Modulation and demodulation

Modulation is the process of impressing information to be transmitted onto an RF ‘carrier’ wave, in such a way that it can be retrieved again in more or less undistorted form at the receiver. Figure 8.1a shows how information is transmitted by CW (continuous wave) using the Morse code, once widely used on the HF band (1.6–30 MHz) for commercial marine traffic and still used by amateurs for world-wide DX-ing on a few watts. Broadcasting on the long, medium and short wavebands uses AM (amplitude modulation) (Figure 8.1b). The amplitude of the RF carrier wave changes to reflect the instantaneous value of the modulating baseband waveform, e.g. speech or music. The baseband signal is limited to 4.5 kHz bandwidth, restricting the bandwidth occupied by the transmitted signal to 9 kHz, centred on the carrier frequency. With maximum modulation by a single sinusoidal tone, the transmitted power is 50% greater than with no modulation; this is the 100% modulation case.

Note that the power of the carrier is – even less during average programme material. For this reason, single sideband (SSB) modulation has become very popular with military, commercial and amateur users for voice communication at HF. In SSB (Figure 8.1c), only one of the two sidebands is transmitted, the other and the carrier being suppressed. Spectrum occupancy is halved and all transmitted power is useful information. At the receiver, the missing carrier must be supplied by a carrier re-insertion oscillator at exactly the appropriate frequency; an error of up to 10 Hz or so is acceptable on speech, less than 1 Hz on music.

In the early days of SSB this was difficult and a very fine tuning control called a clarifier was provided, but with synthesized transmitters and receivers this is no longer a problem. In commercial and military SSB applications USB (upper sideband) operation is the norm, in amateur practice USB is used above 10 MHz and LSB below. ISB (independent sideband) operation is occasionally used commercially. Here, one communication channel is carried on the lower sideband and an entirely different one on the upper. At one time, four international telephone trunk channels were carried on a single suppressed carrier using ‘2 + 2 ISB’. Here, each sideband carried two telephone channels, one at baseband and one translated up to the band 4-8 kHz.

Figure 8.1d illustrates frequency modulation. FM was proposed as a modulation method even before the establishment of an AM broadcasting service, but it was not pursued as the analysis showed that it produced sidebands exceeding greatly the bandwidth of the baseband signal [1]. FM is used for high fidelity broadcasting in the internationally allocated VHF FM band 88-108 MHz, using a peak deviation of ±75 kHz around the RF carrier frequency and a baseband response covering 50 Hz to 15 kHz. Figure 8.1 shows the characteristics of AM and FM in three ways: in the frequency domain, in the time domain and as represented in vector diagrams. Note that in Figure 8.1d a very low level of modulation is shown, corresponding to a low amplitude of the baseband modulating sinewave (frequency f_m). Even so, it is clear that if only the sidebands at the modulating frequency existed, the amplitude of the RF signal would be greatest twice per cycle of the modulating frequency, at the instants when the phase deviation of the RF from the unmodulated state was greatest. It is the presence of the second order sidebands at 2f_m that compensates for this, maintaining the amplitude constant.

At wider deviations, many more FM sidebands appear, all so related in amplitude and phase as to maintain the amplitude constant. Note that the maximum phase deviation of the vector representing the FM signal will occur at the end of a half-cycle of the modulating frequency, since during the whole of this half-cycle the frequency will have been above (or below) the centre frequency. Thus the phase deviation is 90° out of phase with the frequency deviation. For a given peak frequency deviation, the peak phase deviation is inversely proportional to the modulating frequency, as is readily shown. Imagine the modulating signal is a 100 Hz squarewave and the peak deviation is 1 kHz. Then during the 10 ms occupied by a single cycle of the modulation, the RF will be first 1000 Hz higher in frequency than the nominal carrier frequency and then, during the second 5 ms, 1000 Hz lower. So the phase of the RF will first advance steadily by five complete cycles (or 10π rad) and then crank back again by the same amount; i.e. the peak phase deviation is ±5π rad relative to the phase of the unmodulated carrier. Now the average value of a halfcycle of a sinewave is 2/π times that of a half-cycle of a squarewave of the same peak amplitude; so if the modulating signal had been a sinewave, the peak phase deviation would have been just ±10 rad.

Note that the peak phase deviation in radians (for sinewave modulation) is just f_d/f_m, the peak frequency deviation divided by the modulating frequency: this is known as the modulation index of an FM signal. If the modulating frequency had been 200 Hz (and the peak deviation 1 kHz as before), the shorter period of the modulating frequency would result in the peak-to-peak phase change being halved to ±5 rad; so for a given peak frequency deviation, the peak phase deviation is inversely proportional to the modulating frequency.

For monophonic FM broadcasting the peak frequency deviation is ±75 kHz, so the peak phase deviation corresponding to 100% sinewave modulation would be ±5 rad at 15 kHz and ±1500 rad at 50 Hz modulating frequency. Thus on reception, 1 rad of spurious deviation at 50 Hz due to noise will have much less effect than 1 rad of deviation at 15 kHz, giving rise to the well-known triangular noise susceptibility of FM. It also explains the greater signal to noise ratio required for stereo reception, since the left minus right difference signal is a 15 kHz double sideband signal occupying the spectrum 23–53 kHz, modulated on a suppressed 38 kHz sub-carrier. Quite apart from the slightly wider IF bandwidth compared with mono needed to receive stereo FM transmissions, the difference signal is inherently more susceptible to noise degradation as indicated by the triangular noise susceptibility characteristic of FM reception. The noise susceptibility in the upper part of the baseband mono compatible sum signal is reduced by applying a 6 dB per octave pre-emphasis above 3.2 kHz, which effectively produces PM (phase modulation) at the higher audio frequencies. A corresponding deemphasis is applied in the receiver. The pre-emphasis breakpoint corresponds to a time constant of 50 μ (2.1 kHz and 75 μs are the values used in the USA).

If the modulation index is small compared with unity, the second and higher order sidebands are negligible, but if it is very much larger than unity there are a large number of significant sidebands and these occupy a bandwidth virtually equal to 2f_d, i.e. the bandwidth over which the signal sweeps. The usual approximation for the bandwidth of an FM signal is BW = 2(f_d + f_m). Note that if one of the first-order FM sidebands in Figure 8.1d were reversed, they would look exactly like a pair of AM sidebands; this is why one of the first-order FM sidebands in the frequency domain representation has been shown inverted. A spectrum analyser is not sensitive to the relative phases of the signals it encounters during its sweep, so it will show the carrier and sidebands of an AM or of a low-deviation FM signal as identical. However, if the first-order sidebands displayed are unequal in amplitude, this indicates that there is both amplitude and frequency modulation present on the carrier; this is illustrated in Figure 8.2.

Figure 8.1e shows the spectra of high modulation index FM for both sinewave and triangular wave modulation with a frequency f_m. In both cases, the overall shape of the power distribution versus frequency is shown. It consists of discrete spectral lines spaced at intervals f_m, with an overall envelope the same shape as the power density plot of the modulating waveform. The flat power density plot with triangular modulation is useful in a jammer application and a very high modulation index ensures a rapid fall away in power outside the intentionally jammed band, avoiding interference with own communications. However, to jam a bandwidth of many megahertz with lines close enough to ensure jamming even a narrow band target, will require a low modulating frequency. This means that the ‘revisit time’ for a channel, especially one near the edge of the jammed bandwidth, may become overlong. A narrow band of noise may therefore be added to a rather higher frequency triangular wave modulating signal, to spread out the modulation, filling in the gaps between spectral lines.

Many modulation methods have been employed for the transmission of digital data, or of information in digital form such as teleprinter traffic. They are all variations of AM, FM or PM, or of a combination of these. One of the earliest is FSK (frequency shift keying) which is widely used for the transmission of text in ITA2 (international teleprinter alphabet No. 2) by national news agencies (see Figure 8.3a). A commonly used standard on HF is 850 Hz shift (±425 Hz on the suppressed carrier frequency). If the change from one frequency, representing a zero, to the other, representing a one, is abrupt, then the signal will occupy a greater bandwidth than is necessary for its successful reception: the excessive OBW (occupied bandwidth) may interfere with other stations. Several means are used to avoid this, such as band-pass filtering the FSK signal in the exciter before passing it to the PA (power amplifier), shaping or low-pass filtering the data stream and its inverse before applying to two amplitude modulators (this method is known as FEK, frequency exchange keying – Figure 8.3b) or generating the FSK signal by feeding the data stream into an FLL (frequency lock loop). In this latter method, there are no phase discontinuities so it is known as CPFSK (continuous phase FSK). Typically, the transition is arranged to occupy about 10% of a bit period and the data rate with 850 Hz shift would usually be 50 baud.

The baud is the unit of signalling rate over the communications link, and the useful bit rate may be lower or higher than this. For example, in ITA2, each character of the message is transmitted as a start bit followed by five data bits followed by one and a half stop bits, giving a bit rate of two-thirds of the baud rate – or rather less in practice. As the code incorporates start and stop bits it operates asynchronously; one character does not need to follow the next immediately, it can dwell on a stop bit until the next character arrives, e.g. from a typist at a keyboard. The five data bits permit 32 different characters to be encoded, so that figure shift and letter shift characters are used to accommodate the alphabet (capitals only), numerals, punctuation and control symbols. ASCII code (American Standard Code for the Interchange of Information, also known as ITA5) uses seven data bits per character giving 128 possibilities and so can support upper and lower case, without needing shift characters. Often an eighth bit is added for parity, a character thus occupying exactly one byte, and many modems accommodate data with one, one and a half or two stop bits – so there may be up to eleven bits to a character.

FSK/FEK may be very simply demodulated using a frequency discriminator and this was originally the usual method, but it is not optimum. A better scheme is to make use of the fact that the signal effectively uses frequency diversity, in that all the transmitted information could be extracted from either the mark frequency or the space frequency (each regarded as OOK: on-off keying) alone. This is very beneficial for traffic on the HF band, where selective fading may cause one of the frequencies to fade out completely while the other is still usable. Using this characteristic to the full, it is possible to receive the data correctly when one tone is unavailable due to fading (using a ‘slideback’ detector), or even when it is being jammed by a strong continuous signal (using a ‘Law assessor’ [2]). Reliability of HF communications can be improved using an ARQ (automatic repeat request) system, such as that defined in Reference 3.

The need for higher signalling rates on long-haul routes using the HF band brought problems when using FSK. An HF signal received at a distance of several thousand kilometres may be received via several different paths, for which the spread of propagation time may be several milliseconds. Thus increasing the baud rate could result in the early path version of one symbol overlaying the late path version of the preceding one, resulting in ISI (intersymbol interference). One solution introduced by the UK Foreign and Commonwealth Office [4] used MFSK (multifrequency shift keying) at a 10-baud signalling rate. In each 100 ms symbol, it transmitted one of 32 different tones, each one representing an ITA2 character. Thus the character rate equalled the baud rate and the system provided a throughput equivalent to an FSK ITA2 system operating at 75 baud. In a later improvement [5], each character was transmitted as a sequence of two tones at a 20-baud rate. The tones were selected from a group of 6 (or 12) giving operation equivalent to ITA2 at 75 baud (or ITA5 at 110 baud).

FSK/FEK are early forms of digital modulation and although simple to implement and robust, they are not bandwidth-efficient, the OBW being many times the useful bitrate. Other more efficient modulation methods have been developed, e.g. phase shift keying (widely used at VHF where propagation characteristics are rather more stable than at HF) and combined phase-and-amplitude keying (used in terrestrial microwave telephony links where conditions are usually very stable). In FSK there is no ambiguity as to whether a given tone represents a mark or a space, since one is higher in frequency than the other. However, in phase shift keying, the only thing that changes is the phase of the single RF carrier. At the receiver there is no way of knowing the transmitted phase. Even if the transmitter and receiver each had an ideal clock, the number of wavelengths in the over-the-air path is unknown. Consequently, PSK (phase shift keying) systems always use differential encoding (decoding may be either differential, or absolute, i.e. synchronous).

Differential encoding means that a phase change from one symbol to the next indicates a one, and no phase change indicates a zero, or vice versa, depending upon the particular system. A transmission consequently needs a preamble of some sort, e.g. a series of ones, and this serves two purposes. Firstly, it enables the receiver to acquire symbol sync and secondly, the first zero following the ones can signal the start of the transmitted message. The simplest form of phase shift keying is BPSK (binary phase shift keying), often simply called PSK (see Figure 8.4a). The symmetrical form has the advantage that there is always a phase change so symbol sync (the same as bit sync for a binary modulation system) can always be maintained; in the unsymmetrical form a long string of zeros would result in no phase changes, so that the receiver’s bit sync could drift out of synchronism. However, in the symmetrical form, a noise-induced phase shift at the receiver of only 90°(or less with differential decoding) will cause an error, whereas twice as large a phase shift is needed to give an error in the unsymmetrical form. Therefore, twice the received signal to noise ratio is necessary to prevent a noiseinduced error, or put another way, half of the transmitted power is effectively dedicated to maintaining bit sync. On account of the 3 dB power advantage, unsymmetrical forms of PSK may be preferred (depending on the application), the modulation usually being of such a nature that long sequences of zeros do not occur. The receiver decides whether the phase of the signal during one bit is the same as or opposite to that in the preceding bit. The phase is sampled in the middle of the bit period, which is known from the bitsync extraction circuit. Up to 90° difference counts as the same phase, more than this as the opposite phase. In differential decoding (DPSK), the bit phase is measured relative to the phase of the preceding bit, which may of course itself differ from the true phase due to noise. A further 3 dB reduction in the signal to noise ratio required for a given error rate is obtained if the measurement is made relative to true phase, i.e. synchronous decoding. This is possible if the phase of the original carrier is extracted, by doubling the frequency of the IF signal. Phase changes of 180° thus become 360° changes and an oscillator can then be phase locked to this signal (‘Sunde’s method’). If the time constant of the phase lock loop filter is many times the bit period, the phase of the carrier is accurately recovered with minimal jitter, due to the averaging process.

