Usually called D/A converter or DAC (digital-to-analog converter).
We know that scores can be weighted or unweighted. The so-called weighted number means that every digit has a coefficient. For example, 4 out of 45 decimal numbers are 4× 10.
And 5 is 5× 1, that is, the coefficient of 4 is 10, and the coefficient of 5 is 1. In a sense, digital-to-analog conversion is to convert binary numbers into decimal numbers.
The original DAC circuit consists of the following parts: reference voltage source, summing operational amplifier, weight generation circuit network, register and clock reference generation circuit.
The function of register is to register the input digital signal at its output, and the change of input voltage will not cause its output instability during conversion.
The clock reference generation circuit mainly corresponds to the reference voltage source to ensure that the phase characteristics of the input digital signal will not be confused during the conversion process.
The jitter of clock reference will produce high frequency noise.
The weight coefficient of binary data depends on the resistance, and the CD format is 16bit, that is, 16 bit. Therefore, 16 resistors are used, corresponding to each of 16 bits.
The analog signal can be obtained by the weighted sum of the current passing through the reference voltage source of each resistor and the current of each bit of input data in turn.
This is a multi-bit DAC. The difference between multi-bit and 1 bit is that multi-bit compares potentials through an internal precision resistor network and finally converts them into analog signals.
The advantages are high dynamic following ability and high dynamic range, but the accuracy of the resistor determines the accuracy of the multi-position converter. In order to achieve 24-bit conversion accuracy, the requirement for resistance is as high as 0.00005438+05.
Even if the resistance is ideal, the resistance fluctuation caused by thermal noise will be greater than this value. At present, R-2R ladder resistor network is widely used in multi-bit systems, which can reduce the precision requirements for resistors. Even so,
The conversion accuracy of the ideal resistor will not reach 24 bits, and 23 bits is already the limit. The advantage of multi-bit system lies in its simple design, but it is limited by the precision of resistance and its cost is high.
Single-bit principle: By means of mathematical operation, oversampling points are inserted into the pulse coded signal (PCM) of CD, and inserting 7 sampling points is 18 times oversampling.
These inserted sampling points are compared with the original signal through the integration circuit, and the larger value is set to 1 and the smaller value is set to 0, so that the original PCM signal becomes a data stream with only 1 and 0.
1 indicates dense data flow, and 0 indicates sparse data flow. This is a pulse density modulated signal (PDM), which passes through a low-pass filter consisting of a switched capacitor network.
1 is converted into a high voltage signal, 0 is converted into a low voltage signal, and then it is finally converted into an analog signal through cascade integration.
Inserting sampling signal will produce a lot of high-frequency noise, so it needs to be processed by noise shaping circuit to push the noise to the frequency domain that human ears can't hear.
The advantage of 1bit is that the conversion accuracy is not limited by resistance, and the conversion accuracy can exceed 24bits, so the cost is low. However, it is very difficult to design oversampling and noise shaping circuits.
Because the resistance has relatively little influence on the sound quality in terms of precision (lithography) and thermal noise (material), the capacitance and integration circuit of 1 bit have relatively great influence on the sound quality. For the data format of the optical disc,
In terms of sound quality alone, it should be said that multi-bit is better than 1 bit, and multi-bit directly converts the CD signal of 16 bit, while single bit has to go through a process of converting PCM signal into PDM signal and charging and discharging the switched capacitor.
Although theoretically, the speed of the final analog signal will not be slower than that of multi-bits,
But in fact, a single bit is not as dynamic as Dobby, which looks slower, has a thicker intermediate frequency and richer timbre.
1bit was founded in Philips and divided into three factions.
One school is Bitsream, represented by Philips.
One school is MASH represented by Panasonic, but the founder of MASH is NTT Company.
Another school is Delta Sigma, which is very popular today.
Bitsream adopts the most traditional third-order or fourth-order noise shaping, and MASH (multilevel noise shaping) is multilevel noise shaping.
It retains the error between the initial quantization value and the original signal, and subtracts the previous quantization value and error from the original signal in the next quantization, and repeats it several times.
Binary signals can be converted into pulse width modulation (PWM) signals (PWM and PDM are almost the same), and the noise generated by quantization can be pushed to a very high frequency band, thus reducing the noise in the audible frequency band.
