MOBILE DVB-T RECEPTION: QUALITY OF STREAMING OVER IP OF AUDIOVISUAL SERVICES

MOBILE DVB-T RECEPTION: QUALITY OF STREAMING OVER IP OF AUDIOVISUAL SERVICES A. Bertella, M. Rossini, P. Sunna, L. Vignaroli RAI Radiotelevisione Ital...
2 downloads 0 Views 434KB Size
MOBILE DVB-T RECEPTION: QUALITY OF STREAMING OVER IP OF AUDIOVISUAL SERVICES A. Bertella, M. Rossini, P. Sunna, L. Vignaroli RAI Radiotelevisione Italiana CRIT − Research & Technology Innovation Centre, Italy ABSTRACT DVB-T mobile reception has been evaluated through laboratory tests and field trials with different modulation schemes; among these the more robust one is QPSK, rate 1/2, Tg 1/32, FFT 2k, that provides the best performance in terms of C/N but the least bit-rate. In fact the channel capacity drops at 6 Mbps, usable only to transmit an MPEG-2 program with a near good quality. The number of available programmes in mobile reception, keeping the same robustness, could be increased transmitting several audio/video streams encoded at about 500 kbps in CIF format to obtain a good quality on small LCD display; in this case the audio/video stream would be encapsulated over IP, using the Multiprotocol Encapsulation, and transmitted on DBV-T. The paper describes the laboratory tests performed to verify the feasibility of using DVB-T to convey IP media services; a survey also has been carried out using a van to collect data regarding quality of service information and subjective comparison between MPEG-2 and IP-over-DVB-T reception.

INTRODUCTION RAI, the Italian public broadcaster, is directly involved in the activity on digital terrestrial television since 1998, when an experimental plan based on several DTT pilot trials was presented to the Ministry of Communications (1). The first pilot trials started in Turin in February 1998 by RAI−CRIT, with the primary goal of assessing, in the field, the technical performance of DVB-T and identifying the best service configuration to meet a large audience interest. DVB-T (2) has been developed for fixed reception, with a roof-top antenna, and for portable indoor – outdoor reception (3) but not for mobile reception. However, some of the operating modes allow the receiver to track the faster channel variations associated with mobile reception providing an acceptable service quality. The feasibility of mobile reception was demonstrated with simulations, laboratory tests and field trials carried out in the last few years by RAI-CRIT and other partners involved in the MOTIVATE project (4). Field trials performed by RAI-CRIT using 16QAM, rate ½ modulation scheme (5) highlighted that the mobile reception is possible but the coverage probability, at least with the power transmitted on channel 28 on Turin, was not satisfactory. Hence, a more robust modulation scheme (QPSK, rate ½) was adopted to increase the coverage probability. In this case, the useful bit rate became very low and so only one good quality MPEG-2 programme can be transmitted. A key issue for a successfully introduction of the DVB-T services, for both fixed and mobile reception (buses, taxi, trains, cars) is the content and the attractiveness of the offer provided

by the new terrestrial channels in competition with the satellite digital services already in operation. In particular for mobile reception, to increase the number of programmes several IP audio/video streams encoded at low bit-rate (about 500 kbps), in CIF format to obtain a good quality on small LCD display, have been encapsulated over DVB Transport Stream using the Multiprotocol Encapsulation and transmitted on DBV-T channel. A laboratory survey and field trials have been carried out in Turin by RAI-CRIT and the major results and conclusions are reported in this paper. SERVICE CONFIGURATION DVB Platform In February 1998 the RAI Research and Technology Innovation Centre (CRIT) in Turin started up the first experimental DVB-T transmissions in Italy. The digital platform was further enriched to include a Data Broadcasting Centre providing multimedia applications and data broadcasting services. The source encoding and multiplexing chain consists of six statistical multiplexing encoders, one re-multiplexer to provide EPG and MHP applications, and an SI-PSI generator. The multiplexer receives the Packetised Elementary Stream (PES) from the encoders and the multimedia data from the Data Broadcasting Centre, and generates up to four different and independent Transport Streams (TS). At the moment two TS outputs are used to deliver two DVB-T bouquets of services and applications. A PC controls the parameters of the encoding and multiplexing chain as shown in Figure 1. The Transport Streams are transferred to the main transmitter − situated on a hill 500 m a.g.l., 5 km away from the CRIT − by means of a 45 Mbps tributary of an SDH digital radio link. Two ATM Network Adapters are used to convert the multiplexer outputs to the transmission bit-rate.

