What is MPEG

Introduction

At the start of the satellite broadcast era, all radio and television programs were transmitted in an analog format. Only
the last
couple of years one has started to transmit in a digital format. This was made possible by establishing digital
transmission
formats. The advantage of digital data transfer is the high quality of the picture and the low losses involved. Also,
thanks to the
used compression techniques, more programs can be transmitted over the available distribution channels.

Introduction

At the start of the satellite broadcast era, all radio and television programs were transmitted in an analog format. Only
the last couple of years one has started to transmit in a digital format. This was made possible by establishing digital
transmission formats. The advantage of digital data transfer is the high quality of the picture and the low losses
involved. Also, thanks to the used compression techniques, more programs can be transmitted over the available
distribution channels.

Digital television systems

Using digital television, the amount of data when no compression is used, is very high. For digital television, the
following sample frequencies are used according to ITU-R recommendation no. 601 :

13.5MHz for the luminance signal (Y) and 6.75MHz for both color difference signals. Using a 8-bit quantising method,
we end up with a bitstream of 216Mbit/s. The required bandwidth for a signal like this is so big, that even a satellite
transponder can not cope with it. The technique used in digitising and compressing digital broadcasts has been
developed by the Motion Picture Expert Group (MPEG). Digital Video Broadcasting (DVB) in Europe uses the MPEG-2
format. Using this format and a modem, a variety of extra services becomes available, like extensive interactive
services.

Compression converts the analog video signal in a digital signal with a bitstream varying between 2 and 15Mbit/s
(MPEG-2 video). The audio signal is compressed between 32 and 448Kbit/s. Multiplexing (MPEG-2 systems) combines
video and audio, but can also add multiple AV signals together in a single Transport Stream (TS). - Modulation takes
care of a transparent bitstream of say 38.1Mbit/s in a single 8MHz channel. For cable systems, QAM is used, for
satellite transponders QPSK is used. By using compression, it is now possible to transport more than a single channel
over either cable or satellite transponder.

Within DVB, the scrambling of the signals is standarized. Various Conditional Access (CA)-systems are offered. The
use of these enables services like pay-TV and pay per view. The DVB service information (SI) offers the possibility to
add special information to the datastream to describe the contents of the program transmitted. It enables the set top
box to configure itself and aid the viewer in finding TV or radio programs.

Digital signal processing

Audio and video signals are essentially analog signals, that is, signals of which shape, amplitude and frequency
continuously changes. Until recently, processing of these signals could only be accomplished in an analog way. The
characteristics of analog signals, when processed in electronic circuits, will be influenced. At every processing stage,
the quality of the signal degrades. Just think about copying videotapes. A copy of a copy posesses less quality than
the original tape.

By applying digital techniques, these disadvantages have come to an end. Using this technique, we can now keep the
quality of the processed signals at a constant quality and level. A digitised signal can be displayed, transported and
processed completely free of distortion.

A digital signal no longer exists of a continuously varying signal but of several individual signals. Every momentarely
signal is represented by a digital code. We call it digital signal processing when all binary representations of the
original signal are processed following the rules of digital technique.
Source coding of high quality picture- and sound signals plays an important role. One of the reasons is the high
transferbandwidth and enormous storage capacity needed for linear coded signals.

Digitising

Digitising means the conversion of an analog signal into a digital signal consisting of zeroes and ones. Converting
analog signals to digital (A/D conversion in short) is done in three steps :

 Sampling
 Quantising
 Conversion to digital numbers

To convert a digital signal back into its analog format, all three stages of the process are carried out in reverse order.
One than uses a D/A convertor.
Sampling means that the source signal is sliced into equal sections over a time period of 1 second. The audio on a
Compact Disc e.g. is divided in 44100 sections per second. Therefore, the sampling frequency used is 44.1KHz.
For a video signal, the signal is divided into 13500000 sections per second. This explains the sampling frequency of
13.5MHz. For this process, a couple of conditions apply. The most important being that the sampling frequency has to
be at least twice the highest frequency present in the source signal.
Quantising means that every section has its own scale with which the amplitude of the signal is measured. This scale
is notated in bits. For video, an 8-bit quantizing method is used. Using 8 bits, one can have a total of 256 variances of
a signal. This is more than sufficient for the human eye, since we can only recognise about 200 variances in luminance
anyway.
For audio, normally 16 bits are used. Using 16 bits, a total of 65536 different sound variances are possible. In other
words, the resolution is 65536, which is more than sufficient for the human ear.
Conversion to digital numbers means, that a measured audio value of say 32768 is not represented as a number but
as a binary value of ones and zeroes (in this example as 0111111111111111).
Following this, the digital signal is coded and given an error coding to enable us to correct errors at an later stage. The
amount of data is given in the number of bits per second. A digital signal contains a fair amount more information
than an analog signal. To be able to store all this information, datacompression is used.
DIGITIZING AND COMPRESSING AUDIO
Introduction
To digitise audio signals, a couple of different sampling frequencies (depending on the application) are used : 32KHz,
44.1KHz or 48KHz. The quantising scheme is normally 16 bits. On an audio CD, all information is registered and
therefore a lot of bits are used. This form of coding is called linear coding.
By using compressing techniques, the amount of data can be strongly reduced. These forms of datareduction are used
in e.g.
 Digital Compact Cassette (DCC)
 Mini Disc
 Digital Audio Broadcasting (DAB)
 Astra Digital Radio (ADR)
 Digital Video Broadcast (DVB)
 Digital Video Disc (DVD)

