The development of CD offered truly digital technology for the first time and set new standards in sound quality. Digital recording achieves high quality with low cost by using measured values instead of the analogue signal manipulation.
Digital technology describes the music signal in the form of series of numbers in binary notation, called bits. The bits are recorded on the disc in small grooves, called pits and lands. Pits and lands represent 0's and 1's, respectively. From the digital numbers, the CD computer reconstructs the original audio signal. The measuring of the audio signal bit by bit is called sampling.
Bits make up digital words (numbers in binary notation). Binary notation has the base 2, and uses only 0's and 1's. Computers use binary notation since they can read a circuit as "on" or "off", as a "0" or as a "1".
16-bit coding means that 16 digits are used in forming the digital words. The sampling frequency is how often this value is measured.
The Compact Disc player.
The laser beam reads
coded information from the compact disc. The reading is kept accurate by
the servo processor. The data from the reading is sent to the decoder, where
it is converted to regular digital information. A digital
filter then removes noise. The DAC, the most important part of the CD player,
converts the digital data to an analogue audio wave. After an analogue filter
(not shown in the figure) removes noise, this wave is sent to the loudspeakers
(L and R) for reproduction as sound. The microprocessor controls features,
such as volume, balance, tone, etc.
The Digital-Analogue Converter (DAC) takes the digital word as input. It uses a series of comparators to measure the value of the word. One comparator (or a combination of comparators) comes "on" at a given signal strength. The signal measurement must be very accurate, since any error in the Least Significant Bit (LSB) means a large deviation in the Most Significant Bit (MSB). Two DACs may be applied to end small phase errors in the signals of different channels.
A compact disc is scanned by a laser beam. The laser's light is reflected from a land to a photo-electric cell.. The cell emits (give or send out) current and a 1 is registered. When the beam shines on a pit, half of the light is reflected from the surface and half from the depth of the pit. The interference between the two reflected beams eliminates the original beam. The photo- electric cell emits no current, and a 0 is registered.
Error correctionworks through parity bits. Groups of bits are such that adding certain series of them has only one result, eg, a 1. If a 1 drops out, the result of the addition is different. A parity bit is then added to correct the error.
Oversampling is a simple, inexpensive device to filter out conversion side effects. Oversampling takes the sampling farther away from the audible range (from 44.1 to 176.4 kHz), so that ringing is eliminated. Oversampling is another technology from Philips which has helped to achieve audio excellence. CDs may be enhanced by features such as track, time and index indications.
The disc is read on the reflecting side, but the more vulnerable side is the label side. Care should always be taken when handling compact discs. Some CDs are recorded and/or processed with analogue methods. The codes DDD, ADD, and AAD are used to indicate (respectively) completely digitally made discs, discs with analogue recording and digital editing and dubbing, and discs with analogue recording and editing, and only digital dubbing. ADD and AAD discs are not necessarily of inferior quality compared to Ddd discs.
VideoCD is an extension of the capabilities of CD, used for video.
For the HiFi dealer and the consumer, the compact disc player is the first piece of equipment to which the term "digital" may be applied correctly and fully.
Since the introduction of the CD, "digital" has become the vogue word. It is even used loosely of amplifiers and tuners. Digitalisation is coming, slowly but surely. What does the term "digital" actually mean? To find the answer, let's look at compact disc technology.
Digital recording of sound is intended to achieve superb sound reproduction with less cost. Superb reproduction has been possible using analogue methods, but only with a great deal of cost and expertise. The secret of the digital advance in quality is in the fact that digital recording does not manipulate the music signal itself, as is the case with analogue processing. Digital technology instead uses measured values to record sound.
Figure 1: Digital technology uses measured values to record sound.
An analogue signal takes the form of a wave. At different times therefore, its amplitude will vary. When an analogue signal is converted into a digital one, the signal is measured at regular intervals. These measurements (or measured values) are then converted into digital (binary) numbers (a series of '0's' and '1's').
Example of a fragment
of CD music encoded
There are in fact a number of different kinds of digital-analogue conversion in various CD players: 16-, 18-, 20-, or just 1-bit con- version.
The sampling frequency is the number of times a signal is measured per second. The number of bits determines the accuracy with which the measured value is converted into binary notation.
The number of bits used determines the accuracy of measurement, since it determines how much information is in each sampling. The more bits, the greater the accuracy. With an 8-bit system, the waveform is rough. Therefore, music is commonly coded on CD in 16 bits.
That is why the CD
player operates with the higher value of 16 bits, and why nearly all CD players
are equipped with a 16-bit Digital Analogue Convertor (DAC). CD players with
16-bit DACs can detect 65,536 different voltage values (0 + 20 + 21 + 22
Figure 2: The CD player operates with the higher value of 16 bits.