Figure 1 – DVB-T encoding and multiplexing chain

At the transmitter site the two digital streams are re- converted to the original bit-rates and sent to the COFDM modulators of two independent UHF transmission chains, one operating on channel 28 (530 MHz) and the other one on channel 66 (834 MHz). Table 1 gives the modulation and radio frequency parameters (RF) for both chains. Ch. Modulation 28 66

QPSK - 16QAM 2k; 1/2; 1/32 64QAM 8k; 2/3; 1/32

Bit-rate (Mbps)

TX (W) ERP (W)

Pol.

6.03 ÷ 12.06

40

300

V

24.13

350

2000

H

Table 1 – Modulation and RF Parameters The modulation schemes used on channel 28, in both QPSK and 16QAM, are very robust and therefore appropriate for mobile reception trials and reliable indoor portable reception. Of course the robustness of the modulation scheme results in a poor available useful bit rate. Hence, the number of MPEG-2 programmes that can be reasonably transmitted for mobile reception is between one in QPSK, rate ½, and up to three in 16 QAM, rate ½.

The modulation scheme used on ch. 66 is instead very efficient in terms of capacity and allows transmitting up to 6 TV programmes. This configuration, which provides a wide and diversified offer of services, is then particularly oriented to fixed and portable reception in urban areas. The two UHF transmitters on channels 28 and 66 use different antenna systems, which are installed on the same mast; vertical polarisation is more appropriate for mobile services. Fig.2 shows the encoding and multiplexing chain installed at the RAI CRIT premises. Since the radio frequency spectrum adjacent to channels 28 and 66 is occupied by analogue TV services, both DVB-T transmitting chains include suitable RF output filters to guarantee adequate protection versus the analogue services (DVB-T vs. PAL). The same RF filters ensure adequate rejection on the taboo channels, relevant to both channels 28 and 66, which are currently occupied by analogue services in the same area.

Figure - 2 – DVB-T control room at RAI CRIT

DVB-T MOBILE RECEPTION FIELD TRIALS USING MPEG-2 DVB-T, although primarily developed for fixed reception − with roof-top directive antenna − and portable reception, may also be used for mobile TV services (4). In year 2000 some mobile reception tests on Ch. 28 have been carried out in urban and suburban areas, adopting 16QAM, with code rate 1/2, guard interval 1/32 and FFT mode 2k as modulation parameters. The results (for a complete report see (5)) − supported by further laboratory investigations − were encouraging and confirmed the feasibility of mobile DVB-T services provided that a robust modulation scheme as well as a suitable receiving system, e.g. based on space diversity, are adopted in the car. Anyway, the overall coverage probability was not greater than 80% of the locations and so, at least with the ERP (300 W) transmitted on channel 28, the result was not satisfactory. Therefore a more robust modulation scheme was used during the field trials carried out in Figure 3 – Mobile reception field trials in 2002, to increase the coverage probability in Turin mobile reception, since at the moment receivers specially designed for mobile reception (based on antenna diversity and with a faster channel estimation) are not on the market. The modulation scheme used for the Turin test bed was QPSK with code rate ½, guard interval 1/32 and FFT mode 2k. The useful bit rate is about 6 Mbps and so only one good

quality programme in MPEG-2 format, together with some radio channels (not more than 5-6 at 192 kbps), can be transmitted. The transmitter characteristics were the same above reported. More than 150 km were covered in urban and sub urban area on a car equipped with a λ/4 whip antenna mounted on the roof and a commercial receiver directly powered with the car’s 12V. The receiver used during the tests (Figure 3) was a digital terrestrial receiver prototype specially designed for indoor portable and mobile reception. It is equipped with 7’’ LCD display in 16:9 format. If the signal strength drops for a short time, for example in indoor portable reception when people walking around antenna, the receiver loses the signal but it re-locks very quickly. This is very important especially in mobile reception where the interruptions could occur very frequently. The purpose of the trial was to evaluate, with a subjective criterion, the mobile reception service coverage. The results of the mobile reception survey are reported in the map of the Turin area shown in Figure 4. In urban area the reception was satisfactory in about the 95% (green line in Figure 4) of the locations; including the suburban area (motorway around Turin and the airport) the coverage decreases to about 85%. The test showed that the reception was very reliable (about 100% of locations) in almost every road of the town and only in particular place (under large bridge made of concrete or along small roads surrounded by very high buildings) there were some brief video artefacts. Nevertheless, owing to the dramatic useful bit-rate reduction that drops at 6 Mbps, the offer of audio/video programmes could be less attractive for the users. In a context where the radio spectrum is extremely busy and where the availability of a UHF channel is so complicated and expensive, transmitting only one program, even if receivable in mobile reception, is definitely not acceptable for broadcasters. To Figure 4 – Mobile reception coverage in Turin overcome this problem, keeping into (QPSK, rate1/2, guard interval 1/32, FFT 2k) account that for mobile reception small Green: good reception; Red: no service LCDs are usually adopted (for example mounted in the cars head-rests) several audio/video streams have been encoded at low bitrate (i.e. about 500 kbps) in CIF format (352X288), IP encapsulated over DVB Transport Stream, using the Multiprotocol Encapsulation, and transmitted on DVB-T channel. The experiment aimed at verifying the feasibility of using DVB-T to deliver IP media services and to compare MPEG-2 versus IP-over DVB-T, in mobile reception.

