CDB6403 データシートの表示（PDF） - Cirrus Logic

部品番号

コンポーネント説明

一致するリスト

CDB6403

Echo-Cancelling Codec

Cirrus Logic 

CS6403

Several echo-canceller controls are provided in

Mode 2 (16 bit) operation via the SSI control

registers. In addition, visibility and control func-

tions are provided through the general-purpose

I/O pins (see the pin-definition section).

Graded Beta

"Graded Beta" is a performance enhancement

used to improve the convergence speed and as-

ymptotic ERLE performance of the acoustic

echo canceller on the CS6403. Given a lower

limit to the decay rate of the expected echo re-

sponses, the CS6403 will adjust the updates to

the acoustic echo canceller to take advantage of

that information.

By default, the CS6403 assumes a decay rate of

at least 0.38 dB per ms. Based on experiments

performed at Crystal, the acoustic echo-path de-

cay rate for car interiors tends to be at least one

dB per ms. Offices, on the other hand, tend to

be more "live", with decay rate potentially below

0.38 dB per ms. Note that the minimum-ex-

pected decay rate can be set via SSI_CR0 bits 3

and 2.

Half-Duplex Suppression

After the CS6403 is powered up or reset, 2-3

seconds of far end and near end speech must be

processed by the echo cancellers to sufficiently

reduce loop-gain and therefore prevent acoustic

howling. To prevent howling while the echo

cancellers are not properly trained, a half-duplex

echo suppressor is enabled. Once the echo can-

cellers are properly trained, the half-duplex

suppressor is automatically disabled. (Note that

whether the CS6403 is operating in half or full

duplex at any particular time is indicated via the

GPOUT1 output pin).

This half-duplex suppressor works like the sup-

pressor in a half-duplex speakerphone; i.e., it

allows signals to pass through the CS6403 in

only one direction at a time. The talker at one

DS192PP6

end of a conversation cannot be heard at the

other end until the talker at the other end is si-

lent.

The half-duplex algorithm in the CS6403 has

been designed to discriminate between noise and

speech, and should provide good performance in

noisy environments.

Full-Duplex Suppression

After the echo cancellers have been trained, the

half-duplex suppressor is automatically disabled.

This transition occurs when the ERLE perform-

ance of the echo cancellers exceeds a fixed

threshold.

In some cases, due to impairments like non-

linearities, the echo cancellers may not provide

sufficient ERLE. To accommodate such situ-

ations, the CS6403 provides supplementary

full-duplex suppression. This full-duplex sup-

pression technique provides additional ERLE

using dynamic gain-control and accounts for the

"Non-linear Echo Control" block in Figure 3.

Operation in Noise

The CS6403 echo cancellers have been designed

to give good performance in noisy environments.

However, for best performance, the echo cancel-

lers should be trained in a low-noise

environment. If the noise level is high during

the training interval, the echo canceller may not

be able to achieve enough ERLE to transition

out of half duplex, regardless of how much

speech is received.

If the noise level subsequently drops sufficiently

while a speech training signal is present, the

echo canceller can train, allowing the CS6403 to

transition to full duplex. If the noise level then

increases, the echo canceller will use the path-re-

sponse estimate calculated while the noise level

was low, allowing the echo canceller to remain

in full duplex.

25

Share Link: