An important parameter of message relaying between a source and a destination is whether the message is analog or digital. These terms relate to the nature of the message and can characterize either the transmitted data or the form of the actual signal used to carry the message. Thus, we have analog and digital data, as well as analog and digital signals. Analog and digital signal representations are shown in Figures 2.12 and 2.13, respectively. The
Wireless Communications Principles and Fundamentals 41
Figure 2.11 Transition diagram of a Markov chain
difference is obvious: analog signals take continuous values in time whereas digital ones change between certain levels at specific time positions. In the following we discuss and compare analog and digital data representation, while the basic modulation methods for wireless networks, which are used to transmit the signal over the wireless medium, are discussed in Section 2.5.
The vast majority of the early radio communication systems concerned sound transmis- sion. Television transmission comprises two analog components, corresponding to sound and image. Moreover, the only service offered by early cellular systems (e.g. Advanced Mobile
Figure 2.12 Analog signal
Figure 2.13 Digital signal
Phone System, AMPS) was voice conversation. Thus, all these systems represented the information to be transmitted in an analog form since the physical nature of both sound and image is analog. However, modern wireless systems are increasingly being used for computer data communications, such as file transfer. The natural form of such data is digital, thus digital representation is used. There is a trend towards digital representation of analog data, which stems from the inherent advantages of digital over analog technology. These advantages are briefly summarized below:
† Transmission reliability. Transmission of a message through a medium is generally degraded by noise, which is more or less present in all communication mediums. As mentioned earlier, noise causes bit errors and BERs of wireless channels are significantly higher than those of wired channels. The digital representation of a message increases the tolerance of a wireless system to noise. This is due to the fact that, as seen from Figure 2.13, a digital signal is not continuous but rather comprises a number of levels. As a result, in order for noise to alter the message content, it has to be strong enough to change the signal level to another one. Furthermore, digital messages can be accompanied by addi- tional bits, called checksum bits. The actual content of these bits is based on error detect- ing/correcting algorithms and the procedure is known as Forward Error Correction (FEC).
An error detection algorithm works by appending extra bits to a binary message in a way that the receiver can use the received bits and determine whether or not a bit error has occurred and thus, request a retransmission if needed. Error correction algorithms work in the same way, however, in this case the receiver has the ability not only to detect but also to correct bit errors. The Hamming code is a widely known technique used both for error correction and detection.
† Efficient use of spectrum.The above mentioned increased noise tolerance of digital repre- sentation helps increase the amount of information that can be transmitted using a wireless channel. This is because less errors are likely to occur due to the applied coding. Thus, for a given amount of spectrum and a certain time period, more information can be transmitted by using digital representation – a fact that results to a more efficient use of the spectrum.
Furthermore, digital data can be compressed easily which increases spectrum efficiency even more.
† Security.Wireless channels are probably the most easy to eavesdrop on, therefore security is a crucial issue in such systems. Analog systems can be provided with a certain level of security, however, these have proved easy to crack. Digital data, on the other hand, can be easily and efficiently encrypted even up to a point that makes unauthorized decryption of the message almost impossible. Furthermore, encryption does not come at any expense to the spectral efficiency of the system, meaning than an encrypted message can be trans- mitted over the same bandwidth required for unencrypted transmission of the same message.
2.4.1 Voice Coding
While the trend in modern wireless networks is towards data communications, the demand for voice-related services such as traditional mobile phone calls is expected to continue to exist.
Thus voice needs to be converted from its analog form to a digital form that will be trans- mitted over the digital wireless network. The devices that perform this operation are known as
Wireless Communications Principles and Fundamentals 43
codecs (coder/decoder) and have been used mainly in mobile phones. Codecs aim to convert voice into a digital bit stream that has the lowest possible bit rate while maintaining an acceptable quality.
A codec can convert an analog speech signal to its digital representation by sampling the analog signal at regular time intervals. This method is known as Pulse Code Modulation (PCM) and is used in codecs of PSTN and CD systems. There is a direct relationship between the number of samples per second, W, and the width,H, of the analog signal we want to digitize. This is given in the following equation, which tells us that when we want to digitize an analog signal of width,H,there is no point in sampling faster thanW:
W ẳ2H bps ð2:7ị
The process of PCM conversion of an analog signal to a digital one comprises three stages:
† Sampling of the analog signal. This produces a series of samples, known as Pulse Ampli- tude Modulation (PAM) pulses, with amplitude proportional to the original signal. The PAM pulses produced after sampling of an analog signal are shown in Figure 2.14.
† Quantizing.This is essentially the splitting of the effective amplitude range of the analog signal to V levels which are used for approximating the PAM pulses. These V levels (known as quantizing levels) are selected as the median values between various equally spaced signal levels. The quantization of the PAM pulses of Figure 2.14 is shown in Figure 2.15. Quantization obviously distorts the original signal since some information is lost due to approximation. The more the quantizing levels, the less the distortion since the approx- imation with many levels is more precise. Good voice digitization by PCM is achieved for 128 quantization levels. The distortion due to quantization is known as quantizing noise and is given by the following formula [5]:
S
N ẳ6V11:8 dB ð2:8ị
Figure 2.14 PAM pulses created by sampling of the analog signal
† Binary encoding.This is encoding of the quantized values of PAM to binary format, which forms the output of the PCM system and will be used to modulate the signal to be transmitted. For the quantized PAM pulses of Figure 2.15 four bits are used per PCM sample coding (since nine levels can be encoded by four bits) the binary output is 0110011001000011010001111001100 00011.
PCM demands relatively high bit rates and is thus not very useful for wireless commu- nications systems, such as mobile phones. A number of techniques exist that are refinements of PCM and try both to increase voice quality and decrease the output bit rate. PCM with nonlinear encoding takes into account the fact that PCM will produce a largely distorted signal when the effective amplitude of the sampled analog signal is relatively small compared to the amplitude covered by the PCM quantizing levels. Therefore, nonlinear encoding use more levels for such signals – a fact that reduces quantizing noise. For voice signals 24–30 dB S/N improvements have been achieved. Differential PCM (DPCM) outputs the binary repre- sentation of the difference between consecutive PCM samples rather than the samples them- selves. When x bits are used for encoding the differences, the method is known as x-bit DPCM. The method for xẳ1 is known as Delta modulation. DPCM schemes obviously reduce the bit rate produced if the differences between samples can be encoded using less bits than those required for encoding the actual samples. However, DPCM techniques have poor performance when steep changes occur in the analog signal. Adaptive DPCM (ADPCM) tries to predict the value of a sample based on previous sample values. ADPCM helps reduce the bit rate down to 16 kbps while still maintaining acceptable voice quality. The following chapters show that 16 kbps is still a large value for mobile phones, however, prediction is used in conjunction with other techniques in mobile phone codecs to lower the bit rate.
2.4.1.1 Vocoders and hybrid codecs
In an effort to reduce the bit rate required for voice transmission, engineers have exploited the actual structure and operation of human speech production organs and the devices that work
Wireless Communications Principles and Fundamentals 45
Figure 2.15 PCM pulses produced by quantization
based on this are known as vocoders. Vocoders, which were initially only an attempt to synthesize speech, work by encoding not the actual voice signals but rather by modeling the mechanics of how sounds are produced (such as mouth movement, voice pitch, etc.). By encoding and transmitting this information the signal can be reconstructed at the receiver.
A simple vocoder diagram is shown in Figure 2.16. It comprises three parts:
† the part responsible for coding vowel sounds, which are attributed to the vocal cords;
† the part responsible for coding consonant sounds, which are produced by lips, teeth, etc.;
† the part that is responsible for coding the effects of the throat and nose on the speech signal.
Vocoders are very useful since they achieve voice transfer with a low bit rate. ‘Full-rate’
vocoders produce a compressed voice signal of 13 kbps while half-rate vocoders sacrifice some quality and achieve a rate of 8 kbps. Furthermore, there are vocoders that can serve bandwidth-limited scenarios, such as military and space communications. Over the low bandwidth channels of such applications, these vocoders can achieve voice transmission with very low bit rates, as low as 1.2–2.4 kbps. However, the voice produced is not very
‘natural’ and has a somewhat ‘artificial’ quality. In some cases it is even difficult to tell who is actually speaking. Hybrid codecs try to overcome this problem by transmitting both vocoding and PCM voice information while also making sure that sounds that are inaudible to the human ear are not transmitted. An example of such a sound is that of a quiet musical instrument in the background of a loud one. Furthermore, codecs that vary the bit rate according to the characteristics of speech sounds have been produced.