5th etsi speech Quality Test Event Anonymous Test Report

Download 0.63 Mb.

Page	8/13
Date	06.08.2017
Size	0.63 Mb.
	#27239

1 ... 5 6 7 8 9 10 11 12 13

DECT Echo Performance

3.6.8Gateway No.1

The extreme echo reflected from DECT phone 1 lying on a hard surface (transducers down) is not suppressed. The echo level is probably too high to be detected as echo instead of double talk. The echo loss result (see table) and the echo attenuation analysis in fig. 5.99 indicate this. The echoes from the two other DECT scenarios are handled by the implementation. The echo attenuation is sufficiently high and constant vs. time (see fig. 5.100 and 5.101).
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	26 dB			18.1 dB
	DECT Phone 1 @ HATS		31.0 dB	26 dB			45.5 dB
	DECT Phone 2 @ HATS		36.3 dB	26 dB			45.5 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.99: Phone 1, 5 dB_m0		Fig. 5.100: Phone 1 @ HATS, 5 dB_m0				Fig. 5.101: Phone 2 @ HATS, 5 dB_m0

The near end double talk bursts are nearly completely transmitted for the two DECT phones mounted to the HATS. However, echo components occur under double talk conditions between the near end signal bursts.
		Fig. 5.102: Double Talk, Phone 1 @ HATS				Fig. 5.103: Double Talk, Phone 2 @ HATS

3.6.9Gateway No.2

The implemented signal processing suppresses the echoes from the three DECT scenarios. The echo loss is very high (see table) and indicates an aggressive echo suppression (NLP). The echo attenuation vs. time analysis meets or only slightly violates the tolerance in figure 5.104 to 5.106.
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	26 dB			60.6 dB
	DECT Phone 1 @ HATS		31.0 dB	26 dB			70.5 dB
	DECT Phone 2 @ HATS		36.3 dB	26 dB			62.1 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.104: Phone 1, 5 dB_m0		Fig. 5.105: Phone 1 @ HATS, 5 dB_m0				Fig. 5.106: Phone 2 @ HATS, 5 dB_m0

The strong influence of NLP can also be analyzed during the DECT double talk scenarios. The near end bursts are partly clipped under extreme test conditions (high receive level, low near end level). Vice versa, echoes between the near end signal bursts do not occur.
		Fig. 5.107: Double Talk, Phone 1 @ HATS				Fig. 5.108: Double Talk, Phone 2 @ HATS

3.6.10Gateway No.3

The strong DECT echo reflected with an attenuation of 18 dB is not cancelled by this implementation. However, if a high receive test signal level of -5 dB_m0 is used for testing (see figure 5.109) the gateway suppresses even this echo signal. The echo loss results and the echo attenuation vs. time measurements further show, that lower level DECT echoes reflected under realistic conditions (DECT phone mounted to a HATS) are sufficiently attenuated.
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	26 dB			18.1 dB
	DECT Phone 1 @ HATS		31.0 dB	26 dB			45.6 dB
	DECT Phone 2 @ HATS		36.3 dB	26 dB			45.4 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.109: Phone 1, 5 dB_m0		Fig. 5.110: Phone 1 @ HATS, 5 dB_m0				Fig. 5.111: Phone 2 @ HATS, 5 dB_m0

The implementation provides strong echo suppression. The near end signal bursts are partly clipped, the main focus is the avoidance of residual echoes.
		Fig. 5.112: Double Talk, Phone 1 @ HATS				Fig. 5.113: Double Talk, Phone 2 @ HATS

3.6.11 Gateway No.4

The extreme DECT scenario represented by the DECT phone 1 on a hard surface leads to residual echo disturbances (see table). The echo level probably activates double talk detection. The echo loss is sufficiently high for the two other realistic scenarios. The echo attenuation vs. time violates the tolerance for the combination of a high test signal level of -5 dB_m0 and two of the three DECT scenarios (see fig. 5.114 and 5.115). The tolerance is met for a DECT phone providing slightly higher acoustic echo attenuation (DECT phone 2 @ HATS).
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	26 dB			17.6 dB
	DECT Phone 1 @ HATS		31.0 dB	26 dB			48.6 dB
	DECT Phone 2 @ HATS		36.3 dB	26 dB			47.7 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.114: Phone 1, 5 dB_m0		Fig. 5.115: Phone 1 @ HATS, 5 dB_m0				Fig. 5.116: Phone 2 @ HATS, 5 dB_m0