Decimal notation uses the base of 10, and is therefore also called Base 10 notation. The base is the exponent which increases with each place to the left. In decimal notation, each place has a value which corresponds to a power of the base 10. Numbers are formed by positioning powers of the base 10. Compare the decimal notation numbers 1, 10, 100 and 1,000. The figure "1" means "1 x 100" when placed on the far right (= 1). In the second place from the right it means "1 x 101" (= 10). In the third place from the right it means "1 x 100" (= 100), and in the fourth place from the right it means "1 x 103" (= 1,000). The power, represented by the exponent, increases by 1 with each place to the left (101, 102, 103, etc.). In decimal (base 10) notation, the power is a power of 10. The base is 10. So when the figure moves one place to the left and the power increases by 1, the figure is worth 10 times more.
The calculation of
the number 231 in decimal notation would therefore be:
(2 x 10E+02) + (3 x 10E+01) + (1 x 10E+00) = (2 x 100) + (3 x 10) + (1 x 1) = 200 + 30 + 1 = 231
For our discussion of binary notation, we will call the power which increases by position the weight of the figure. Consider the binary figure: 1011. Converting this number into our common decimal notation is helpful for understanding what it means. The weights of the figures now are not powers of 10, but powers of 2. So when the power increases by 1, the figure is worth 2 times more.
To convert the number 1011 from binary notation into
decimal notation, use this formula for each figure:
(the figure: "1" or "0") x (the base "2" to the power of the place) 1011 (1 x 23) + (0 x 22) + (1 x 21) + (1 x 20) = (1 x 8) + (0 x 4) + (1 x 2) + (1 x 1) = 8 + 0 + 2 + 1 = 11.
In the binary number 1011, the right "1" has the lowest
weight and the left "1" has the highest weight. They are therefore called
the Least Significant Bit (LSB) (the bit carrying the least weight) and the
Most Significant Bit (MSB) (the bit carrying the most weight).
1 0 1 1 | | MSB LSB
Computers, including the compact disc player, read electrical pulses. They interpret an electrical pulse to mean one of the figures of binary notation, a "1". They understand a lack of an electrical pulse to mean a "0". The consequence of having to convert these measured values to "1's" and "0's" is that long series of figures are needed. However, modern computer technology has no difficulty with long series of figures. All these series of data are recorded on the CD in the form of microscopically small pits and lands. At playback the pits and lands are scanned by a laserbeam. Pits are interpreted as "0's" and lands as "1's".
Suppose that, during recording, a volt metre is used
for measuring the analogue signal. This volt metre only indicates whether
the voltage is "negative" or "positive". This is a 1-bit system, the simplest
system possible. This is a very coarse and slow system which is unable to
transmit any signal details. A wavelike signal, of any form whatsoever, becomes
a block wave with such a simple converter.
Figure 3: A 1-bit system transmits only block waves
Instead of a volt metre, digital comparators are used in the digital- analogue converter (DAC). Each comparator tells whether the voltage being measured is higher or lower than a set value. The comparators, each adjusted to a different level, are connected in series. To illustrate, imagine four comparators called A, B, C and D. A is adjusted to 1 V (=20). B is adjusted to 2 V (=21), C to 4 V (=22) and D to 8 V (=23). (Remember that we are working with base 2 notation.)
The value of the signal is 1 V at the moment the measuring sample is taken (the moment of the sample pulse). The output of the 1 V comparator (A) is high: it emits a signal, a "1" in binary notation. The output of the next few comparators (B, C and D), which are all adjusted to a somewhat higher level, is 0. A digital signal is formed--a digital word, 0001.
At a signal value of 2 V, the comparator B has a high output. The digital word 0010 is therefore formed. At a voltage of 5 V (4+1), a high output appears at comparator A (for the "1") and C (for the "4"). The other two remain 0. The digital word is 0101. Remember that, in binary notation, each place to the left is worth one power of 2 more than the value immediately to its right. (The base is 2.)
That is how the A-D converter works during recording. In this example, there are four comparators. This is a 4-bit system for measuring different voltage values. Only steps of 1 V can be measured. The converter cannot be adjusted for smaller differences. The 4-bit system is too inaccurate to produce enjoyable music, so more than four digital comparators are commonly used. The system then has more than four bits. The absolute minimum for reproducing music is eight bits. The 8-bit system can measure 256 different voltage values (0 + 20 + 21 22 + ...27). Quite a few more than the 4-bit system!
