A VLSI DPCM encoder/decoder chip set for extended quality digital TV

Abstract

An experimental Differential Pulse Code Modulation (DPCM) coder and decoder chip set for digital television has been designed using symbolic layout tools and was fabricated in 2-micron CMOS, Both the coder and decoder have 3 independent DPCM loops on-chip to accommodate various types of sampling and component video formats. Each DPCM loop has 4 different quantizers to permit control of the encoded bit rate. The chips operate to a 16 MHz sample rate. I . INTRODUCTIONWith the increased interest in Advanced Television (ATV) systems and the move toward increased digital traosmission of television signals, the availability of economical digital television coding hardware becomes more important. An experimental prototype Differential Pulse Code Modulation (DPCM) encoder and decoder chip set has been designed to demonstrate that high quality digital television transmission can be realized with low-cost technology. The chip set comprises a single-chip DPCM encoder and a single-chip DPCM decoder. Three identical DPCM loops are designed into each IC to accommodate various component coding and decoding formats, such as RGB, YIQ, and YW. Different types of component sampfing schemes tut possible: [44:4], [42:2] and [4:1:1]. The chips can also be used for coding three NTSC channels. The uncoded data words m eight bits wide and the coded DPCM words are five bits wide. The coders employ a previous pixel predictor.A special high speed DPCM architecture combined with high speed circuit techniques allowed the chips to achieve sample rates exceeding the required 14.3 MHz in the 2 micron CMOS VLSI technoIogy employed. DPCM CODING FOR VIDEO COMPRESSIONDifferential Pulse Code Modulation (DPCM) is a predictive coding technique widely used for the efficient digital traosmission of television signals, providing a useful degree of bit rate compression with respect to ordinary Pulse Code Modulation (PCM) [1,2]. Figure 1 illustrates how DPCM coding works. In essence, the next sample value to be coded is predicted by some combination of previously-decoded samples. In the absence of transmission errors, the identical prediction is made in both the encoder and the decoder. Then only the difference between the next sample and its prediction is sent. One of the most widely used predictors is merely the value of the previous sample (pixel) as decoded. That is the type of predictor employed in this coder. Latch Tp generates this prediction by simply providing a one-sample delay. Notice that a decoder is effectively embedded in the feedback path of the encoder. The decoder acts much like an integrator.To achieve further compression, the value of the difference is quantized, usually in a nonlinear way, to a limited number of values. For example, with 8-bit samples ranging in value from 0 to 255, the differences might be ENCODER DECODER Figure 1. DPCM encoder/decoderquantized to only 31 discrete values for differences ranging from -255 to +255. This would be called 5-bit DPCM because the quantized difference can be coded into a 5-bit word.The integra...