January 21, 2016 at 12:04Digital Sound: DSD vs PCM
Digital sound. How many myths revolve around this phrase. How many disputes arose between lovers of the convenience and quality of numbers and adherents of the “live air” vinyl sound multiplied by the “warm tube” sound. In addition, there are many disputes between fans of the "numbers": is 16x44.1 enough or 24x192 needed? Which is better: multibit or delta sigma? CDDA or SACD? PCM or DSD? In this article I will try to explain the basics of digital sound in a simple language, and I will also dwell in more detail on comparing two types of encoding of an analog signal in digital: DSD and PCM .
First, answer the question, what is digital sound? How is it different from analog? In short, in mathematical language, an analog audio signal is a continuous function, digital audio signal is a discrete function . What does it mean?
If we draw in our imagination a graph of a sinusoid (this is how a sound wave is most often depicted): then, no matter how we enlarge it, trying to consider all the details, we will always see a smooth smooth line: this is an analog sound signal (Fig. 1).
Fig. 1. Analog signal
An analog sound (recording) has many parameters with which you can evaluate its quality. Consider the three most important: frequency range, dynamic range, distortion.
frequency range- a set of frequencies contained in the sound. It is generally accepted that the frequency range of human hearing is 20 ... 20,000 Hz (16 to 22,000 Hz are sometimes indicated). The frequency range of the music itself is of no interest in terms of quality assessment (for example, the frequency range of the same take-off plane will be very wide, and the tenor's vocal part will be much narrower). A qualitative parameter, say, of headphones is the potential frequency range, and it is estimated using the amplitude-frequency characteristic (AFC). The ideal frequency response - a straight line over the entire range of hearing frequencies - means that the sound source does not amplify or attenuate any individual frequencies, which means that the extracted sound matches the original.
Fig. 2. Frequency response of an MP3 file 256 kbps
Dynamic range(DD) - The difference between the quietest and loudest sounds. The volume in decibels (dB) is measured. It is generally accepted that the maximum volume that does not injure a person is 130 dB - the sound of an airplane taking off, and the minimum audible volume - 5 ... 10 dB - at the level of leaf rustling in light weather. Naturally, the rustling of leaves against the background of a take-off plane will be impossible to make out, and listening to music with a level of 130 dB is extremely unpleasant. Therefore, it is generally accepted that a comfortable DD for listening to music is 80 ... 100 dB.
Distortion is nothing more than a deviation of the signal from the original.
What happens when digitizing an analog sound? We will not delve into the technical aspects, we will analyze everything, as they say, on paper: for this we will draw our imaginary “ideal” sine wave and we will measure the signal magnitude at regular intervals (this process is called discretization or quantization ): we will get a certain sequential set of values - this will be our digital signal obtained by pulse-code modulation (PCM) (Fig. 3).
Fig. 3. Convert analog signal to PCM
The two main quality parameters of a PCM signal are frequency and bit depth. The frequency is the number of measurements in one second, the more there are - the more accurately the signal is transmitted. The frequency is measured in hertz: 44100 Hz, 192000 Hz, etc. Bit depth - the number of possible values of the signal value (transmission accuracy). The more options, the greater the accuracy of the signal. Bit depth is measured in bits: 16 bit (65.536 possible values, DD 96 dB), 24 bit (16.777.216 values, DD 144 dB), etc.
But this is not the only option for representing a sound wave in digital form. There is a way to get rid of such a parameter as bit depth, leaving only two levels of amplitude: -100% and + 100% (0 or 1). To achieve this without losing quality, it is necessary to repeatedly increase the frequency of reading the magnitude of the signal (Fig. 4).
Fig. 4. Converting an analog signal to DSD
This type of representation of digital sound is called pulse density modulation, most often the abbreviation DSD is used for it. In fact, the only qualitative parameter of such a signal is frequency. But since the frequencies used are very high (from 2.822.400 Hz), such numbers are difficult to remember; it is customary to divide the DSD signal frequency by 44.100 Hz. The resulting number is an indicator of quality: DSD64 (DD 120 dB), DSD128, DSD256, etc.
