN (Rayleigh channel)        2.1       2.2        (**)       3.5        2.6       2.8
   Regional SFN (Rayleigh hannel)      2.1       2.2        3.4        3.4        2.5       2.8
Note (*): computer simulation results
Note (**): not applicable, since there is an echo outside the guard interval

The test results are in good agreement with the simulation results (see Table A1 in
the Annex).
On the Ricean channel (main path dominating on the echoes of about 10 dB, fixed
reception), the system performance is about 0.5 dB worse than on the AWGN
channel. On the Rayleigh channel (no dominant path, portable reception) the
degradations are significantly smaller than with a single 0 dB echo (see Table 3).
       5.1.3. Spectrum tilt

The dTTb system noise margin loss has been tested in the presence of a spectrum
tilt (with constant slope across the signal bandwidth BW), which is a typical distortion
for master antenna systems and for active indoor repeaters.
In the laboratory, a spectrum tilt of T [dB] (peak-to-peak) was simulated, by setting
up a channel with a single 0 dB echo, 50 ns delay (producing spectrum notches with
periodicity of 20 MHz) and suitable phases to achieve the desired tilt. Table 5 shows
the maximum spectrum tilts T to achieve noise margin losses lower than 0.5 and 1
dB with respect to the AWGN channel. The tested modes were 2K and 8K, 64QAM
rate 2/3, =1/4.

       TABLE 5 - Spectrum tilt T [dB ] for a given noise margin loss C/N [dB]
                               for 64QAM 2/3 (=1/4)
                                              T [dB]
                  mode        C/N = 0.5 dB            C/N = 1 dB
                           slope +       slope -   slope +     slope -
                   2K          3            9          5         12
                   8K         2.5           8         4.5        11

The maximum acceptable spectrum tilt heavily depends on the sign of the tilt slope
(positive or negative). Considering the worst case (positive slope),the maximum tilt in
the 7.6 MHz signal bandwidth is 2.5-3 dB for 0.5 dB degradation and 4.5-5 dB for 1
dB degradation.
A similar spectrum tilt was applied to a signal already affected by the multipath echo
configuration of the “Ricean” channel (Fixed channel, as defined in section 5.1.2),
which is typical of roof-top reception and signal distribution via a master antenna
system. In the case of positive slope (worst case) similar degradations (reference:
C/N on the Ricean channel, without tilt) were measured for the same spectrum tilts.
Conversely, for negative slope (best case), the high tilts which were possible on the
AWGN channel (e.g. 8-11 dB) were no longer applicable, and figures similar to that
of positive slope are to be adopted.
In conclusion to assure both on the AWGN and Ricean channels noise margin
losses lower than 0.5 dB and 1 dB, the maximum spectrum tilt (positive or negative)
in BW should be kept lower than 2.5-3 dB and 4.5-5 dB, respectively.

5.2.     Time varying echoes

Although the DVB-T system has been designed for static reception, slowly varying
channel frequency responses are possible, due to the movement of the reflecting
objects around the receiving antenna, in particular for portable receivers. In addition,
using the 2K mode and the most rugged modulation/coding configurations, mobile
reception cannot be excluded, even if the system does not include time interleaving
to overcome short-term signal fluctuations.
The system has been tested in the presence of single and multiple echoes affected
by Doppler frequency shifts. The channel response is therefore characterised by
notches shifting in the frequency domain at constant rate, producing amplitude and
phase variations on the OFDM constellations. The total received power remains
constant, while in mobile reception the power can fluctuate rapidly.
The tests identified the maximum Doppler shift f d (peak-to-peak) acceptable by the
system, for a noise margin loss C/N =4 dB 4 (reference: C/N on the relevant channel
with fd=0 Hz). Two echo configurations have been tested, with echoes within the
guard interval:
 single echo; delay 0.9; Doppler: 0 (main path), 2fd ; echo attenuation: 0 or 3 or
   5 or 10 dB;
 4 paths; delays: =0 (main path), Doppler: +fd, -fd , +fd, -fd;
   attenuation of each echo with respect to the main path of 0, 8.77 dB
   (corresponding to a total C/I=4 dB) or 14.77 dB (corresponding to a total C/I=10
   dB).
Table 6 shows the main results of these tests.

     TABLE 6 - Maximum Doppler shift fd for a noise margin loss C/N=4 dB
                                                                         fd [Hz]
    modulation                                      single echo                           multiple echoes
   and code rate           mode          C/I=0     C/I=3    C/I=5      C/I=10      C/I=0     C/I=8.8 C/I=14.8
                                                                                            (total: 4) (total 10)
 QPSK 1/2 (=1/8)            2K           150       160       >210      >210        115        280       >425
                                                                                   (320*)
                             8K           24         50        75       >210         27         70        >425
     16QAM 3/4               2K          23(**)      58        88        165         37         80        155
      (=1/32)                                                                     (120*)
                             8K            5         15        21        40          10         19         45
     64QAM 2/3          2K (=1/4)        14         19        33        95          15         28        100
                                                                                   (120*)
                        8K (=1/8)         4         8         10        23           4        8.5         21
Note (*): computer simulation results
Note (**):=0.8 was used, since the system could not operate at =0.9 even with
fd=0

The performances with a single echo and with multiple echoes are comparable and,
as expected, the 2K configuration is about four times faster than the 8K
configuration, because of its shorter symbol duration. The maximum acceptable
Doppler shift depends heavily on the echo amplitudes.
In the presence of 0 dB echoes, the 64QAM 2/3 8K mode can follow only slow
channel variations (few Hertz), while the QPSK 1/2 2K mode can track Dopplers
higher than 100 Hz, typical of mobile reception.
It should be noted that the simulation results predict Doppler tracking capabilities for
the DVB-T system which are 4 to 6 times higher than the measured results. This is
due to the difference of the algorithms implemented in the dTTb system, that must
be able to handle also other channel distortions (e.g. narrow band interferences)
which are not considered in the simulations. Further studies are ongoing to identify

4 Since the curves of C/N versus f are very steep, the choice DC/N=4 dB is not crucial for the results
                                   d
of the tests.
  optimised receiver algorithms.
  6.     PERFORMANCE WITH INTERFERENCES

  6.1.    DIGITAL TV UNWANTED
  The co-channel (CCI) protection ratio PR is defined as the carrier-to-interference
  power ratio producing a noise margin loss C/N=###. Table 7 gives the measured
  co-channel Protection Ratios of the three modulations under test, in 2K, =1/4
  mode, on UHF channel 28.
  Since a single OFDM modulator was available at RAI, the interfering signal was
  derived from the useful signal, delayed by 360 s (more than the total OFDM symbol
  duration of 280 s) and with a frequency shift f (with respect to the centre of the
  RF channel). The frequency shift was set at 17 different values (between -8 kHz and
  +8kHz, step 1 kHz), and the worst PR was identified.
  When other DVB-T modulators will become available, further tests will be carried out
  with independent interfering signals, to validate and complete these preliminary
  results.

       TABLE 7: Co-channel Protection Ratios for DVB-T interfered by DVB-T
                        (2K mode, preliminary results)
                            Modulation     Code rate          PR
                              QPSK             1/2            5.1
                             16-QAM            3/4           14.5
                             64-QAM            2/3           19.5

  The results show that the digital interference from an uncorrelated DVB-T signal has
  similar effect as a Gaussian noise of equal power in the receiver bandwidth.

  6.2.    ANALOGUE TV UNWANTED
  A number of co-channel protection ratio measurements with DVB-T interfered by
  analogue TV have been carried out by BBC (PAL-I), by CCETT (L/SECAM) and by
  RAI (PAL-G). The measured results should be regarded only as provisional figures
  at the moment, as the equipment being tested is only first generation equipment.
  Table 8 summarises the laboratory test results of co-channel protection ratios of the
  DVB-T signals interfered by unwanted analogue TV signals, modulated by 75%
  colour bars still picture. The measured protection ratios are the ratio of the mean
  power of the wanted DVB-T signal to the peak sync. power of the unwanted
  analogue signal. The results are measured without noise, and at a bit-error-rate of
  2.10-4 after the Viterbi decoder, except where otherwise indicated.
  The columns L and U have the following meanings: L corresponds to the lower part
  of the measurement envelope; U corresponds to the upper part of the measurement
  envelope. It should be noted that the lower part of the measurement envelope
  corresponds to better performance.
  The following conditions apply to the respective unwanted analogue signals and
  wanted DVB-T signals:
  1. PAL-I, with FM sound unmodulated, with NICAM. DVB-T signal: guard interval
  1/32
  2. PAL-G, FM sound modulated by 1 kHz tone, with NICAM. DVB-T signal: guard
  interval 1/4
  3. L/SECAM, AM sound unmodulated, no NICAM. DVB-T signal: guard interval 1/4

