Inside DSP on Audio: Inside a Modern Digital Audio Product

Submitted by BDTI on Mon, 01/05/2004 - 19:00

The last decade has seen consumer audio products from home theater systems to car stereos and portable players go digital. These complicated devices play back compressed audio formats, compensate for room acoustics, and add effects such as reverberation, equalization, and dynamic bass, thanks to the power of digital signal processing.

How do manufacturers pack the DSP punch these applications require into small, affordable, and power-efficient systems? A portable audio player, for example, contains a processor, volatile and non-volatile memory, analog and power management components, and more. These components need to talk to each other and to the outside world. And there is more to these devices than hardware: complex signal processing software, user interface software, and device drivers are also needed.

This article charts the anatomy of a typical consumer digital audio product, focusing on portable players and network-enabled devices. Closely related to this topic, the "Under the Hood" column in the January 5 issue of EE Times provides a tear-down of one handheld audio product, the Apple iPod.

Processors
At the heart of today's digital audio product is a programmable processor or processor core. Roughly speaking, processing requirements vary from roughly 20 MIPS for MP3 decoding to well over 100 MIPS for products that feature sophisticated audio encoders or combine high-end multi-channel decoders with numerous other audio processing functions. These computational loads are well within reach of many DSPs and general-purpose processors, giving designers a wide range of options to choose from. The article, "Digital Audio Technology Guide," discusses considerations and tradeoffs in selecting a processor for a digital audio product.

Some products use a DSP processor, or incorporate multiple processors; for example, a microcontroller for connectivity, storage management, and user interface functions, and a DSP for audio processing. The latter approach is particularly attractive for convergence products such as audio-enabled digital cameras where a microcontroller or general-purpose processor is already present, but does not provide sufficient performance for implementing the necessary audio functions. Network-enabled A/V receivers, PVRs, home media servers and set-top boxes are likely to utilize a general-purpose processor for all networking functions, storage management, user interface, and audio processing functions.

Often at least one processor core is hidden inside an ASSP (Application-Specific Standard Product chip) such as the PortalPlayer chip used in the Apple iPod (See the "Under the Hood" column in EE Times for details). ASSPs combine processor cores with appropriate peripherals and sufficient on-chip memory for implementing a particular consumer audio product. The cost, size, and power consumption advantages of integrating many functions on one chip make ASSPs particularly attractive in portable products, although ASSPs also make their way into many other applications. Figure 1 compares key components of Texas Instrument's TMS320DA250 DSP chip and SigmaTel's STMP3560 ASSP, both targeting portable audio applications.

figure 1

While ASSPs targeting decoding of compressed audio are common, some system developers prefer to design their own custom ASICs—if they can tolerate the long and costly development cycles involved. (See "Digital Audio Technology Guide" for an in-depth discussion of processor selection.)

Memory
ASSPs targeting digital audio products typically include sufficient on-chip memory for audio decoding and additional processing such as equalization and dynamic bass processing. Many packaged DSP processors also offer sufficient on-chip memory for these functions. Avoiding the use of off-chip memory lowers the system component count, reducing cost and simplifying hardware design. In general, avoiding the use of off-chip memory also lowers power consumption by reducing the activity on off-chip buses. However, in hard-disk based portable players, a large off-chip memory buffer can actually reduce power consumption. Compressed audio data is read into memory from the hard drive in quick bursts, allowing the hard drive to be powered down for long periods while the contents of the buffer are decoded and played. This power saving technique is illustrated in the design of the Apple iPod. Off-chip SDRAM is typically used for this purpose, invariably leading the designer to select a processor or ASSP with an integrated SDRAM controller.

Internal connections
Serial buses play an important role in consumer audio product designs: synchronous serial I2S buses are typically used to connect digital-to-analog (DAC) and analog-to-digital (ADC) converters to the chip performing the audio processing. Serial I2C buses are often used to send volume control, equalization control, and other messages from a microcontroller handling user interface functions to the chip handling audio processing. A microcontroller may also use a dedicated serial connection to send compressed audio read from a hard disk drive, compact flash card, or a network port to a DSP or ASSP for decoding. Serial ports typically have ample bandwidth for these purposes, and because they require fewer signals to connect chips together, serial ports can simplify hardware design and circuit board layout compared to parallel connections. 

