|
Why do we need digital radio?
1.1 The VHF FM service
1.2 Digital transmission 1.2.1 Digital radio: the need for frequency sharing 1.2.2 - Differential quadrative phase shift key (DQPSK) modulator 1.2.3 - Choice of symbol length and transmitter spacing 1.2.4 - Digital radio ensemble data capacity 1.2.5 - How many services? 1.2.6 - Spectrum Allocation
1.3 - Summary
1.1 the VHF FM service
Most engineers will be familiar with the practical problems associated with radio wave propagation at VHF and UHF frequencies. At these frequencies propagation is by space wave or direct wave and is affected by objects or terrain along its path. Depending on the frequency, waves are attenuated by hills and buildings which cause shadow areas where the reception is poor. The higher frequencies suffer more severe attenuation, for example UHF TV signals are more prone to shadow effects than VHF Band II signals. This is due to diffraction where longer wavelength signals (lower frequency) are diffracted more, thereby bending around objects and filling in shadow areas. This is why the sky looks red when the sun is below the horizon, as the visible red end of the spectrum (low frequency) is bent around the earth’s surface.
FM transmissions in the band 88 MHz to 108 MHz do, however, suffer from quite severe problems in the form of multipath propagation which causes frequency selective fading. See figure 1. The effect of multipath propagation is to produce geographical areas of frequency dependent fading. This means that some FM stations can be received with good audio quality whilst others are noisy. But, move the receiver a few metres and the good station becomes bad and the noisy station becomes good. When this scenario is transferred to the motor car things become worse. Moving the receiving antenna through areas of multipath produces a fluttering effect on the received audio.
back to top
1.2 digital transmission
1.2.1 digital radio: the need for frequency sharing
Figure 2 shows the effect of multipath on adjacent FM services: some are faded up while some are faded down. Move the receiver antenna and the pattern will change and different carriers will fade up and down. To protect a radio service from multipath it would be far better to code the audio service and to have segments of data on each carrier and use error protection to reconstitute the lost (faded) data.
This is exactly what digital radio does. It uses 1536 carriers in mode 1 (Ref. 1), which are collectively known as an ensemble, multiplex or block. Using 1536 carriers for one service is spectrally inefficient. So, many services are brought together into one multiplex. Each service is error protected and then split into small data units called capacity units. These capacity units are then multiplexed together to form one single data stream. This stream is then split into bit pairs and modulated onto the 1536 carriers, so sharing data from all services across the frequency block. This makes each service (audio or data) protected from selective frequency fading.
back to top
1.2.2 - differential quadrative phase shift key (DQPSK) modulator
Differential quadrative phase shift keying (DQPSK) is a very simple digital modulation system. Figure 3 shows a sine wave suddenly changing phase. The phase change can be measured in degrees. In DQPSK four phase states are allowed, thus four phase differences can be measured on symbol boundaries. This allows two bits of data to be carried per symbol boundary.
| Phase Change | Data Pair | | | 0 | 00 | | 90 | 01 | | 180 | 10 | | 270 | 11 |
back to top
1.2.3 - choice of symbol length and transmitter spacing
The length of the symbol (i.e. the length of time the phase is held constant on each carrier) must be long enough for the receiver to measure the phase of the symbol so the change in phase can be calculated on the symbol boundaries. With a mode 1 signal (Ref. 1) the digital radio modulation system uses carriers that are 1kHz apart, thus a 1ms measurement time is required. To ascertain the phase of a carrier one complete cycle is required to make the measurement. This is not the end of the story.
The next requirement for digital radio is to be able to use the same frequency across the whole of a country or region to carry the same programmes. This requires transmitters to be on the same frequency. So at the point of reception there may be signals from many transmitters. This can be treated as long distance multipath. See figure 4. In the UK the digital radio frequency allocation is in band III. Band III frequencies propagate well and band III networks lend themselves to transmitters spaced at about 75 km. So a receiver must be able to withstand signals coming into the antenna delayed by up to 75km.
Figure 5 shows a set of symbols arriving at the receiver when they have travelled over two different paths. Because the system uses DQPSK the two voltage vectors, one from each path, can be added in the receiver and the resultant vector has the correct phase change when both symbols have the same symbol count. To remove the intersymbol interference the 1ms phase measurement must be made where the symbols overlap for 1ms. In a 75km spaced network the symbols must be 1.246ms long to ensure this.
| 1ms for phase measurement | | 75000 m | | | max. delay or guard interval | = -------------------- | 0.25ms | | 3 x 108 m/s | | | So total symbol length is 1 ms (phase measurement) + 0.25ms (max. delay) = 1.25ms |
back to top
1.2.4 - digital radio ensemble data capacity
In a digital radio mode 1 signal, there are 1536 modulated carriers and the symbol length is 1.246ms.
| The symbol rate is | | 1 | | | | --------------- | = | 800 symbols per second | | 1.246ms | | |
The symbols are split into groups to make frames and synchronisation and control channels are added. The digital radio transmission frame is 96ms long in mode 1, and contains 77 symbols. Symbols one and two are used for receiver synchronisation. Symbols three, four and five are used for the fast information channel (FIC) which contains all the information about how the multiplex is set up. This leaves 72 active symbols. Each symbol carries two bits per carrier.Therefore the total data rate available for audio or data services is:
| 72 x 2 x 1536 | | | | ------------------------- | = | 2.34Mbits/s | | 96 x 10-3 | | |
back to top
1.2.5 - how many services?
The 2.3Mbits/s cannot all be used for data and audio as overhead for error protection is necessary. In a bad multipath environment many carriers will be faded down into the noise and the data will not be recoverable without a “strong” forward error correction system (FEC). Viterbi error correction, a convolutional coding system, is used in digital radio. This uses a shift register to generate additional data as in figure 6.
As the additional data stream contains knowledge of each original bit of data three times, because of the three taps in the polynomial, if one bit gets corrupted in transmission, the receiver can correct it. This is because the receiver knows the polynomial used to generate the protecting data and it can compute the most likely data string to have generated the additional error correction data.
At the output of the FEC system in the coder the two data streams are multiplexed together to form a data stream with twice the data rate of the original audio or data service. This is called half rate coding, not to be confused with half rate coding in Musicam. More polynomials can be added to increase the error correction. For an audio service in digital radio there are five allowed error protection levels. Using a heavier protection level increases the services ability to withstand fading and interference. This will increase the service area, but more data capacity is required.
| Error Protection Level | FEC Rate | Capacity required for a 192 kbits/s Musicam channel | | 1 | 0.34 | 568 kbits/s | | 2 | 0.43 | 448 kbits/s | | 3 | 0.51 | 384 kbits/s | | 4 | 0.62 | 312 kbits/s | | 5 | 0.75 | 256 kbits/s |
So how many services? There is a 2.3Mbits/s pie to share out between the services. ntl’s field trials have shown that good coverage can be achieved using half rate FEC. So there is now 1.15Mbits/s net data to share and that is not enough for one linear 16 bit, 48kHz, sampled stereo programme. The answer is audio bit rate reduction and digital radio uses Musicam (ISO layer II) coding with a modified header. The Radio Authority have set minimum audio data rates (Ref. 3) for mono and stereo programme genres. For an audio service the data rate chosen is very much dependent on the trade off between audio quality, the amount of programme associated data (PAD) and the cost per bit.
| Some examples of data rates: | | | Mono music service | 96 kbits/s | | Mono speech service (R. A. min) | 64 kbits/s | | Stereo music service (R. A. min) | 128 kbits/s | | Recommended stereo music service | 128-192 kbits/s |
For a data service, the data capacity required is completely dependent on the maximum data rate and protection required for the application. See Station 5 on data services. At the time of writing the present legalisation allows only 10% of the available total data capacity to be used for non programme related data services. Two examples are given below of ways of making up digital radio multiplexes assuming half rate coding on all services. See section 3.4.1.2 on the calculation of capacity units (CU).
| Example 1 | | Service Name | Data Rate (kbits/s) | Protection | CUs | | Classic | 160 | 0.5 | 128 | | Virgin | 160 | 0.5 | 128 | | Data | 64 | 0.5 | 48 | | Magic | 160 | 0.5 | 128 | | Data | 32 | 0.5 | 24 | | Talk | 96 | 0.5 | 70 | | Jazz | 160 | 0.5 | 128 | | News Channel | 96 | 0.5 | 70 | | Soft Rock | 160 | 0.5 | 128 | | Total number of CUs | 852 |
| Example 2 | | Service Name | Data Rate (kbits/s) | Protection | CUs | | Classic | 160 | 0.5 | 128 | | Virgin | 160 | 0.5 | 128 | | Data | 64 | 0.5 | 48 | | Magic | 160 | 0.5 | 128 | | Data | 64 | 0.5 | 48 | | Talk | 96 | 0.5 | 70 | | Jazz | 160 | 0.5 | 128 | | News Channel | 96 | 0.5 | 70 | | Schools Channel | 96 | 0.5 | 70 | | Announcement Channel | 48 | 0.5 | 35 | | Total number of CUs | 853 |

back to top
1.2.6 - spectrum allocation
In the UK the Government have allocated between 217.5 and 230 MHz for digital radio. The spectrum allocated is part of Band III which was formally used for VHF 405 line television. Within this allocation there is enough available bandwidth for seven digital radio ensembles. Two ensembles have been earmarked for national radio coverage, one for Independent National Radio (INR) and the other for BBC national services. The remaining five ensembles have been allocated to Independent Local Radio (ILR) and BBC local services. See figure 7. Note that all channels except 12B could be used for ILR in the UK, although there are national and international restrictions. It is envisaged that the majority of areas will receive the two national ensembles and one local ensemble which will carry ILR and BBC local services. In addition, main city areas will receive a second local/regional ensemble providing listeners with around 12 local services and 12 national services.
back to top
1.3 - summary
2. The new digital radio service uses a 1536 carrier modulation scheme. The individual carriers are modulated at 800 symbol/s and the overall bit rate is 2.3 Mbit/s.
3. Each symbol is 1.25ms long of which 0.25ms is considered to be a guard interval. This enables a single frequency network (SFN) to be built with transmitter separation of less than 75km as long as all data is transmitted at the same moment in time from each transmitter in the SFN.
4. In the UK the spectrum allocation allows for seven digital radio ensembles. In most areas there will be two national ensembles and two local/regional ensembles.
back to top
|