  Table 8 Co-channel protection ratio (dB) for DVB-T interfered by analogue TV
  signals
Wanted     Inter-                               DVB-T MODE
signal      fering          QPSK, r=1/2                   16-QAM, r=3/4                 64-QAM, r=2/3
            signal         8K         2K                 8K         2K                 8K          2K
                         L    U     L    U             L     U    L     U            L    U    L                U
BBC*        PAL-I      N/A N/A -11       -9          N/A N/A -4        -0.5        N/A N/A    -1.5              +1
dTTb       PAL-G        -2    +4   -3    +4           -2    +4   -2     +5         +2     +6   +3               +6
dTTb       SECAM       N/A N/A N/A N/A               N/A N/A N/A N/A                -5    +5   +5               +8
 Note (*): failure point measurements (probably these measurements give slightly
 lower protection ratios than BER=2.10-4 measurements). The channel estimation in
 the BBC modem uses temporal filtering, which gives better measurements on static
 channels, but worse temporal performance.

 The protection ratios show a periodicity which corresponds to the DVB-T carrier
 spacing (i.e. 4.4 kHz in the 2K mode and 1.1 kHz in the 8K mode). The amplitude of
 this variation is indicated in the table by the columns “L” and “U” in the results. In
 some cases a more random shape is observed. If in the future, with more mature
 technology, the periodic shape is confirmed, it could be exploited to optimise the
 required protection ratios in the service area, using precision offsets.
 The figures for 2K and 8K are not significantly different, except for the SECAM
 measurements. However in this case, the AM sound carrier was unmodulated, and
 no NICAM signal was used.
 In the dTTb demonstrator, the results are much less dependent on the DVB-T mode
 than expected. The reason for this is still being investigated, but the performance in
 the 16-QAM and QPSK modes could be expected to improve following further
 hardware developments.
 The 64-QAM mode is identical to the mode M3 in the draft ITU-R WP11C
 recommendation [XYZ] (TG11-3). The measured protection ratios for this mode are
 very close to the figures currently in this draft recommendation.

 Very significant performance degradation has been observed when the analogue
 vision carrier is close to a DVB-T continual pilot. With 64-QAM a degradation of 18-
 21 dB has been observed when this effect occurs. The results in Table 1 have
 excluded this effect, which is not thought to be fundamental to the DVB-T system.
 However, receiver designers should make any effort to improve the equipment with
 respect to these artefacts (5). In the Informative Annex some more technical
 background to this problem is given.

 7.       SENSITIVITY TO TRANSMITTER NON-LINEARITY
 The sensitivity of a digital system to non-linear distortions is a key parameter to
 determine the possibility to exploit at best the RF power of the transmitters, which
 are intrinsically non-linear. The OFDM modulation has a quasi-Gaussian amplitude
 distribution, therefore a non-linear high power amplifier (HPA) causes peak clipping
 and intermodulation among the various carriers, with the following effects:
  in-band interference generation, with noise-like distribution, affecting all the
     OFDM constellations (while the constellation shape distortion, which takes place
     in single carrier systems, does not occur)
  out-of-band spectrum "shoulders" regeneration, with interference effect on the
     adjacent channels

 While the levels of the OFDM spectrum shoulders are directly dependent on the
 particular HPA technology, on its back-off (power reduction with respect to a

 5 Although the frequency position of the DVB-T signal could be carefully selected in order to avoid
 interactions between the vision carrier and the continual pilots, this could be a significant constraint for
 planning, and would not solve the problem of man-made CW interferences that can be occasionally
 observed in the service area.
reference level, e.g., the saturation level) and on the use of HPA non-linearity pre-
correctors, in first approximation the noise margin loss C/N only depends on the
spectrum shoulders and on the system ruggedness against noise. In order to
demonstrate this, the dTTb system has been tested at various levels of HPA out-put
power levels and C/N was measured versus the spectrum shoulders, defined as the
power density ratio [dB] between the in-band and out-of-band 6 maxima of the
spectrum. It should be noted that the evaluation of the shoulders on a spectrum
analyser is affected by an uncertainty of at least 1 dB.
Table 9 gives the test results achieved with a 50 Watt solid state HPA on UHF
channel 43, limited to the 2K mode, since the 8K mode gave the same results (within
0.1 dB). In the back-off determination, the reference HPA power is chosen arbitrarily
at the level corresponding to a shoulder level of about 20 dB.



  TABLE 9 -Noise margin loss C/N [dB] versus spectrum shoulders SS [dB]
                     (reference: C/N in IF(tx)- RF(rx) configuration)
                                              Noise margin loss C/N [dB]
output back-off [dB]    5.9      5.2        4       3.1      2.4      1.7  1   0.2                    0 (**)
      SS[dB]             38       37       32        29      26       24  22.5 21                      20
modulation & code
QPSK 1/2 (=1/8)           0       0 0    0.1     0.1                      0.1       0.1      0.2      0.2
                          (0*)    (0*)
                                    (0*)  (0*)    (0*)                     (0*)     (0.1*)   (0.1*)   (0.1*)
    16QAM 3/4              0       0 0    0.1     0.2                       0.3       0.5      0.9     1.1
      (=1/32)            (0*)   (0.1*) (0.1*) (0.3*)
                                    (0*)                                  (0.5*)    (0.7*)   (1.0*)   (1.3*)
64QAM 2/3 (=1/4)  0.2     0.2    0.3     0.5     0.8                      1.3       2.0      3.9       -
                   (0*)   (0.1*) (0.2*) (0.5*) (1.0*)                     (1.7*)    (2.6*)   (4.5*)
Note (*) from approximated theory (see below); note (**):                  back-off reference
arbitrarily chosen

The results are in good agreement with the approximated theory which assumes that
the in-band intermodulation is Gaussian noise-like, with a C/I equal to the spectrum
shoulder.
Under this hypothesis, the degradation produced by the non-linearity is         = -
1/(-1 --1), where  is the C/N required by the system without non-linear distortion
and  is the shoulder level (natural numbers). The theoretical results in Table 9 are
obtained by this formula, with C/N=4.8, 14.0 and 19.1 dB, for QPSK, 16QAM and
64QAM, respectively.

From these non-linearity tests it is possible to deduce that, to achieve a degradation
lower than 0.2 dB, the spectrum shoulders should be kept at least 13 dB over the
C/N required by the modulation/coding scheme to be adopted.

9.      CONCLUSIONS

This contribution has reported the laboratory test results obtained by RAI on the
dTTb demonstrator, compliant with the DVB-T specification. The tests investigated
the ruggedness of the OFDM system in presence of a variety of typical channel
distortions, such as Gaussian and impulsive noise, tuner phase noise, echoes with
various delays and amplitudes, Doppler frequency shifts, interference from an other

6 measured around 500 kHz from the edge of the main lobe; the worst case between the upper and
lower shoulders is used
DVB-T signal and from analogue PAL/SECAM TV signals, transmitter non-linearity.
The test results have been largely in line with the performance predicted by
computer simulations, and allow to gain an important insight on the characteristics of
the sophisticated modulation, channel coding and equalisation techniques of the
DVB-T system. Nevertheless the equipment being tested at the moment is only a
first generation of equipment, and consequently the tests allowed to identify few
channel configurations in which the performance was not as good as expected. On
the basis to the test results, further activity is ongoing to improve the receiver
algorithms. When a variety of DVB-T compliant receivers will become available, the
test results will be compared to identify a “reference receiver model”, which should
become the basis for radio-frequency planning and service coverage predictions.

10.    REFERENCES

1. ETSI, "Digital broadcasting systems for television, sound and data services;
   framing structure, channel coding and modulation for digital terrestrial television",
   ETS 300744, 1997

2. Moller: “COFDM and the choice of parameters for DVB-T”, Proceedings of the
   20th International Television Symposium, Montreux 1997

3. Stott: “Explaining some of the magic of COFDM”, Proceedings of the 20th
   International Television Symposium, Montreux 1997

4. Oliphant: “ VALIDATE - verifying the European specification for digital terrestrial
   TV and preparing the launch of services”, Proceedings of the 20th International
   Television Symposium, Montreux 1997

5. Morello, V.Mignone, M.Visintin “A High SFN coverage algorithm for DVB-T
   receivers” to be published in the Proceedings of the International Broadcasting
   Convention, IBC’97, Amsterdam

11. ANNEX 1:                   AN OVERVIEW OF THE DVB-T
SYSTEM
The DVB systems designed for the various media show a great commonality of sub-
systems, including video and sound coding, multiplexing, error protection and
channel coding. The only sub-system which is optimised for each specific channel
characteristic (satellite, cable or terrestrial) is the “channel adapter”, including the
channel coder for error protection and the modulator. The main parameters of the
DVB-T system are described in [2], while the basic concepts of C-OFDM are
explained in [3].
The conceptual block diagram of the DVB systems is shown in Fig. A1.
  Source coders Programme M UX