But digitizing an analog signal is half the battle. To listen to digital music, you need to perform the inverse conversion. To begin, consider how to turn a digital DSD stream into sound. As we already know, this stream is a high-frequency (2.8 MHz or more) two-level signal, the average value of this signal changes with the sound frequency. That is, if approaching the solution of the problem is as simple as possible, you need to filter out all the high-frequency components of the DSD stream, leaving only a useful audio signal (frequencies up to 20 ... 22 kHz). This is done using an analog low- pass filter (low-pass filter). The simplest low-pass filter is an RC chain . The signal received, after passing this chain, is shown in Fig. 5.
Fig. 5. Recovery of an analog signal from DSD
As you can see, the resulting graph only vaguely resembles the original sine wave. But do not forget that we “applied” the simplest filter, improving the filter circuit we can achieve an almost complete absence of high-frequency noise and get analog sound with good quality indicators.
To restore an analog signal from a digital PCM, just an analog low-pass filter is not enough, you need to decrypt the digital data first, for this digital-to-analog converters are used(DACs). They are of different types, but to describe them all is not included in the objectives of this article. Let us dwell on the 2 most common types in sound technology. Firstly, this is the so-called ladder-type DAC (it is also called multi-bit). As you probably guessed, such a digital-to-analog converter converts a PCM digital data stream into a stream of sound signal values, which on the graph look like a ladder (Fig. 6). As in the case of DSD, it is mandatory to use an analog filter to smooth the "steps".
Fig. 6. Recovery of an analog signal from PCM
Often, such converters use an intermediate oversampling of a digital PCM signal to higher frequency values (for example, 192 kHz): this reduces the “steps”, which simplifies the analog filter circuit.
The second type of DAC - delta-sigma - uses oversampling to even higher frequencies with a simultaneous decrease in bit depth to one bit. Doesn’t resemble anything? This is a familiar DSD signal! How to further process such a signal and turn it into an analog one, we have already considered above.
Where can we find each of the coding methods? The PCM format is very common: CDDA discs, DVD Audio, MP3, FLAC, ALAC, AAC files, movie sound, and on and on, it’s easier to say when non-PCM. Super Audio CDs, DSDs, DSF, DFF files are DSD format. What is still better? When playing what format do we get a better sound?
The articles on the DSD format describe many advantages over PCM, but are all the described advantages true or are these myths invented for ordinary people who do not understand the technical component in order to conquer the market densely occupied by the PCM format? Let's go through the list briefly.
So what to choose DSD or PCM? There is no definite answer and cannot be: PCM 24 bit 92 kHz and DSD128, for example, are very similar in quality characteristics, and these characteristics are better than the equipment on which these formats will be played, which means a further increase in the quality of digital formats for playback on This stage is impractical. When assessing the sound quality of different high-definition formats, subjective sensations come to the forefront, because the human brain does not eat quality alone: the design of the equipment, its cost, and, most importantly, the well-being and mood of the listener give a much greater effect on the sensations of listening to music. So, choose what you personally like, and do not impose your opinion on others. Enjoy listening to everyone!
First, answer the question, what is digital sound? How is it different from analog? In short, in mathematical language, an analog audio signal is a continuous function, digital audio signal is a discrete function . What does it mean?
Analog signal
If we draw in our imagination a graph of a sinusoid (this is how a sound wave is most often depicted): then, no matter how we enlarge it, trying to consider all the details, we will always see a smooth smooth line: this is an analog sound signal (Fig. 1).
Fig. 1. Analog signal
An analog sound (recording) has many parameters with which you can evaluate its quality. Consider the three most important: frequency range, dynamic range, distortion.
frequency range- a set of frequencies contained in the sound. It is generally accepted that the frequency range of human hearing is 20 ... 20,000 Hz (16 to 22,000 Hz are sometimes indicated). The frequency range of the music itself is of no interest in terms of quality assessment (for example, the frequency range of the same take-off plane will be very wide, and the tenor's vocal part will be much narrower). A qualitative parameter, say, of headphones is the potential frequency range, and it is estimated using the amplitude-frequency characteristic (AFC). The ideal frequency response - a straight line over the entire range of hearing frequencies - means that the sound source does not amplify or attenuate any individual frequencies, which means that the extracted sound matches the original.
