Consumer IC Advances -> MIPS, software make for smooth DVD decoding

MIPS, software make for smooth DVD decoding
By Alson Kemp, Senior Field Applications Engineer, Cirrus Logic, Fremont, Calif., EE Times
December 18, 2001 (1:12 p.m. EST)
URL: http://www.eetimes.com/story/OEG20011218S0049

Digital videodisk playback is a complicated process that requires decoding many different complex data types and coordinating the flow of time-sensitive data through a complex series of transformations. One of the major functions of a DVD player is the playback of compressed audio and video from the DVD disk. Using complex graphics, the menu display allows users to navigate the disk and select playback options and system configurations. Dramatic price decreases in VLSI silicon have created room for innovation for this rising consumer star, with cost-effective chips incorporating the large, complicated functional blocks required for DVD playback.

At the heart of a DVD player is the multifunction DVD chip. This block typically incorporates functions such as an MPEG-2 video decompressor, a DSP, a 32-bit microprocessor, I/O for audio, infrared for control and a built-in TV encoder or digital video output for a TV encoder. The processor may be a 32-bit RISC device with performance of approximately 100 Mips with separate instruction and data caches. Some microprocessors may have a memory-management unit, but some DVD chips have removed the MMU to reduce cost. Most microprocessors include integer multipliers and some, including the Cirrus/CS98100, include a multiply-accumulate unit to speed complex math operations.

The DVD loader includes the laser, pickup and servos. The loader also has a DVD controller to drive the servos and spindle, read data from the optics and present a suitable data interface to the DVD chip. Although there is no rule defining which loader interface a particular DVD player will use, market segments favor certain interfaces: Higher-end, multispeed or PC-focused loaders use an ATAPI/IDE interface; lower-end, single/double-speed or consumer-focused loaders use proprietary serial or 8-bit interfaces to reduce cost. DVDs have 2,048 byte sectors and like CDs, are encoded at a constant linear velocity, so that as the pickup's radius on t he disk is increased, the drive's spindle slows down.

Several I/O and memory buses are used. Typically, SDRAM is used as a data and program store for the processor, DSP and other functional blocks. Flash is used to store program and persistent data. Simple serial interfaces, such as the Sony/Philips digital interface, ports and UARTs, along with a port to provide infrared remote control functionality are included. Complex direct memory access hardware assists the CPU in moving blocks of data both around the chip and between external memory and devices.

Some DVD video packets are encrypted with the Content Scrambling System. DVD chips incorporate a CSS decryption block to offload decryption processing from the microprocessor. The video processor off-loads common graphics chores from the microprocessor. De-interlacing duties for progressive-scan systems, video scaling, frame-rate conversion, video mixing and zoom are all handled by the video processor. Subpictures, a custom DVD graphics format that is used for overlays, subtitles, menu highlighting and screen text, and on-screen display graphics are decoded by separate hardware modules and blended with the video processor output.

By using a combination of perceptual, predictive and entropy-encoding algorithms, MPEG compresses video by approximately 95 percent. Decompressing the resulting video is very compute-intensive, and so most DVD chips incorporate hardware MPEG video decoding. Although the decoder must only support a subset of MPEG-2's Main Profile @ Main Level (specifically geared to television resolutions and frequencies), most MPEG video decoders support many more resolutions and frequencies than the MPEG-2 standard specifies.

Most MPEG video decoders, for example, will decode MPEG video from video CDs and supervideo encoders and convert CCIR-601 or -656 digital video from the DVD chip's video output into analog signals for a television. The digital video data is presented to the encoder as 8-bit data, but many video encoders oversample and interpolate the data to 10 bit for improved video quality before converting it into composite, S-video or component video signals. The encoder must support Macrovision copy protection for DVD compliance. Progressive-scan video outputs are not yet common among mid- to low-end DVD players, but some DVD chips, such as the CS98100, support progressive-scan video capabilities.

All DVD players must have analog audio outputs, either stereo or 5.1 channel digital audio outputs, S/PDIF and standard optical TOSLINK . Some multichannel audio D/As are specifically tailored for the consumer DVD market and others are tailored for the high-end DVD player markets. The digital audio outputs are encoded using IEC 60958 for uncompressed pulse-code-modulated audio or IEC 61937 for AC3, Digital Theater Sound (DTS) or otherwise compressed audio.

An embedded real-time operating system runs on the microprocessor to provide basic functionality for a range of functions including booting, memory management, I/O, task cont rol, messaging, semaphores, timers, peripheral drivers and communications stacks. Support for the DVD and CD file systems, generally UDF and ISO9660, respectively, must also be provided by the operating system. A real-time operating system is necessary because of the time constraints imposed by the need to quickly respond to requests from the various functional blocks.

The microprocessor also handles demultiplexing audio, video and subpicture data. Each DVD packet/sector can contain a variety of data types that must be pulled apart and passed to specific decoders. Each sector can also carry a time stamp for synching audio and video. Audio and video decompression are handled by separate decoders, but decompressed streams must be presented to the user according to the data's time stamp.

Compressed audio on a DVD is stored as MPEG-2, Dolby Digital (AC-3) or DTS. Some DVD players can decode other audio formats such as MPEG-1 Layer 3, WMA and AAC. All of those encoding methods are perceptual coders and g ive compression ratios of around 5 to 10:1. The coders use knowledge of human auditory perception to remove sounds below the ear's threshold (a very soft 80-Hz tone), frequency-masked sounds (a 1-kHz tone might mask a 1.1-kHz tone that is 40 dB down), channel-masked sounds (a loud sound from the center channel can mask a soft sound in another channel), and temporally masked sounds (a quiet sound that follows a loud sound). Also, unused bandwidth from one channel can be allocated to a busier audio channel. The resulting bit stream is then entropy-encoded to further reduce the bandwidth.

Because the software is specific to the chip and to a single application (the DVD player), the DVD chip vendor will often develop much of the required software and license it to the DVD system maker. The system maker can then tailor and brand the software to meet specific player needs. The DSP software is also responsible for post-processing of the audio. This post-processing can include effects such as pitch shifting for karaoke, the virtualization of stereo to multichannel, loudness, equalization and bass management.

Today's large DVD manufacturers use proprietary virtualization algorithms or preferred equalization algorithms, so the DSP must be programmable and the DVD chip vendor must either provide tools for OEMs to develop DSP software or in-house programmers to adapt software to the OEM's needs.

See related chart