Ideally, the OBW of the transmitted signal would be limited to ±f_b/2 about the nominal carrier frequency, where f_b is the bit rate. However, if the phase changes in BPSK are instantaneous, there will be higher order sidebands (sidelobes), the first sidelobes being only 13 dB down. Filtering may be used to reduce the amplitude of these, but will have the effect of introducing amplitude variations into the envelope of the signal, which creates difficulties if the transmitter uses a class C power amplifier. It will also introduce ISI, resulting in a finite irreducible error rate on reception, even in the absence of noise. The ISI introduced by filtering can be largely corrected by a suitable all-pass filter or phase equalizer, but the problem of envelope variations remains. It can be minimized in some forms of QPSK (quadrature phase shift keying), also known as 4-level PSK. Here, there are four possibilities for each phase change, so each symbol conveys two bits of information (Figure 8.4b). The UK developed NICAM-728 (Near-Instantaneously Companded Audio Multiplex, providing digital-audio quality stereo or dual-language mono sound, adopted by the European Broadcasting Union for PAL and SECAM systems) uses asymmetrical QPSK. In other QPSK applications, the symmetrical form may sometimes be preferred, since then there is always an obvious minimum phase change to get from one symbol to another. In the unfiltered asymmetrical form, as in unfiltered asymmetrical BPSK, instantaneous 180° phase changes occur. Instead of filtering, the phase transition can be arranged to occur smoothly, occupying an appreciable fraction of a symbol period, giving a much faster fall-off in sidelobe level without introducing envelope variations. SQPEK (four-level symmetrical differential phase exchange keying, Figure 8.4b) is produced by baseband filtering and pre-equalizing the data fed to I and Q (in-phase and quadrature) modulators and combining their IF outputs. It is a non-constant envelope scheme, exhibiting occasional dips in the envelope of up to 10 dB, depending upon the preceding bit sequence. To minimize both OBW and the receiver noise bandwidth, the overall filtering is equally split between transmitter and receiver. In the receiver IF the signal may be hard limited, but only after filtering to final bandwidth, otherwise excessive ISI is re-introduced.

Bit rates up to 2400 bits/s are possible over HF paths using parallel tone modems. Reference 6 describes one such system, where 16 data tones and two special-purpose tones are transmitted continuously. Each data tone is BPSK or QPSK modulated at a 75 baud rate giving up to 2400 bits/s throughput in good conditions, with fall-back using increasing levels of diversity via 1200, 600 bits/s, etc., right down to 75 bits/s at 32 level diversity. However, with this scheme, the power available to each tone is very limited. Interest has therefore turned to serial tone modems for HF use, operating typically at 2400 bits/s. These use sophisticated filtering and training techniques to overcome the effect of ISI experienced due to the high baud rate, which is typically in excess of the effective bit rate to allow for periodic filter-training sequences, checkcodes, etc. Various formats are used, Reference 7 being one.

OQPSK (offset keyed QPSK, also known as OK-QPSK) and MSK (minimum shift keying, also known as FFSK and fast FSK) are important variants where the bit timing in the I and Q channels is offset by half a symbol period [8]. If either is band limited in the exciter to narrow the OBW and then hard-limited for the benefit of a class C power amplifier, the degree of regeneration of the filtered sidelobes is less than with filtered QPSK. Furthermore, MSK can be economically non-coherently detected using a discriminator, although a rather higher signal to noise ratio is then required. In unfiltered OQPSK (the asymmetrical form is usual), the maximum instantaneous phase change is 90°, since the component 180° I and Q channel phase changes are staggered. MSK and OQPSK may be coherently demodulated using the recovered carrier. This is obtained by quadrupling the IF signal, phase locking an oscillator to this and dividing its output by four. In MSK, as in CPFSK, there are no instantaneous phase transitions, so it offers low side sidelobe levels without the need for filtering, combined with a constant envelope. MSK can be viewed either as FSK where the frequency shift is ± 1/(4T), T being the bit period, or as OQPSK where the pulses in the I and Q modulator channels are shaped to a half-sinusoid instead of square. For a continuous stream of ones (or zeros), the phase of MSK advances (retards) linearly by 90° per bit period: for reversals (alternate 0s and 1s), it describes a triangular waveform of 90° peak-to-peak phase deviation. QMSK (quaternary MSK) is the symmetrical version, with phase changes of ±45° or ±135°: GMSK (Gaussian-filtered MSK) offers reduced sidelobe levels and these are even lower in QGMSK, which has been proposed for land mobile secure voice communications systems.