But it seems that only Panasonic has adopted this technology in large numbers. MASH is rare now, but it is excellent in theory.
1987, Philips first introduced a single-bit DAC chip using Bitsream technology, which laid a solid foundation for the emergence of high-performance and low-cost CD players.
199 1 DAC-7, which was launched in September, has developed the bit stream technology to the maximum extent while maintaining a reasonable price. There are many famous machines using DAC-7 in the history of sound.
Such as Philips LHH-900R, 800R, 300R, 95 1.
Maranz's CD-72, CD- 17, CD-23
MCD- 7007 of Maijingtu.
Pioneer early flagship PD-T07.
602/603 of the meridian,
And almost all European digital audio source manufacturers such as Rotel, Altis, Deltec, Revox and Student. All flagship systems use DAC-7.
After entering 2 1 century, TDA 1547 has not lost its advantages. At present, SA- 1 of marantz, the most advanced SACD player in the world, is still using DAC-7, which makes the world look at DAC-7 again.
Up to now, DAC-7 is still the most advanced bit stream DAC chip of Philips.
In Philips' product manual, DAC-7 is evaluated like this; Dual-channel digital bit stream DAC chip with top performance,
1Bit digital-to-analog converter is dedicated, and high-fidelity digital audio reproduction can be easily obtained by using DAC-7.
DAC-7 is very suitable for high-quality CD and DAT players, or digital amplifiers and digital signal processing systems. This evaluation is very pertinent.
DAC-7 includes TDA 1547 and SAA7350, because a large number of high-frequency digital signals generated by oversampling and noise shaping circuits will interfere and modulate the analog circuits in TDA 1547.
Therefore, the third-order noise shaping and 24-fold oversampling circuits of TDA 1547 are designed separately in SAA7350. This is also the key to the success of TDA 1547.
Now Philips has comprehensively improved SAA7350 and integrated digital filters. The new model is TDA 1307, or a chip specially designed for TDA 1547.
However, TDA 1547 and TDA 1307 are collectively called DF7.
TDA 1547 adopts bipolar composite metal oxide semiconductor process. In digital logic circuits, the optimal clock frequency can reduce digital noise.
Using bipolar transistors in analog circuits can make operational amplifiers achieve higher performance.
In terms of power supply, TDA 1547 has made great efforts. First of all, analog circuits and digital circuits are powered separately.
In the digital circuit, the high-level logic circuit and the low-level logic circuit are powered separately, and the left and right channels are powered independently.
In terms of internal structure, TDA 1547 adopts dual mono design, which is completely separated, and the output is also independent of the left and right channels.
TDA 1307 can receive signals in 16, 18 and 20bits formats and output audio in 32bits format.
Built-in receiving interface, de-emphasis filter, 8 times oversampling finite impulse response (FIR) filter and 3 or 4 order optional noise shaping circuit.
The signal-to-noise ratio of standard chip is as high as 142dB, and the dynamic range is as high as 137dB.
Marantz's SA- 1 makes full use of DAC-7. It uses four TDA 1547 and TDA 1307 to form a fully balanced circuit.
The analog amplification part adopts HDMA, which is widely used in advanced cars in marantz.
Delta-Sigma 1bit is very popular nowadays. It includes two parts. One part is a Delta circuit, which compares the quantized signal with the initial signal and calculates the difference. These interpolated signals then enter Sigma circuit.
The circuit adds the errors of these interpolated signals, and then adds them to the signals before quantization. Then quantify it.
Dynamic Element Pairing (DEM) quantization technology developed by Philips is usually adopted. This quantization includes a very high-precision current source and multiple 1/2 mirror current sources. Because integrated circuits are best at mirroring current source circuits,
Therefore, the requirements for component accuracy can be reduced and the cost performance can be improved.
The quantized signal is converted into an analog signal through a switched capacitor network.
It should be pointed out that not all δ-σ conversions are single-bit. Delta-sigma has the advantage of high cost performance, so it is very popular in the middle and low-end digital audio source market.
Even those manufacturers who insist on using multiple vendors, the low-priced ones have to adopt Delta-sigma.
I'm afraid CRYSTAL is the one who insists on using Delta-sigma. Crystal's CS 4390 and 4396 are also widely used in the industry.
There are also some excellent products such as MBL1611hr.
And meridian 506.20 with fever gobbledygook a level,
Meridian 508.24, meridian 506.24
There is also a domestic new deco DAC- 1.
CS4390 was released in1June, 1998, and it is the first Delta-sigma DAC chip of CRYSTAL.
This is a complete stereo DAC decoding chip. The signal first enters the interpolation circuit with 128 times, and then undergoes the oversampled Delta-sigma digital-to-analog conversion with 128 times.
Then output the analog signal and the modulated reference voltage, and finally enter the superlinear analog low-pass filter.
Among them, the Delta-sigma digital-to-analog conversion part has not adopted the DEM technology of Philips.
The signal-to-noise ratio of CS4390 is 1 15dB, the dynamic range is 106dB, the total harmonic distortion plus noise is -—98dB, the conversion accuracy is 24bits, and it is not sensitive to time-base jitter.
Later, the volume control was added on the basis of CS4390, and it was renamed CS439 1.
One year later, in July, 1999, Crystal launched an upgraded product of CS4390-CS4396. The biggest difference between CS 4396 and CS4390 is that DEM technology is adopted.
CS4396 is also a complete stereo DAC chip. After interpolation and Delta-sigma transformation, the signal enters the DEM program block, then passes through the switched capacitor network, and finally passes through the analog low-pass filter.
The output stage adopts high-quality difference channel. After using DEM, the distortion and noise of CS4396 are reduced to-100dB, and the dynamic range is also increased to 120dB.
The conversion accuracy is still 24bits, and the highest sampling frequency rises to 192KHz, but the signal-to-noise ratio parameter is no longer provided.
Meanwhile, CS4397 is an interpolation filter that supports external PCM (corresponding to DVD-AUDIO) and DSD (corresponding to SACD) on the basis of CS4396.
After more than half a year, Crystal Company launched an upgraded product of CS4396-CS 43122.
One of the differences with CS4396 is that the second generation DEM technology is adopted.
The other is that Delta-sigma modulator uses 5-bit third-order modulation instead of 1 bit.
The interpolation circuit is also improved, and the stopband attenuation performance of 102dB is realized. The performance parameters of CS43 122 and CS4396 are basically the same, but the dynamic range reaches 122dB, which is also the highest dynamic range DAC chip at present.
On September 20th, 2000, CRYSTAL Company introduced the DAC chip CS4392 corresponding to DVD-AUDIO and SACD, with a dynamic range of 1 14dB and a total harmonic distortion plus noise of-100dB.
But it's only OEM, and it's not in circulation for the time being. The price of each piece is only $2.8.
(Note that CRYSTAL didn't mention the signal-to-noise ratio from beginning to end, because its signal-to-noise ratio reached 1 15dB with CS4390).
Japan's NPC company is also famous for its sigma-delta transformation technology. We must be familiar with NPC's high-performance digital filters. The most famous SM5842 is recognized as the best.
Similarly, SM5865 is the best sigma-delta decoding chip. Although unknown, SM5865 will be recognized as the best in the near future.
SM5865 was launched in February this year. First of all, it is a single chip with a truly completely balanced circuit inside. The signal first passes through the interpolation circuit, and then enters the third-order multi-bit sigma-delta transform program.
Then it is quantized by 3 1 DEM, and finally it becomes an analog signal through a switched capacitor network.
SM5865 DEM quantization series is extremely high and very successful, which can completely ignore the audible frequency domain noise caused by quantization, so the analog low-pass filtering of the last stage can be omitted, and the distortion and noise amount of the ideal state can be obtained.
SM5865 is the lowest distortion and noise DAC chip in the world, and the total harmonic distortion plus noise is only 0.0003%, that is,-1 10.5 dB.
At the same time, it still achieves the signal-to-noise ratio of 120dB and the dynamic range of 1 17 dB. The accepted data format is between 20-24 bit, and the highest sampling frequency is 192 kHz, thus becoming the king of DAC today.
Multi-bit DAC is divided into two famous companies, one is UltraAnalog and the other is Burr-Brown.
