Digital audio processors aim for high-performance filtering

Digital audio processors aim for high-performance filtering
Although CDs have been in use for nearly twenty years, audio has remained mostly analog. Until now, digital music on the CD has been converted to analog inside the CD player and taken through analog signal processing and amplification before being applied to the speakers. But with the recent increase in such digital infrastructure as the Internet, digital networks in buildings and automobiles, and wireless digital communications, along with new digital sound sources like MP3 and DV, an unprecedented opportunity to process digital audio data has been created. And increasingly, there will be a need to do that kind of processing.

Digital systems allow the use of digital signal processing, which has the power and flexibility to address, in digital, problems that were too expensive or too difficult to address well in analog. Unfortunately, even in those cases where digital signal processing has been used in audio, it has until recently remained necessary to convert this digital signal to analog form.

That conversion, which was needed to perform amplification via Class AB amplifiers, or analog-input Class-D amplifiers, is no longer necessary.

In addition to the increase in digital media and infrastructure, the market is seeing an unprecedented trend toward minimization and sleeker-than-ever audio equipment styling. The convergence of these trends indicates that the market is now ripe for fully digital audio systems. Such systems, which take a digital source all the way to the speaker without ever converting it to analog, have been the dream of system designers for some time, but have not been realized until the recent advent of new technologies and newly available devices.

In a fully digital audio system, digital data from a digital audio source is sent using some digital bus and interface protocol to digital processing, where audio effects such as tone controls, volume control, equalization, 3-D effects and sound effects are applied. Once this processing is completed, the signal is sent to a digital amplifier, which converts the digital signal to a digital format that can supply adequate current and voltage to drive the speaker. This signal flows through a simple, passive filter network connected directly to the speaker.

The loudspeakers at the end of the chain are best known as digital audio sources. But what is less widely known are the other elements of the fully digital audio system: the digital interface, digital audio processing and the fully digital audio amplifier.

There are a variety of digital buses available to transmit and receive audio data. Each has its own special requirements. Some of these are standard system buses, but other markets are adopting entertainment buses whose high data rates support audio and video data along with system clocks, system control and status information. In any case, the important aspect of the interface is that the data and the clocks must reach the system in the correct format and in good condition. Of particular importance to high-quality audio reproduction is the accuracy of the recovered master clock, since low clock jitter is necessary for excellent audio performance.

Recently, excellent transceivers have entered the market, making this part of the system readily available using off-the-shelf devices. And the road maps for these devices indicate that they will continue to improve. The market can expect jitter performance to reach unprecedented levels as new technologies are applied to attack the persistent problem of reclocking serial data.

Audio data can be processed using such standard digital signal processing devices as microprocessors and application-specific ICs. However,the unique needs of audio processing (high precision and high throughput) have driven the market to produce cost-effective devices specifically for that purpose. Of particular interest is a new line of devices known as digital audio processors. These devices have been built around the philosophy of minimizing end-product design effort and cycle time while maximizing sound per pound.

These processors have efficient, built-in, configurable algorithms. This eliminates the need for user-written code while still allowing significant configurability. Design tools are provided to bridge the gap between the audio engineer — especially the loudspeaker designer — and a digital system.

The processors are optimized to specific audio-processing tasks. Therefore, the data paths are sized according to anticipated audio-processing needs.

The first generation of audio processors received IIS inputs of 16-, 18- or 20-bit audio data at 44.1 or 48 kHz. The stereo mixer uses 24-bit linear-gain words and allows gains between zero gain (a true mute condition) and +18 dB. The second-order IIR filter sections are a limit-cycle-free design based upon the canonical Direct Form I filter structure. Twenty-four-bit filter coefficients are used, and the accuracy of the 56 bits resulting from the 32 x 24-bit multiply is retained for critical filter computations.

This structure allows the user to download an extremely wide range of first- or second-order filters, such as bass shelves, treble shelves, high-pass, low-pass, parametric EQ, notch, Linkwitz-Riley, etc., for equalization or other filtering needs. The treble and bass controls, here, use second-order shelf filters and have a range of 18 dB. The user downloads an index associated with a desired gain, on half-dB increments, and the processor will automatically change to the new level without creating audible artifacts. The volume control is commanded using a 20-bit linear gain word and transitions to the new gain without audible artifacts. The dynamic range compressor has an adjustable threshold with enable/disable capability.

Next-generation audio processors will include a full 48-bit data path, 28-bit coefficients 76-bit accumulator for high-performance filtering even at low frequencies and high sample rates. They will accommodate sample rates between 8 and 192 kHz, with 20 filters per channel, for three channels — allowing precision digital crossovers along with equalization.