All a matter of timing...

The digital transmissions employed for UK digital audio and TV broadcasting use Coded Orthogonal Frequency Division Multiplexing (COFDM) to communicate the information. As the name indicates, COFDM is based on OFDM. The addition of the term ‘Coded’ is to indicate the way OFDM is used to deal with some of the practical problems which arise when we transmit information. In effect, the data we wish to communicate is ‘coded’ into a form that is designed to make the actual process of communication easier and more reliable.

The systems used for practical applications are very complex, and vary from one application to another. However, typical features of COFDM, and how they help with information transmission can be understood from the simplified explanations given below. Please note that the following omits or simplifies a great deal, though!

When we try to receive EM signals we encounter various problems. These include:

Thermal Noise, and other types of continuous background noise.
Multipath
Interference from other transmissions (often unintentional)
Ignition or Impulse Interference.

In addition, when using OFDM we have timing problems because the receiver has to able to Synchronise with the transmissions, both to be able to recognise the phases of the transmitted symbols, and to determine which symbols have which meanings in the arriving stream.

Continuous background noise is also called Stationary noise as its statistical properties remain the same as time passes. For a given carrier to noise ratio (CNR) there is a statistical chance that each of the transmitted symbols will not be correctly identified when it is received. The lower the CNR, the more often such mistakes/errors tend to occur. If we do a careful analysis then we find that no matter how high the signal power, there is always a non-zero chance that a symbol may be misunderstood when received. Such errors are rare when the CNR is high, but they still happen.

Impulse Interference (II) can occur when some device is producing ‘bursts’ of interference. The most well known sources of this are the sparks which occur in some electric motors, and also in the ignition of internal combustion engines. These can act like ‘spark gap transmitters’. Any rapid change in current and voltage may radiate EM energy. Spark gap transmissions from faulty motors, etc, can radiate pulses of EM energy that only last for a tiny fraction of a second, but since they pack that energy into a very short time, the power can be very high. As a result, the sparks may briefly produce far more power than the wanted transmissions when picked up by a nearby receiver. A particular problem with II is that it does not obey the same statistics as stationary noise. This, combined with the high peak powers, means that it can sometimes be a serious problem for some receiver locations, even if the received carrier power is well above the natural stationary background noise level. Since car/motorbike ignition systems have often been a source of this type of interference it is also called Ignition Interference.

The basic method for dealing with the risk of errors is Redundancy. This is considered in more detail in Information Theory. However it essentially means that we send ‘copies’ of the information and/or additional data which can be used to help detect or avoid errors.

Digital TV and Radio

In the UK, the Digital Terrestrial TV (DTTV) system is officially referred to as Digital Video Broadcasting - Terrestrial (DVB-T) to distinguish it from satellite and cable systems. The trade name FreeView is also used to mean the DTTV broadcasts which are freely available without any encryption or the need to pay for viewing. Hence these terms are often used interchangeably in the UK. DTTV in the UK carries some ‘sound radio’ stations as well as TV stations.

The UK system used for digital sound broadcasting is called Digital Audio Broadcasting (DAB). This has some similarities with DTTV, but is broadcast at different RF frequencies from its own set of transmitter locations, and there are various differences between the COFDM systems employed for DAB and DTTV broadcasts.

Both systems are transmitted as Multiplexes. This means that – unlike conventional ‘analogue’ TV and audio broadcasts, each digital transmission carries the information for a number of TV or sound radio stations. In analogue broadcasting the terms ‘station’ and ‘channel’ are often used as if they mean the same thing. But with digital multiplexes we have to be careful as the channel means the frequency band allocated for a given RF transmission, but that transmission may then convey many TV or radio stations. The table below summarises some of the main differences between the UK DAB and DTTV systems.

System (UK)	RF band allocated	Channel Width	Useful Symbol Length	Guard Interval	Number of Carriers	Modulation
DAB	174·928 to 239·200 MHz	1·75 MHz	1 ms	0·246 ms	1536	form of QPSK
DTTV	470 - 580 and 615 - 850 MHz	8 MHz	224 s	7 s	1705	16QAM or 64QAM

COFDM and Multipath

Although the use of a suitably long guard interval helps us to avoid each symbol pattern interfering with the next, this isn’t the only problem multipath may cause. It also tends to alter the amplitudes and phases of the received signals. The intended amplitude and phase (i.e. what we’d receive in the absence of multipath) for a symbol can be represented as a complex value

To assess the effect of multipath we can consider an additional contribution due to a path of extra length,

. This provides a delayed and attenuated copy of the intended symbol. We can represent this by a contribution

where

is the relative amplitude of this contribution, and

is the change in phase produced by the longer path it follows from transmitter to receiver. Combining these we can say that the actual signal we receive, including the multipath contribution will be

Using the above expressions we can now say that the change in the received power level will be

and that the phase of the received signal will be changed by an amount

Expressions 22.4 & 22.5 have two interesting properties. Firstly, they do not contain either the amplitude,

, or the phase,

, of the intended symbol. Secondly, they do contain the carrier frequency,

.

