Digital to analog converter
The term "digital to analog converter" is used to describe a device that accepts a digital audio input and outputs an analog audio signal that is re-constructed from the digital code. This code is typically linear PCM format; but may also be other formats such as DSD or I2S (typically used internally in digital to analog converter units). A digital to analog converter must be used to listen to a digital audio signal. The term can be used to describe the actual digital to analog converter IC or circuit, or an entire unit that incorporates all of the necessary support circuitry to accept the encoded digital audio signal in one or more formats and output line level analog audio signals.
Prior to the development of practical digital audio recording systems; DA converters were used primarily in industrial control applications. These early converters were limited either by the converter technology at the time or by the amount of data that associated system could handle to much lower resolution than typically used to encode audio. The resolution both in the amplitude domain (typically the voltage of the output waveform) and time domain of these converters was often quite limited when compared to contemporary digital audio standards.
Before storage of the huge amount information generated by CD quality AD converters became practical, the earliest application of DA converters in music recording was in "outboard" equipment such as digital delay or effects processors. Largely because the output of these early units was mixed in with the original (unprocessed) source at a low level as an ambient effect; the less-than high fidelity quality of the converters was acceptable. Even with the noise and distortion present in analog recordings, the perceived quality of the analog tape recordings was far better than the signal processed through these early converters. One of the more popular early digital delay units employed a novel form of digital encoding "sigma-delta" where, in contrast to the "linear PCM" format where each "sample" of the analog input waveform is represented by a digital word made of a number of bits; sigma-delta encoded only one bit at a relatively high sample frequency. Compared to the relatively inaccurate PCM-based units, most recording engineers felt that the sigma-delta digital delay unit sounded closer to the source.
With the introduction of Compact Disc technology by Sony/Phillips in the early 1980's came the standard of recording audio in 16 bit linear PCM format. DA converter technology was still evolving at the time and even though many DA converters were nominally "16 bit" they were not truly accurate to 16 bit resolution. Contemporary DA converters are typically "24 bit." The sample frequency capability of DA converters has also increased since the original CD format of 44.1 kHz was introduced; with contemporary DA converters supporting output sample frequencies as high as 384 kHz. Although there are a number of advantages to DA conversion at sample frequencies higher than 44.1 kHz, these advantages are gained at sample frequencies of 88.2 or 96 kHz. Increasing the sample frequency beyond 96 kHz will degrade the conversion accuracy in the audio frequency range, while the only advantage is the ability to reproduce hypersonic frequencies beyond the range even dogs can hear.
In order to make a useful digital audio system; the method used to encode and decode the analog audio signal must:
- Be reciprocal for encoding (recording) and decoding (playback).
- Must be able to "re-construct" the original analog information to a minimum level of accurately.
- Ideally incorporates a "standard" that facilitates interchange between systems made by different manufacturers.
To achieve (1), contemporary digital audio systems use a method referred to as "sampling" which, in a manner analogous to film or video cameras, takes a contiguous series of "snapshots" of the audio waveform at a specific frequency (the sample frequency). Analog audio derives its name from the manner in which the acoustic pressure variation of the original sound is represented by a voltage waveform with the same variations- the voltage variation is "analogous" to the pressure variation at every point in time. Although it is possible that at specific points in an audio system the signal is represented by current variations as versus voltage variations; the analog signal is typically a voltage waveform when it is transmitted from one piece of audio equipment to another.
The digital "words" are recorded in sequence as a file, and can be stored or transmitted without change to the information. In order for the playback DA to accurately reconstruct the voltage waveform; it must output the voltage of each sample at exactly the same voltage level and exactly the same relative time. This means the sample frequency must be very close to the same frequency and, more importantly, the sample clock must have extremely even time periods for each sample. This where the discussion of "jitter" comes in- jitter is the term that is used to describe short-term variations in the clock cycle period caused by real-world issues common to the transmission of very high frequency signals over signal conductors (cables or even signal "traces" on printed circuit boards). Although voltage (amplitude domain) accuracy has increased dramatically since the early days of digital audio; the performance of even extremely accurate converters can be compromised by inaccurate clocking of the conversion either during AD conversion, during DA conversion, or both.
