The Story of Digital Transmission
Section 5: Sound into Pulses: The Benefits of Digital Transmission
One of the main advantages of digital technology (see section 4) is in the area of the transmission or communication of information from one point to another. In the sense we will use it here, information can be any value, measurement or signal we want to transfer. It can be, for example, the human voice or the sound of music.
Keen stereo listeners like to talk about ''hi-fi'' - short for "high fidelity'' - by which they mean the reproduction of music while faithfully preserving its quality as much as possible. When you are sitting in front of a band or an orchestra, you are hearing the music directly, just as it comes from the instruments, in the highest fidelity possible. The problem faced by a recording company is how to preserve this fidelity and re-create it in your living room when you play a CD or record. Ideally, you should be able to sit back and close your eyes and hear the sound exactly as you would if the band was there in front of you.
In reality, there are a number of problems that prevent the achievement of this ideal - noise, distortion, compression of dynamic range and so on. For our discussion, we'll concentrate on the problem of noise.
Noise is any sound on the CD or record that wasn't there at the performance during the recording session. More generally, it is any unwanted signal that adds on to the information being transmitted. When a vinyl record is being made, noise is introduced at every step of the recording process, although of course the company makes every effort to reduce such noise to as low a level as possible. The sound that reaches the microphones is converted into an electrical signal which is then recorded on a wide magnetic tape moving at high speed. This tape is then used to control the cutting of a master disc, from which moulds are then made. These in turn are used to mass-produce the records which are eventually sold in shops. Noise is produced at every step, not forgetting that introduced by your own stereo equipment. It can never be entirely eliminated.
The same problems of noise are shared by any method of transmitting information, and certainly by telecommunications, including telephone calls.
In the production of vinyl records, companies have used purely analogue (see section 4) means to transfer the information representing the sound of the music from one point to another. That is, they use an electrical signal that changes smoothly in strength, exactly modelling the smooth but complex changes in the sound.
When noise is created in the recording process - because of tape hiss, dust on the master disc, electrical interference or any other cause - this is added on as a random signal on top of the complex electrical signal representing the sound. There is no way that electronic equipment can tell such random noise from the original electrical signal, so there is no way it can be removed again without removing some of the original signal.
We can see this more clearly if we draw a graph of the level of the analogue audio signal over a period of time (diagram 1a). The shape of this graph represents both the changes in the electrical sound and the changes in the electrical signal that model it. Now if we add to this audio signal some random noise, this affects the shape of the signal, and this degrades the sound that your stereo reproduces (diagram 1b).
The trouble with an analogue audio signal is that its exact shape has to be preserved if you are to hear the music exactly as it was when it was played. If there was a means of transmitting the signal so that only the overall shape of the signal mattered, then noise would not be so important.
Such a method exists in digital encoding. One popular form of digital encoding is called pulse code modulation (PCM)( see section 7), for reasons we will discuss later. In this method, the original sound is monitored by electronic equipment and at very short intervals of time, a sample is taken of the level of the audio signal at that moment. This information is then converted to a binary number (see section 3) which is then transmitted as a series of on-or-off pulses.
To make this easier to understand, consider this analogy. Let's imagine the way that the port authorities used to find the shape of the bottom of the harbour, so that ships could navigate more safely. It certainly wasn't possible to drain the harbour and take a photograph of it, so what they did instead was send out a boat which travelled slowly across the harbour. Every few metres a person at the back of the boat dropped down a plumb-line (a weight at the end of a rope), until it reached the bottom of the harbour. The line had knots tied in it at regular spaces and the person called out the number of knots under water, so indicating the depth of the harbour at that point. A clerk wrote these down, and eventually it was possible for him to draw a graph of the shape of the harbour by using these numbers.