      Video                                                                                          (*)        (**)
                                                  M UX                           Outer                       Inner
                             1                                                               Inner
                                                Adaptation        Outer          Inter-                     Inter-
      Audio          M UX                                                                 Convolutional                  M odulator
                             2                     &              Code          leaver                      leaver
                                   M UX                                                     Coder
                                                 Energy         RS(204,188)
       Data                                                                    (I=12)     (1/2, ...,7/8)
                             n                  Dispersal
     SI
                            Transport M UX                                                (*) absent in DVB-C    (**) only in DVB-T
  Source coding         M PEG-TS multiplexing                    Outer adapter                          Channel adapter
                  (common sub-systems)                       (common sub-systems)              (optimised to specific channels)
                         Fig.A1: Basic block diagram of the DVB Systems

The DVB systems are based on MPEG-2 vision and sound coding. The MP@ML
(Main Profile at Main Level) image coding algorithm is adopted, operating at bit-rates
up to 15 Mbit/s, but the introduction of higher MPEG-2 profiles and levels potentially
could allow for future evolution towards HDTV. The MPEG-2 Transport Stream (TS)
Multiplexing is adopted to merge in a single transmission stream a large number of
video, audio and data services. The MPEG transport packets have 188 bytes length
and are delimited by a sync byte.
The outer adapter (Fig.A1), common to all the DVB systems, provides signal
randomisation and a basic level of error protection by a Reed-Solomon outer code
RS(204,188), with correcting capability of T=8 random byte-errors. This error
                                                          .  -4
correction scheme provides, for an input BER of about 2 10 (independent errors), a
Quasi Error Free (QEF) quality target, i.e., less than one error-event per
transmission hour at the input of the MPEG-2 demultiplexer in the receiver.
To overcome the problem of the burst error statistic after Viterbi decoding, a
convolutional interleaving process (depth I=12 bytes) is applied, which multiplies the
burst-error correcting capability of the RS code by a factor of 12.

The DVB-T channel adapter, providing convolutional inner coding, inner interleaving
and modulation, allows to adapt the digital signals to the terrestrial channel
characteristics. It is optimised for 8 MHz channels (European UHF channellisation),
but it can be easily adapted to 7 MHz and 6 MHz channels by adjusting the receiver
sampling frequency.
The DVB-T system has been designed in order to cope with short “natural” echoes
due to multipath propagation, as well as with relatively long “artificial” echoes due to
self-interference occurring in SFNs. The system also provides good protection
against high levels of interference emanating from PAL/SECAM TV services. These
characteristics are achieved by using an OFDM modulation system associated with
convolutional error correcting coding [3], and by separating adjacent OFDM symbols
by means of a “guard interval”. Two modes of operation are defined: a “2K mode”
with guard intervals up to 56 s and a “8K mode” with guard intervals up to 224 s.
The “2K mode” is suitable for single transmitter operation and for “dense” SFN
networks with limited transmitter distances, of the order of 10 to 20 Km. The “8K
mode” can be used both for single transmitter operation and for large SFN networks,
with transmitter distances of the order of 40 to 80 Km.
The system allows different levels of QAM modulation (4, 16 and 64) and different
convolutional code rates (1/2, 2/3, 3/4, 5/6 or 7/8) to be used to trade bit rate versus
ruggedness.
The system also allows two level hierarchical channel coding and modulation,
including uniform and multi-resolution constellations, to improve the ruggedness
against channel impairments of part of the transmitted bit-stream. A low-bit-rate
  programme service can thus be received under severe reception conditions, while
  the other programmes in the multiplex can be correctly decoded only under less
  critical conditions.
  The transmitted signal is organised in “frames” of 68 OFDM “symbols”. Each OFDM
  symbol is constituted by a set of K carriers (1705 for 2K and 6817 for 8K) with a
  minimum frequency separation to avoid inter-carrier interference (4464 Hz for 2K
  and 1116 Hz for 8K) and transmitted simultaneously with a symbol duration T s. The
  symbol is composed of two parts: a “useful” part with duration Tu (224 s for 2K,
  896 s for 8K), and a “guard interval” with a duration T g (where Tg/Tu can be 1/4,
  1/8, 1/16 or 1/32). Not all of the carriers are modulated with data, since some of
  them (the “pilot carriers” or “pilots”) are used to transmit reference information
  required by the receiver for synchronisation (frame, frequency, phase), channel
  estimation, transmission mode identification. There are three types of pilots:
  scattered, continual, TPS (transmission parameter signalling).
  The spacing between first and last carriers of the spectrum is 7.61 MHz,
  approximately corresponding also to the total spectrum occupation because of the
  steep roll-off of the OFDM signals.
  11.1. System capacity and Simulated performance