Figure 22.1 illustrates the effect this has on the received signal. On the left of 22.1 the plot shows the change in symbol amplitude and phase for an individual carrier. (QPSK is assumed for the sake of example.) Here each intended symbol is represented by a circle, and this is altered into the multipath modified result represented by a square.

Since for a given carrier, all the symbol phases are rotated by the same amount, and the amplitudes are all changed by the same amount, the result is much the same as if the distance between the transmitter and receiver had been changed. Just as in the absence of multipath, if the receiver can correctly distinguish the phases, etc, and is given calibration information, it can demodulate the results. However this requires two assumptions to be correct:

That the received signals continue to have a power level high enough compared with any other noise (or other interference) to be reliably distinguishable.
That the details of the multipath don’t vary rapidly or unpredictably during reception since this would mean the alterations produced will change from one symbol interval to the next.

The first of the above assumptions may not be satisfied in all cases. Sometimes we may find that the direct and delayed contributions are almost 180 degrees out of phase, and tend to cancel each other. This may weaken the received signal, making it harder to correctly receive. The second assumption may not be reliable if the transmitter or receiver are mobile.

The plot displayed on the right of Fig 22.1 shows how the change in received level varies with frequency. (For the sake of example, a value of

= 0·5 and

= 90 metres were used.) It can be seen that the level changes with frequency in a cyclic manner. Looking back at expression 22.4 we can see that the change in carrier frequency,

, between an adjacent peak and dip will be

This result is an important one. It tells us that for broadcasts that have a bandwidth or channel width,

, the multipath will tend to Lift (increase) or Fade (reduce) the received level almost unformly across the broadcast bandwidth. A reduction across the entire broadcast channel is called a Flat Fade and ideally we would wish to avoid this occurring as it may cause the entire signal to become unreceivable. However for broadcasts that spread their information content over a bandwidth similar to or larger than,

, we find that although some carrier steams are faded, others will be lifted. Since

falls in value as the multipath delay,

, rises, we can avoid long distance multipath problems by deliberately choosing a large COFDM bandwidth.

DAB COFDM and Interference

The UK DTTV and DAB systems use different COFDM schemes. Here we will examine the type of arrangement employed for DAB.

Figure 22.2 is a simplified diagram of the COFDM arrangement used for DAB. The carrier frequencies,

, used for the OFDM are shown vertically. Time is shown horizontally. Each symbol is represented by a circle. We can now imagine that we have some symbol values,

, to transmit. Instead of simply sending these, just once, in their original sequence, the COFDM approach is to send this information more than once, and to redistribute the ‘copies’ so that they don’t appear at the same time, or use the same carrier. Given this arrangement, if one particular symbol is ‘lost’ as a result of random noise, then the receiver can use the ‘copy’ to read the required information.

The data is arranged into a Transmission Frame (TF), also sometimes called an OFDM Frame. To indicate the start of a frame we arrange for it to start with a unique pattern. The most obvious and convenient choice is for all the symbols at that time to be Null Symbols. i.e. symbols of amplitude,

. This makes a very obvious ‘start’ as the total carrier power briefly goes to zero, making it easy for a receiver to identify that a frame is starting. The next set of symbols all have a pre-defined phase which is not affected by the actual information we have to transmit. These are called Phase Reference Symbols. These gives the receiver the information it needs to define the phases of the following Data Symbols.

The nulls lasts for 1·279 ms and are then followed by 76 symbols on each carrier for a complete frame. Each of these symbols (inc. guards) has a duration,

, of 1·246 ms. The time interval between successive frames is therefore 96ms, and the data carried must therefore be sufficient for 96ms-worth of information for every audio station being conveyed. Note that the null symbols are also longer in duration than the data symbols that follow. This helps make them easier to distinguish from data symbols, and helps the receiver synchronise to the frames.

Figure 22.3 illustrates the effect on the COFDM of the various kinds of noise.

In each of the above cases we can see that the interference or noise prevents one copy of the data ‘x’ from being received, but another copy of the data is received correctly. There is always a chance that we may be ‘unlucky’ and that multiple noise sources or events will lose both of the symbols that carry a given set of bits, but multiple events like this will be rarer than a single symbol loss. By arranging for copies to be spread over a range of different times we can help avoid loss due to II. By using different carrier frequencies for the copies we can avoid loss due to interference by a sinewave (c.w. or continuous wave) source which blocks one carrier of the ensemble from being correctly received.

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.