Fig. 2. Frequency response of an MP3 file 256 kbps
Dynamic range(DD) - The difference between the quietest and loudest sounds. The volume in decibels (dB) is measured. It is generally accepted that the maximum volume that does not injure a person is 130 dB - the sound of an airplane taking off, and the minimum audible volume - 5 ... 10 dB - at the level of leaf rustling in light weather. Naturally, the rustling of leaves against the background of a take-off plane will be impossible to make out, and listening to music with a level of 130 dB is extremely unpleasant. Therefore, it is generally accepted that a comfortable DD for listening to music is 80 ... 100 dB.
Distortion is nothing more than a deviation of the signal from the original.
Principles of digital sound presentation
What happens when digitizing an analog sound? We will not delve into the technical aspects, we will analyze everything, as they say, on paper: for this we will draw our imaginary “ideal” sine wave and we will measure the signal magnitude at regular intervals (this process is called discretization or quantization ): we will get a certain sequential set of values - this will be our digital signal obtained by pulse-code modulation (PCM) (Fig. 3).
Fig. 3. Convert analog signal to PCM
The two main quality parameters of a PCM signal are frequency and bit depth. The frequency is the number of measurements in one second, the more there are - the more accurately the signal is transmitted. The frequency is measured in hertz: 44100 Hz, 192000 Hz, etc. Bit depth - the number of possible values of the signal value (transmission accuracy). The more options, the greater the accuracy of the signal. Bit depth is measured in bits: 16 bit (65.536 possible values, DD 96 dB), 24 bit (16.777.216 values, DD 144 dB), etc.
But this is not the only option for representing a sound wave in digital form. There is a way to get rid of such a parameter as bit depth, leaving only two levels of amplitude: -100% and + 100% (0 or 1). To achieve this without losing quality, it is necessary to repeatedly increase the frequency of reading the magnitude of the signal (Fig. 4).
Fig. 4. Converting an analog signal to DSD
This type of representation of digital sound is called pulse density modulation, most often the abbreviation DSD is used for it. In fact, the only qualitative parameter of such a signal is frequency. But since the frequencies used are very high (from 2.822.400 Hz), such numbers are difficult to remember; it is customary to divide the DSD signal frequency by 44.100 Hz. The resulting number is an indicator of quality: DSD64 (DD 120 dB), DSD128, DSD256, etc.
Recovering an analog signal from a “digit”
But digitizing an analog signal is half the battle. To listen to digital music, you need to perform the inverse conversion. To begin, consider how to turn a digital DSD stream into sound. As we already know, this stream is a high-frequency (2.8 MHz or more) two-level signal, the average value of this signal changes with the sound frequency. That is, if approaching the solution of the problem is as simple as possible, you need to filter out all the high-frequency components of the DSD stream, leaving only a useful audio signal (frequencies up to 20 ... 22 kHz). This is done using an analog low- pass filter (low-pass filter). The simplest low-pass filter is an RC chain . The signal received, after passing this chain, is shown in Fig. 5.
Fig. 5. Recovery of an analog signal from DSD
As you can see, the resulting graph only vaguely resembles the original sine wave. But do not forget that we “applied” the simplest filter, improving the filter circuit we can achieve an almost complete absence of high-frequency noise and get analog sound with good quality indicators.
To restore an analog signal from a digital PCM, just an analog low-pass filter is not enough, you need to decrypt the digital data first, for this digital-to-analog converters are used(DACs). They are of different types, but to describe them all is not included in the objectives of this article. Let us dwell on the 2 most common types in sound technology. Firstly, this is the so-called ladder-type DAC (it is also called multi-bit). As you probably guessed, such a digital-to-analog converter converts a PCM digital data stream into a stream of sound signal values, which on the graph look like a ladder (Fig. 6). As in the case of DSD, it is mandatory to use an analog filter to smooth the "steps".
Fig. 6. Recovery of an analog signal from PCM
Often, such converters use an intermediate oversampling of a digital PCM signal to higher frequency values (for example, 192 kHz): this reduces the “steps”, which simplifies the analog filter circuit.
The second type of DAC - delta-sigma - uses oversampling to even higher frequencies with a simultaneous decrease in bit depth to one bit. Doesn’t resemble anything? This is a familiar DSD signal! How to further process such a signal and turn it into an analog one, we have already considered above.
PCM and DSD application, advantages / disadvantages
Where can we find each of the coding methods? The PCM format is very common: CDDA discs, DVD Audio, MP3, FLAC, ALAC, AAC files, movie sound, and on and on, it’s easier to say when non-PCM. Super Audio CDs, DSDs, DSF, DFF files are DSD format. What is still better? When playing what format do we get a better sound?
The articles on the DSD format describe many advantages over PCM, but are all the described advantages true or are these myths invented for ordinary people who do not understand the technical component in order to conquer the market densely occupied by the PCM format? Let's go through the list briefly.
- The first advantage that DSD proponents like to bring is rather vague - noise immunityand reducing the impact of errors. It is strange to hear about different noise immunity in the digital world: both formats are subject to interference just as much as the digital book is subject to interference. The duration of storage of any digital format or the quality of its transfer between devices depends only on the media / transmission method, and not on the format itself. So, the noise immunity is the same. What about reducing the impact of errors? Suppose we store 2 albums on optical discs (one PCM, another DSD), what will happen if the disc is scratched? Errors will occur while reading a damaged medium, but how critical are they? PCM coding uses multi-digit numbers, the error in the high order is very critical (as an example, the difference between decimal numbers 11 and 91): it will feel like a click by ear.
- The second advantage is described a little more specifically: a greater dynamic range compared to PCM. Well, there is some cunning here, DD is larger only in comparison with the classic CDDA format: 120 ... 140 dB against 96 dB. If we compare, for example, with DVD Audio - DD is about the same.
- Third advantage: DSD is technically simpler. There is nothing to argue with: simpler signal decoding, no need to synchronize and buffer the stream when transmitting a signal from one device to another is a complete victory of DSD. By the way, amid this advantage, it is strange to see sky-high prices for equipment supporting DSD playback.
- Well and another advantage that DSD fans love to bring: music in this format is closest to the original analog sound. This is argued by the fact that modern analog-to-digital converters(ADC) - work on the principle of delta-sigma modulation, that is, these ADCs produce a digital DSD stream. And here again cunning: the recording will be completely original only in the case of a direct recording of a live performance or when digitizing a finished analogue recording from a high-quality medium. The operations of mixing, superimposing effects, mastering, even simply adjusting the volume — all that the creation of a studio album cannot do without — are impossible for a DSD record due to the lack of normal processing algorithms. This means that all these operations are performed with the PCM format, and only after that the finished PCM record is converted to DSD. However, it should be noted that the conversion of PCM> DSD and vice versa is quite accurate: the noise only slightly increases outside the real dynamic range (Fig. 7). Which means it doesn't really matter in what format to listen to the recording: PCM Hi-Res or DSD - both formats are very similar in terms of quality characteristics. Also, in fact, it makes no sense to buy a separate sound card for playing DSD, following the advice of a friend, a fan of this format.
Fig. 7. Dynamic range / noise when converting between DSD and PCM
conclusions
So what to choose DSD or PCM? There is no definite answer and cannot be: PCM 24 bit 92 kHz and DSD128, for example, are very similar in quality characteristics, and these characteristics are better than the equipment on which these formats will be played, which means a further increase in the quality of digital formats for playback on This stage is impractical. When assessing the sound quality of different high-definition formats, subjective sensations come to the forefront, because the human brain does not eat quality alone: the design of the equipment, its cost, and, most importantly, the well-being and mood of the listener give a much greater effect on the sensations of listening to music. So, choose what you personally like, and do not impose your opinion on others. Enjoy listening to everyone!