Introduction

GL's Super T1 and Super E1 Cards can be used to test echo canceller performance in a wide variety of network configurations, including: 

• conventional telephone networks
• networks with satellite connectivity
• wireless networks such as cellular and PCS

Many important performance characteristics discussed and specified in Recommendation G.168 can be easily tested as well as other important multichannel tests.

Overview of Echo Canceller Testing

Echo Canceller function and location in a network (Figure 1):

See Figure 1

Echo cancellers are placed in networks as shown in Figure 1, intended to cancel the echoes introduced by the nearest echo path (the nearest 2/4 wire hybrid), also called the tail. Acoustic echoes could also be introduced by end devices such as speakerphones. Generally, in terrestrial voice networks, there are two echo cancellers in an established voice connection, one at each tail. In cellular networks, there may only be one echo canceller, canceling the terrestrial tail echo. Most recent echo cancellers are generally digital and operate on multiple channels simultaneously. Some overall performance parameters concerning echoes in a network are:

• Any connection with a one way delay of 25 ms or greater requires an echo canceller.
• A talker requires that any echo heard in the earpiece be at least 46 to 54 dB below his/her own speech level.

See Figure 2

An echo canceller operates on the principle that a portion of the signal in the receive path is reflected with some linear distortion into the send path. DC offsets and frequency translation are other distortions that may be introduced by the echo path elements. The linear distortion can be estimated and subtracted from the send path. The receive path is transparent. The send path is digitally processed and not transparent. The echo estimation process is constantly adapting to the echo path (which could be changing with time) through a feedback process. During double talk, the adaptation process is frozen.

See Figure 3

Echo Canceller Testing (Figure 4):

See Figure 4

A generalized block diagram for echo canceller (EC) testing is provided in Figure 4. A variety of stimulus signals can be injected into the Receive Input of the EC. The echo path can be modeled at near end using DSP techniques. The input and output of the EC may be captured for post analysis.

Echo path stimulus with recording (Figure 5a):

See Figure 5a

This script permits the continuous transmission of any file (Composite Source Signal (CSS) is shown) into the receive input of the EC. Simultaneously, the send output and the receive input is continuously recorded for post analysis. A program such as Spectra Plus or Adobe Audition can be used to display the waveform in synchronism. The file transmission can be preceded with signalling, such as may be required for setting up a call.

Echo path stimulus with real time power monitor (Figure 5b):

See Figure 5b

This script permits the continuous transmission of any file (Composite Source Signal (CSS) is shown) into the receive input of the EC. Simultaneously, the send output and the receive input levels are continuously measured and displayed on the screen. This script may be used to observe the EC performance in real time.

Typical echo path impulse response (Figure 5c):

See Figure 5c

An impulse response such as showncan be captured by transmitting into the receive input of an EC, a file with impulse spikes at regular intervals. Such a file, 'impulse.pcm' has been provided.

Simulation of Echo Path (Figure 6):

See Figure 6

Simulation of the echo path involves, simulating the hybrid reflection and loss and the transmission delay of the echo path. This can be conveniently done using buffering and DSP techniques. Scripts that perform these functions on single and multiple channels are provided.

Double Talk Simulation with Single Echo (Figure 7):

See Figure 7

Echo cancellers must be able to detect double talk quickly and disable the non-linear processor in the send path. To simulate this condition, a double talk speech file has been provided, that can summed into the echo path.

Echo path simulation with VF Channel (Figure 8):

See Figure 8

An analog connection can be connected during the simulation of the echo path for testing live analog connections such as speech, modem, fax, etc.

Double talk simulation with multiple echo paths (Figure 9):

See Figure 9

See Figure 10

Description of G.168 TestsScripts are being developed to test the following Recommendation G.168 test. Contact GL Communications for latest status.

See Screenshot of Test No 1

See Screenshot of Test No 1 - pic2

Test 2A

Convergence test with NLP enabled - Essentially, within 1 sec (plus the echo path delay), the return level should be as specified in Test No.1.

See Screenshot of Test No 2A

Test 2B

Convergence test with NLP disabled - Within 1 sec (plus the echo path delay), the loss due to cancellation and the echo path should exceed 16 dB.

See Screenshot of Test No 2B

Test 2C

Convergence test with background noise - A background noise source is added to the return path at a level 15 dB  below the incoming signal level. With the NLP disabled the convergence time should be less than 1 sec with returned levels at least 15 dB less than the applied signal level. A criterion has not been established for this case with the NLP enabled.

See Screenshot of Test No 2C

Test No. 3

Performance under conditions of double talk:During conditions of double talk, the echo canceller should continue cancellation. However, should not change its echo path model so as to introduce distortions such as clipping and noise contrast. Three test have been identified:

Test 3A

Double talk test with low near end levels

See Screenshot of Test No 3A

Test 3B

Double talk test with high near end levels

See Screenshot of Test No 3B

Test 3C

Double talk test under simulated conversation

See Screenshot of Test No 3C

Test No. 4 - Leak rate test:

This test measures how rapidly the H register values (impulse response) decay to zero. This test is performed with the NLP disabled. A signal is applied to R(in) to allow the echo canceller to converge and then is removed (set to idle code ?) for two minutes. The signal is then reapplied and the return level is measured. The steady state level should not have decayed by more than 10 dB.

