![](http://datasheet.mmic.net.cn/330000/PEB2245_datasheet_16443993/PEB2245_20.png)
PEB 2245
Semiconductor Group
20
The block diagram is shown in
figure 10
. The MUSAC is designed to connect any of 512 PCM-input
channels to any of 256 output channels. Any input channel up to a total number of 64 can be
handled in 21 independent conferences simultaneously. Any conference combination from 3
subscribers in 21 conferences up to 64 subscribers in only one conference is possible. Not more
than 8 subscribers should be connected to a single conference, however, in order to ensure an
acceptable speech quality. It can be improved by selecting an additional attenuation and activating
the noise suppression: The input channels can be attenuated by 0, – 3 dB, – 6 dB or – 9 dB and the
output channels by 0 or – 3 dB. Input signals below a threshold programmable to four different
levels are disregarded.
The input information of a complete frame is stored in the on-chip 4-Kbit Speech Memory (SM). The
incoming 512 channels of 8 bits each are written in sequence into fixed positions in the SM with a
repetition rate of 8 kHz. Additionally, in the second half of the frame the 64 conference output
channels of 8 bit each are written into the SM. The memory access is normally controlled by the
input counter in the timing control block when writing into the SM but by the conference unit when
writing the conference output channels. The read access is independent of the write access, so that
both input and conference output channels can be read at any time.
For outputting, the Connection Memory (CM) is read in sequence. Each location in the CM points
to a location in the Speech Memory. The byte in this SM location is read into the current output time-
slot. The read access of the CM is controlled by the output counter also contained in the timing
control block. In addition, in the first half of the frame the input channels connected to a conference
are read in sequence by the Conference Unit (CU).
All connections are set up by an external controller which programs the Connection Memory (CM)
and the Conference Control Memory (CCM) using the microprocessor interface. The CM address
corresponds to one particular output time-slot and line number. The contents of this CM-location
points to a particular input time-slot and line number in the transparent mode. In the conference
mode or multipoint switching mode it contains the conference address and points to a conference
output location in the SM instead. The same conference address is used to access the CCM. The
parameters stored in the CCM include the input time-slot and line number, the associated
conference number as well as the noise suppression thresholds and the attenuation levels. The
conference number defines a unique location in the Conference Sum Memory (CSM) used to store
the accumulated samples for each conference. The Conference Sum Memory is alternately loaded
in the first half of the frame and unloaded in the following second half. In the first half the input
samples are processed to implement the noise suppression, the expansion according to the
European A-law or the US
μ
-law and the attenuation function. The Data Memory (DM) buffers these
samples for output processing. The CSM is used to accumulate these samples and store the
resulting sum. During output processing the input sample is retrieved from the Data Memory and the
appropriate sum from the Conference Sum Memory for subtraction, so that the channel output
signal contains the contribution of all the other channels in the conference except its own. After
output attenuation and PCM compression, the data are written in the Speech Memory for output
switching.
If one result of the subtractions exceeds the full scale value, a saturation appears and the MUSAC
signals this conference overflow condition by an interrupt. The conference number of the
conference in overflow is buffered in the Conference Status Register (CST) which can be retrieved
by the external controller.