TFM (tamed frequency modulation) is a PR (partial response) version of MSK, offering even lower sidelobe levels at offsets from the carrier equal to the bit rate and beyond [9]. In a PR system, decoding one bit demands a knowledge of some other bits. In TFM, the bit information is spread over three adjacent bits, so that, for example, during a sequence of reversals the phase neither advances nor retards (Figure 8.4c). PR systems exhibit error propagation: an error in one bit may affect others also.

Where it is necessary to transmit a higher data rate in a given bandwidth than can be achieved with 4-level modulation, 8-PSK permits the transmission of three bits per symbol (Figure 8.4d) at the expense of requiring a higher E_b/N₀ (energy per bit over noise per unit bandwidth). Similarly, 16-PSK carries four bits per symbol, but as the number of levels increases, phase space positions become very crowded. Over high signal-to-noise ratio links, e.g. terrestrial microwave telephony bearers, the number of bits per symbol can be increased without such crowding by using both phase and amplitude modulation. Figure 8.4d shows 16-ary APK (sixteen level amplitude and phase keying); 64APK and 256APK, carrying 6 and 8 bits per symbol respectively, are used on some links.

Communications systems standards have proved very resilient in accommodating and carrying more information than they were originally intended to. As already mentioned, the broadcast FM standard has been modified to carry a difference signal permitting stereo broadcasting, at some slight reduction in the mono-service area and a more restricted area of satisfactory stereo reception, whilst more recently comparatively low speed Radio Data has been added, using yet another sub-carrier.

A similar evolution has taken place in monochrome television standards, leading to the NTSC, PAL and SECAM standards. Faced with the task of defining a television signal format which would convey a full colour picture and yet provide an acceptable monochrome picture on millions of existing black-and-white sets, the National Television Standards Committee came up with the ingenious NTSC arrangement, using a subcarrier for the colour difference signals. These were carried as in-phase and quadrature amplitude modulation of a suppressed sub-carrier, at about 3.58 MHz near the top end of the video baseband signal. A short burst of this carrier is transmitted during the back porch of the sync. pulse, i.e. at the start of each line, and a phase-locked loop (see Chapter 9) used to recover it. The input to the PLL is enabled only during the colour sub-carrier burst, and a fairly long loop timeconstant is used to ‘remember’ the phase for the rest of the line.

The standard takes ingenious advantage of the characteristics of human colour vision, which is far less sensitive to changes of hue in a scene, than to changes of brightness. Consequently, the two colour difference or chrominance signals only need to be broadcast at a much lower bandwidth than the mono-compatible brightness or luminance signal and are only of significant amplitude in highly coloured areas of the picture, resulting in a 525 line 30 fields/sec signal compatible with American monochrome sets on 60 Hz mains. This is because the luminance information does not completely blanket the video bandwidth, but is concentrated in narrow sidebands around each harmonic of the line timebase frequency. The exact colour sub carrier frequency is carefully chosen to minimize, even in highly coloured areas, effects such as dot crawl on monochrome pictures and ‘cross colour’ or ‘mixed highs’ resulting in false colour on e.g. striped jackets, on colour displays. NTSC is used in North America and some countries of South America, Japan and various other countries.

The later 625 lines/field PAL (phase-alternation line) was designed to minimize the effect of colour phase errors at the transmitter end, over the air and in the receiver, errors responsible for the ‘rainbow round my shoulder’ type of distortion sometimes seen on NTSC, leading to the jibe ‘Never Twice the Same Colour’. In PAL the phase of one of the two chrominance channels is reversed on alternate lines, as signalled by the phase of the colour burst, which is now no longer a constant. In early cheaper PAL receivers, this resulted in the hue errors being positive and negative on alternate lines, so that, viewed from a distance, large flat areas of colour still appeared correct. Nowadays, a glass electro-acoustic delay line providing a delay exactly equal to one line, makes alternative lines of any frame available simultaneously. They can thus be averaged before display, removing the effects of errors up to 40°, at the expense of some slight but unimportant reduction in vertical colour resolution. In PAL, a frame occupies 20 ms (one cycle of 50 Hz mains) and comprises 312.5 lines, leading to a 15.625 kHz line timebase frequency, as against 15.750 kHz for NTSC. In both standards, the odd line per field or half line per frame results in an interlaced picture (unlike the ‘progressive’ noninterlaced display of computers), minimizing flicker, despite the fact that there are only 25 complete pictures (fields) per second. The latest television sets use frame storage techniques to display a ‘progressive’ (non-interlaced) scan, providing 100 complete picture fields per second. Thus, like a computer display, picture flicker is eliminated completely.

