Digital Voice Systems VC-20-ATC-10B Especificaciones descargar pdf (Pagina 48)

FAA20 Embedded NEXCOM Vocoder

Board Manual

This manual has been prepared for the Federal Aviation Administration.

FA100-00107 (March 2004) – Rev 1.0.0 Page 42

THREAD ↓ DESCRIPTION

DEC_SWI 11 Decoder. The function converts linear voice data to compressed

voice data. It typically completes execution within 7.5 ms

(includes the execution time for higher priority threads). It is

triggered once every 20 ms (normal timing).

KNL_SWI* 12 Kernel. The DSP/BIOS

CMD_TSK 13 Command Task. This thread implements the multi-port command

processor. The thread services each COM port in round-robin

fashion. When a complete command line has been received, the

command is executed. A maximum of 1 command is executed

each time a port is serviced.

IDLE_TSK* 14 Idle Task. This idle task runs when no other thread is running.

* This hardware interrupt is not shown in Figure 20.

6.2.3 Queues, Packets and Data Management

Queues are used extensively to pass data between threads. This approach

provides flexible data flows when processes run at different rates. Queues

are sized and controlled to minimize voice delays while still providing some

execution economy.

The processing threads work with data packets, i.e. a set of data. The

queues contain data only, i.e., no headers, footers, or other information.

For the linear voice process (LVP_SWI), the packet size is 5 samples. For

the rate adapter thread (RAT_SWI), the packet size is 5 or 6 samples for

normal or truncated timing, respectively. For the vocoder threads

(ENC_SWI and DEC_SWI), the packet size is 160 samples (nominal). The

COM queues are fairly large to accommodate the amount of binary data

needed to work with the predefined ASCII data formats.

The compressed voice router works with voice frames. Typically, the

compressed frame (packet) size is 12 bytes (96-bits); however, during

BITEXACT testing the frame (packet) size is 48 bytes to allow for up to 4

soft detection bits.

The FAA20 software actively manages queue depths to minimize voice

delays and to compensate for slight timing differences. Both coarse and

fine adjustments are made, as required. Coarse adjustments add/drop data

packets. Fine adjustments add/drop data samples. Together, these events

are referred to as SPAD (sample/packet add/drop) events. Fine

adjustments are only made on linear voice queues, since compressed

queues must work on a complete frames

The management of the QUE_LVOCTX and QUE_LVOCRX queue depths

is the most important since the potential delay impact is large, i.e., the 160

sample frame (packet) represents a 20 ms delay. The queue size is a

multiple of this value. The other queues in the primary voice flow paths are

small, i.e. 13 samples or less.

Queue management also handles slight differences in interface timing via

single sample add/drop events. Figure 20 shows that voice processing

1 2 ... 43 44 45 46 47 48 49 50 51 52 53 ... 108 109

Comentarios a estos manuales

Sin comentarios

Digital Voice Systems VC-20-ATC-10B Especificaciones Pagina 48

Comentarios a estos manuales