See Screenshot of Test No 4

See Screenshot of Test No 5

Test No. 6 - Non divergence on narrow band signals:

This test verifies that the echo canceller remains converged on narrowband signals such as DTMF and single frequency tone after initially converged on wideband signals such as CSS. Criteria have not been established for this test. 

See Screenshot of Test No 6

Test No. 7 - Stability test:

This test verifies that the echo canceller remains stable for narrowband signals such as a single frequency tone. A test tone is applied and the canceller is allowed to converge. The tone is continuously applied, and the residual signal level from the canceller is measured after two minutes.

See Screenshot of Test No 7

Test No. 8

Non convergence of echo cancellers on specific ITU-T No. 5, 6, and 7 in-band signalling and continuity check tones:Test the echo canceller in the presence of single and dual frequency tones used in signalling systems.

Test No. 9

Comfort noise test:Test to verify that the noise applied to the S(out) port of the echo canceller is equal to the noise seen at the S(in) port of the canceller. The adaptation of the canceller to changing levels of S(in) noise is also tested.

See Screenshot of Test No 9

Test No. 10

Facsimile test under call establishment phase:

Test No. 11

Tandem echo canceller test:Test No. 12 - Residual echo test:

Test No. 13

Performance with ITU-T low bit rate coders in echo path:

Test No. 14

Performance with V.Series low speed data modems: Other Tests:

A variety of other test performance tests that Recommendation G.168 does not address can also be performed.

Echo path capacity test

This tests the maximum echo path delay that an echo canceller will cancel

Channel by channel tests

This tests each channel for proper echo canceller operation Scripts. There are some specialized scripts that have been written by GL engineers (using GL's Scripted Control Software Product) that find ready application in the field of echo canceller testing

They are listed in the below table

Script Name	Script Description
complex.scr	Performs complex echo generation
danterm.scr	Receive a call from a DMS-250 and then simulate echo path on a single channel
dandms.scr	Send signalling information prior to sending echo canceller stimuli; Send CSS and capture on one board
danvf.scr	Simulate echo path on all timeslots except for one timeslot on which VF Input and VF Output is permitted
dan3a.scr	Simulate echo path on a single channel with double talk source file
dan4.scr	Send tone capture on one board
dan2.scr	Two board configuration: On the first board, send CSS and capture return signal; on the second board, simulate echo loop
dan1.scr	Send CSS and capture one board. Assumes that echo loop simulations are being done on a diferent PC
dan7.scr	Send CSS or QRSS and monitor power on one board
dan6.scr	Send CSS and monitor power on one board
dan3.scr	Simulate echo path on a single channel
dan5.scr	Send impulse train, capture impulse response on one board
fanhuang.scr
dspatten.scr	Multiple DSP operators; tests multiple channel attenuate operation
dsploop2.scr	Multiple DSP operators; tests multiple channel loop back operation
phaserev.scr
echogen.scr	Performs complex echo generation
dspgenrx.scr	Multiple DSP operators; tests multichannel dsp operations. Receive channels are processed and sent back out. Channels are generated, processed and sent out also
dspdelay.scr	Multiple DSP operators; test multiple channel delay operation
dialtone.scr	Shows how to generate U.S. dial tone
dcoffset.scr	Show how to introduce a DC offset to any signal

Glossary of Echo Canceller Terms:

Echo Canceller Term	Description
acoustic echo	echo caused by some portion of the speaker output being fed back to the microphone input.
automatic level control (alc)	sometimes performed by echo cancellers if an imbalance in speech levels exists.
comfort noise	noise introduced into the send path to compensate for the total silence produced by the non linear processor action. Sometimes the original background noise may be substituted.
composite or complex echo	the addition of multiple echoes
composite signal source	a test signal used to test echo cancellers consisting of noise, speech and silence periods
convergence and convergence time	the process or time required for the echo canceller to adapt to the echo path
echo canceller	a device used to cancel echoes produced by 2/4 wire hybrids in a telephone network
echo path	a reflection path that causes echo and upon which the echo canceller operates
echo path capacity	a cancellers maximum echo path delay limit
echo path delay	the echo delay seen by the echo canceller
echo return loss	the level of the reflected signal compared to the original signal
echo return loss enhancement	the additional loss introduced into the echo signal by the canceller operation
far end	the side on which the canceller does not operate; the opposite of the near end
H register	the impulse response coefficients stored in the echo canceller
leak time	a measure of the decay of a converged echo canceller
near end	the side upon which the canceller is designed to operate
non-linear processor (nlp)	technique used to eliminate the residual echo
receive path	the path through the echo canceller of the original signal (not the echo signal)
residual echo level	the level of the signal after echo cancellation but before nlp application
returned echo level	the output of the echo canceller
send path	the path upon which the canceller operates
sidetone	the instantaneous echo heard on the phone to provide 'the feel of a live connection
talker echo	the echo heard by the talker when he/she speaks into the phone
tone disabler	in band tones that disable the echo cancellers; may be necessary for some modem connections and signalling.

 Back to Echo Canceller Testing Application Notes Main Page