MEASURMENT SET UP For the mobile reception field trials one audio/video MPEG-2 program and four IP audio/video streams were transmitted on channel 28 by means of the encoding and multiplexing chain above detailed. The adopted modulation scheme and the transmission parameters are reported in Table 2. Ch.

Modulation

Bit-rate (Mbps)

TX (W)

ERP (W)

Pol.

28

QPSK 2k; 1/2; 1/32

6.03

40

300

V

Table 2 - Ch. 28 modulation and RF parameter

The car used during the mobile reception (Figure.5) was equipped as follow: • • • • • •

Omnidirectional antenna (G = 2 dBi) installed on the car roof. Personal Computer equipped with a DVB-T board to receive IP audio/video streams. DVB-T set-top-box to receive the MPEG-2 program. DVB-T professional receiver with TS ASI output. Transport Stream recorder. DC/AC inverter to supply all the equipment.

A whip antenna, tuned on channel 28 (530 MHz), was used as omnidirectional antenna.

Figure 5 – Van used during mobile reception field trials in Turin

All the equipment have been installed inside the car used during the field trials. A Personal Computer equipped with a DVB-T board received the DVB-T signal from the antenna and the audio/video streams were decoded using Microsoft® Windows Media™ Player and displayed on a small LCD monitor. A DVB-T set-top-box, equipped with 7’’ LCD display was used to receive the MPEG-2 program, while the TS output of a professional receiver fed a TS analyser in order to record the Transport Stream for further laboratory investigations.

TX EREMO 0

4 Km

Two DC/AC inverters were installed on the car equipped for the mobile reception; the inverters supply (220V, 50 Hz) all the equipment used during the field trials. Keeping into account the coverage area of the Torino-Eremo DVB-T transmitter and previous tests carried out in Turin, we choose a critical urban route (see Figure 6) in order to compare the MPEG-2 program with the audio/video streams.

Figure 6 – Urban route used during mobile reception field trials in Turin

IP AUDIO/VIDEO STREAMS GENERATION In a typical architecture of Webcast Streaming platform (Figure 7) the input (analog or digital) of the audio/video capture card is processed by a real time software encoder running on a personal computer providing an audio/video stream on the LAN. The data flow is then distributed through the LAN to a streaming server. Client machines, connected to the network, equipped with the right streaming player, request the data streams directly from the server through a virtual publishing point to which the stream are assigned. The server provides several client management controls, reporting functions and can operate both in unicast and/or multicast mode. Multicasting [6] Figure 7 –Architecture of Webcasting follows a push model of communications, similar streaming service to a radio or television broadcasting where a user receives a program by tuning the station he wants to receive. In the case of multicasting, the user simply instructs the computer's network card to listen to a particular IP address for the multicast. The computer originating the multicast flow does not need to know which user has decided to receive it. Today audio/video streams are generally transmitted over the Internet through unicast. With unicast, the server produces individual data streams, one for each player requesting it. This is different from multicast, where a single data stream is transmitted only to network routers that show current requests for the data. The architecture used during the tests for Streaming Multimedia content over DVB-T is reported in Figure 8. The streaming server used is based on Microsoft® Windows Media™ Technologies that allow creating, delivering, and playing streaming media files and live Audio/Video contents. The Streaming Server must be able to perform multicast streaming over a connectionless protocol. In this way a back channel link is not necessary and the scenario of a pure broadcast transmission is then possible.

Figure 8 – Architecture used for Multimedia Streaming over DVB-T

The IP multicast stream is encapsulated over a DVB transport stream adopting the Multi Protocol Encapsulation profile (see ETSI European Standard EN 301 192 (7), (8)) by means of an IP-DVB gateway. AUDIO/VIDEO STREAMS ENCODING PARAMETERS Transmitted test sequences, containing different types of audio/video content (News, Sport, Fiction) were real time encoded with Microsoft® Windows Media™ Encoder according to parameters in Table 3.

Bit-rate (Kbps)

Frame rate (Hz)

Buffer (s)

Key Frame Distance (s)

Qual.

Form.

V: 494 A: 32

25

5

1 -3 -5

50

CIF

Table 3 – Encoding parameters

A brief description of each parameter is reported in the following: •

The frame rate specifies the maximum frames per second; the actual frame rate achieved generally is lower, depending on factors such as the video quality setting and the available bandwidth.



