hd2.gif - 22Kb

The UK currently[1] has two ‘national’ DAB multiplexes which are broadcast from a set of transmitters across the UK. There are also various ‘local’ or ‘regional’ DAB multiplexes which only cover limited areas. Each of the national multiplexes operates as a Single Frequency Network (SFN). This means that all the BBC transmitters for their national multiplex use exactly the same channel, and transmit the same modulation pattern and carrier frequencies in a synchronised manner. If this were attempted using conventional analogue modulation systems the result would be that the transmissions would interfere with one another and reception would be impossible in many places around the country. However the same guard interval that allows the digital transmissions to avoid being upset by multipath also allows us to deliberately use many transmitters and combine the results at the receiver.

So, for example, the DAB channel allocated to the BBC for its ‘national’ sound radio broadcasts is centred on 225·64 MHz. This multiplex currently carries all the main BBC sound radio stations – Radio 1, 2, 3, 4, etc. The nominal width of the channel is 1·75MHz. However since the system uses COFDM, the symbol duration of 1 ms means that the 1536 DAB carriers are spaced 1 kHz apart. Hence the actual carriers for this multiplex spread over 1·536 MHz. This leaves around 100 kHz either side for the modulation sidebands, and to make it easier for RF/IF filters in a receiver to accept a wanted transmission whilst rejecting others in adjacent channels.

The UK DAB guard interval of 246 s means that delayed versions of the transmission can arrive via paths which differ by up to the distance EM waves travel in this time. i.e. around 73·5 km. So if your local DAB transmitter is, say, 10 km away, then any transmissions for any other transmitters up to around 80 km from you should combine with it without causing problems! Since the signal level we get tends to fall away with range we can expect this to mean that transmitters close enough to give a useful signals will tend to combine to support one another. This is quite a useful feature of DAB as it saves having to allocate different frequencies for the same stations in adjacent areas to avoid interference. Along with the ability of the multiplex to carry many stations, it allows us to have many more stations in a given spectral bandwidth. In practice, the DAB transmissions don’t bother to actually switch off the transmitted power during every guard interval. Instead the symbol’s amplitude and phase is used during the preceding guard interval.

Each sound radio station first produces the audio in the form of Linear Pulse Code Modulation (LPCM) with a sampling rate of 48,000 samples/sec, with 16 bits per sample. This means that the initial bitrate for such a stream of stereo audio will be = 1·536 million bits per second. Since the UK DAB system uses QPSK there are 2 bits per symbol. The ‘raw’ bitrate each DAB transmission or Multiplex provides is

equation

but for various reasons – some of which have already been outlined – the useful data rate for a DAB multiplex is only around half the above value. i.e. around 1·2 Mb/s. which is not quite enough to carry the LPCM for a single audio station!

In order to reduce the bitrates required for audio the broadcasters employ a form of data reduction. The method chosen is called MUSICAM[2]. Since this isn’t a course on Information Theory I won’t discuss the details, but the result is that the Encoded output has been processed so as to require a bitrate much less than 1·2Mb/s. The MUSICAM system can provide various rates from 384 kb/s down to 32 kb/s, with the audible quality of the results falling with the chosen output bitrate. At a bitrate of 256kb/s per stereo audio stream we could expect to fit four or five sound radio stations onto a single DAB multiplex. Alas, in the UK at present, quantity seems to count for more than quality, hence most of the stations use 128kb/s. This means we can fit around ten stations onto a multiplex, but the level of data reduction is such that audible effects may become noticeable, albeit not objectionable to listeners who have no interest in ‘hi fi’.

The MUSICAM encoded data for the stations on a given multiplex is first arranged into ‘Logical Frames’ which represent sections of the audio signal which are 24 ms long. These are also called a Common Interleaved Frame (CIF). Four CIFs are grouped together to make up a TF. The UK system developed from the Eureka-147 project and is sometimes referred to as Mode I DAB.