Source coding
The MPEG system committee determines the norm for the combination of a large number of coded audio- and
videosignals in a single datastream. This norm guarantees the synchronisation of audio and video and enables the
transfer of combined information by using various digital media. After having tested various applications, the MPEG
experts have established the audio coding algorithm. Depending on the application, a total of three layers with
increasing complexity and reduction capacity can be used. For a recording that can not be distinguished from the
original, this comes to :
 Layer 1, 2x 192Kbit/s
 Layer 2, 2x 126Kbit/s
 Layer 3, 2x 64Kbit/s

Important for an economical use of the number of available bits is the source coding of the signal. Source coding
means the amount of bits that are created after the A/D conversion. As an example : By sampling an audio signal with
a sampling frequency of 44.1KHz, using 16 bits per sample, an audio stream of 44100 x 16 x 2 = 1410000 bits/s is
generated.
By using intelligent algorithms which take the properties of the human ear into account, the amount of data can be
strongly reduced. Bits can be saved by redundancy, or by simply throwing away irrelevant parts of the signal. With
irrelevant, we mean those parts that the human ear does not use. In other words, frequencies that are outside our
hearing capabilities do not have to be registered.
Apart from this, there is another important dynamical effect. This is the phenomena that a very load tone masks a
weaker tone so that it can not be heard anymore. This is a very complex psycho-acoustic effect. By leaving off this
information, the total soundimpression is not effected. To calculate which parts of an audio signal can be heard or not,
the signal is divided into subbands. For Musicam for instance, the digital audio signal is divided into 32 subbands with
an equal width. In the coder, 12 subsequent samples of the subband are combined to a block to calculate the mask
effect.
Every subband is allocated a couple of bits, depending on the need. This way, no more bits than stricktly necessary
are used. Also, this way, the accuracy is as high as possible. By using this method, it is now possible to reduce the
1.4Mbit/s bitstream on a CD to just 200Kbit/s.
The most important data-reduction and coding methods for recordings are :
MUSICAM (Masking Pattern Universal Subband Integrated Coding and Multiplexing PASC (Precision Adaptive Subband
Coding).
Musicam is used in DVB, DAB and ADR, whilst PASC is used on a DCC.
DIGITIZING VIDEO IMAGES
Introduction
Digitising a video image is far more complex than for an audio signal. Because of the far higher frequencies used in
television, the datarate is much higher. For a video image, this is about 100 times than what is needed for audio.
Television pictures consist of lines. In Europe, we use 25 frames per second, each frame consisting of 625 lines. At 25
frames per second, the human eye experiences the frame changes as a flicker. For this reason, interlacing is used.
That is, every 625-line image is divided into two equal 312.5 line frames. The first frame carries the even lines, the
second carries the odd lines. The two frames combined for the complete image again.
To get the proper definition, the television signal should have a certain bandwidth. A complete picture should consist
of 530 x 400 - 212000 pixels. The required bandwidth than is 5MHz. This applies to a black and white picture.
For a color picture, another calculation applies. A PAL color pictures is made up out of 768 x 576 pixels for a complete
frame.
Color television
Color television uses the primairy colors RED, GREEN and BLUE. By adding those primairy colors, other colors,
including white, can be constructed. Those three colors are not transmitted individually by a television transmitter, but
as a luminance signal (Y) and the color difference signals R-Y and B-Y. Both color difference signals R-Y and B-Y are
also called U- and V signals, once adapted in amplitude.
The required bandwidth for the color signals is far less than for the Y-signal. For TV signals, the ratio between
luminance (Y) and color signals is given as 4:1:1. The required bandwidth of the color signal is four times less than
that of the luminance signal. In professional studios, one normally uses a ratio of 4:2:2, but due to the limited
bandwidth of a television transmitter, it can not be broadcasted in this format.
When we want to digitise such a signal, we can only do it on a component level, which means that the video signal
has to be split up.
For a sampling frequency of 13.5MHz and an eight bit quantising scheme, we get the following video bitstream :
 Y-signal : 13.5MHz x 8 bit = 108Mbit
 R-Y signal : 3.375MHz x 8 bit = 27Mbit
 B-Y signal : 3.375MHz x 8 bit = 27Mbit

