file:///W|/A436_summary.html Summary A436Parallel Video Chip

The Datasheet Archive - 100 Million Datasheets from 7500 Manufacturers.

file:///W|/A436_summary.html Summary A436Parallel Video Chip

Datasheet Thumbnail

Top Searches for this datasheet

A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
Summary A436Parallel Video Chip
A436 Video Digital Signal Processor Chip fully C-programmable, universal, real-time image/video compressor/decompressor communicator
This device under development specifications subject change without notice.
Rev. June 2001
Click navigation left.
download.
Contents
Features Applications Enhancements Image/Video Compression/Streaming Decoding Description Benchmarks Block Diagram Related Documents Contact Information
Features
Overview Highly optimized efficient, general purpose, very high performance, 512b advanced imaging parallel RISC processor MMU) single chip single instruction stream Performance 50,000 RISC MIPS motion estimation billion MACS with only clock Achieves very high performance with moderate clock rate main-stream fabrication Fully software-programmable, parallel image processor optimized real-time image/video processing/compression RTOS Much faster, more efficient easier understand, optimize than other fast DSPs Directly software programmable universal compressor (encoder)/decompressor (decoder) multi-format, standards-based proprietary image/video compression streaming decompression
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
Provides fully software programmable video compression real-time Provides powerful, flexible, fully software programmable, autonomous "smart" cameras that "watch" images themselves people don't have providing "scene content analysis" Enhanced version (fourth generation Ax36 core) proven A236 Video Chip Scaled-down reduced pin-count versions built demand satisfy less demanding applications requiring even-lower price points Easy program microcode! Full software development environment includes compiler, assembler, linker, loader, simulator debugger Develop code using parallel-enhanced ANSI-standard compiler with assembly language output Simple parallel programming model supports parallel operations structures using embedded parallel data types what intended provide tight control code generated control need embed assembly language programs rely precompiled subroutine libraries Evaluation board evaluation "smart" camera board integrated with software development environment Supports modular development, licensing protection code third-parties Open architecture with simple pipeline efficient code written easily Simple powerful instruction provides quick interrupt response time Tools automatically pack scalar instructions into word when parallel processing needed, further reducing program size Three internal controllers automatically build circular, multi-frame image/video buffers with programmable sizes memory, providing standardized format video capture, processing display uCLinux RTOS with TCP/IP UDP/IP Internet connectivity, file system device drivers (video input/output, IDE, USB, ethernet, PCMCIA) Enhanced SIMD processor architecture live imaging 32-bit Scalar Arithmetic Unit tightly coupled Parallel Processing Unit with parallel arithmetic units, motion estimation coprocessors Highly efficient memory architecture, Data Cache architecture page replacement algorithm optimized imaging Wide width high capacity, fully non-aligned Data Cache provides single-cycle addressing scalar operands byte addresses, parallel operands bytes byte addresses, with performance penalty non-aligned scalar parallel operands Extensive bit-stream manipulation capability Data Cache supports large search distances motion estimation 1,152 general purpose registers Parallel Processing Unit Fast matrix-transpose capability Automatic conversion between parallel data types common imaging Very high useful performance (figures clock) 16b, 16b, multiply-adds clock with accumulation (3.2B MACS) histogram, table look-up zero-detection operations clock 16b, plus 32b, operations clock bit-realignment storage operations clock 64-point motion-estimation pattern matching operation clock (50,000 RISC MIPS) Eight 4-point matrix-vector multiplies convolutions clock (3.2B MACS)
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
Four pairs convolutions clock implementing wavelets (3.2B MACS) clock, SDRAM clock Enhanced ports imaging, multimedia device control Three independent, asynchronously clocked, glueless, video-aware packet capable, double-buffered parallel ports, each data path programmable 16-bits wide, video sync signals supports live digitized video input output, including video buffering required, also enable multiple A436's work together even higher performance Able receive images from four image sensors, produce video output from them, simultaneously Able provide high speed refresh video output devices including video encoder chips, field sequential color microdisplays Bit-Programmable port bit-programmable pins interface switches, actuators, keypad other devices; interrupts fully programmable Buffered, full-duplex, Stereo Audio Port provides handles 16-bit samples little-endian big-endian formats Buffered, full-duplex, programmable 2-wire/3-wire, Serial Port controls/accesses external speed high speed devices including image sensors, video encoders decoders, serial EEPROMs flash memory cards Supports video encoder/decoder chips audio CODECs simultaneously UART ports fully programmable programmable interrupt timers Real time clock clock counter Internal debug device provides built-in logic analyzer with trace capability program debugging address space; memory configuration dynamically specified boot-up Memory interface provides clock SDRAM 5v-tolerant 3.3v SDRAM interface power dissipation portable devices Optimized design imaging minimizes power dissipation Power management with multiple operating modes including clock enable external SDRAM Output external crystal oscillator up-converted internal provide SDRAM clocks Packages including 208-pin PQFP that signal pin-compatible with A236, 0.35 3.3v CMOS technology high performance cost
Applications
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
MPEG-2, MPEG-4, JPEG, JPEG-2000, motion JPEG, wavelets, "your own", universal image video compression/encoding, streaming decompression/decoding, HDTV high resolution encoding real time using multiple chips parallel
Wireless, consumer, industrial professional digital "smart" still movie cameras, camcorders PDAs