There are half a dozen or more variations on the PAL standard, reflecting different combinations of channel spacing, video bandwidth, width of the vestigial video sideband, polarity of vision modulation and spacing between the vision and sound carriers. In I/PAL, used in the UK and some other countries, these parameters are respectively 8 MHz, 5.5 MHz, 1.25 MHz, negative and 6 MHz. The sound carrier carries a monophonic channel, joined in more recent years by a digital sound channel called NICAM (Near Instantaneously Companded Audio Multiplex, using QPSK modulation of a carrier 20 dB below the vision carrier) at a spacing from the video carrier of 6.552 MHz. In the UK, NICAM carries a near CD quality stereo sound signal, but in some countries is used for broadcasting monophonic sound in two different languages.

The various signals can be seen in Figure 8.5, showing an off-air signal at about 474 MHz, received in the author’s laboratory, at a dispersion of 1 MHz per division, 477 MHz display centre frequency, 10 dB per division vertical. Centred about the vision carrier, which is at three divisions left of centre, is the vision signal. On its left is the vestigial lower sideband, while on the right the full upper video side band appears, with some of its line structure just visible. One and a half divisions right of centre appears the colour subcarrier, 4.5 MHz above video carrier, and its size indicates that the picture content at the time was highly coloured, certainly not black and white. To the right of that is the sound subcarrier at 6 MHz above video, and to the right of that again, the NICAM signal.

The SECAM system (Sequentielle Couleur À Mémoire) used – in various of its subformats – in France and many other countries from Afghanistan to Zaire, is basically different from NTSC and PAL, in that it does not broadcast both colour difference signals on every line. A delay line makes both signals available simultaneously, albeit at the cost of halving vertical colour resolution, although this is not noticeable in practice. The single colour component on each line is broadcast as FM modulation of the colour subcarrier, a ‘cloche’ filter (one with a bell-shaped response curve) picking out the colour component to be fed to the colour demodulator.

All television formats are capable of bearing Teletext information, which is carried in some of the lines of the vertical blanking period. In the UK PAL system, possible teletext lines are 7 to 22 and 320 to 335, although lines 19, 20, 322 and 323 are used for test purposes, using ITS (Insertion Test Signal). Further details can be found in Ref. 10, which is doubtless out of date, but the BBC website proved less than helpful in locating any reference to the subject. Detailed information on the various world-wide TV Broadcasting Standards is given in References 11 and 12.

One of the problems encountered in television reception is ‘ghosting’, due to multipath reception. As well as the direct signal from the transmitter, other versions of it, reflected from large buildings, hills etc. may be received, with a corresponding time delay. The result is a feint second image, slightly displaced to the right relative to the main picture, the offset depending upon the delay. Digital television is in principle capable of giving a picture free from these and other distortions, provided the bit stream can be demodulated with a sufficiently low BER (bit error rate).

To provide adequate picture quality, even allowing for the considerable data compression provided by the various MPEG (motion picture experts group) standards, a high data rate is required. With a modulation scheme such as DPSK, QPSK or even one of the more exotic types, the symbol rate would be so high that inter-symbol interference due to multi-path would be a severe problem. OFDM (orthogonal frequency division multiplex) is a modulation scheme which achieves a high bit rate but a low symbol rate, and is therefore very resistant to multipath problems. Instead of trying to cram more and more bits onto each symbol, as in 64APK or 256APK, a large number of separate carriers are used, each with OOK (on-off keying) or BPSK (binary phase shift keying). Each modulated carrier exhibits a {sin(x)}/x or ‘sync’ spectrum, with frequency sidelobes, alternately positive and negative, and of decreasing amplitude with increasing offset, on either side of the carrier frequency. By choosing the distance between carrier frequencies, relative to the bit rate, the zeros between the sidelobes of any carrier fall on the other carrier frequencies, so that the signals are ‘orthogonal’ – non-interfering. Further details on OFDM can be found in Ref. 13.