which totals to 162Mbit/s. To transmit this amount of data, a very high bandwidth is required. Was the bitstream for
CD audio 1.4Mb/s, for a video signal this is about 100x higher.
MPEG VIDEO COMPRESSION
Introduction
When we have to start using digital source coding in television systems, some international agreements have to be
made. This not only applies to video, but also audio, the multiplexing of video and audio as well as other signals like
teletext.
Between 1988 and 1994 an international standard has been agreed on by MPEG, a subgroup of ISO and IEC.
The goal of MPEG was :
1. To produce a world wide standard for video and audio coding, with options for various applications
2. To define transmissions specifics that can be used for all media, including transmission and recording
3. Defining a compliance procedure by which systems can be evaluated
4. Defining a datastructure that can be used to develop encoders and decoders

The first standard agreed on in 1992 was MPEG-1, used for computers and CD-ROM. The datastream here is
1.15Mbit/s. Picture quality is comparable to VHS recorders.
In november 1994, MPEG-2 was established. This not only enables a datastream of 100Mbit/s but also it created the
possibility to have multiple programs in one single datastream.
MPEG-2 is the basis for Digital Video Disc and Digital Video Broadcast. For the European market, the DVB project has
established almost the complete system for the new generation of digital TV on both cable and satellite. This standard
not only allows data to be transported more efficiently, bit also various new serviced can be exploited. DVB has
standarized the scrambling of the signals and can add special information about things like program contents,
transmission path etc. It not only allows the set top box to configure itself but also aids the viewer in finding
programs.
MPEG-1
 Name : ISO/IEC 11172
 Bit rate : Usually 1.5Mbit/s
 Video : Resolution CIF (354 pixels * 256 lines * 25Hz)
 Audio : 64Kbit/s to 384 Kbit/s (Musicam)
 Systems : Multiplexing 1 * video + stereo audio + data
 Applications : CDI and computer games

MPEG-2
 Name : 13818
 Bit rate : Usually 2 - 15Mbit/s
 Video : Resolution ITU-R 601 (720 pixels * 576 lines * 25Hz) and HDTV
 Audio : 64Kbit/s to 384Kbit/s, stereo + surround (5 channels)
 Systems : Multiplexing video, audio, data, conditional access, multiple video signals, each with their own
timebase
 Applications : Digital TV

To reduce the number of bits in a digital TV system, reduction and compression techniques have been developed to
make this possible. It is called compression once the picture image in the decoder can be perfectly reconstructed.
Reduction will allways show a difference between an original and a decoded image.
Compression techniques
By using the properties of the human senses like eyes and ears, it is possible to apply a hardly visible datareduction
hence is acceptable for certain applications.
The basis is formed by leaving off information that can not be registered by the eye (irrelevance-reduction) because of
which not always the same information has to be transmitted when the picture contents has not changed. This is
called redundancy reduction.
In MPEG-videocompression, multiple methods are used to reduce the number of bits like :
 Compression based on Discrete Cosinus Transformation (DCT)
 Segmentation, splitting the image into blocks
 Movement Compensation
 Temporal prediction and interpolation

Adjacent pictures in a television signal are pretty much the same. Every picture is built out of 2 frames which carry
the same information. This is what we call redundancy information. There are several forms of redundancy :
 Spatial redundancy
 Temporal redundancy
 Static redundancy

To understand redundancy, a couple of agreements have been made within MPEG. A MPEG decoder has various
picture memories in which different frames for decoding are stored and out of which the original picture can be
reconstructed.
This way, the bandwidth necessary for a single analog television channel can now contain 5 - 7 digital television
channels using the MPEG-2 data compression. This technique allows a 4-hour movie to be recorded on a double sided
Digital Video Disc (DVD).
ADR / DMX DIGITAL RADIO
Digital Music Express (DMX) is a digitally coded radio signal complying to the Astra Digital Radio (ADR) specifications.
The applied technique is constructed in a way the standard 180KHz spacing of satellite audio transponders could be
used. This enables the simultaneous transmission of both analog and digital audio to assist in a fluent transition from
analog to digital. Per transponder, a maximum of 12 radio programmes can be put 'behind' the television program, or
a total of 48 radioprogrammes can be transmitted on a single transponder.
Important is, that a current analog system needs two carriers to carry both left- and right channel for stereo
transmissions in contrast to ADR and DMX where only a single carries is needed.
Firstly, the left- and right audio signal are digitised with a sampling frequency of 48KHz / 16 bits resulting in an audio
stream of 1.536Mbit/s. This has to be reduced by a factor 8 to be put in the narrow banded transmitter signal. This is
accomplished using the MUSICAM encoder technique.
After the MUSICAM encoder, extra data like RDS and programme information is added. The encoder output
delivers a