What Is a Bit Stream?

DSD is an abbreviation of Direct Stream Digital, which stands for Direct Bit Stream Digital Coding, and is a SACD (Super Audio CD) coding mode. It is a high-resolution digital audio specification announced by Sony and Philips in 1996.

DSD

(Direct bitstream number)

DSD ( Direct Stream Digita l) "Direct Bit Digit", it is

Difference between DSD and linear PCM

PCM (Pulse-Code Modulation, Pulse Code Modulation) is the most common audio encoding format now, almost all common audio such as wav, ape, flac, mp3, etc. are PCM encoding format.

The principle is simple, first prepare a set of specified level values (for the concept of level, it can be simply equivalent to voltage), such as -3, -2, -1,0, 1, 2, 3, etc Each value is assigned a number, like ABCD, but we now use binary numbers for these level values (that is, 000 001 010, etc.).

Then from the previous set of levels, use the method of rounding to find the value closest to the sampling level, then record the number of the closest value, and then perform the next sampling .... The serial number (that is, the digital signal) records the original analog signal from the microphone. The recorded digital signal is PCM.

The whole process above is often referred to as the ADC coding process, and this is the recording process in the recording studio.

The difference between the output signal and the input signal during this entire process is called the quantization error . The quantization error is a kind of noise to the signal, so it is also called quantization noise.

PCM is like this, each sampling point is to measure an absolute value, and the sampling points are independent and unrelated to each other.

For the 16bit 44.1kHz PCM used in the CD, the signal is sampled 44100 times per second, and then the sampling level is measured with a set of 65536 (that is, 16bit, 2 to the power of 16) values. With a high sampling frequency and 16-bit specified level accuracy, the recorded signal is very close to the original signal (at least most human ears cannot tell the difference).

We can also better record the original signal by increasing the sampling rate and increasing the accuracy of the specified level, such as the common 24bit 88.2kHz, 96kHz, 192kHz, and 32bit 96khz.

However, this method of PCM still has bottlenecks. The quantization noise is evenly distributed on all frequency bands. Even if the accuracy and sampling rate continue to be greatly improved, it is difficult to reduce more noise.

In order to comprehensively improve the pulse code modulation digital audio technology and obtain better sound quality, we need new technology to replace it, so we have DSD.

For 16-bit PCM, 16 bits of data are required for each sampling point, but DSD can record 1 bit for each sampling point. 1 "to record the level value of this sample point.

In the DSD encoding process, the way to quantize signals is completely different from PCM.

The concept of modulation is introduced here. Instead of using a set of specified level values to measure as in PCM, only a fixed value "" is used to measure the original signal. It is still taken once every fixed period of time, each time The obtained level will be compared with the last sampled signal. If the interpolation is greater than , it will output "1". If the interpolation is less than or negative, it will output "0". So, every sampling point It can be recorded in 1bit format.

One disadvantage of delta modulation is that as the frequency of the input analog signal increases, the signal-to-noise ratio will decrease sharply. We can control the quantization noise by reducing the value of delta and increasing the sampling rate.

The main idea of DSD is this, the value of each sample is the relative value of the previous sample, and the front and back sampling points are inseparable. The idea of this quantization method is closer to the sound in nature because of its continuity (the sound signal is A series of single points is meaningless).

Sigma-Delta Modulator

In order to overcome the shortcomings of delta modulation, a sigma-delta modulator (Sigma-Delta Modulator) was developed
As shown in the figure, if we add a differentiator to the input end of the signal, the signal is input from the normal phase of the differentiator, then passes through an integrator, and then to the delta modulator (A / D). / A conversion, and the delay input to the inverting terminal of the differential as feedback, this is a complete sigma-delta modulator.
The overall quantization method is still similar to modulation, but the level returned to the inverting input of the differentiator is the maximum or minimum value of the entire signal (that is, if modulation output 1 is returned, it is returned to Vmax, and output 0 is returned to Vmin , Both are fixed values), that is, the integrator integrates the difference between the input level and the highest / low level, and then we perform a delta modulation on the integrated result (the original signal can be regarded as a function f (x) derivative, and then we modulation quantize f (x), which may be better understood).
In this way, the object of quantization becomes the difference between the current signal level and all previous differences. The quantization level is no longer affected by frequency. The maximum quantization range depends directly on the level value.
The delay circuit added to the feedback makes the sigma-delta modulator have the characteristics of noise shaping. The noise shaping effect of the first-order sigma-delta modulator is not obvious, but we can superimpose the multi-sigma sigma-delta modulators together so that The noise shaping effect reaches a high level. The specific result of this noise shaping is that the overall amount of quantization noise has not changed, but it is not evenly distributed across all frequency bands. There will be less quantization noise in low frequency bands, and quantization noise in high frequency will be lower. That is to say, quantization noise is "pushed" into high frequencies. In audio applications, most of the quantization noise is pushed to high frequencies far beyond 20kHz, which is a frequency band that cannot be heard by the human ear. Low-pass filtering can simply remove these noises.
This is the biggest advantage of DSD over PCM.It has minimal quantization noise and ultra-high signal-to-noise ratio.

DSD is a digital signal obtained through the above sigma-delta modulation. If this series of digital signals is placed on the same scale and compared with the original signal, the digital "0" and "1" will be found as the signal frequency increases The degree of decrease causes a corresponding change in density, so DSD is also known as Pulse Density Modulation

For example, PCM draws points toward the original image, but how accurate you are at this point will always have a small error, and DSD is directed at the contour of the original picture, but this contour is more accurate than the point of PCM.

Although DSD has various advantages over PCM, there is a flaw. DSD cannot be used in the post-recording mixing production. Only PCM can perform mixing processing .
So now almost all recording studios use the PCM format. After the mixing is completed, it is compressed into the DSD format and made into SACD. This process has actually lost most of the advantages of DSD.
Unprocessed pure DSD direct recording audio is really very rare, mostly audition products for recording studios.
Therefore, the road to DSD is still very long.