PEB 4364 T V1.2 Infineon Technologies, PEB 4364 T V1.2 Datasheet - Page 64

• Manufacturer Part Number: PEB 4364 T V1.2
• Description: IC SLIC VOICE ACCESS PDSO-36
• Manufacturer: Infineon Technologies
• Series: DuSLICr

Specifications of PEB 4364 T V1.2

• Function: Dual Channel Subscriber Line Interface Circuit (DuSLIC)
• Interface: IOM-2, PCM
• Number Of Circuits: 2
• Voltage - Supply: 3.3 V ~ 5 V
• Current - Supply: 105mA
• Operating Temperature: -40°C ~ 85°C
• Mounting Type: Surface Mount
• Package / Case: 36-PDSO
• Includes: DTMF Generator and Decoder, Line Echo Cancellation (LEC), Teletax (TTX) Generation, Universal Tone Detection (UTD)
• Lead Free Status / RoHS Status: Lead free / RoHS Compliant
• Power (watts): -
• Other names: PEB4364TV1.2XT; SP000007728

DuSLIC
Functional Description
provided by external circuitry. If the adaptation of the shadow filter is performed better
than the adaptation of the actual filter by a value of more than DeltaQ, then the shadow
filter coefficients will be copied to the actual filter. At the start of an adaptation process,
the coefficients of the LEC unit can be reset to default initial values or can be set to the
old coefficient values. The coefficients may also be frozen.
2.8.4
Non Linear Processor (NLP) in DuSLIC-E/-E2/-P
In SLICOFI-2 Version 1.5 a Non Linear Processor (NLP) in addition or as an option to
the existing Line Echo Canceller (LEC) is implemented. Please note that the NLP is not
available with SLICOFI-2 Version 1.5.
The principle of the NLP is based on a limitation of the input signal. This means, that all
samples which are below a limit (in the case of a negative sample above the negative
limit), can pass the NLP without any modification. All samples which are above the limit
(in the case of a negative sample below the negative limit), will be set to this limit (or
negative limit). The value for the limit is the estimated background noise. The advantage
of the limitation is, that the background noise can pass the NLP unchanged. Therefore
the far end talker can not hear the NLP.
The decision when the NLP should be activated, is based on the estimated residual echo
after the LEC. If the signal after the LEC is higher than the estimated residual echo, the
NLP is bypassed.
When both the transmit and receive speech detectors are detecting speech (double talk),
the NLP is also bypassed. An overhang counter ensures, that the end of the speech
signal can pass the NLP unchanged. If end of speech is detected, the overhang counter
starts counting from a predefined start value down to zero. This time ensures that silent
parts at the end of the speech are not ignored.
Important coefficients:
– BN-LEV-X, BN-LEV-R, BN-MAX and BN-ADJ for the background noise estimation
– RE-EST-ERLL for the NLP behavior in the simple mode
– SD-LEV-R, SD-LEV-X and SD-LEV-RE for the speech detection
– CT-LEV-RE for early double talk and near end speech detection
For all other coefficients the default values (see
Table
5.2.3.4) should be chosen.
LEC and NLP implementation
The NLP is integrated in the LEC unit. That means when the LEC and NLP are active,
the output signal of the LEC is influenced by the NLP.
If the DTMF receiver and/or the Universal Tone Detection (UTD) unit are active, the input
1)
signal for both is the LEC and the NLP output signal
. Usually this is no problem, as the
1) For the UTD this is only valid when the sum signal of receive and transmit is fed to the UTD (see
Figure
29).
Preliminary Data Sheet
64
DS3, 2003-07-11