Different Types of DAC explained
Posted: Mon Mar 05, 2018 1:56 pm
This article is a review of a DAC but it contains a description of the different types ...
http://enjoythemusic.com/superioraud ... iew.htm
"......
Digital: Perfect Sound Forever
Back when CD first came on the market, it was toted as producing "Perfect Sound Forever," which is wonderful advertising lingo, but doesn't actually address the reality of making the all important jump from binary numbers back into analog sound. As it turns out, there are several different kinds of DAC technologies that have been created over the years. And each has it's own strengths and weaknesses, based on cost, technical limitations, and perceived sound quality.
The two basic categories are "Multi-Bit" or "Ladder" DACs and "Single-Bit" or "Sigma-Delta" DACs. These two categories of DACs can be implemented into more advanced and hybrid designs that are used in most modern DACs. And any of these technologies can incorporate different types of noise shaping, upsampling, oversampling, digital signal processing (DSP), digital filters, and format transcoding. Below are simplified descriptions of some of the more popular DAC technologies:
A) Multi-Bit R-2R Ladder DAC
The first and most basic type of DAC. It comprises three parts: a series to parallel shift register, an R-2R resistor ladder, and a summing amp. The R and 2R stand for "resistance" and "two times resistance" which comprise the two values of precision matched resistors that make up the ladder. Each "rung" on the ladder corresponds to a bit, and each time the resistance is doubled it decreases the reference voltage by half, which corresponds to the binary relationship between bits. A stream of digital words is fed in series into the input of the shift register. The series digital word is "shifted" into the series to parallel register, which then outputs each bit simultaneously in parallel with each pulse of the "word clock" (WCLK). Each parallel output from the shift register corresponds to a rung in the R-2R ladder. And the output drivers from the shift register push a reference voltage representing each bit through each rung in the R-2R ladder. Then all the voltages from all the rungs in the ladder are added together by the summing amp. A simple and elegant solution. If you are playing a 44.1 kHz Red Book CD this process takes place 44,100 times per second. And to decode a 16-bit Red Book CD file, a 16-bit shift register and 16 precisely matched pairs of R and 2R resistors is required. Great sound quality and high value have been achieved using this design that is featured in the original Philips TDA1540 DAC chips from the 1980's. One downside to multi-bit DACs is that they can only decode PCM formats, such as Red Book CD, AIFF and WAV, or decompressed PCM CODECs such as FLAC, ALAC, and MQA. Another downside is that even with the most advanced modern technology they can not match the resistors in the ladder closely enough to produce better than a 20-bit resolution (although Zwickel and I disagree on this point).
B) Multi-Bit Binary Weighted DAC
Similar to an R-2R Ladder DAC, but simpler and larger. It requires the same three parts: shift register, resistors, and summing amp. But in the case of a binary weighted DAC, instead of creating a resistor value to correspond to each bit by summing multiple resistors in a parallel/series ladder configuration, it uses dedicated resistors for each bit (e.g. R, 2R, 4R, 8R, 16R, 32R, 64R, etc.) in a parallel configuration. In modern monolithic chip manufacture each resistor is laser matched on the chip using automated technology. So since it doesn't require more precision to match these resistors than an R-2R, and since it has half the number of resistors to precision match, it is actually less expensive to manufacture than an R-2R DAC chip. But since each resistor value is independent instead of the summation of several lower value resistors, it is significantly larger. Some theorize that because each resistor value is discrete as opposed to a combination of multiple smaller value resistors, this decoder design can display tremendous sonic accuracy and an incredibly low noise floor. Famous high-performance DACs produced by the original Theta Digital and Wadia used this technology. Of course being a multi-bit DAC, it is only capable of decoding PCM file formats, and being made from laser matched resistors, it is only capable of 20-bit resolution. (Again, Zwickel and I disagree on this point.)