Most people may not be familiar with UltraAnalog, because it was acquired by Wadia in February 1998, and we haven't heard from it since. But its position in DAC history is far from Burr-Brown.
The flagship decoder DA- 10 adopts UltraAnalogDAC chip, with Conterpoint,
Polaroid Parasound's flagship decoder D/Ac-2000,
Mark Levinson's early flagship decoders N0.30.5 and No.N0.30.5.
There is also the decoder DAC-x 1 of Stax, a famous Japanese static earphone manufacturer.
Advanced decoder kcd-55 of KinergetICs
The flagship decoders of Manleylab, Sonic Forntiers, Camelot, Entech, Aragon and Audio Synthesis all use super analog chips.
Basically, decoders using super analog chips will be A-class products of fever books. Moreover, almost all the top American decoders before 1998 used super analog chips.
Although UltraAnalog's products are very good, the profit is low, because UltraAnalog has only one product, which is simply unsustainable for IC manufacturers. UltraAnalog will live to 1998.
After Wadia's acquisition, it did not absorb and transform the technical resources of UltraAnalog. At the same time, Wadia also thinks that UltraAnalog is a burden, and gradually UltraAnalog has disappeared.
Up to now, there are still diehards who are super-analog, such as Manleylab, Sonic Forntiers, Camelot, Entech, Aragon and Audio Synthesis, and they still insist on using super-analog chips.
There may be a lot of inventory, and Sonic Forntiers also has a cooperative relationship with UltraAnalog. Superanalog chips can also be produced.
UltraAnalog is the first manufacturer in the world to seriously study the timing jitter, and the timing jitter of UltraAnalog products is also the lowest in the world.
UltraAnalog also puts forward a digital audio signal interface, which can greatly reduce the time base jitter.
1993 UltraAnalog also invented a very cheap time-base jitter analyzer.
UltraAnalog's chip is mainly D20040, and we know little about it, except that it has a conversion accuracy of 20bits. Internally, it consists of two parallel 19-bit dacs. I don't know the rest.
I believe that after 10 years, who else will know about UltraAnalog? Technology and business are definitely not for a while.
Burr-Brown occupies a large share in the DAC chip market and enjoys a long-standing reputation. Burr-Brown, founded in 1993, is a diehard faction like UltraAnalog.
PCM58 and PCM63 were introduced at the beginning of the factory, which received rave reviews, but they were still inferior to UltraAnalog.
The introduction of PCM 1702 in 1995 can finally compete with UltraAnalog. Until today, there are not a few advanced CD players using PCM 1702.
The Sondek CD player launched by Linn in 2000 uses PCM 1702, and the price is as high as $20,000. The fever book was rated as Grade A.. After four years of silence,
1704, the final product of multi-bit DAC, was launched in February, 1999. At this time, UltraAnalog has been acquired by Wadia and gradually declined. Burr- Brown was also acquired by TI (Texas Instruments).
Relying on the strong strength of TI, Burr-Brown has developed well and become the leader of DAC chip market today.
PCM 1702 went online in June, 1995. At that time, 1bit had a strong reputation in the market, and Burr-Brown challenged 1bit.
Burr-Brown pointed out that inserting 1bit at the sampling point would cause a lot of high-frequency noise. Although the frequency of these noises is relatively high, the auditory frequency domain can still be modulated.
In addition, these artificial noises need to be eliminated by noise filters. The addition of the filter greatly attenuates the signal-to-noise ratio, and the low-level response is not good enough. Burr-Brown thinks that this characteristic of signal-to-noise ratio is almost the most important.
The only disadvantage of multi-bit is zero-crossing distortion. PCM 1702 adopts the sign amplitude structure to solve this problem perfectly.
A pair of DACs are connected in parallel in 1702. The advantages of parallel connection are improved signal-to-noise ratio and conversion accuracy. 1702 connects two 19-bit DACs in parallel, and the conversion accuracy is 20 bits.
These two DACs * * * * use reference voltage, * * use R-2R ladder resistance network, and the bit current source of ladder resistance network is powered by double balanced current stage, which ensures the perfect tracking characteristics of the bit current source.
Each DAC uses a laser to adjust the molybdenum-chromium resistance to ensure high accuracy, and the two DACs are precisely adjusted to ensure the same phase. Finally, the positive and negative half-cycle conversion of two DACs perfectly solves the zero-crossing distortion problem.
However, the traditional R-2R resistor digital-to-analog conversion has achieved high signal-to-noise ratio and low distortion, as well as nearly ideal low-level performance and high-current output capability.
The signal-to-noise ratio of PCM 1702 is 120dB, which has not been broken by anyone so far, and it was even unimaginable at that time. The total harmonic distortion plus noise of 1702 is -—96dB, which was also a very good characteristic at that time.
PCM 1704 was launched in February 1999, which is the ultimate product of multi-bit DAC. I'm afraid there will never be more multi-bit DACs than it.
Burr-Brown uses its best resistor manufacturing process to manufacture resistors with ideal accuracy, thus obtaining the world's most accurate multi-bit DAC, up to 23 bits. After two parallel connections, it reaches 24 bits.
As for the internal structure, it is basically the same as PCM 1702.
The signal-to-noise ratio of 1704 is still 120dB, the dynamic range is 1 12dB(K level), and the total harmonic distortion plus noise is-10 1dB(K level).
Since 1704, Burr-Brown has never introduced a multi-bit DAC with a higher level than 1704, and Burr-Brown cannot break his own record.
On April 30th, 20001,Burr-Brown introduced a new generation of top-level DAC-PCM1738, which adopted advanced graded DAC. Burr-Brown also knows that the traditional multi-bit has come to an end.
The advanced hierarchical structure divides digital signals into high 6-bit signals and low 18-bit signals by using digital interpolation filters with sampling frequency of 24 bits and 8 times.
The high 6-bit signal is converted into a 62-level digital signal by inverse complementary shift binary decoding, and the low 18-bit signal is modulated by the third 15-level δ-σ.
The modulation frequency is 64 times of the sampling frequency, and finally it is converted into a four-level digital signal.
Then the two are added to form a 66-level digital signal, and the LSB signal of 1 level is added, making a total of * * * 67-level digital signals.
Then, the 67-level digital signal is processed by a data weighted average (DWA) program to reduce the noise caused by mismatched analog components.
In fact, DWA is the second generation DEM. After DWA processing, it finally enters the current-mode digital-to-analog converter to convert the binary pulse signal into a pulse current signal.
Then the off-chip operational amplifier converts the current into voltage, and finally gets the analog signal. It should be said that this kind of DAC is not single bit or multi-bit, and it should be called current pulse DAC.
The signal-to-noise ratio and dynamic range of PCM 1738 are 1 17dB, and the total harmonic distortion plus noise is-108dB, which should be said to be better than PCM 1704, but its price is far lower than PCM1704 (.
Analog devices is also very good at making top-notch DAC chips. For example, Golden Voice always uses only analog device chips.
In the theoretical design of DAC chips, analog devices play a supreme role. Analog devices invented multi-bit Delta-sigma modulation as early as 1998.
Due to the traditional single-bit Delta-sigma modulation, the size of the discrete-to-continuous boundary of each step is too large, which requires the stability of the master clock to be very high.
For example, if the signal-to-noise ratio (SNR) in the audible frequency domain is above 100dB, the time base jitter of the master clock should not be greater than 10PS, but this is impossible, so the single-bit Delta-sigma modulation must be abandoned to achieve high SNR.
The disadvantage of multi-bit Delta-sigma modulation is that it is inconvenient to use DWA program and the noise generated by analog components is inevitable.
If the DWA program is adopted, the format of the input signal is required to be lower than18 bits, and now it is 24bits. Obviously unacceptable.
ADI takes a different approach and adopts segmented noise shaping technology to solve this problem. Burr-Brown separated the signal from the beginning.
The traditional single-bit decoding must use switched capacitors, and the capacitance will increase by 4 times for every bit of precision conversion.
To know that each capacitor will produce noise, large capacitors require higher conversion rate for operational amplifiers with switched capacitor networks.
Therefore, the conversion accuracy of DAC chip using switched capacitor network is high, which will cause a certain degree of sound quality degradation. If the design is not good, it is possible that the higher the conversion accuracy, the worse the sound, and the sound is too nice and too thin.
The analog device adopts current pulse DAC, and the pulse current output of current mode DAC has uneven rise and fall time. Using the general voltage-current conversion operational amplifier will lead to the linear decline of conversion and is also very sensitive to time-base jitter.
The analog device is solved by double rotary zero switch circuit. This technology was jointly developed by Sony and was first used in Sony's top ES series.
Because the current pulse adopts an extremely pure instantaneous current source, the current pulse will not have any ripple, which is almost equivalent to a perfect square wave. The sound quality will be very pure.
After 1999, ADI found that the audio market was shrinking, so it turned to the development and research of SHARC general DSP chip, without further research on DAC.
Nevertheless, the DAC chip AD 1853 introduced by ADI in 1998 is still the most advanced DAC chip at present, which is no worse than PCM 1738 or SM5865. Although these chips were introduced at 200 1,
But in terms of performance and technology, AD 1853 is not bad.
Moreover, AD 1853 is also the world's first DAC chip with a sampling frequency of 192KHz, and it is also the world's least sensitive DAC chip to time-based jitter.
Its signal-to-noise ratio is 120dB, its dynamic range is 1 17dB, and its total harmonic distortion plus noise is-107 dB. Compared with SM5865, it should be said that it is equal.
You should also know something about the emerging DAC chip with audio format.
The format of DVD-AUDIO still uses pulse code modulation, so the principle of DAC decoding chip of DVD-AUDIO is the same as that of CD.
Only higher conversion accuracy, sampling frequency and input format width are needed.
SACD is different. When recording, it converts the input analog signal into a binary digital signal with a single-bit sampling frequency of 2822.4kHz by Delta-sigma modulation.
Moreover, the digital signal at this time is already a pulse density modulated signal (PDM), so there is no need to add sampling points and noise shaping circuits when performing single-bit decoding.
As long as it passes through the switched capacitor network and the analog low-pass filter, the analog signal can be obtained.
Therefore, the circuit is very simple. There is no digital operation circuit and no clock reference generation circuit in the digital-to-analog conversion stage, so there will be no digital noise mixed in, and the sound purity is extremely high.
Sony's SACD machine does not use switched capacitor network, but uses the highest level of current pulse digital-to-analog conversion.
Incidentally, the CD signal is also a binary signal with a sampling frequency of 44. 1kHz, which is converted from an input analog signal by delta-sigma modulation and then passed through a digital decimation filter.
Any digital filter will produce noise that can not be ignored, as well as ripple and ringing in the passband, which will reduce the purity of sound.
SACD has no digital filter in both recording and playback systems, while CD not only exists in both recording and playback systems. Single-bit system needs interpolation sampling point filter.
The purity of sound quality cannot be compared with SACD at all. SACD is the recording medium and playback system with the highest sound purity at present, which is closest to the real sound.
At present, there are three kinds of DAC chips for SACD in the world.
One is DSD 1700 used by Sony SACD, which belongs to Burr-Brown Company.
The second is SM5866 from NPC.
The third one is CRYSTAL's CS4392, but it is not for public sale.
Because SACD thinks that the sound performance is the best at present, it generally adopts current pulse digital-to-analog conversion circuit.
Generally, these circuits are composed of discrete components, so DSD 1700 and SM5866 are mainly analog low-pass filters.
Strictly speaking, DSD 1700 and SM5866 are not DAC chips, but analog low-pass filter chips.
DSD design can only be used in SACD system, which is mainly composed of four groups of analog low-pass filters, namely, forward and backward filtering on the hot side and forward and backward filtering on the cold side.
There are eight three-terminal infinite impulse response filters in each group of filters. Four groups of filters finally output double-difference channels.
The dynamic range of DSD is 1 10dB, the signal-to-noise ratio is 1 10dB, the total harmonic distortion is-100 dB, and the high frequency response is 100 kHz (-3 dB).
SM5866 of NPC Company was launched on September 22nd, 2000, which can be used in SACD and DVD audio systems. Its internal information has not been published.
Its signal-to-noise ratio is 120dB, total harmonic distortion plus noise is-109 dB, and its high frequency response is 100 kHz (- 1 dB). Obviously higher than DSD 1700.