Stereo cameras, night vision systems, Real-time image/video capture, correction video surveillance, video recording lens errors, formatting display video security, scene content analysis Video editing special effects real time Digital office imaging equipment including scanners, printers, cameras machines Multimedia Internet video applicances 360-degree panoramic vision capture, merge display video streams from four color digital image sensors simultaneously Videoconferencing, H.261, H.263 Smart cards, biometrics, touchless fingerprints, neural networks, pattern recognition Communications, encryption, decryption Decompress live video using format, scale recompress using another format
Enhancements Image/Video Compression/Streaming Decoding
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
Easily programmed (not microcode) good control code generated provided 1,152 16-bit general purpose registers Parallel Processing Unit store many coefficients motion targets 32-bit general purpose registers Scalar Arithmetic Unit store many addresses offsets Performs 8-pixel pixels total) motion-estimations single cycle, including keeping track best matches, detects spatial temporal patterns Performs multiply-adds single cycle DCT, convolution, wavelets scaling Performs four pairs 4-point convolutions single cycle wavelets
Data Cache architecture, page size, capacity write policy optimized imaging Fully "non-aligned" Data Cache provides access 16-byte parallel operands single cycle regardless memory address Data Cache large enough handle large search distances motion estimation Memory written without reading first quickly make large scratchpads fill display buffers, saving memory bandwidth Video-aware parallel ports capture place data memory addresses that optimize performance Data Cache Efficiently packs, unpacks, decomposes, decimates, interpolates reconstructs color video data
Reads writes successive Performs matrix-transpose (zero) alternating 8-bit values from/to cycles quantizes entire memory single cycle matrix cycles quickly handle SRGB image/video data Converts 8-bit signed unsigned memory operands 16-bit signed Checks 16-bit values simultaneously values read saturates quickly detect block operands store single cycle wavelets efficiently handle SRGB data
Description
Oxford Micro Devices' Ax36 family video digital signal processor chips highly optimized handling live images being programmed directly "C". API's, "canned" libraries microcode required. Extremely high performance achieved with moderate clock rates conservative, low-cost fabrication using efficient, highly parallel design. architecture exploits patterns with which image data
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
stored memory. These patterns represented series hardware/software templates parallel data structures. Many levels optimization result much higher performance, smaller programs, easier I/O, improved memory utilization, smaller system size, lower power dissipation lower cost than alternative processors. Highly efficient programs that short, quickly have small memory footprint easily written using Oxford Micro Devices' ANSI-standard, parallel-enhanced compiler. Time-critical loops easily re-coded using parallel enhancements existing functions that time-critical simply recompiled. Instructions that perform parallel processing only bits long, reducing program size ROM-based applications. A436 large (20x) increase image processing power compared A236. This provided combination increased clock rate, increased parallelism, improved Data Cache, implementation parallel data types. A436 eight instances improved version A236's Parallel Processing Unit. Among many other uses, cost, universal, real-time, fully software programmed, standards-based proprietary, wavelet- DCT-based, video compressor implemented with A436. most significant improvements A436 compared A236 are: number parallel arithmetic units increased from number multipliers from number motion estimation coprocessors from number general-purpose registers from 1,152 width Data Cache increased from bits sizes parallel operands have been increased take full advantage Parallel operands have many adjacent bytes, placed memory byte address, accessed, sign- zero-extended 16b, processed single cycle, with performance penalty non-aligned operands. Two-dimensional, imaging-oriented parallel data types added. Special instructions parallel implementation matrix-vector multiply, convolution, matrix transpose, added. Bit-realignment storage parallel operands done single instruction. full-duplex Stereo Audio Port, Bit-Programmable Port, second interval timer internal Debug Port added. fully programmable, 2-wire/3-wire, full-duplex, buffered Serial Port added
Benchmarks
clock DCT: only cycles (340 ns), including reading packed alternating bytes from memory, sign- zero-extending them 16b, matrix transpose bit-realignment adjust multiplying bytes words, precision; entire 64-word result quantized only additional cycles motion-estimation: only cycles average, since eight, 8-point motion estimation calculations performed simultaneously matrix transpose words: only cycles ns), starting with matrix registers,
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
(zero) cycles internal part convolution signed unsigned values using coefficients: only approx. 0.45 cycle/block (4.5 average, since convolutions performed simultaneously, including bit-realignment adjust multiplying bytes words prior storing results memory Erosion/dilation overlapping blocks pixels memory using arbitrary mapping function requires only cycles, average only 0.55 cycle/block since filters implemented simultaneously JPEG compress color image 4:1:1 resolution, approx frames/s), 3,000 2,000 color image approx. (2.5 fps)
Block Diagram
Notes: Ports that labeled "8b/16b indicate that there data bits plus control bits Clock Generator crystal external clock input
Related Documents
A436Programmers Reference Manual, requires password A436Hardware Reference Manual, requires password
A436 Parallel Video Chip from Oxford Micro Devices
file:///W|/A436_summary.html
Novel architecture programmed Slide Show: OMDI: Providing Universal Video Solution Application Note: Advantages A436 Video Chip Image Compression Ax36logic symbols Orcad format, requires password Articles Slide Shows Developers' Program Application Notes Software Tools, Libraries Applications
Oxford Micro Devices, Inc.
Ax36Parallel Image Video Digital Signal Processors Chips Lantern Ridge Office Park; Main Street, Building Monroe, 06468 tel. 203-445-0562; 203-445-0564; e-mail: info@omdi.com; Web: http://www.omdi.com/ Copyright 1995 2001. rights reserved. Patents pending. trademarks property their respective owners used product identification purposes only.

file:///W|/A436_summary.html Summary A436Parallel Video Chip

Top Searches for this datasheet

Other recent searches