A typical DA converter is actually a system made up of a number of stages:
- a.) The digital input circuitry
- b.) The clock circuitry
- c.) The digital signal processor
- d.) The DA converter
- e.) The level-shifting stage
- f.) The output filter
- g.) Line output stage
To accept an external digital audio signal. the DA converter must have at least one digital audio input.The AES (Audio Engineering Society) began the process of standardizing the format of transmission for digital audio- both digital coding and the physical/electrical connections, in the 1980's. Most contemporary digital audio devices incorporate the AES3 standard and the corresponding IEC consumer standard which is nearly identical in coding. The primary difference is the professional AES3 standard employs either "balanced" XLR connections carrying differential "TTL" 5 volt signals or BNC coaxial single-ended ("unbalanced) TTL level signals. The consumer formats are either RCA coaxial 0.5V unbalanced signals or optical signals typically employing "Toslink" connectors. In some cases BNC connectors are substituted for RCA connectors or other physical forms of optical connectors are sued in place of Toslink.
The digital audio input signal is a "self-clocking" code that contains both the digital audio data and an embedded "bit clock". The DA converter must synchronize to the incoming digital audio both to accurately accept the incoming data and generate the clock signal used for the actual DA conversion.
Internally, the DA converter may employ the I2S format for transmission between IC’s. The I2S format is common and typically consists of three signals: I.) The "Bit clock" which has one cycle for each "bit" in the serial data output of the AD converter. II.) The "Word Clock" which is at the sample frequency and each half cycle is used to define whether the serial data is the left channel or right channel data (most contemporary converters are "stereo" two channel units). III.) The "Serial Data" which is the digital code containing each sample's voltage level information.
This format has advantages for transmission of the digital audio information between IC's located in close proximity to each other on the same PC board; but is subject to the same quality issues as any other high frequency signal traveling down a conductor. It was not intended for transmission between pieces of equipment.
The DA converter in (d) is typically an IC that incorporates inputs for the digital audio, clock, and digital and analog power supply. It also can contain output “de-glitching” circuitry to eliminate distortion caused when the output of the DA converter circuit is changing. Contemporary DA converters typically employ over-sampling to increase accuracy and ease output filter requirements. This is why most DA converters have some form of DSP; to perform the calculations necessary for over-sampling.
Most digital circuitry operates on a "single-ended" power supply; which means that one of the two power supply connections is ground or "0 volts" and the other connection is a "+" voltage (typically 5 or 3.2 Volts). Because a digital audio converter must "bridge" between digital and analog circuitry; the analog output of a DA converter IC is typically also single-ended (varies between 0 and 5 volts). Most contemporary analog audio is "bi-polar" and operates on "plus and minus" voltages equally above and below ground. So a level-shifting stage (e) is used to "shift" the audio waveform from a positive-only voltage to a bi-polar voltage.
The output of the DA converter will contain square-edged "pulses" that effectively contain energy above the Nyquist frequency (supersonic energy). Although inaudible; most audio circuitry is not designed to operate at these frequencies and will quite often operate improperly if hypersonic frequencies are present in the audio signal. Output filters (f) are required to eliminate this hypersonic energy from the output of the DA converter. After passing through the output filters; a nearly exact replica of the originally recorded analog waveform is re-constructed from the digital code.
For transmission to other pieces of audio equipment, the analog signal is amplified and "buffered" by a line level output stage. The line output is typically either in the form of an RCA connector for consumer -10dBV line level or XLR for +4dBu line level. In some cases, outputs may have variable level controls for calibration of the output level to match other equipment or Volume control. Additional outputs specifically for headphones may also be present in a Digital to analog converter unit.