Question 7/XV - Characteristics of network echo control equipment
(Continuation of Question 10/XV 1985-88)

Considering

(1) that the current tests in Recommendation G.165 do not take into
account many real-world signal conditions such as:

1.1 high level speech,

1.2 signalling tone,

1.3 call-processing tone,

1.4 voice band data signals,

1.5 FAX transmission,

(2) that there is a tradeoff between echo path delay (td), convergence
quality and convergence time that is inadequately explored in
Recommendation G.165;

(3) that modifications, if made, to Recommendation G.165 may improve
echo canceller subjective performance;

(4) that double-talk break-in and amplitude clipping can occur in echo
cancellers;

(5) that there may be advantages in the use of noise injection;

(6) that uncertainties over the precision required of the tone disabler
still remain;

(7) that the canceller echo path (ROUT to SIN) may include ADPCM
encoders as well as tandem PCM codings;

(8) that ISDN and ISDN/PSTN interworking issues may impact on the
performance and control requirements of echo control devices,

1. What tests should be added or changed in Recommendation G.165 to
guarantee acceptable behaviour with high level speech, signalling tones,
call- processing tones, voice band data signals, FAX transmissions and
the like? (see annex);

2. What should be the recommended convergence time for echo cancellers
as a function of echo path delay?

3. What signal(s) could be used instead of or in addition to white noise
to better ensure that an echo canceller that passes tests with these
signals results in a good Mean Opinion Score?

4. What signal or test could be used to test for situations where
double- talk break-in and amplitude clipping could occur?

5. What should be recommended for noise injection in echo cancellers and
what should be recommended for any related tests?

6. What degradation is there when ADPCM encoders or multiple PCM coders
are used in the echo path of an echo canceller and what should be
recommended in Recommendation G.165 to verify that adequate cancellation
is obtained?

7. What should be recommended for network echo control devices to take
account of the requirements of the ISDN?

Points for study

1.Whether the performance of echo cancellers which conform to
Recommendation G.165 can adversely affect narrow-band signals present on
their transmission paths and, if so, how this should be taken into
account.

2.The performance of network echo cancellers with speech signals that
have been processed other than by being encoded according to
Recommendation G.711, e.g. by low bit rate ADPCM or wideband encoding.

3.The correct operation of tone disablers - conditions under which
correct operation should be expected.

4.Network echo canceller performance in the presence of signals
corresponding to very loud speech.
5.ISDN requirements.

Notes

1. Account may need to be taken of the study of Questions 8/XV, 21/XV
and 23/XV.

2. Account should be taken of the work of Study Group XVIII on ISDN and
general network issues.

3. Account should be taken of the work of Study Group XII on subjective
testing issues, transmission plan aspects and actual speech level in
modern networks.

ANNEX

(to Question 7/XV)

Proposed tests currently under study for Recommendation G.165

Two tests to ensure stability with narrow-band signals (like some voice
band data sets) are under study:

1) Old provisional test No. 6:

The test procedure is to clear the H register and disable the nonlinear
processor. Apply a sine wave (1 300 Hz + 50 Hz) at Rin, connect a
detector at Sout and provide an echo loss between Rout and Sin.

Requirement (provisional, under study)

With the H register initially set to zero, for all values LRin > 15 dBm0
and < 10 dBm0 and present for three minutes, and for all values of echo
loss > 6 dB, the residual echo level should be less than LRin 18 dB
throughout the test after an initial convergence period of < 1 s.

2) Proposed new test No. 6:

This test has the object of verifying that the echo canceller will stay
stable for narrow-band signals. Measure the maintenance of the
cancellation during and after the application of a sinusoidal wave or a
wave composed of two frequencies.

The method consists of completely converging the echo canceller. A mono
or bifrequency signal is then applied to Rin. After three minutes the
adaption is inhibited and the returned echo is measured. The non-linear
processor is disabled.

Conditions

With the echo canceller fully converged, after application at Rin for
three minutes of a mono or bifrequency signal (f1 + f2 with |f1 f2| >
170 Hz) such that LRin > -30 dBmO and < -10 dBmO and for all values of
echo loss > 6 dB and echo path delay, td <ms, the residual echo level
should be less than or equal to that shown in Figure 7/G.165.

EREC MONO OR BIFREQUENCY SIGNAL

TEST SIGNAL

N

INHIBITION
ADAPTION
ENABLE MEASURE OF THE RESIDUAL
ECHO LEVEL

Echo cancellers which have not been disabled, and which are located on
the line side of Signalling System No. 5 in international exchanges, or
which are associated with national exchanges must operate properly with
signalling tones. The following test to ensure proper operation is under
study:

Proposed test No. 7:

Test No. 7 - Non convergence of echo cancellers on mono or bifrequency
signals transmitted in a handshaking protocol

Test No. 7 is meant to ensure that echo cancellers will not dump a mono
or bifrequency signal transmitted in a handshaking protocol on the
transmit direction after having received an identical signal (except for
level and phase) on the receive direction. This is intended to allow a
correct transmission of certain types of signalling tones without
disabling echo cancellers.

Requirement:

With the echo canceller initially in whatever possible convergence state
(for simplification, the fully converged state for an echo loss of 6 dB
may be chosen) the level of a mono of bifrequency signal (f1 + f2 with
|f1 f2|> 170 Hz) applied at the send in port 90 ms after having applied
the same signal (except for level and phase) at the receive in port,
should not vary by more than 2 dB when compared to the nominal level of
the injected signal. The level "N" of each frequency applied is so that
-18 dBmO < N < O dBmO and the echo loss is infinitive.