CDB6420 Ver la hoja de datos (PDF) - Cirrus Logic

Número de pieza

componentes Descripción

Lista de partido

CDB6420

Full-Duplex Speakerphone Chip

Cirrus Logic 

CS6420

are the heart of most echo canceller implementa-

tions.

The worst case situation for the CS6420 is when

parties at both ends are speaking and the person at

the near-end is moving. In this case, the echo can-

celler will cease to adapt because of the double-

talk, but the echo will not be optimally reduced be-

cause of the change in path.

Adaptive Filter

The adaptive filter in the CS6420 uses an algorithm

called the “Normalized Least-Mean-Square

(NLMS)” update algorithm to learn the echo path

transfer function. This Finite Impulse Response

(FIR) filter has 508 taps, which can model up to

63.5ms of total path response at a sampling rate of

8kHz. The coverage time is calculated by the fol-

lowing formula:

8----k--1-H-----z x 508 = 63.5 ms.

The CS6420’s adaptive filter, like all FIR filters,

only models Linear and Time Invariant (LTI) sys-

tems. So, any non-linearity in the echo path can not

be modeled by the adaptive filter and the resulting

signals will not be cancelled. Signal clipping and

poor-quality speakers are very common sources of

non-linearity and distortion.

A common integration problem for echo cancellers

is signal clipping in the echo path. For example, if

a speaker driver is driven to its rails, the distortion

of the speech may be hard to perceive, but it is very

bad for the echo canceller. This technique has been

used in half-duplex phones to provide good low-

level signal gain at the expense of distortion with

high amplitude signals. Since this does not work

for the CS6420, an AGC mechanism has been in-

troduced to provide equivalent behavior without

clipping. See the section on AGC for more details.

Another common problem is speaker quality. A

poor quality speaker which is perfectly acceptable

for a half-duplex speakerphone, may limit the echo

canceller’s performance in a full-duplex speaker-

phone. The distortion elements will not be modeled

by the adaptive filter and so limit its effectiveness.

Speakers should have better than 2% THD perfor-

mance to not impede the adaptive filter.

Volume control should be implemented only using

the CS6420 Microcontroller Interface. A real-time

external change in the gain of the speaker driver,

for example, would result in a change in the trans-

fer function of the echo path, and so would force

the adaptive filter to readapt. If the volume control

is done before the input to the adaptive filter, the

echo path does not change, and no retraining is nec-

essary. Another side benefit of the CS6420 volume

control is that it transparently provides dynamic

range compression.

Pre-Emphasis

The typical training signal for the adaptive filter

will be speech, but most adaptive filters work opti-

mally with white noise. Speech has very different

spectral characteristics than white noise because of

its quasi-periodic nature.

Research at Crystal has shown that quasi-periodic

signals cause the formation of spurious non-zero

coefficients within the adaptive filter at tap inter-

vals determined by the periodicity of the signal.

This results in small changes in period being very

destructive to the adaptive filter’s performance.

One mechanism the CS6420 uses to prevent this

filter corruption with speech is to pre-emphasize

the signal sent to the adaptive filter so that much of

the low frequency content is removed.

The CS6420 works very well with a speech training

signal because of the pre-emphasis filter. White

noise training signals, however, will result in sub-

optimal performance, so when testing, white noise

is not recommended as a training signal.

Graded Beta

The update gain of an adaptive filter, sometimes

called the “beta”, is the rate at which the filter co-

DS205PP2

23

Share Link: