digital-to-analog converter (DAC) is one of the core components (although not necessarily expensive) for digital audio sources such as various CD players, SACD players, portable players and mobile phones. The DAC is responsible for converting the decoded digital audio stream into an analog audio signal. In this process, the chip plays a key role.
This series of articles aims to introduce each company's DAC chips, parameter indicators, etc., and try to introduce its sound characteristics. I hope you will like it.
It should be noted that chip-only theory is common among enthusiasts and non-enthusiasts alike. But in fact, the DAC chip is only one of the key factors that determine sound quality, not the only factor, or even the most important factor. Peripheral circuit design, clock accuracy, analog circuit design, component selection and power supply processing play a vital role in the sound of the final audio equipment. Readers must be aware of this! Many of the decoding equipment that were great in the past are still great now. The only difference they have is in the formats they support rather than the sound. Nowadays, many decoders use much better chips than before, but the sound is still terrible. The reason is very simple, it's just the bucket effect. When choosing peripheral circuits and components, analog circuits are decisive!
d/a digital/analog converters can be roughly divided into two mainstreams: one is multi-bit, and the other is 1-bit. In terms of development time, multi-bit products appeared on the market when the CD turntable came out, while the first generation of 1-bit products only appeared on the market in the early 1990s. Multi-bit and 1-bit structure analysis, which one is better and which one is worse? For example, a string of beads strung together. We use two methods to remove the beads from the string. The first method is: take it several times and remove a fixed number of beads each time; the second method is: take as many beads as there are, each time. Take only one bead at a time. In fact, the first method is equivalent to the multi-bit method. Only after receiving all 16-bit numbers, the decoding process is performed once. The second method is equivalent to the single-bit method, decoding one digital digit one digital digit continuously. As far as multi-bit is concerned, its advantage is that there is no so-called re-quantization process, so the noise is lower; in addition, it also has better dynamic performance. However, traditional multi-bit low-level situations have problems with nonlinear distortion and zero cross distortion. To overcome these problems, a very complex circuit structure is required. This has resulted in multi-bit DAC chips in the pursuit of high quality. goals, but also at a high cost. In contrast, the advantage of 1-bit DAC is that it inherently does not have the problem of zero-crossing distortion, the linearity of the analog waveform is good, and the production cost is low. This is why 1-bit DAC is widely used by mid- to low-priced CD players on the market. reason. However, 1 bit DAC requires a higher frequency clock, and it will cause some information loss during the "re-quantization" process, which is why it is difficult for it to be used in hi-end equipment.
The first two articles in this series have introduced the two current mainstream decoding companies and products (ESS and AKM). This article introduces some other chip companies and products. In fact, in addition to these,
also has a special decoding method, that is, the decoding method of R2R discrete components. There are very few companies in the world that adopt this method. The famous British dCS and the domestic Holo Audio are two representative ones. This series of articles will talk about them and related products in the next article.
1. Introduction to multi-bit chips
1, Burr Brown (Burr Brown)
Burr Brown (Burr Brown) company belongs to Texas Instruments (TI), a well-known heavyweight manufacturer in the semiconductor industry. Its most well-known DAC chip is the PCM1704. Many Hi End manufacturers have praised it, including many Hi End manufacturers that insist on traditional two-channel audio. For example, Mark Levinson's top decoder NO.360 (US$4495) uses PCM1704. It is a precision 24bit D/A conversion chip with ultra-low distortion and low-level response linearity. It uses a 2μm BICMOS manufacturing process and a very unique Sign Magnitude architecture. Two 23-bit fully complementary D/A converters are designed inside it to achieve 24-bit accuracy.These two D/A converters share a clock reference and a R2R type resistor ladder network, and obtain accurate digital current source signals through continuous voltage division. The double balanced current loop used by the R2R resistor ladder network ensures ideal tracking of voltage signals at any level. These two D/A converters are completely independent in internal data calculations and can have a very linear level response, especially at low levels (ie, small volumes). The resistors in the R2R ladder resistor network are all made of nickel-chromium film resistors through laser trimming, so the accuracy is high enough. In addition, the two D/A converters are also used after precise matching. The signal-to-noise ratio of to of
PCM1704 reaches an astonishing 120dB, and it is a standard K-class chip. Its total harmonic distortion and noise reached 0.0008% (-101.94 dB), which is also a standard K-class chip. The dynamic range of the standard K-class reaches 112dB. The sampling frequency range of PCM1704 is 16~96kHz, and the oversampling frequency is 8 times oversampling of 96kHz. In addition, the input audio data format is 20bit or 24bit, the fast current output is ±1.2mA/200ns, and the power supply voltage is ±5V.
PCM1704 is a product launched in February 1999. From the perspective of today's continuous development, it seems a little backward, especially its sampling frequency is only 96kHz. Burr Brown launched a new generation DAC chip PCM1738 that can replace PCM1704 on April 29 this year. It is designed using the Advanced Segment chip architecture, which can achieve higher dynamic range and tolerance of time base jitter. Although the signal-to-noise ratio is slightly reduced to 117dB, the dynamic range is widened to 117dB, the total harmonic distortion and noise are also reduced to 0.0004% (-107.96 dB), and the sampling frequency range is 10 ~ 200kHz. PCM1738 can additionally connect the digital filter and the DSD decoder corresponding to SACD through the optical fiber interface. At the same time, it has built-in 8 times oversampling digital filter, digital compensation, digital de-emphasis and soft squelch. Its full-scale output voltage is 2.2Vrms and its differential current output is ±2.48mA. This is a DAC chip specially designed for the multi-channel amplifier .
In addition, some old-fashioned BB chips still sometimes appear, such as the 20BIT PCM63, which should be the most classic DAC chip of BB Company, with excellent performance indicators and a good listening experience. It has always been seen on many high-end aircraft in the past. PCM1702 of
20BIT, this is the product used by BB Company to replace PCM63. It has better performance indicators than PCM63, but for some unknown reason, many people would rather like the sound of PCM63 than accept the sound of PCM1702. It may be that PCM63 is easier to get a thicker and warmer sound than PCM1702.
PCM1702, PCM63 and PCM1704 all adopt BB's unique COLINEAR structure. It uses two 19-bit DACs in each channel to be responsible for the positive and negative half-cycle signals respectively, and then combines the current output parts of the DACs into one. This structure not only has the high dynamic characteristics of multi-bit DACs, but also successfully solves the problem of zero-crossing distortion that occurs in general multi-bit DACs, improving the linearity of the output. Most of the DAC chips of
BB company are usually divided into three levels, namely: standard P suffix, higher-level PJ suffix, and high-level PK suffix. The internal structure is exactly the same, but there are slight differences in performance.
PCM1792: The technical specifications, performance and pin positions of 1794 and 1792 are exactly the same. The difference between them is that 1794 supports hardware control and 1792 supports software control. Since the 1792 design requires additional software writing, warranty replacement is troublesome, so generally HIFI designers prefer to use 1794 or 1704. As a result, the sales volume of 1794 was several times that of 1792. There were many manufacturers using it, and the peripheral circuit design was more mature.
PCM1794: A new chip after 1704, its performance indicators were recognized as the best in the world (at that time). The details, dynamics, and resolution are as impeccable as the 1704, and it also has a pretty good musical taste. The sound effect is almost exactly the same as the PCM1704. It belongs to the type that is loose and atmospheric, light and easy, very relaxed and leisurely, the dynamic performance of the bass is also very good, the listening experience is relaxed, and it can be both static and dynamic. It is more expensive and requires more complex supporting circuits to achieve the highest and strongest effect. After listening to the good 1704/1794, the sound is unforgettable.Many foreign countries use 1704/1794 as the core, and there are many decoders worth more than 10,000 yuan. What's more, 2 parallel connections or 2 differential + 2 parallel connections (a total of 4) are used as decoders, so that its dynamics can be doubled and the current driving capability can be doubled to achieve extremely high effects. The sound is also thicker.
PCM1798: 1794, 1792, and 1798 are three brothers. PCM1798 is a simplified version of 1794/1792, which is more cost-effective.
PCM1793: According to the positioning of PCM, it is two levels lower than 1798. It is also a cost-effective product, but the surrounding supporting circuits are simple. The
PCM series of chips has excellent effects, but it requires extremely high skills from the designer. If the design is not good, it will easily cause bad noise. If you want to buy works from the DIY circle, it is not recommended to choose this one.
2, ADI's AD1862
This is a 20BIT DAC chip produced by Analong Devices. The signal/noise ratio of AD1862 is very good. With two external bypass capacitors, the signal/noise ratio can be as high as 119dB, and the 20BIT resolution The efficiency also gives it a dynamic range of 120dB. A typical application is DENON's DG2560. It should be noted here that some semiconductor manufacturers test the finished products and classify their performance before higher-end chips leave the factory. Therefore, even DACs of the same brand and number have differences in performance and price. AD1862 is divided into two grades: AD1862N and AD1862N-J. The appearance of
ad1862 is a 16pin plastic dip package, and the circuit structure is roughly composed of three parts. The serial signal is input to a 20-bit decoding section and converted into a parallel signal and sent to the 20-bit DAC. The digital signal is converted into an analog current signal for output. There is also a set of reference voltage sources supplied to the 20-bit DAC section.
In addition to being used in digital audio, it is also used in electronic synthetic instruments, professional recording digital mixers and various effects devices. The range of applications is quite wide.
3, PHILIPS's TDA1541 and TDA1543
TDA1541 and TDA1543 are all 16BIT DAC chips, and TDA1541 can be called the king of 16BIT chips. Its excellent THD+N performance index cannot even catch up with many so-called 18BIT DAC chips. In particular, its external ladder bypass point has become a favorite among DIY players to adjust the sound. MARANTZ and PHILIPS are the main users of this DAC chip, especially MARABTZ, which also uses it on its top-level CDs.
TDA1543 is also a 16BIT DAC. It has only 8 pins and can be regarded as a very simple DAC chip. But its high-frequency distortion can be considered appalling. TDA1543 has TDA1543A with suffix A and TDA1543 without. The data formats accepted by these two models are different and cannot be directly replaced. The differences are: TDA1543A accepts MSB 16BIT data signals and cannot accept IIS data format. The TDA1543 only accepts IIS data format. There are 4 types of
TDA1541. The grades from low to high are: TDA1541A, TDA1541A/R1, TDA1541A/S1, TDA1541A/S2. Their performance data are different, which are mainly reflected in S/N and -60DB respectively. THD+N distortion, these 4 models only accept the IIS data format. Although
TDA1541 is 16BIT, the effect is superb, the midrange is warm and charming, and the music has a rich flavor. It is a warm and sweet type, suitable for small and medium-sized classical and vocal arrangements.
4, SANYO's LC7881 and LC78820, LC7881 is 16BIT, and LC78820 is 18BIT. The DAC chip used by the previous generation of affordable CD player CEC891 is LC78820. The LC7881 is mostly used in affordable computers, and its performance is not as good as the PCM56.
html From 0 to multi-BIT DAC chips, there is another one that must be mentioned, which is the D20400A of ULTRA ANALOG. It has good performance indicators and a good listening experience. It is a 20-BIT DAC chip. At that time, it could only be seen in high-priced equipment. figure.The current DAC chip production camp is much stronger than the multi-BIT ones. These include PHILIPS' bit stream technology, Panasonic's MASH, CS's DELTA SIGMA, SONY's PULSE, and Japanese AK and NPC.
2. 1-bit chip introduction
1, philips saa7350
saa7350 is a 1-bit DAC produced by Philips. Each saa7350 can provide two-channel balanced output.
When it is connected to the 20-bit eight-fold supersampled digital filter , the digital signals of the left and right channels are sent to the noise shaping filter through the input interface, and come out as a 1-bit signal, and then enter the 1-bit signal composed of a cmos gate and a capacitor. Bit d/a turns the 1 bit signal into a continuous signal, and finally outputs it through the analog low-pass filter . The dac7 decoding module of
philips uses saa7350. In fact, it should be said that it only uses the first half of saa7350 plus two tda1547 1bit converters. The function of tda1547 is roughly equivalent to the 1 bit output of the noise shaping filter of saa7350. bit signal, and then passes through 1 bit of tda1547 d/a to the integral line, and then output through the low-pass filter from saa7350+tda1547 plus npc's sm5803 and other peripheral components to form the dac7 digital module, which is widely used in 1-bit d/a systems of Philips and other brands middle. Although
dac7's method only uses the second half of the saa7350 separately, it prevents the analog part from being seriously interfered by the digital part and makes the linearity more perfect. Otherwise, Philips would not be so wasteful to use the tda1547, which has the same function as the saa7350 analog part. wafer.
In order to achieve a considerable degree of multi-bit accuracy, the early 1-bit DAC had to use a very high-frequency colck. If 1bit d/a is used to represent 16bit data with one lsb, the clock frequency will be as high as 44.1khzx (216-1) = 2.8900935ghz. To be equivalent to 16bit, a clock frequency of nearly 3ghz is required. From this point of view, it seems that 1bit dac has no future. Fortunately, the △Σ modulation method was gradually developed later, and it evolved into the "multi-stage noise quantization suppression technology" mash (multistage noise shaping), so that 1bit dac can achieve higher accuracy than before using a lower clock speed. . Today, most 1-bit decoding units have applied the above technology, so whether it is mash, △Σ, or Σdeco, they are actually 1 bit.
2, Crystal's CS4396/4397 and CS43122
Crystal also has a high status in the digital audio industry. The CS4396 and CS4397 launched in July 1999 are enough to be on an equal footing with PCM1704. CS4396 adopts the very popular multi-bit Delta Sigma decoding method, with 24-bit decoding accuracy and built-in digital filter. CS4396 converts the pulse code modulation (Pulse Code Modulation) signal into the pulse density modulation (Pulse Density Modulation) signal based on the Dynamic Element Matching technology developed by Philips, and finally converts the digital signal through a low-pass filter composed of switched capacitors. is an analog signal. This multi-bit chip structure has lower out-of-band noise and is much less sensitive to time base jitter. At that time, the highest sampling frequency of CS4396 had reached 192kHz, the signal-to-noise ratio had reached an astonishing 120dB, the total harmonic distortion and noise had reached -100dB, and the dynamic range had reached 120dB (exceeding PCM1704). The performance parameters of CS4397 are not much different from CS4396. The only difference is that CS4397 can provide DSD interpolation filter with external PCM or SACD. In October 1999, Crystal launched the more highly integrated CS4391/4392, which not only contains fourth-order Delta Sigma decoding, but also includes digital filters including analog output filters and volume controls. Its maximum sampling frequency is still 192kHz, but the dynamic range is reduced to 107dB, and the total harmonic distortion and noise reach -97dB.
In December 2000, Crystal Company launched the DAC chip CS43122 with superior performance, which adopted the second generation dynamic unit matching technology and achieved a dynamic range of up to 122dB. Its signal-to-noise ratio is still an astonishing 120dB, and its total harmonic distortion and noise reached -102dB. It was one of the best-performing DAC chips of the year.
CS4397 has a neutral sound, good musical appeal, higher resolution and dynamics than TDA1541, and higher musical taste than AD1955, which is between the styles of 1541 and 1955. This characteristic creates a thicker sound, the most delicate treble, a neutral to warm sound, and a sweet sound. A large number of famous machines are selected, such as the machines of American MBO, which have many reference circuits and mature designs.
CS43122: Better than CS4397, the sound is slightly different. It was the top model of Crystal Company at that time. However, the chip has now been discontinued.
3, ADI's AD1853
Analog (analog device) company also has a place in the field of DAC chips. At the end of 1998, the company launched the top DAC chip AD1853. It is the world's first chip that can adapt to the DVD Audio 192kHz sampling frequency and has multi-bit Delta Sigma decoding function. It is fully adapted to all DVD Audio formats, supports 48kHz, 96kHz, 192kHz sampling frequencies, supports 24bit data width, supports the Red Book standard, and has a built-in speed of 50 μs/15 μs digital de-emphasis filter (sampling frequencies are 32kHz, 44.1kHz and 48kHz), and its cutoff passband in-band compensation is as high as 115dB. AD1853 contains perfect differential linear restoration technology, which can significantly reduce noise. Its built-in data irregular orientation technology can minimize the possibility of infection with time base jitter (Jitter). The best thing is that its total harmonic distortion plus noise is -107dB in mono and -104dB in stereo . Although it has been more than two years, there is still no DAC chip with such low noise and distortion. The performance of the AD1853 is not bad in other aspects. When the A-weighted mono unmuted sampling frequency is 48kHz, the signal-to-noise ratio is 120dB, and in stereo it is 117dB; when the A-weighted mono unmuted sampling frequency is 48kHz, the dynamic range is 119dB, and in stereo it is 116dB. . Judging from data alone, it can also become one of the best DAC chips in the world (and one of the manufacturers with the most stringent data specifications), and has maintained this record for more than two years. Analog later launched a higher-end DAC than the AD1873, the AD1955.
Such a good chip will not go unappreciated. Golden Throat uses AD1853 in all two-channel decoders and CD players it produces, and Denon also uses AD1853 extensively in all its top-level multi-channel amplifiers and CD players. The
AD1853 chip and AD1955 have very good details and dynamics, as well as a good sense of energy. Except for PCM1704/1794, their details/dynamics/resolution are quite high and they are fierce. But some people feel that the sound is like boiled water and lacks musicality. The design of AD1955 requires software programming. It is difficult to fully utilize the advantages if only hardware is used.
4, NPC company's SM5865
The only DAC chip that exceeds the AD1853 is NPC company's SM5865. NPC also has a high reputation in the industry. Marantz's high-end CD players use NPC's SM5872 extensively for D/A conversion, but NPC's presence in multi-channel surround sound amplifiers is not common. However, the quality of NPC's DAC chips is not inferior to any other manufacturer in the world. SM5865 adopts Philips' second-generation dynamic element pairing technology, third-order Delta Sigma noise shaper, and 31 levels of quantization, thus achieving the world's lowest total harmonic distortion characteristics of DAC chips. Its total harmonic distortion plus noise is only 0.0003%, which is -110.5 dB. It also has very low in-audio quantization residual noise, even eliminating the need for a low-pass filter. The output stage also uses a differential circuit, which can output very accurate analog signals. SM5865 still uses 24bit conversion accuracy, the highest sampling frequency is also 192kHz, the dynamic range is 117dB, and the signal-to-noise ratio is 120dB. It should be pointed out that SM5865 is a mono chip.
5, Wolfson - WM8741
WM8741 is a representative product launched by Wolfson. Wolfson is currently the leading manufacturer of high-performance mixed-signal chips in the consumer electronics market. It is committed to designing and supplying breakthrough products, including ultra-low power consumption and high-performance audio components. , microelectromechanical microphones, power management ICs and noise cancellation solutions, forming a complete Wolfson The AudioPlus product portfolio helps Wolfson's current customers' digital consumer electronics products establish competitive advantages and enable new end-user experiences. The
WM8741 combines ultra-high performance out-of-band noise and linearity with a unique programmable high-quality digital filter to produce a more natural "analog feel" sound than traditional digital-to-analog converters of the past. Low out-of-band noise and high-quality linearity are achieved through a unique low-order modulator and multi-bit digital-to-analog converter (DAC) architecture, resulting in improved sound quality. The WM8741 uniquely handles the unnatural pre-reverberation and delay effects found in traditional DACs, providing a clearer, more natural sound for music playback.The index of
WM8741 is very high, but the actual listening effect is inferior to its technical index. It is still incomparable to PCM1794/PCM1792. Generally speaking, it has a delicate and sweet sound, a relatively gentle sound, with some European and British flavor. The disadvantage is that the compatibility is worse than CS4398. If the hard control and software design are not good, some strange problems may easily occur. This chip is the same as the AD1955. The product circuit design and tuning have a greater impact on the sound. The British "Lotus" prefers to use WM8741, which has a European analog music taste.
6, Cirrus Logic - CS4398
CS4398 is manufactured by Cirrus Logic, a supplier of precision linear circuits and high-quality mixed-signal chips. Selling its products under its own brands and trademarks, the products it produces are used in fields such as mass storage (magnetic storage and optical storage), audio equipment (professional, consumer fields and personal computers), precision data conversion, etc. In the past ten years, Cirrus Logic has won more than 60 industry firsts, many of which have established new corresponding industry standards for its technologies.
CS4398 is a complete stereo 24-bit/192 kHz digital to analog (D/A) conversion system. The D/A system features digital de-duplication, 0.5 dB step volume control, ATAPI channel mixing, selectable fast and slow digital interpolation filters, and an oversampling multi-channel filter that uses mismatch shaping technology to eliminate capacitive mismatching and eliminate distortion. Delta-Sigma regulator. The system also features a multi-element switching capacitor and low-pass filter with differential analog output. The expressed sound: warmer, more delicate, full of emotion, and with excellent details. CS4398 is a new chip launched after CS43122 was discontinued. It is the highest level model of Crystal Company. The sound style is similar to CS43122, and now it competes side by side with CS4397 in the DAC decoding chip market. The surrounding circuits are well designed, and the dynamics are by no means inferior to those of PCM1704/1794 and other higher-index chips. The sound displayed: warm and delicate, probably the most delicate among these top-level chips. It is very good to listen to singing, full of emotion, good details, and very high dynamic indicators, soft and strong.
The decoding module of the hot-selling Marantz decoding and headphone amplifier all-in-one HD-DAC1 uses CS4398. When paired with the Marantz HDAM-SA2 module, the HD-DAC1 not only drives headphones with excellent performance, but can also be used as a pure decoder to connect to other power sources. amplifier or a pair of powered speakers.
Final summary:
Under the same peripheral design level and basically the same conditions, the DAC chip sound effect, comprehensive ranking (actual sound performance has a huge relationship with the circuit, for reference only):
No. 1: PCM1794/PCM1792/PCM1704. (The chip with the best performance in the world at present, the best chip in the 24BIT era, is several times more expensive than the following models)
No. 2: ESS9018/9028/9038 (the chip with the highest index in the world at present) chip, supports 32-bit/500KHZ, 1BIT high-speed chip, and is a strong competitor of BB Company)
No. 3: CS43122/CS4398/CS4397/CS4396. (The best CS43122 has been discontinued, otherwise it would be able to compete with PCM1794. Unfortunately, the only one that can be purchased in large quantities now is CS4398)
No. 4: AKM4497/4490/4395, a cheap and affordable decoding chip with low quality indicators. It's a bit lacking but sounds good. Among them, 4497 can be ranked 4th, and the rest can only be ranked ahead of 8th.
No. 5: AD1955 (Since the performance of AD1955 varies greatly with the design level)
No. 6: WM8741 (The sound effect of this chip is no worse than CS4398, but the compatibility is inferior to CS4398)
No. 7: PCM1702 (20BIT era One of the better chips)
No. 8: TDA1541/TDA1547. (The best chip in the 16BIT era. After all, the old Huang Zhong was 20 years ago. In terms of indicators, the 16BIT lags 1024 times behind the 24BIT.)
