From professional studios to car radios to home entertainment systems, the demand for high-quality audio is on the rise. The increasing performance of low-cost programmable DSPs is helping makers of these systems meet customer demand.
In fact, digital techniques have reinvigorated audio processing, bringing professional-quality surroundsound and 3D effects to consumer products. They're spearheading the convergence of PC and consumer audio products in applications like Internet audio by using MP3-like compression techniques. Moreover, the addition of digital amplifiers creates complete digital solutions, starting from the source and going all the way to the speakers.
To keep the external component count low, developers are bundling optimized peripherals and functions, including interfaces, around high-performance DSP cores. Highly integrated system-level audio processors are the result. Also, advances in process technology permit designers to pack Class D digital amplifiers on the same audio processor chip. Unlike their analog counterparts, digital amplifiers eliminate radiation and dramatically cut power dissipation. Placing them inside speakers creates a new class of audio transducers called digital speakers.
Embarking on that all-digital audio paradigm, Texas Instruments (TI) has bundled a fully digital audio amplifier with its TMS320C54x DSP, and a portfolio of algorithms to offer a complete hardware/software solution that includes digital audio decoding at the speakers (Fig. 1). Some standard formats handled by the DSPs include Dolby Digital, 3D Surround Sound, DTS, MP3, and Advanced Audio Coding (AAC).
There's enough horsepower available from the DSP chip to perform other tasks, such as filtering and high-precision parametric equalization, as well as artifact-free volume, bass, and treble control. So, a host system can send digital data directly to the speakers from a CD, DVD, TV, game console, or the Internet.
"By having a digital connection, designers avoid the power and signal losses associated with transmitting analog signals in audio systems," notes Chris Schairbaum, worldwide marketing manager for TI's Internet audio products. "Also, the system noise and other faults inherent to analog transmission are eliminated to greatly improve sound quality. Plus, integrating codecs with audio processors on the same silicon simplifies design and cuts board space requirements."
This all-digital audio chain is making the speaker enclosure an attractive place for amplifying the sound and for achieving a balanced mix of signal-to-noise ratio (SNR), dynamic range, and frequency response at a reasonable cost. According to TI, audio designers can now enclose a complete system inside a speaker with higher than 90% efficiency and 95-dB SNR while eliminating bulky and expensive heatsinks.
With the ability to sample at rates as high as 192 kHz, and a dynamic range of 110 dB, TI's TAS5015 digital audio amplifier can now support sound reproduction that's compatible with DVD-audio standards. Additionally, the amplifier dissipates about one-tenth the power of conventional Class A and Class AB amplifiers.
Combining audio-friendly peripherals around its cores, TI continues to offer power-efficient, application-specific DSPs for a variety of consumer and PC applications. The company is looking into bringing the digital audio-power amplifier on board too.
"End-to-end digital solution" also is the buzz term at Cirrus Logic, a major contender in the audio arena. Em-ploying DSP-controlled digital-switch techniques, the company has developed a fully digital Class D amplifier, dubbed TrueDigital. To offer a platform ap-proach to building audio systems, Cirrus has combined its True-Digital power amplifiers with DSP-based audio processors and created a total entertainment (total-e) solution.
Moreover, the company has integrated this technology with a DSP/RISC core to create a highly integrated Maverick processor, the EP7409. The processor is aimed at Internet audio players and many other portable digital audio applications. Some key functions combined on this processor include a TrueDigital Class D headphone amplifier, the AMR7TDMI CPU, a 24-bit audio DSP, MavericKey security technology, 104 kbytes of SRAM, and 60 kbytes of ROM (Fig. 2).
"This integrated Internet audio processor approach provides cost savings due to a lower bill of materials (BOM), and substantially reduced power consumption," says John Marc Woosley, business development manager for Cirrus. The Maverick processor dramatically reduces the number of external parts required to build a player, thereby cutting the system cost, he asserts. It eliminates the need for external NOR flash to program and store data and requires a low-cost 32-kHz crystal clock. An internal PLL generates all clock frequencies needed by the system. The EP7409 is designed to bring the cost of flash-based Internet players below $100 (retail price).
"Three things contribute to the overall reduction in system power consumption," notes Woosley. These are a 90% efficient on-chip TrueDigital Class D headphone amplifier, low-voltage operation, and fewer clock cycles to perform MP3 tasks. While the employed 0.25-µm CMOS process lowers the supply voltage, the dual-MAC-based 24-bit DSP architecture uses an 18-MHz clock to perform MP3 tasks.
Compared to first-generation products, power requirements were slashed in half to 170 mW in the latest third-generation players, claims Cirrus. That translates to double the run time in portable Internet audio players.
For those requiring a separate Class D amplifier, Cirrus offers these parts in standalone packages, with output power levels from 5 to 50 W, and a dynamic range of up to 100 dB. For a complete audio amplifier solution, the company recommends International Rectifier's application-specific power-chip set at the output of its power amplifier.
At the heart of Cirrus' digital pulse-width-modulation (PWM) amplifier is a DSP that controls a switched power output stage, monitors the amplifier's load, and handles audio-processing tasks such as equalization gain control and crossover separation. It doesn't require any filtering at the output, and it eliminates EMI/RFI.
According to Cirrus, its proprietary Class D approach overcomes the limitations of conventional PWM methods. It delivers higher SNR, lower total harmonic distortion (THD), wider dynamic range, greater efficiency, and better frequency response. This approach has made digital Class D amplifiers a commercial reality.
TI and Cirrus Logic aren't the only players in the market with this capability. Other companies, like Franco-Italian supplier STMicroelectronics, also back this all-digital audio paradigm.
Using its leading-edge smart power bipolar-CMOS-DMOS (BCD) technology, STMicroelectronics is exploring an integrated audio processor. But the lack of flexibility in an audio solution has pushed that level of integration further out in time for STMicroelectronics, according to Bob Samson, marketing manager for audio products. He adds, however, that the company will continue providing DSP-derived audio processors and Class D amplifiers as separate solutions for now.
Like TI and Cirrus, STMicroelectronics has taken the system-level approach to solving audio problems. Uniting its direct-digital amplifiers with DSP-enabled audio processors, the company has crafted all-digital audio solutions for car radios, Internet audio players, and various consumer products. Meanwhile, STMicroelectronics continues to enhance the performance of its audio wares by integrating multiple DSP cores, codecs, more memory, interfaces, and other application-specific peripherals to handle more channels from a single device, and to directly drive the digital power amplifiers in these channels.
Present digital Class D amplifiers need power MOSFETs at their outputs. By applying its proprietary BCD2 technology and expertise in Class D circuits, designers at STMicroelectronics have simplified that task, integrating such power MOSFETs and digital amplifiers on the same die, and creating a direct-digital amplification (DDX) solution for audio applications.
The DDX capability was acquired in an exclusive licensing deal with Apogee Technology, the original developer of the damped ternary DDX technique. Hence, Apogee also offers this patented solution separately. In fact, it's readying a 24-bit version that can handle 96-kHz and higher sampling rates. Presently, Apogee's DDX amplifier is designed for 20-bit data and 44.1- or 48-kHz sampling frequencies.
"It eliminates the DAC and analog processing, while increasing the efficiency by over a factor of three compared to conventional Class AB designs," Samson notes. Based on a unique, three-stage or damped ternary modulation, the DDX amplifier produces less carrier energy than conventional switching amplifiers that use two-state or binary modulation, Samson explains. The result is higher overall efficiency than binary designs. At 10% of the output power, a DDX-based digital amplifier outperforms conventional Class D and Class AB amplifiers. By comparison, the DDX amplifier is shown to offer over 85% efficiency at 10% output power, whereas others fall steeply (Fig. 3).
In addition to higher overall efficiency, the DDX technique cuts unwanted EMI. All switching amplifiers produce unwanted carrier-band energy at and above their switching frequencies. Therefore, reducing this interference is a key design issue with digital amplifiers.
According to STMicroelectronics, the damped ternary modulation scheme produces much lower carrier energy, approximately 16 dB lower than conventional methods. Other features include higher SNR, insensitivity to supply variation, and scalability.
DDX is said to be scalable in terms of both signal processing and output power stages. DDX processing can be upward-scaled to 24 bits for high-end audio applications, or reduced to 8 bits for voice or lower-performance audio needs. Moreover, the same DDX controller can drive power stages that provide from a few milliwatts to over 120 W of full-bandwidth audio power. "Eventually digital amplification will become a dominant mode," Samson concludes.
While major suppliers of DSP-enabled audio processors prefer digital Class D amplifiers at the output of their solutions, SigmaTel continues to tap the benefits of analog Class AB amplifiers. That support is evident in SigmaTel's STMP3400 monolithic audio decoder with an on-chip headphone Class AB amplifier. This chip competes head-on with Cirrus Logic's EP7409. SigmaTel agrees that Class AB's efficiency is much lower, but the BOM also costs less, resulting in cost and board space savings, argues Debby Clarke, SigmaTel's product marketing manager.
This issue has been addressed by ensuring that the overall consumption of the STMP3400 is lower and comparable to competing decoders. The STMP3400 is rated for 150-mW overall power consumption, which is about 20 mW less than competing devices. SigmaTel estimates the total BOM cost at $20.00, without the external flash memory.
