New technologies are making it possible to build all-digital audio systems with excellent performance, and many benefits that promise to change the face of the industry. Rusty Allred and Mike Tsecouras report.

Although audio systems have been gradually going digital for some time, fully digital audio playback systems, such as the one shown in Fig. 1, have not been available in the consumer market until quite recently.
In 1998 TacT Audio developed its Millennium product, the world's first fully digital amplifier, proving that a digital amplifier could achieve high sound quality. David Smith, vice-president of Engineering for Sony Music had this to say of the Millennium: "The next era has arrived." The Millennium uses digital amplifier technology from Toccata Technology (now Texas Instruments - Copenhagen).
Digital amplifiers (roughly the bottom path in Fig. 1) allow significant miniaturisation of form factor since their high power efficiency results in less thermal power dissipation, allowing the use of power supplies that are about one-third the size, and heatsinks that are about one-tenth the size of those used in equivalently-rated analog power amplifiers. The result is lighter weight products with minimal supplemental cooling requirements as compared to the cooling requirements of the traditional analog class A/B amplifiers often found in racked systems.
Although many varieties exist, a typical digital amplifier operates in the following manner:
First, the PCM (pulse code modulation) signal from the digital audio processor section (top half of Fig.1) is interpolated to increase the sample rate. This allows the amplifier to perform the PCM-to-PWM (pulse width modulation) conversion without signal degradation.
Next, feedforward correction is applied at the increased sample rate. This step is crucial for excellent measured performance and sound quality. After the correction is applied, a noise shaper is used to push the quantisation noise error, which occurs in the PCM-to-PWM conversion, outside the audio band.
The output of the noise shaper is now mapped to a PWM signal, which drives the H-Bridge. The H-Bridge is a combination of power MOSFETs used in switching mode to provide the current needed to drive the speaker load.
The PCM-to-PWM conversion is depicted graphically in Fig.2. Notice that the PCM depiction of a sine wave, in the upper diagram, is a uniformly time sampled waveform with amplitudes corresponding to the amplitudes of a continuous sine wave at the various time points. When this is mapped to PWM, the signal amplitude remains constant while the widths of the various pulses change to represent the amplitude information of the sine, as seen in the lower diagram in Fig.2.
Since a PCM to PWM mapping is a non-linear transform, distortion will naturally occur. However, this undesirable effect is mitigated by the feed-forward correction block.
The final stage is a simple network of low-cost, passive components (two inductors and two capacitors) that combine to demodulate the PWM signal and present a high-quality audio signal to the speaker.
Since the release of the Millennium there has been a flurry of activity surrounding deployment of this new technology at different performance levels for various market needs. Higher-performing systems implement what is known as the Equibit grade of this digital amplifier technology. Figs. 3 to 5show the performance of this technology: The THD+N vs. frequency curve is flat, at around 0.015per cent, and this performance is maintained nearly up to the peak output power. In addition, the noise floor has been measured to be in the -140 to -150dB range.
For the consumer market, cost competitive chip sets implemented with consumer-grade components still show very good performance as seen in Fig. 6 to 7. These chip sets have the same flat THD+N curves, with values right at 0.08per cent for 30W, 0.02per cent for 10 W, and 0.035 per cent for 1 W operation. Here a noise floor in the -125dB range is more common.
Recently, systems implementing these new chips have come into the market. Consumer products that use this technology include the Panasonic SA-XR10A/V receiver and the Panasonic SC-DA100 and 300home theatre in a box systems .

As has been true in other end equipments, digitalisation offers many advantages to audio systems. Among these are simpler design and verification, greater flexibility, greater reliability, elimination of costly low tolerance parts, improved testability, simplified production, lower cost per feature, ability to rapidly prototype, reduced time to market, and performance stability with temperature fluctuation and other environmental effects. In addition, the configurability and flexibility of digital systems offer advantages such as the ability to use the same hardware in multiple systems and the ability to reconfigure systems according to application, even in the field and on the fly, according to source material selection or user preferences.
In the past the disadvantages for digital audio systems included limit cycles, audible quantisation noise, pops, clicks and zipper noise. But modern processing techniques have eliminated these problems.
The front-end audio processing, shown in the top half of Fig. 1 was originally done in analog. But that has changed due to the advantages of digital processing, the fact that the disadvantages have been eliminated, and now especially due to the advent of a true digital amplifier. Now it is desirable to also do this processing in the digital domain, preserving a digital path all the way from the digital source to the speaker terminals.
In response to that desire, the audio industry has seen a wide variety of processing products offered. Fully programmable digital signal processors (DSPs) have been mainstays and will continue to be due to their absolute flexibility and their ability to meet the changing needs of audio standards.
In addition to DSPs, second-generation digital audio processors are now available. While less flexible than fully-programmable DSPs, they are attractively priced and specifically designed to excel at audio processing tasks. Many typical audio functions are built in. Examples include second-order ±18 dB bass and treble controls with selectable corner frequencies; volume controls operable between +24dB and true mute; dynamic range control with two thresholds and a full range of compression and expansion in each of the three resulting regions; volume-tracking loudness compensation; spectrum analyser/VU meter; delay for sound effects, lip synchronisation, and speaker time alignment; scalable dither with selectable probability density function; and numerous, cascaded filters for equalisation and sound effects.
Philosophically speaking, the goals of the second-generation DAP devices are:
* Efficient, configurable, built-in algorithms.
* Elimination of the need for user-written code while still allowing significant configurability.
* Availability of design tools to bridge the gap between the audio engineer, especially the loudspeaker designer, and a digital system.
* Optimisation to specific audio processing tasks.
* Datapaths sized according to anticipated audio processing needs.
* New architecture built from the ground up solely for audio processing.
These high precision (48-bit single precision, 76-bit accumulator) devices handle low-frequency-high-Q filtering even at high sample rates. Furthermore, all of the built-in functions are configurable by downloading coefficients. In addition, while the processors are technically fixed-function devices, not only can the functions themselves be configured, but also the connecting paths. This flexibility easily allows implementation of yet-to-be-invented audio functions.
Currently known-implementable functions include: mixing/muxing, a wide variety of third-party virtual 3D algorithms, centre or subwoofer channel synthesis and sound effects. Since these functions are configurable, the system designer has complete control over the features and the flexibility to voice systems as desired.
To eliminate the need for user-developed software, configuration is achieved through coefficients that are simple to generate and download using manufacturer-provided software tools. This allows rapid prototyping, reduced inventory through use of the same hardware in multiple end products, reduced end product development time, and reduced scrap since the hardware is coefficient-reconfigurable to accommodate upgrades and changes.
The tools include the functionality to design and download a wide variety of filters for use in equalisation, sound effects, loudness compensation, and so forth. In addition, a speaker response can be read into the tool and automatically equalised to the user's specification. Or, a desired equalisation curve can be read in and matched with the available resources. In addition, the tools will plot expected loudness compensation characteristics, and include everything necessary to graphically set up the dynamic range compression and expansion function.
Taken together, these new technologies provide all the tools needed to configure a fully digital audio playback system. That in itself is interesting, but perhaps the most exciting aspect of these new technologies is their promise for the future. These advances put high-quality, feature-rich, small-scale audio technology within the economic grasp of mainstream audio end equipments. The future is bright with possibilities as equipment designers exercise their creativity to bring us audio equipment of which we have yet to dream.

ENQUIRY No 58

Rusty Allred and Mike Tsecouras are with Digital Audio Solutions, Texas Instruments Incorporated, Freising, Germany www.ti.com