The person in the boat had been taking samples of the depth of the harbour at frequent intervals, so that the graph would accurately describe the ups and downs of the harbour bottom. In just the same way, the electronic equipment used in pulse code modulation takes a sample of the level of the audio signal, and converts these measurements into numbers, or digits. See Diagram 2. It is important to note at this point that because the measurement is represented by a limited number of digits, there may be some slight loss of accuracy (in our analogy, the depth of the harbour is indicated by whole numbers of knots under water, not by fractions or parts of the distance between knots).
These numbers are then coded into a series of on-or-off electrical pulses (in some systems they may be strictly high-or-low pulses), in very much the same way that letters of the English language were coded into the dots and dashes of Morse Code. These pulses are then transmitted.
Now, although noise will still occur during this transfer of information, exactly as it does in analogue transmission, it has little effect unless the noise level is very high. The reason for this is that now the receiving device only has to examine the incoming signal to determine if a pulse is present at a particular moment or not. The exact level of the signal is not important at all, so long as an "on" condition can still be clearly distinguishable from an "off" condition. (See Diagram 3a and 3b.) The original signal can then be reconstructed exactly. (See Diagram 3c.)
However, we must remember that the information transmitted by digital methods can only approximate the original analogue sound. There are two reasons for this. Firstly, as we have mentioned, the level of the audio signal can only be represented by a limited number of digits, restricting the accuracy with which it can be measured. Secondly, the original signal is only sampled at the end of discrete intervals of time, and we know nothing of what is happening to the signal between these intervals. For this reason, the sampling must take place at very short intervals.
It turns out that the sampling must occur at least twice as often as the fastest changes in the audio signal. In technical terms, the frequency of the sampling must be at least twice that of the highest frequency in the audio signal to be transmitted. For "hi-fi", this means the sampling must occur 50,000 times a second! This is possible with modern electronic methods and, in practice, the errors owing to digital encoding can be reduced to a negligible amount, and the problem of noise is almost entirely eliminated.
Record companies increasingly use digital methods for the recording process. Of course, with an audio signal, these digital pulses must eventually be converted back to an analogue form, because this is how your ears work! This is not too difficult. The quality of such "digital recordings" can exceed that of more conventionally recorded records. Now, home stereo equipment can reproduce discs which are entirely digital in manufacture. Compact discs do not have the familiar wiggly spiral of vinyl records, but have a series of tiny pits in their surface, scanned by lasers.
Pulse code modulation also has many benefits in the telephone system, where noise can substantially degrade the quality and understandability of telephone conversation. For voice transmission, the sampling rate need only be about 8000 times a second. Telstra now installs digital transmission systems between exchanges. This means it is likely that when you next make a phone call, your voice will be conveyed at least part of the way by a digital system, sampling the electrical signal which represents your voice, breaking it up into binary numbers, and transmitting it as a series of electrical pulses.
Apart from the benefits of reducing noise and cross-talk (the interference between two neighbouring telephone lines), pulse code modulation allows many more telephone conversations to be transmitted along the same set of wires.
This is possible using "time division multiplexing". Although this sounds rather formidable, it is not difficult to understand. You will recall that your voice is sampled 8000 times a second in pulse code modulation. This means, of course, that a period of 1/8000th of a second goes by between the time it is sampled once and the next time it is sampled.
Now, although 1/8000th of a second is a very short time to humans, for modern electronic systems it is quite a long time. In that time, in a basic PCM system, it is possible not only to sample your voice and transmit a binary number representing that sample, but also to transmit another 31 binary numbers before the system needs to come back again and sample your voice. This means that the binary numbers representing samples of your speech can be interleaved with 31 others. Two of these numbers carry special signalling information, but the rest can be used to carry telephone conversations. So including your own conversation, one telephone line can be used to carry 30 conversations simultaneously.
Imagine 30 people all talking on the phone at the same time, with the electronic system going around to them each in turn, taking a sample of their speech and sending it on. At the receiving end, the interleaved binary numbers are analysed each in turn, the different voices reconstructed and sent to 30 different listeners, all without ever mixing up the conversations!