Audio inputs and outputs

Some DSPs and ASSPs targeting digital audio applications do not include on-chip DACs or ADCs, requiring separate converter chips—but allowing system designers to select converters to suit their needs and budgets. Off-chip converters may or may not include additional analog circuitry for functions such as volume control, headphone amplifier, and microphone preamplifier. In contrast, some vendors, such as SigmaTel, strive to pack as much functionality as possible onto a single chip, and offer ASSPs that integrate line- and mic-level analog audio inputs, DACs and ADCs, and a headphone amplifier along with the DSP and other peripherals. Digital and analog subsystems often require separate power regulation and separate analog and digital power and ground circuit board traces in order to prevent pesky digital switching noise from making its way into analog signals via the power supply.

Some products (most notably high-end home stereo and home theater equipment) offer digital audio input and/or output in S/PDIF format, in addition to analog input and output options. This feature is directly supported by some ASSPs, and can also be implemented by connecting an S/PDIF receiver and/or transmitter chip to the same serial I2S port used to connect DACs or ADCs.

Products that include power amplifiers—such as DVD receivers and automotive audio systems—may substitute digital amplifiers for the more traditional combination of separate DACs and analog power amplifiers. See "Ear to the Ground" for a brief discussion of the advantages and disadvantages of digital amplifiers.

Mass storage
Another key component in many consumer audio products is a mass storage device for storing the user's compressed audio files. Hand-held players use a variety of flash memory card formats, including CompactFlash, MultiMediaCard, SmartMedia, and Secure Digital cards. Flash memories provide limited capacity, however: capacities greater than one gigabyte per flash memory card are still rare, while hard disk drives used in portable products hold as much as 60 gigabytes of data. Flash-based portable players are therefore gradually giving way to hard disk-based products like the Apple iPod. Portable players generally utilize ASSPs that include an on-chip flash memory interface and/or an IDE interface for connecting a hard disk drive. MP3-capable CD players are often powered by an ASSP that integrates MP3 decoding with the optical decoding required for CD playback.

Connecting to the world
Portable players typically connect to a PC via a USB port—often integrated into an ASSP, but occasionally requiring an additional chip. USB version 1.1 is more commonly supported in ASSPs than the significantly faster version 2.0. Firewire (IEEE 1394) interfaces are less common, and typically require an additional chip, unlike USB ports.

Network-enabled products typically feature an Ethernet port for connection to a home network or to the Internet via a home broadband link. These products are not as mature as portable players and as a result, Ethernet ports frequently require separate chips. Network-enabled products may also include a dial-up modem for connection to the Internet or to product-specific data services, requiring an additional DSP or ASSP to implement modem functions.

Display
Portable audio products and many other audio products such as Internet radios require an LCD display and associated controller. While many portable products rely on a display controller provided in an audio decoding ASSP, other products such as multimedia jukeboxes or audio-enabled PDAs may utilize a discrete display controller chip.

Power supply
Battery-powered products often require DC-DC converters to provide the various supply voltages needed by different chips or subsystems, and to maintain appropriate operating condition as battery charge decays. DC-DC converters are often built into ASSPs, simplifying hardware design. Products with rechargeable batteries may also require a separate battery charger controller chip.

Software modules
Typically, much of the software that powers a digital audio product is provided by the processor vendor. DSP processor vendors such as Analog Devices provide optimized software modules for MP3 decoding, WMA decoding, AAC decoding, equalization, dynamic bass, and so on. Products capable of recording audio may include an MP3 or other high-quality audio encoder, an ADPCM encoder for speech compression, or both. ASSPs typically include these software modules in on-chip ROM, which helps reduce cost.

Often, sample rate conversion software is also required. For example, if audio processing modules such as equalization or reverberation cannot operate at all of the sampling rates supported by the audio decoding module, or if the DACs do not support these sampling rates, then rate conversion will be needed. 

