User`s manual
AMBE-2000™ Vocoder Chip
User’s Manual Version 4.9
DVSI Confidential Proprietary, Subject to Change Page 36
Visit us at www.dvsinc.com
Sleep (SL):
A 1 in this field indicates the device has been put into sleep mode.
Decoder Output Volume Control: Indicates the current decoder volume.
5.3.10 Framed Output: Words 12-23 : Channel Data
This is the field that contains the actual coded bits. Output of the data begins with the MSB of the first word in this field and
continues through with the final bit output being the LSB of the final word. If the data rate selected is less than 9600bps then
the unused bits in each frame are zero and populate the end of the field. As is noted in the Channel Interface definitions, these
unused bits must still be clocked out of the AMBE-2000™. The packet must always consist of 24 words.
5.4 Unframed Serial Format
The Unframed Format for the channel data is useful for applications which desire minimal glue logic between the AMBE-
2000™ and the channel hardware. The use of minimal hardware in place of a microcontroller can be realized using this data
format. Another distinct difference in this data format is that framing information (data which carries the positional
information relating to the coded bits) is embedded into the data stream itself. Using this data format, the system designer need
only transfer the coded data itself. A single bit from each frame is ‘borrowed’ from the voice data to embed the framing
information. Keep in mind that this ‘borrowed’ bit reduces the effective voice coding rate quality by 50 bits per second. For
example, a system with no FEC running at 2450 bps in Unframed mode will sound equivalent to one running at 2400 bps in
Framed mode.
The designer should also be aware that it takes approximately 15 frames (300 milliseconds) for the decoder to attain
synchronization with the incoming stream before it can output synthesized speech. Systems which are attempting to save
power by shutting down transmission during periods of silence, and then resuming during periods of speech can not handle this
300 millisecond delay for each synchronization, and thus should use Framed mode with a more sophisticated framing method.
The 16 bit per word format, pictured in Section 5.2, is maintained in this mode but only a fraction of the full 16 bits is used to
transfer the coded data. The user selects whether 1, 2, 3 or 4 bits will be transferred in each word based on pins
BAUD_SEL [0:1] and Table 4-B. IMPORTANT: The voice coding data rate selected must be evenly divisible by the number
of voice data bits per word selected.
5.4.1 Unframed Serial Output Format
The Unframed output format contains 1 to 4 bits within each 16-bit serial output word. The formats that contain more than one
bit for each word require the MSB of the data bits to be considered first in the transmission. In Unframed mode, only the
coded voice data bits are output. None of the superfluous information that exists in framed mode is available in this mode.
The number of words that need to be transferred out of the encoder for each 20-millisecond frame will be the number of bits
per frame divided by the number of bits per word. So a system coding at 4800 bps with 3 bits per word will need to read 32 (
[4800 ÷50] ÷3 = 32) words each frame. The serial clock rate is computed by1/ [20 msec /(32 x 16)] = 25.6 kHz. If passive
unframed mode is selected, the data strobe will be computed by 1/[20msec / 32] = 1.6 kHz.
Table 5-P Unframed Serial Output Data Format
Data Unused
Bits per Word
See Table 4-A
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
1 bit per Word Format
D
msb
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 bits per Word Format
D
msb
D
0 0 0 0 0 0 0 0 0 0 0 0 0 0
3 bits per Word Format
D
msb
D D
0 0 0 0 0 0 0 0 0 0 0 0 0
4 bits per Word Format
D
msb
D D D
0 0 0 0 0 0 0 0 0 0 0 0
5.4.2 Unframed Serial Input Format
The Unframed Input format contains 1 to 4 bits within each 16 bits serial output word. For the formats that contain more than
one bit in each word, the MSB of the data bits is considered first in the transmission. In Unframed mode, the header data from
Framed mode is dropped and each 16-bit write contains 1 to 4 coded voice data bits. The number of words that need to be