At the receiver, the data on each carrier is recovered by performing a DFT (discrete Fourier transform) on the received signal, which was created in the first place, by the inverse process, an IDFT (inverse discrete Fourier transform) at the transmitter. At the transmit end, error correction coding is added to data, which is then interleaved between time slots and carriers for immunity to impulsive and CW interference, a signal format described as COFDM – coded orthogonal frequency division multiplex. European terrestrial television uses the DVB-T (digital video broadcast – terrestrial) standard, which specifies either 2048 or 8196 COFDM carriers within a standard 8 MHz TV channel. More recently – June 2004 – the DVB-H specification was developed. DVB-H stands for Digital Video Broadcasting to Handhelds and is specified in EN 302 304. It is basically an extension to the DVB-T standard and details can be found at www.DVB-H-ONLINE.org, while for further information on DVB-T itself visit www.dvb.org.

OFDM is also used in new digital radio systems. In Europe, new frequency allocations have been provided, and six stations or programmes are carried by a single transmitter. A major driving force behind digital radio has been the poor reception of FM usually encountered in moving vehicles, since the majority of radio listening is done in cars.

This arrangement, requiring new frequency allocations, is not suitable in the fragmented radio market in the USA, so OFDM is used, at a low signal level, for IBOC operation – the OFDM signal is transmitted ‘in band, on channel’ together with the existing analog signal, either AM on medium wave or FM on VHF. New receivers will receive the high quality digital signal when conditions permit, otherwise falling back to the analog signal, to provide ‘graceful degradation’. It is planned that when digital receivers achieve 85% market penetration, the analog component will be discontinued, and the full transmitter power made available to the digital signal.

OFDM is also used, under the name DMT (discrete multi tone) to provide ADSL (asymmetrical digital subscriber line) high speed modems for use over domestic phone lines. Another OFDM variant, using 16 carriers with modulation ranging from BPSK to 64-QAM per carrier, is used for high speed 5 GHz wireless networks, to the American IEEE 802.11a and European ETSI Hyperlan/2 standards.

For each type of modulation an appropriate demodulator is required in the receiver. Figure 8.6a shows a simple diode detector circuit for AM signals. The diode charges the RF bypass capacitor up to the peak voltage of the IF signal. A path to ground (or − V_s) is necessary to enable the voltage to fall again as the RF level falls on negative-going slopes of the modulating waveform. The detector circuit provides the demodulated audio frequency baseband signal varying about a dc level proportional to the strength of the carrier of the received signal. A capacitor blocks the dc level, passing only the audio to the volume control. The dc component across the RF bypass capacitor is extracted by a low-pass CR filter with typically a 100 ms time constant, and used as an AGC (automatic gain control) voltage to control the gain of the IF stages. This automatically compensates for variations of signal strength due to fading, and also ensures that weak and strong stations are all (apparently to the user) received at the same strength. Figure 8.6b shows one of the many forms of detector used for FM signals. A small winding closelycoupled to the primary of the discriminator transformer injects a signal V_ref, in phase with the primary voltage, at the centre tap of the secondary circuit, which is also tuned to 10.7 MHz. The secondary is very loosely magnetically coupled to the primary, so that the voltages V₁ and V₂ are in quadrature to the reference voltage when the frequency is exactly 10.7 MHz. As the frequency deviates about 10.7 MHz, V1 and V2 advance or retard (shown dotted) relative to V_ref, so the voltages VR₁ and VR₂ applied to the diodes become unequal, but R₁ and R₂ensure that the average of VR₁ and VR₂ is held at ground potential. Thus the recovered audio appears at point A – note that the capacitor to ground at A is a short circuit to IF but an open circuit at audio frequency. (This circuit, known as the ratio detector, was popular in valve receivers in the early days of FM broadcasting as it provides a considerable degree of AM suppression. Thus if the level of the IF signal were suddenly to rise and fall (e.g. due to reflections from a passing vehicle or plane), the damping imposed upon the secondary would rise and fall in sympathy as the make-up current required to keep CA charged to a higher or lower level varied. Modern FM receivers incorporate so much gain in the IF strip that they always operate with a hard-limited signal into the FM demodulator.) The recovered audio is deemphasized to provide the mono-compatible sum signal; the stereo decoder extracts the difference signal from the raw recovered audio at point A. Figure 8.6c shows an FM quadrature detector. Here again the signal across the tuned circuit is in quadrature with the drive voltage when the frequency is exactly 10.7 MHz and varies in phase about this in sympathy with the deviation. The phase detector output voltage thus varies about a steady dc level, in sympathy with the modulation.