Operating systems and drivers
Network-enabled products, multimedia jukeboxes, and similarly sophisticated products often use a multitasking operating system such as a Linux variant. An operating system can significantly ease the work of scheduling software tasks with different rates of operation. For example, a network de-packetizer task that is invoked as Ethernet packets arrive, an audio decoder task that operates at the audio frame rate, and a user interface task that operates as time allows can all be scheduled automatically by an operating system. Several vendors offer variants of Linux that are tailored to the needs of embedded systems.

A simple portable MP3 player, on the other hand, may get by without an operating system. In many audio products, the audio frame rate provides the system's only hard real-time constraints and determines the operation rates of all tasks except for the user interface. In such products, a multitasking operating system is often unnecessary. In this case, developers may choose to use a rudimentary statically-scheduled or interrupt-driven control flow. The advantages to this approach include reduced processing and memory requirements compared to a multitasking operating system, and no OS licensing fees.

Figure 2 provides a simplified illustration of software scheduling in a simple portable player product with no operating system and in a network-enabled product using a multitasking operating system.

Figure 2

Products that feature USB or Firewire ports require appropriate device drivers and protocol stacks. Similarly, products that provide an Ethernet connection require network protocol stack software. These drivers and stack are often—but not always—available from the processor vendor or a third-party supplier.

Software development
For many products, much of the software development effort is focused on implementing a convenient user interface that is easy to navigate, and integrating that user interface with off-the-shelf audio decoding and processing modules, the operating system, and device drivers. Major software components, including signal processing modules and the operating system, are usually acquired from the processor vendor or third party developers, but the integration effort can still amount to a sizeable project.

Some chip vendors provide a complete "turn-key" software package, saving product designers valuable time and money by solving most of the integration challenges in advance. Even when significant customization of the chip vendor's software is required, a "turn-key" solution is usually a convenient and cost-effective starting point. The software development effort increases considerably if new signal processing modules, such as decoders for emerging compressed audio formats, need to be created. (See "Building Good Audio Software" for more on audio software development.)

Trends to Watch
As storage capacities of hand-held devices continue to increase and the portable player market shifts from flash-based products to hard disk-based devices, new and better user interfaces will become increasingly important. The user's ability to quickly and intuitively locate and play songs from a vast collection can easily make the difference between a cherished source of entertainment and a forgotten gadget collecting dust on a shelf. To accommodate state-of-the-art user interfaces, LCD displays are growing larger (and more power hungry) with each product generation.

Increasing integration is a clear trend in modern digital audio product design. Flash, SDRAM, IDE, and USB interfaces, display controllers, and memory are commonly integrated on the same chip with DSP and/or microcontroller cores, providing most of the major digital components of a hand-held digital audio player in a single device. The chip vendor frequently also supplies most of the associated software. This simplifies product design and frees digital audio product developers to add sophistication and improve functionality. For example, additional chips can be added to support new cutting- edge features, and software development effort can be directed toward developing better, more intuitive user interfaces.

The line between ASSPs and generic processors is rapidly blurring as processor vendors such as Texas Instruments strive to provide the right combination of processing resources, memory, peripherals, and software for digital audio products in chips based on their existing DSP processors, and ASSP vendors strive to provide increased programmability and flexibility by utilizing common processor cores such as ARM's ARM7.

With the increasing consolidation of major hardware functions into a single chip and the consolidation of software functions into "turn-key" solutions provided by the chip vendor, brand-name product manufacturers will increasingly be able to purchase entire system designs and even fully manufactured systems. With this business model, "original design manufacturers" (ODMs) can amortize design and manufacturing start-up costs over several competing but essentially identical products that are packaged and distributed by different companies with brand-name appeal and established distribution channels. This option is particularly attractive for inexpensive products where cost is more important than cutting-edge features.

Brand-name manufacturers that opt to design their own digital audio products will continue to provide increasingly sophisticated products, enabled by a growing selection of feature-laden chips. In addition to better user interfaces and higher storage capacities, many new products will incorporate video playback or combine digital audio with largely unrelated functions in convergence products such as audio-enabled digital cameras, PDAs, and cell phones.

Users are already able to transport and access vast music collections using a variety of products. As storage capacities, connectivity options, and features improve and multiply, new digital audio products will continue to revolutionize the way that people share and enjoy music.

Add new comment

Log in to post comments