C) Multi-Bit Segmented DAC
Most complex and most expensive of the multi-bit DACs. Similar to both the R-2R and Binary Weighted DACs in many ways. The major difference is that in order to achieve lower noise and higher resolution they use multiple "segments" of resistor networks, one or more for the Most Significant Bits (MSB), and one or more for the Least Significant Bits (LSB). Then different segments are summed, multiplied by different values, and the segment summations are combined to reconstruct a resolution greater than 20-bits. The famous PCM1704 DAC chip is a good example of a modern 24-bit multi-bit segmented DAC. And there are even companies like DaVinci that boast about their proprietary 32-bit multi-bit segmented DAC technology. Of course unless a transcoding algorithm is added to one of these modern multi-bit segmented DACs, they still can only decode PCM formats.
D) Single-Bit Oversampling Delta-Sigma DAC
Simplest, least expensive, and most modern of the DAC technologies. Why? Because it doesn't require a laser matched shift register or laser matched resistor for each bit. A 1-bit "switch" processes the digital words in an oversampling "Delta-Sigma" loop. Delta modulation encodes the stream of numbers (PCM values) into a stream of pulses known as Pulse Density Modulation (PDM). The accuracy of Delta modulation is improved by passing the encoded output through a 1-bit DAC and adding the resulting analog signal, known as Sigma, back into the PCM input signal. This process is often oversampled (x2, x4, x8, etc.) to improve statistical accuracy. But most often, a 1-bit MASH (Multi-Stage Noise Shaping) circuit running at very high speed (MHz) has become the most commonly used DAC in modern consumer electronics. PDM is the native format of DSD64 a.k.a. SACD. A DSD64 file is 1-bit at 64 times the 44.1KHz sampling frequency of a Red Book CD, which translates to sampling 2,822,400 times per second. And that's just Single-Rate DSD: most modern Delta-Sigma DAC chips can decode Double-Rate (5,644.8KHz) and even Quadruple-Rate DSD (11,289.6KHz). To give you some perspective on what this all means, DSD64/SACD is roughly 33 times the resolution of a Red Book CD, and roughly equal resolution to a 24-bit/96kHz PCM file. By oversampling and running Delta-Sigma DACs at these insanely high sampling frequencies it puts the quantization noise high above the audible band. This allows them to use less extreme and more sophisticated post conversion high-frequency analog filtering that puts far less digital artifacts into the audible band than the "brick wall" high-frequency analog filters used in earlier types of multi-bit DACs. This type of DAC can decode both PCM and DSD, but since its native format is DSD, they usually perform better when converting Double- or Quad-Rate DSD.
E) Single-Bit And Multi-Bit Hybrid DAC
A combination of one or more of the above DAC technologies combined with algorithmic methodologies used together to save cost, increase versatility, increase accuracy, and/or increase mass appeal. This can include single and multi-bit converters, oversampling, noise shaping, conversion to PWM or PDM, as well as adding negative feedback to correct errors. In more advanced Delta-Sigma DAC chips 4- to 8-bit Delta-Sigma oversampling cascade several lower bit decoders in order to develop a more precise interpretation of the signal. This topology is used in the many fine DACs at affordable prices from PS Audio, Mytek, Bryston, AMR, and iFi Audio, just to name a few. This type of DAC can include all sorts of features, such as built in MQA decoders, built in digital volume controls, selectable output filtering, selectable dither, and digital EQ.
F) FPGA Hybrid DAC
Versions of the above DAC topologies executed in a Field Programmable Gate Array (FPGA). This is actually not a different type of DAC, but more a way companies can engineer their own unique algorithms, noise shaping, oversampling, upsampling, error correction, firmware, transcoding, and filtering technologies, in a powerful, flexible, modest cost, integrated circuit (IC). The exact same circuits could be put into a dedicated IC module like the ones used by all of the above DAC technologies, but the initial investment and minimum production run quantities would be far beyond the means and needs of one manufacturer. Using an FPGA not only allows for more processor intensive algorithms, it also allows these DACs to be updated/upgraded at any time, making them potentially more flexible, while preventing obsolescence. Some excellent examples of FPGA DACs would be the more advanced DACs designed by companies like PS Audio and Chord.
....."
http://enjoythemusic.com/superioraud ... iew.htm
"......
Digital: Perfect Sound Forever
Back when CD first came on the market, it was toted as producing "Perfect Sound Forever," which is wonderful advertising lingo, but doesn't actually address the reality of making the all important jump from binary numbers back into analog sound. As it turns out, there are several different kinds of DAC technologies that have been created over the years. And each has it's own strengths and weaknesses, based on cost, technical limitations, and perceived sound quality.
The two basic categories are "Multi-Bit" or "Ladder" DACs and "Single-Bit" or "Sigma-Delta" DACs. These two categories of DACs can be implemented into more advanced and hybrid designs that are used in most modern DACs. And any of these technologies can incorporate different types of noise shaping, upsampling, oversampling, digital signal processing (DSP), digital filters, and format transcoding. Below are simplified descriptions of some of the more popular DAC technologies:
A) Multi-Bit R-2R Ladder DAC
The first and most basic type of DAC. It comprises three parts: a series to parallel shift register, an R-2R resistor ladder, and a summing amp. The R and 2R stand for "resistance" and "two times resistance" which comprise the two values of precision matched resistors that make up the ladder. Each "rung" on the ladder corresponds to a bit, and each time the resistance is doubled it decreases the reference voltage by half, which corresponds to the binary relationship between bits. A stream of digital words is fed in series into the input of the shift register. The series digital word is "shifted" into the series to parallel register, which then outputs each bit simultaneously in parallel with each pulse of the "word clock" (WCLK). Each parallel output from the shift register corresponds to a rung in the R-2R ladder. And the output drivers from the shift register push a reference voltage representing each bit through each rung in the R-2R ladder. Then all the voltages from all the rungs in the ladder are added together by the summing amp. A simple and elegant solution. If you are playing a 44.1 kHz Red Book CD this process takes place 44,100 times per second. And to decode a 16-bit Red Book CD file, a 16-bit shift register and 16 precisely matched pairs of R and 2R resistors is required. Great sound quality and high value have been achieved using this design that is featured in the original Philips TDA1540 DAC chips from the 1980's. One downside to multi-bit DACs is that they can only decode PCM formats, such as Red Book CD, AIFF and WAV, or decompressed PCM CODECs such as FLAC, ALAC, and MQA. Another downside is that even with the most advanced modern technology they can not match the resistors in the ladder closely enough to produce better than a 20-bit resolution (although Zwickel and I disagree on this point).
B) Multi-Bit Binary Weighted DAC
Similar to an R-2R Ladder DAC, but simpler and larger. It requires the same three parts: shift register, resistors, and summing amp. But in the case of a binary weighted DAC, instead of creating a resistor value to correspond to each bit by summing multiple resistors in a parallel/series ladder configuration, it uses dedicated resistors for each bit (e.g. R, 2R, 4R, 8R, 16R, 32R, 64R, etc.) in a parallel configuration. In modern monolithic chip manufacture each resistor is laser matched on the chip using automated technology. So since it doesn't require more precision to match these resistors than an R-2R, and since it has half the number of resistors to precision match, it is actually less expensive to manufacture than an R-2R DAC chip. But since each resistor value is independent instead of the summation of several lower value resistors, it is significantly larger. Some theorize that because each resistor value is discrete as opposed to a combination of multiple smaller value resistors, this decoder design can display tremendous sonic accuracy and an incredibly low noise floor. Famous high-performance DACs produced by the original Theta Digital and Wadia used this technology. Of course being a multi-bit DAC, it is only capable of decoding PCM file formats, and being made from laser matched resistors, it is only capable of 20-bit resolution. (Again, Zwickel and I disagree on this point.)
C) Multi-Bit Segmented DAC
Most complex and most expensive of the multi-bit DACs. Similar to both the R-2R and Binary Weighted DACs in many ways. The major difference is that in order to achieve lower noise and higher resolution they use multiple "segments" of resistor networks, one or more for the Most Significant Bits (MSB), and one or more for the Least Significant Bits (LSB). Then different segments are summed, multiplied by different values, and the segment summations are combined to reconstruct a resolution greater than 20-bits. The famous PCM1704 DAC chip is a good example of a modern 24-bit multi-bit segmented DAC. And there are even companies like DaVinci that boast about their proprietary 32-bit multi-bit segmented DAC technology. Of course unless a transcoding algorithm is added to one of these modern multi-bit segmented DACs, they still can only decode PCM formats.
D) Single-Bit Oversampling Delta-Sigma DAC
Simplest, least expensive, and most modern of the DAC technologies. Why? Because it doesn't require a laser matched shift register or laser matched resistor for each bit. A 1-bit "switch" processes the digital words in an oversampling "Delta-Sigma" loop. Delta modulation encodes the stream of numbers (PCM values) into a stream of pulses known as Pulse Density Modulation (PDM). The accuracy of Delta modulation is improved by passing the encoded output through a 1-bit DAC and adding the resulting analog signal, known as Sigma, back into the PCM input signal. This process is often oversampled (x2, x4, x8, etc.) to improve statistical accuracy. But most often, a 1-bit MASH (Multi-Stage Noise Shaping) circuit running at very high speed (MHz) has become the most commonly used DAC in modern consumer electronics. PDM is the native format of DSD64 a.k.a. SACD. A DSD64 file is 1-bit at 64 times the 44.1KHz sampling frequency of a Red Book CD, which translates to sampling 2,822,400 times per second. And that's just Single-Rate DSD: most modern Delta-Sigma DAC chips can decode Double-Rate (5,644.8KHz) and even Quadruple-Rate DSD (11,289.6KHz). To give you some perspective on what this all means, DSD64/SACD is roughly 33 times the resolution of a Red Book CD, and roughly equal resolution to a 24-bit/96kHz PCM file. By oversampling and running Delta-Sigma DACs at these insanely high sampling frequencies it puts the quantization noise high above the audible band. This allows them to use less extreme and more sophisticated post conversion high-frequency analog filtering that puts far less digital artifacts into the audible band than the "brick wall" high-frequency analog filters used in earlier types of multi-bit DACs. This type of DAC can decode both PCM and DSD, but since its native format is DSD, they usually perform better when converting Double- or Quad-Rate DSD.
E) Single-Bit And Multi-Bit Hybrid DAC
A combination of one or more of the above DAC technologies combined with algorithmic methodologies used together to save cost, increase versatility, increase accuracy, and/or increase mass appeal. This can include single and multi-bit converters, oversampling, noise shaping, conversion to PWM or PDM, as well as adding negative feedback to correct errors. In more advanced Delta-Sigma DAC chips 4- to 8-bit Delta-Sigma oversampling cascade several lower bit decoders in order to develop a more precise interpretation of the signal. This topology is used in the many fine DACs at affordable prices from PS Audio, Mytek, Bryston, AMR, and iFi Audio, just to name a few. This type of DAC can include all sorts of features, such as built in MQA decoders, built in digital volume controls, selectable output filtering, selectable dither, and digital EQ.
F) FPGA Hybrid DAC
Versions of the above DAC topologies executed in a Field Programmable Gate Array (FPGA). This is actually not a different type of DAC, but more a way companies can engineer their own unique algorithms, noise shaping, oversampling, upsampling, error correction, firmware, transcoding, and filtering technologies, in a powerful, flexible, modest cost, integrated circuit (IC). The exact same circuits could be put into a dedicated IC module like the ones used by all of the above DAC technologies, but the initial investment and minimum production run quantities would be far beyond the means and needs of one manufacturer. Using an FPGA not only allows for more processor intensive algorithms, it also allows these DACs to be updated/upgraded at any time, making them potentially more flexible, while preventing obsolescence. Some excellent examples of FPGA DACs would be the more advanced DACs designed by companies like PS Audio and Chord.
....."