Both the ratio and the quadrature FM detectors provide a dc output level which is proportional to the standing frequency offset of the IF signal from 10.7 MHz. This voltage is usually fed back to control a varicap diode in the receiver’s local oscillator circuit, in such a sense as to move the IF towards 10.7 MHz. This arrangement forms an AFC (automatic frequency control) loop, and if the loop gain is high, any residual mistuning is minimal. With the AFC in operation, as the receiver is slowly tuned across the band, it will snap onto a strong station and hold onto it until the receiver is tuned so far past it that the AFC range is exceeded, when it jumps out to the currently tuned frequency. It may thus be impossible to tune in a weak station on the adjacent channel to a strong one, so a switch is usually provided permitting the user to disable the AFC if required.

Detectors for QAM and other signals using both phase and amplitude modulation are designed to be sensitive to both amplitude and phase variations. They also incorporate symbol timing extraction circuitry to determine exactly when in each symbol period to sample the signal. If operating as coherent detectors, they also need a carrier regeneration circuit.

Spread spectrum (SS) is a term indicating any of several modes of modulation which may be used for special purposes. Conceptually, the simplest form of SS is FH (frequency hopping), where the transmit frequency is changed frequently, usually many times per second. The transmit frequencies are selected in a pseudo-random sequence either from a predefined set of frequencies or from a block of adjacent channels. There is a dead time between each short transmission or hop, typically of 10% of the hop dwell time, to allow the power to be ramped down and up again smoothly (avoiding spillage of spectral energy into adjacent channels) and to allow time for the synthesizer to change frequency. To minimize dead time, two synthesizers may be used alternately, allowing each a complete hop period to settle to its next frequency.

The main purpose of an FH system is to provide security of the link against eavesdropping and exploitation, typically in an ‘all-informed net’ structure for tactical communications. Every station in the net will know the set of frequencies to be used and the PRBS (pseudo-random bit sequence); they also have pre-synchronized clocks driven from accurate frequency references, giving them a guide to the phase of the PRBS to within a few bit periods at worst. Periodic transmission of timing signals enables a late entrant to acquire net timing. By contrast, an adversary trying to penetrate the net does not know the set of frequencies in use and does not know the PRBS (which may be changed frequently for further security), let alone its phase.

An FH system typically uses digital modulation, even though the traffic may be speech, which will be digitized and probably also encrypted. The bit rate over the air will be a little faster than the voice digitization rate, to allow for the dead periods; a FIFO (first in – first out memory) at the receiver reconstituting the original data rate. In order to receive the data transmitted during any one hop, the received signal to noise ratio in that particular channel must be at least as good as in a non-hopping link. Interference or jamming may wipe out any particular hop, but speech contains so much redundancy that up to 10% blocked channels is no disaster, especially at VHF where a higher hopping rate of several hundred per second (compared to nearer 10 hops/s at HF) can be used. Even jamming an FH system poses problems for an adversary; not knowing the exact channels in use, let alone their sequence, he must spread his available jamming power over the whole band. It will thus be much less effective than if he had been able to concentrate it on a single channel transmission.

The other type of SS is DS (direct sequence) spreading. This is used at VHF and UHF and is more versatile than FH. Whereas FH uses only one channel at a time, SS uses the whole band the whole of the time. This is achieved by deliberately increasing the bit rate and hence the bandwidth of the transmitted data. For example, the baseband bandwidth of a 100 kb/s data stream is 50 kHz, giving a minimum bandwidth needed for the PSK modulated transmission of 100 kHz. However, if each successive data symbol (bit) is exclusive ORed with a 10 Mb/s PRBS prior to PSK modulation, the transmitted bandwidth will now be 10 MHz. The PRBS does not repeat exactly each symbol; each symbol is multiplied by the next 100 bits of a very long PRBS. The PRBS is called the ‘chipping sequence’ and in the example given there are 100 chips per symbol. In the receiver, the signal is multiplied by the same PRBS in the correct phase, e.g. at IF using a double balanced mixer or a SAW convolver. This has the effect of de-spreading the energy and concentrating it all back into the original bandwidth. The received signal strength is thus increased by the amount of the ‘processing gain’, which in the example given is ×100 or 20 dB. By constrast, any interference such as a large CW or narrow band signal is spread out by the chipping sequence. Thus the signal can be successfully received even though the RF signal at the antenna is many decibels below noise and interference. The receiver in a DS spreading system has to acquire both symbol and bit (chip) sync in order to recover the transmitted data, by means much as described above for an FH system. Eavesdropping is even more difficult, since an adversary will not even know that a transmission is taking place if the signal in space is below noise.