  In principle the C/N required by the DVB-T system is a random variable depending
  on the channel response and on the adopted transmission mode (coding rate,
  modulation, guard interval). Since a statistical characterisation of the system in the
  various reception environments is too complex, only two “representative” channels
  have been chosen in the specification (see Annex A and B of the ETSI specification)
  for computer simulations, one for the fixed reception with directive antenna (F1,
  Ricean channel) and one for portable reception (P1, Rayleigh channel). It should be
  taken into account that these channels are not a worse case, since a single 0 dB
  echo introduces higher degradations. It should also be noted that these channels
  include only relatively short natural echoes (up to 5.4 s), well within the guard
  interval, and do not represent a SFN situation. When an echo delay exceeds the
  correct equalisation interval TF (corresponding usually to Tg), a steep transition
  occurs and the echo effect becomes similar to that of an uncorrelated Gaussian
  noise interference.
  The required C/N (computer simulation results, ideal synchronisation and channel
  estimation, excluding any implementation margin) for non-hierarchical transmission
  for all combinations of coding rates and modulation types is given in Table A1, taken
  from Annex A of the DVB-T specification [1]. The net bit rates after Reed-Solomon
  decoder are also listed.

                                         TABLE A1
            REQUIRED C/N FOR NON-HIERARCHICAL TRANSMISSION TO ACHIEVE A BER = 2
            10-4 AFTER THE VITERBI DECODER FOR ALL COMBINATIONS OF CODING RATES
              AND MODULATION TYPES. THE NET BITRATES AFTER THE REED-SOLOMON
                                  DECODER ARE ALSO LISTED.
                       Required C/N for
                          . -4
                    BER=2 10 after Viterbi                            Bit rate (Mbit/s)
                   QEF after Reed-Solomon
                     (IDEAL RECEIVER)
Modu-    Code Gaussian Ricean        Rayleigh
lation   rate channel channel        channel    Tg/Tu = 1/4 Tg/Tu = 1/8 Tg/Tu = 1/16 Tg/Tu = 1/32
                          (F1)          (P1)
QPSK      1/2   3.1        3.6           5.4       4.98        5.53          5.85          6.03
QPSK      2/3   4.9        5.7           8.4       6.64        7.37          7.81          8.04
QPSK      3/4   5.9        6.8          10.7       7.46        8.29          8.78          9.05
QPSK      5/6   6.9        8.0          13.1       8.29        9.22          9.76         10.05
QPSK      7/8   7.7        8.7          16.3       8.71        9.68         10.25         10.56
16-QAM   1/2     8.8      9.6        11.2          9.95     11.06       11.71       12.06
16-QAM   2/3    11.1     11.6        14.2         13.27     14.75       15.61       16.09
16-QAM   3/4    12.5     13.0        16.7         14.93     16.59       17.56       18.10
16-QAM   5/6    13.5     14.4        19.3         16.59     18.43       19.52       20.11
16-QAM   7/8    13.9     15.0        22.8         17.42     19.35       20.49       21.11
64-QAM   1/2    14.4     14.7        16.0         14.93     16.59       17.56       18.10
64-QAM   2/3    16.5     17.1        19.3         19.91     22.12       23.42       24.13
64-QAM   3/4    18.0     18.6        21.7         22.39     24.88       26.35       27.14
64-QAM   5/6    19.3     20.0        25.3         24.88     27.65       29.27       30.16
64-QAM   7/8    20.1     21.0        27.9         26.13     29.03       30.74       31.67

  As derived by a theoretical study, for Tu/Tg=1/4 the system implementation loss due
  to non-ideal channel estimation and equalisation in the receiver is in the range 1.6 to
  2.1 dB (including 0.3 dB loss due to pilot boosting), depending on the adopted
  algorithms (indicated as 2-D or 1-D). This C/N loss could be easily reduced by
  averaging in time the channel estimation, in order to filter-out the noise components,
  but this would cause a reduction of the channel tracking capability in the presence of
  time varying channels.

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:93
posted:2/25/2010
language:English
pages:18
Description: PERFORMANCE ASSESSMENT OF A DVB-T TELEVISION SYSTEM