In digitising an audio signal of which the maximum voltage is 1 V, the 8-bit system can only measure differences larger than .003906 V (1/256). This is still coarse, although it is effective in practice. With the 8- bit system, the wave form is rough after digital-analogue conversion.
Figure 5: With the 8-bit system,
the waveform is rough after digital-analogue conversion.
As we saw earlier, the LSB is the bit carrying the least weight and the MSB is the bit carrying the most weight. The MSB is the first (on the far left) and most important bit of a digital word. In a 16-bit binary number, the MSB is 215 (=32,768) times bigger than the LSB (remembering that each place to the left in a binary number is worth 1 power of 2 more than the place to its right). An error of only 0.01% will shatter the value of the 15th and 16th bits completely. That is why every measurement must be very accurate. (see figure 2 above).
Just as the analogue-digital converter determines the quality of the digital recording, the DAC determines the quality of playback. With the standard 16-bit converter, the DAC converts the 16 bits of coded information into a series of voltages through a 16-step resistor ladder network to which a reference voltage is connected. This series of voltages is added by an "operational amplifier", which generates a certain output voltage.
Because of the analogue character of the resistors in the DAC, the accuracy is limited. Such a DAC, when going from a "0" to a "1" and vice versa, generates zero-crossing (or cross-over) distortion (distortion caused by amplifying the positive and negative half cycles separately), non-linear distortion and harmonic distortion. As a result, the sound loses part of its clarity and, especially in the silent passages, irregularities such as noise are heard.
The number of bits determines what the highest attainable, or theoretical, S/N ratio is. With 16 bits, this ratio is 96 dB S/N, allowing the CD to have a larger dynamic range than any other medium.
The S/N ratio is not a constant; it varies with the signal level. At low sound levels, the S/N ratio is lower; there is more noise. This can be heard in CD players of inferior quality. They produce a soft background noise as soon as they amplify even very weak signals. This is quantisation noise. Quantisation is the conversion of digital information to an analogue signal. Quantisation noise is the result of the digital system working with voltage steps.
With the analogue
tape or cassette recorder something similar happens. When the system is recording
a tone, small side bands may be generated to the left and to the right of
the actual frequency. These side bands cause modulation noise. With
the digital signal, the many more voltage sources cause more side bands to
be generated. Some side bands are generated at very low levels, yet nevertheless
they may become quite audible, if not as noise, then as a certain coarseness
Music is recorded on compact disc with 16 bits, so it may seem impossible to play the music back more accurately than with a 16-bit CD player. But in practice it is not even possible to use this 16-bit information capacity to a full 100%.
This is the reason why some converters are designed to work with more than 16 bits. These converters represent many manufacturers' attempts to reach the theoretical 100% use of information capacity as closely as possible. Most manufacturers try to do this by developing 18-bit, 20-bit or even 22-bit converters, and by oversampling (a technique to filter out noise, to be discussed later in this chapter). These refinements in the form of 18 or 20 bits aim at enlarging the S/N ratio and at constructing even more accurately the waveform to be transmitted by the DAC.
In theory, converters with more capacity and/or oversampling allow some gains in quality. In practice, however, it appears that the performance of these converters is often below that of a good 16-bit converter, because of their complexity. Even though Philips CD has obtained very good results with a 16-bit converter, Philips has developed a new converter to take one step closer to perfection. This new technique is called 1-bit Bitstream conversion.
The 16-bit DAC requires 16 transistors (current dividers), one for each bit. The currents are switched "on" or "off" according to the digital values (1 or 0) of the bits. With 16 bits, there are 65,536 possible current values. The current values range from half the total current for the MSB to 1/32, 768th of the total for the LSB.
Any variation of more than a fraction of the LSB in any of these current values causes distortion and non-linearities. Noise is generated when the switches fail to operate in perfect synchronisation. Philips TDA 1541 series of multi-bit converters use the patented Dynamic Element matching principle to adjust the bit current values constantly and automatically, in order to maintain maximum performance.
But the Philips Pulse Density Modulation (PDM) Single-Bit D-A Conversion is an even better solution. In Philips PDM conversion, the 16-bit digital samples read from the CD are transformed into a high-speed, one-bit data stream (with 256 times oversampling). This stream is then converted into an analogue signal by a digital single-bit converter. This technique eliminates the non-linearities and distortion for which there is no ultimate solution in multi-bit converters.
The Philips single-bit converter generates positive and negative current values more than 11 million times per second, in accordance with the 1's and 0's of the high-speed bitstream. The ratio of 1's to 0's determines the actual level of the current. A data stream of all 1's produces the maximum positive current, and a data stream of all 0's results in the maximum negative current. Alternating 1's and 0's results in no current. The high speed of this single-bit conversion is what makes bitstream much better than the coarse, slow 1-bit system we mentioned earlier.