The improved power consumption comes from the integration of a dc-dc converter on the same die. On-chip registers control the converter, ensuring that the Class AB amplifier is on only when necessary. Other on-chip functions include a 56004-compatible DSP core, a USB interface, an LCD interface, a memory interface that supports NAND flash, and a codec (Fig. 4).
"With this compact and inexpensive solution, SigmaTel has the potential to expand beyond MP3 audio players into other systems like cell phones, PDAs, and digital cameras," says Will Strauss, president of Forward Concepts, a market research firm. The flexibility and programmability of an on-board DSP allows the STMP3400 to handle other decoding formats, like Microsoft's Windows Media Access (WMA) and Dolby's AAC.
Meanwhile, Motorola has taken a new direction with its audio solutions. Until now, the supplier had focused on solutions based on its 24-bit DSP architecture, the DSP56300. Combining this powerful DSP core with specialized peripherals, software, and development tools, Motorola offered its Symphony audio processors for a myriad of audio applications. Now, it's complementing this line with a new family based on a 32-bit microprocessor architecture, called ColdFire. By running both types of code efficiently, ColdFire enables the replacement of two separate processors by a single CPU, Motorla contends.
Aiming for higher levels of integration and lower system power, Motorola is readying an MP3 decoder using the ColdFire architecture. Incorporating an enhanced multiply-accumulate (MAC) unit, custom audio peripherals, and system software, this new audio decoder uses just 36 kbytes of memory and a 19-MHz clock. The decoder can run from on-chip memory and requires fewer cycles than typical decoders to provide longer battery life for consumer products, says Flip Lockhoof, marketing manager for Motorola's audio and gaming products.
Because external memory and other interface devices aren't required, the ColdFire-based decoder also cuts system costs. "We could have chosen to implement it on our 56300 Symphony DSP family. But we selected the ColdFire microprocessor as the optimum architecture, due to a large amount of control code that must be dealt with," says Ken Obuszewski, Motorola's audio solutions operation manager.
In short, it's a popular embedded controller that's easier to code for audio algorithms and supports high-level languages like C and C++, Lockhoof says. It's also being targeted for applications that require significant control processing for file management, data buffering, and user interface, along with signal processing, he adds. However, there are a number of applications where DSPs offer a better solution. And, Motorola will continue to serve those systems with 24-bit DSP cores in its arsenal. Even so, integrating a Class D amplifier on an audio processor chip is not yet on the supplier's drawing board.
The first ColdFire audio derivative should be unveiled later this year, with general sampling beginning in the first quarter of 2002. In addition, Motorola is readying an MP3 encoder employing the ColdFire microprocessor.
While most suppliers have moved from 16-bit resolution and a 44.1-kHz sampling frequency to 24 bits and 96 kHz or higher, Analog Devices (ADI) continues to field 32-bit DSP solutions. With the advent of DVD audio, 20- or 24-bit resolution and 96- or 192-kHz sampling rates are becoming the norm. Advances in data converters are helping designers to go for 20- and 24-bit ADCs and DACs.
"Digital processing of these higher-resolution audio signals is requiring the use of 32-bit processing to ensure that quantization noise artifacts won't exceed the 20- or 24-bit input signal," explains Colin Duggan, product manager for ADI's digital audio group. "Plus, the 32-bit DSP provides sufficient headroom to achieve the desired dynamic range, and it doesn't limit the SNR of data converters. In addition, it uses less MIPS to do the decoding job."
Therefore, it's not surprising that ADI is pushing the 32-bit single instruction, multiple-data (SIMD) SHARC DSP into this space. Also, according to ADI, SHARC prices have dropped in the last few years, making this DSP architecture attractive for low-cost consumer applications.
SHARC was first introduced into professional audio equipment and has now gotten into high-end audio due to price cuts. ADI is planning further reductions in cost to make it attractive for price-sensitive consumer applications. The company's goal is to bring the price of 32-bit floating-point SHARC DSPs to under $5.00 within three years.
All of this activity in audio chips comes against a background of optimistic market forecasts by audio chip manufacturers. Despite this year's decline in PC sales and associated audio chips, they see the market for audio chips as "booming." A recent report by Forward Concepts shows that the total market for all varieties of audio chips reached nearly $1.7 billion last year, and it's projected to go over $4.9 billion by 2005. This represents an annual compound growth rate of about 20%.
|Companies Mentioned In This Report|
Analog Devices Inc.
Apogee Technology Inc.
Texas Instruments Inc.
(800) 477-8924, ext. 4500