The PRS symbols are followed by three symbol periods which provide some special data called the Fast Information Channel (FIC). This FIC contains data that is not time-interleaved, so can be read immediately by the receiver. The FIC provides information on topics like which of the following symbols relate to which stations, and what stations are being broadcast, etc. This then allows the receiver to pick up and decode only the data from the rest of the frame which relates to the station which the user has chosen to hear. The FIC also allows the tuner to detect when the details of the multiplex alter – e.g. when a ‘new’ station starts broadcasting, or the one being listened to changes its bitrate in response to some other station starting or finishing for the day. Thus the system is designed to be flexible, and allow the receiver to adapt to changes in the transmissions as they occur.

Although grouped into 96 ms frames for transmission purposes, the actual audio data is interleaved and spread over more than one frame. In practice, UK DAB interleaves audio data over a range of 360 ms. This means that even when some audio data is completely lost and can’t be recovered, the result should be a slight inaccuracy spread over more than a quarter second. Not simply a ‘click’ or momentary silence. The drawback of this extended interleaving is that the receiver has to wait this long for all the relevant information to arrive before it can reconstruct the sounds. Hence it adds an extra delay into the communication process. In practice, however, there will already be delays due to the time needed to MUSICAM encode/decode, etc. so delays of the order of 1·5 seconds during the entire transmission and reception process are fairly normal. These delays may also very from station to station, and with the design of the receiver being used.

The data stream can also include other forms of data – e.g. a text commentary on what the selected station is laying, and the time, date, etc.


Digital Terrestrial TV in the UK
DTTV in the UK also uses a form of COFDM. However since the required data rates for video are much higher than for audio-only it has to provide a much higher information capacity. This means that the details differ in various ways. The basic parameters were listed in the table in an earlier section along with those for DAB. Here I will just outline the main features of how UK DTTV works, but as with DAB please note that the explanation given here is simplified and omits many details. Note that I will describe the ‘2k mode’ system in use in the UK, but that there are alternative modes available which are not currently in use.

The UK DTTV system had a ‘false start’ because the original system was not a commercial success. However the in-place hardware was used for a ‘re-launch’ in 2002. This took the opportunity to make some changes on the basis of practical experience. As a result of this history, the DTTV multiplexes in the UK are of two types. Some of the initial multiplexes use 64QAM modulation with a level of redundancy indicated in terms of a Rate specified as 2/3. This indicates

a method of encoding that uses 3 transmitted bits to carry just 2 bits of actual information. i.e. the information rate is only 2/3rds the bitrate. The later system uses 16QAM with a Rate of 3/4. 16QAM only requires about half the power that 64QAM needs for good reception. The sacrifice is that the 16QAM system carries data at a lower rate. Hence 16QAM multiplexes give more reliable reception, but either provide fewer stations, or offer poorer picture/sound quality.

16QAM means 4 bits/symbol, and there are 1705 carriers in the ensemble. The nominal symbol rate per carrier is 1/(231s) leading to a raw capacity of 29 Mb/s. 64QAM means 6 bits/symbol, increasing this nominal rate to around 45 Mb/s. In practice, however, just as with DAB the useful bitrate is much lower than this due to the methods used for error protection, synchronisation, etc. As with DAB, DTTV groups the data into frames, but the details are different to those for DAB.

The DTTV frames are groups of 68 symbols per carrier. The highest and lowest frequency carriers in the ensemble always carry Pilot information. In effect, they are continuous sinusoids of a defined phase and frequency to assist the receiver with synchronising. For this reason, they are referred to as Continuous Pilot symbols. In addition, a series of symbols distributed across the rest of the ensemble in a pre-defined pattern also contain Pilot phases and amplitudes to help the receiver assess and correct for effects such as multipath.

The pilot symbols are all transmitted with an amplitude that is 1/0·75 (+2·5dB) of the maximum size of the data symbols. This means that stand out in a clear pattern above the rest of the stream of symbols and the receiver can easily identify them and use them to align its detection to the symbol stream. Frame synchronisation is then achieved using a specific sub-set of the carriers which are NOT modulated with QAM data. These ‘special’ carriers are the Transmission Parameter Signalling (TPS) streams, and are BPSK modulated. As a result, the TPS carriers only convey 1 bit per symbol, but are much less prone to errors due to noise, etc, than 16/64QAM.

fig4.gif - 44Kb

In the UK DTTV system there are 17 TPS carriers in the ensemble. To make them even easier to recognise and use, they all carry exactly the same data pattern during a given frame. The 1st bit is an initialisation bit. The next 16 TPS symbols are a frame synchronisation pattern. This is then followed by 37 bit/symbols of information about the transmission, and the last 14 are redundancy checking bits which the receiver can use to detect and correct errors in the rest of the TPS frame sequence. The pattern of the 16 TPS synch symbols is pre-defined and the rest of the TPS information arranged never to give this pattern. Hence the receiver can watch for this pattern, and when it is detected, use it for the frame synchronisation.

The use of the TPS/Pilots is a very effective way to obtain synchronisation and phase reference. However the DTTV system does not employ interleaving of the transmitted data at the transmission frame level. This means that II can more easily produce an audible ‘click’ on DTTV reception. For video images, the equivalent may be that the image briefly freezes or shows visible ‘blocks’. For the above reasons DTTV require a higher carrier/noise and carrier/interference level than DAB to achieve similar levels of audible/visible effects due to noise, etc. However the effects of II on video are more varied than on audio. Fortunately, the higher carrier frequencies used for DTTV in the UK should mean that the levels of Impulse Interference are less than for DAB. Also, our eyes react differently to our ears to such problems. Hence the developers of DTTV decided the chosen system was acceptable in order to help achieve the information capacity needed for video.

The relatively short guard interval (7ms) means DTTV is potentially more sensitive to multipath than DAB. The good news here is that the higher carrier frequencies used for DTTV tend not to propagate so far, so there should be less chance of interference from distant transmissions. Despite this, however, the short guard interval means that a DTTV SFN is not practical using the system as described, and there is a greater risk of reception problems due to multipath than with DAB.

DTTV uses a data reduction system based on MPEG-2 to reduce the amount of information required to communicate the video and audio for each station on a DTTV multiplex. The audio process is similar to that used for audio on DAB. The video system starts by representing the input images as a series of bitmaps with 720 columns and 576 horizontal lines or rows. (Note that the UK PAL ‘analogue’ images – although described as ‘625 line’ actually devotes some of the scan lines to other functions, and hence only around 576 lines are actually used for the image.)

720 576 = 414 720 pixels. If we use 8 bits to represent each of the three colour levels in each pixel this comes to 9 953 280 bits for each image. In the UK TV systems we then require 25 such images per second, which implies a ‘raw’ information rate of 248 832 000 bits/second! In practice the data reduction systems employed allow the video to be conveyed with bitrates of the order of 10 Mb/s or less! Hence the level of data reduction employed is quite dramatic, and allows us to use a DTTV multiplex to typically convey over 5 TV stations per multiplex.


Summary
You should now know how the UK DAB and DTTV transmissions use different forms of COFDM to broadcast audio and video information in digital form. That both systems use Frames to group the data, and various methods to provide synchronisation. That the data requirements are reduced by employing data reduction systems on the input audio and video information. That the DAB system allows the use of SFNs, and a high level of protection against II as well as continuous noise. But that the DTTV, although it carries much higher data rates than DAB, requires higher carrier/noise levels, and is less well protected against problems like II.


[1]  2006
[2]  Masking pattern Universal Sub-band Integrated Coding and Multiplexing

Content and pages maintained by: Jim Lesurf (jcgl@st-and.ac.uk)
using TechWriter Pro and HTMLEdit on a RISCOS machine.
University of St. Andrews, St Andrews, Fife KY16 9SS, Scotland.