The single bit bitstream converter eliminates zero crossing distortion, quantisation noise and other distortion producing better quality sound. What these converters produce is a wide stereo image and almost tangible depth. Philips bitstream technology is a new level of excellence in D-A conversion.
technology is used to prevent problems like component non-linearities and
cross-over distortion (zero crossing). 1-bit DAC uses a high oversampling
rate of 256 times. In the process the normal 16-bit input signal with a sampling
frequency of 44.1 kHz is converted into a 1-bit data stream with a frequency
of 11.2896 MHz. Since this 1-bit DAC must choose from one of only two possible
values, instead of 65,536 which are possible in a 16-bit converter, there
can be no errors.
In order to avoid these kinds of side effects, oversampling is now used. Sometimes two, three, or even four oversampling devices are applied in one machine. The great value of oversampling lies in the possibility of applying a simple filter that leaves the high quality of the signal intact. Due to oversampling, a low-priced digital filter can be applied.
As a CD player reads information from the disc, the information is converted from its digital form ("1's" and "0's") to its analogue form (a series of electrical current values). This conversion, called quantisation, causes noise. So before sending the signal to the loudspeakers, it is necessary to remove the quantisation noise with a filter. However, some types of filters affect the quality of the music reproduction.
The brick-wall filters employed in many first-generation CD players cut off any information above a certain point, 24 kHz. Unfortunately, these filters were themselves found to cause ringing, a type of distortion. Ringing is actually a remainder of the sampling frequency in the output signal. Ringing causes the stereo image to become unclear.
The technology of oversampling, developed by Philips, virtually eliminates this problem through ultrasonic noise reduction. If the problem of ringing with brick-wall filters lies in their proximity to the audible range, why not move the filter further away from the audible range? By oversampling the digital data four times (at 176.4 kHz instead of 44.1 kHz), the quantisation noise is moved further away from the audible range. The noise is then removed by an analogue filter that also eliminates ringing. The rough analogue wave form of simple sampling is refined by oversampling.
Figure 11: By oversampling the digital data four times (at 176 kHz instead of 44.1 kHz), the quantisation noise is moved further away from the audible range. The noise is then removed by an analogue filter.
Figure 12: The rough analogue waveform of simple sampling is refined by oversampling.
In practice, 2-time oversampling works very well, as many CD players prove. Even better is 4-time oversampling which most higher-quality CD players have. The more expensive models aim at 8-time or even 16-time oversampling, to attain even further refinement. Some CD players even have a 256-time oversampling system!
Higher oversampling levels are not, however, always better. Sampling systems which make use of higher rates can actually degrade the quality of music produced, since many DACs cannot "settle" from one input value to the next. Converters can only process a certain amount of information in a given period of time. Beyond a certain point, the faster the process is attempted, the more errors are introduced. Imagine that a person is told to compute the equation 2 + 2 in ten seconds. This is a simple matter. But if the same person must calculate a series of 20 equations in the same ten seconds, the chance of error is much greater. So it is with the hundreds of thousands of calculations which the DAC must perform. Beyond a certain point, sound quality is diminished rather than increased.
The digital signal transmitted by the CD contains the complete stereo information, both the left (L) and right (R) channels. This information is passed on alternately. After oversampling, the digital signal reaches the DAC, which converts it into an analogue signal. If only one DAC is applied, as is the case with many simple CD players, the signal is very quickly switched from L to R behind the DAC. This system appears to work well at first sight, but it results in a small time difference between L and R. This small time delay leads to phase shifts. Human ears are very sensitive to these differences, however small they may be. Stereo infor- mation is located precisely in these phase relations of the high tones. Through faults in the conversion procedure, a considerable part of the real depth of the stereo image is lost.
If, on the other hand, two DACs are applied, things will be quite diffe- rent! At playback, the same phase relation as was recorded by the microphones will be preserved exactly. Then the playback will have real spaciousness, depth and complete naturalness. A double DAC is an essenti- al feature of a good CD player.
Left and Right signals do not reach the DACs at the same time. By inserting a delay circuit in one of the two channels, the delay of the second signal is perfectly counterbalanced. Because this is done in the digital section, the approach is the same for all frequencies, so that higher frequencies are never delayed more than the lower frequencies.
Behind the single or double DAC there is a single or double analogue filter, as we saw above. We have the intact sound signal at our disposal, after it passes through this filter. We only have to connect the CD output to the AUX or CD input of the stereo amplifier.