The buffer size specifies the amount of time that content is stored before encoding starts; the larger is the buffer, the better is the quality but the longer is the delay before Windows Media Player runs the content.



The key frame interval defines the number of seconds that separate key frames; all frames between key frames are “delta” frames, i.e. contain only the differences from the key frame. The player needs a key frame before starting decoding and displaying.



The image quality can be tuned using a video quality slider to emphasise either smoothness or clarity of images; increasing the clarity the video becomes less smooth and the frame rate decreases. This value would be chosen according to the kind of the content.

Regarding the choices indicated in Table 3, the following considerations have to be kept into account: •

the bit-rate and the format resolution assure a good quality of vision on small size screen of about 8’’.



The buffer length was chosen to contain at least 1 key frame to avoid the stream to be interrupted for a period too long when transmission error occurs.



Different values of key frame distances were chosen to understand the player ability to recover the audio/video content in case of packet losses due to the DVB-T channel.



For the quality, an intermediate value was selected as a good compromise given the variability of the sequences content.

EXPERIMENTAL RESULTS For the tests, the decoded MPEG-2 programme was compared with the same audio/video material decoded with Microsoft® Windows Media™ Player; an expert viewing was carried out on 8’’ LCD monitors. The results of the subjective evaluation are reported below: •

if a short interruption of the DVB-T signal occurs, the quality of the audio/video content over IP appears better than the MPEG-2 quality; more annoying blockiness affects the MPEG-2 signal.



If the interruptions introduced by the DVB-T channel are longer, both the systems crashes; the MPEG-2 IRD looses the signal and Microsoft® Windows Media™ Player rebuffers more than once according to the interruption length; the greater is the distance between the key frames, the longer will be the waiting, because the player needs a key frame before start decoding.

CONCLUSIONS In this paper DVB-T has been evaluated as a possible candidate for the delivery of IP-based multimedia services to mobile terminals. For mobile applications the reduction of the transmission capacity necessary to increase the channel robustness would allow the transmission of only one MPEG-2 program, encoded at about 6 Mbps, to provide the users with a good quality. Adopting IP over DVB-T, much more programmes (from 10 to 12) encoded at lower bit-rate, about 500 kbps each, can be transmitted offering a good quality of vision on small size displays. According to the results reported in the article the following conclusion can be drawn: •

DVB-T can be successfully used for the delivery of mobile multimedia services.



The quality of the audio/video content over IP appears better than the MPEG-2 quality when short interruption of the DVB-T signal occurs.



More generally the approach suffers of the same limitations as the MPEG-2 reception but it allows better exploitation of the available bandwidth.



Thanks to the flexibility of the system, different bouquet scenarios can be expected: the first one considers that all the available bandwidth is assigned for several IP audio/video streams, the second one provides a single MPEG-2 program of fair quality together with some IP audio/video streams in the residual transmission capacity.

Figure 9 - Part of Data Broadcasting Centre

By adding to the system a return channel, based on new mobile technologies (i.e. GPRS, UMTS etc.), a number of enhanced features become available (for example reliable file transfer applications) allowing the development of added value services. The integration of IP and DVB technologies, combined with low bit-rate encoding techniques alternative to MPEG-2 (e.g. MPEG-4), allow the exploitation of a new range of television services and value added services to mobile terminals with narrow display size. ACKNOWLEDGEMENTS The authors would like to thank their colleagues G. Alberico, G. De Biase, P. Casagranda, B. Sacco, M. Stroppiana and M. Tabone for their contribution to the paper, for car set-up and for performing the laboratory survey and field trials. REFERENCES 1. M. Cominetti, A. Morello, R. Serafini: “Current plans for DTT implementation in a densely utilised frequency spectrum“. Montreux TV Symposium ’99. 2. ETSI: “Digital Video Broadcasting (DVB); framing structure, channel coding and modulation for digital terrestrial television”. European Standard EN 300 744. 3. CEPT: “The Chester 1997 Multilateral Coordination Agreement relating to Technical Criteria, Coordination Principles and Procedures for the Introduction of Terrestrial Digital Video Broadcasting (DVB-T)“. Chester, July 1997.

4. MOTIVATE Report: “Using DVB-T standard to deliver broadcast services to Mobile receivers”. 36th DVB-TM, January 2000 5. A. Bertella, M. Cominetti, S. Ripamonti, M. Visintin: “The RAI DTT pilot trials in Turin”. IBC 2000. Amsterdam, September 2000. 6. IETF: “Host extensions for IP multicasting”. S.E. Deering. Aug-01-1989 RFC1112. 7. ETSI: “Specification for data broadcasting”. European Standard EN 301 192. 8. ETSI: “Specification for data broadcasting; Guidelines for the use of EN 301 192”. Reference No. prTR 101 202.

Suggest Documents