CT8022 Datasheet, PDF (87/194 Pages) List of Unclassifed Manufacturers – VOIP/VON G.723.1, G279AB TRUESPEECH CO-PROCESSOR

Version: 1.18
PRELIMINARY/CONFIDENTIAL
TrueSpeech® Co-Processor
The 8-bit PCM µ-law codes produced by the CODEC when converted to a linear representation occupy 14 bits
(including the sign bit). The CT8022 handles these 14-bit codes using a 16-bit word representation such that the 14
µ-law bits are left-justified within the 16 bits. The lower (least significant) two bits are padded with zeros. This
ensures compatibility with digital recordings originating from external systems where a full 16 bits of significant
input are encoded or compressed. When the CT8022 is used to decode or de-compress such recordings for output to
a µ-law CODEC, the two least significant bits of the de-compressed 16-bit signal are discarded before conversion to
µ-law encoding and output to the CODEC occurs. For 8-bit PCM A-law, the linear representation occupies 13 bits.
This is also left-justified within 16 bits and the lower bits are filled with zeros.
Q8 format is a notation used to represent fractional numbers. In normal integer arithmetic, it is not possible to
represent numbers less than zero or any fractional number. To overcome this limitation, multiplication by fractional
numbers is performed by up-scaling the multiplier by a fixed base number and then downscaling the result by the
same base number. The base number provides an effective decimal point placed at the base number. That is, when
the multiplier equals the base number, the effective multiplication is by 1.0 (multiply by base number, then divide by
base number). To keep the downscaling process simple, the base number selected is always a multiple of 2. This
makes downscaling (dividing) by the base a simple binary bit-shift. In Q8 arithmetic, the decimal point is placed
between bits 8 and 7 of the 16-bit binary multiplier word (least significant bit is bit 0, ms bit is bit 15). The base
number used is 2 to the power of 8 or 256 = 100H. Therefore, to multiply by 0.5, the Q8 number 80H (= 0.5 *
100H) is used thus: 400H * 0.5 is performed as 400H * 80H/100H. The divide by 100H is performed not as a
division but as a binary 8-bit right shift.
The record and playback volume controls described here do not affect the operation of the CT8022 when in
standalone speakerphone mode (analog telephone line speakerphone). Refer to Section 9.21 for a description of
volume control in Speakerphone Mode.
CT8022A11AQC FW Revision 0118
DSP GROUP, INC., 3120 SCOTT BOULEVARD
87
SANTA CLARA, CA 95054 PH: 408 986 – 4300 FAX: 408 986 – 4490
All specifications are subject to change without prior notice.