AN643 Adaptive Differential Pulse Code Modulation ® Using PIC Microcontrollers 3.00. The ITU (formerly CCITT) G.721 ADPCM algo- Author: Rodger Richey rithm is well known and has been implemented on Microchip Technology Inc. many digital signal processors, such as the TMS320™ DSP family of devices from Texas Instruments™ and the ADSP-2100 family of devices from Analog Devices. INTRODUCTION ITU G.721 uses floating-point arithmetic and logarith- In the past, adding speech recording and playback mic functions, which are not easily implemented in the capability to a product meant using a digital signal 8-bit microcontroller world. The IMA Reference processor or a specialized audio chip. Now, using a Algorithm significantly reduces the mathematical simplified Adaptive Differential Pulse Code Modulation complexity of ITU G.721 by simplifying many of the (ADPCM) algorithm, these audio capabilities can be operations and using table look ups where appropriate. ® microcontroller or DSC digital signaladded to any PIC controller. This application note will cover the ADPCM COMPRESSION compression and decompression algorithms, and an application using a PIC18F67J10 microcontroller. The input, s , to the encoder routine must be 16-bit i two’s complement speech data. The range of allowable is 32767 to -32768. Figure 1 shows a blockvalues for s DEFINITION OF TERMS i diagram for ADPCM compression and Appendix B: step size – value of the step used for quantization of “Generic ADPCMEncoder() Function” lists the analog signals and inverse quantization of a number of generic C code for the ADPCMEncoder () function. steps. , and the quantizer step sizeThe predicted sample, s p index (q) are saved in a structure for the next iteration quantization – the digital form of an analog input of the encoder. Initially, the quantizer step size index signal is represented by a finite number of steps. ) are set to zero. Theand the predicted sample (s p adaptive quantization – the step size of a quantizer is encoder function takes a 16-bit two’s complement dramatically changed with time in order to adapt to a speech sample and returns an 8-bit number containing changing input signal. the 4-bit sign-magnitude ADPCM code (t). inverse quantizer – a finite number of steps is converted into a digital representation of an analog FIGURE 1: ADPCM ENCODER BLOCK signal. DIAGRAM adaptive predictor – a time varying algorithm that Output of Encoder calculates an estimate of the input signal from the quantized difference output. s + d i dt q Inverse Quantizer Quantizer THEORY OF OPERATION s r Step Size Adaptive The ADPCM algorithm takes advantage of the high PredictorAdapter correlation between consecutive speech samples, q s p which enables future sample values to be predicted.