Echo components can also be detected under double talk conditions. Furthermore the near end double talk bursts are partly clipped under the extreme test conditions analyzed in both figures.
		Fig. 5.117: Double Talk, Phone 1 @ HATS				Fig. 5.118: Double Talk, Phone 2 @ HATS

3.6.12 Gateway No.5

The echo loss results given in the table below are measured with an average test signal level of –16 dB_m0. The results are sufficiently high under all test condition. However, the echo attenuation vs. time measurements indicate strong level variations. The residual echoes are not sufficiently suppressed. These tests are carried out with higher test signal levels of -5 dB_m0.
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	26 dB			43.7 dB
	DECT Phone 1 @ HATS		31.0 dB	26 dB			46.7 dB
	DECT Phone 2 @ HATS		36.3 dB	26 dB			48.9 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.119: Phone 1, 5 dB_m0		Fig. 5.120: Phone 1 @ HATS, 5 dB_m0				Fig. 5.121: Phone 2 @ HATS, 5 dB_m0

The near end signal bursts are transmitted, but impaired by initial clipping. Furthermore echo components occur between the near end bursts.
		Fig. 5.122: Double Talk, Phone 1 @ HATS				Fig. 5.123: Double Talk, Phone 2 @ HATS

3.6.13Gateway No.6

The implemented signal processing reduces the residual echoes under the three DECT test conditions. The echo loss is sufficiently high for the realistic test cases (see table). Even the worst case DECT echo is significantly reduced. The implemented NLP suppresses most of the echo signal energy. Some residuals still lead to an echo tolerance violation. The echo attenuation vs. time analyses show a slight violation at the beginning of each signal burst.
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	Internal			38.7 dB
	DECT Phone 1 @ HATS		31.0 dB	Internal			49.4 dB
	DECT Phone 2 @ HATS		36.3 dB	Internal			51.0 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.124: Phone 1, 5 dB_m0		Fig. 5.125: Phone 1 @ HATS, 5 dB_m0				Fig. 5.126: Phone 2 @ HATS, 5 dB_m0

The strong influence of NLP leads to an efficient echo suppression even under double talk conditions. On the other hand, the near end bursts simulating near end speech are partly clipped.
		Fig. 5.127: Double Talk, Phone 1 @ HATS				Fig. 5.128: Double Talk, Phone 2 @ HATS

3.6.14Gateway No.7

The two realistic DECT scenarios with two phones mounted to the HATS are appropriately handled by the implementation. The echo attenuation expressed by the echo loss results in the table below is high. The worst case echo condition (DECT phone 1 on a hard surface) leads to strong residual echoes. The echo attenuation vs. time significantly violates the tolerance for the DECT phone on a hard surface. The two other test conditions slightly violate or meet the tolerance.
	DECT Test Scenario		DECT ERL	Hybrid ERL			Gateway
	DECT Phone 1 (hard surface)		17.8 dB	Internal			19.1 dB
	DECT Phone 1 @ HATS		31.0 dB	Internal			52.9 dB
	DECT Phone 2 @ HATS		36.3 dB	Internal			52.9 dB
	Echo loss (G.122) for different simulated DECT phones


Fig. 5.129: Phone 1, 5 dB_m0		Fig. 5.130: Phone 1 @ HATS, 5 dB_m0				Fig. 5.131: Phone 2 @ HATS, 5 dB_m0

Strong echo components appear under double talk conditions. The NLP is not activated, double talk is detected. Consequently the near end double talk bursts are completely transmitted in both figures.
		Fig. 5.132: Double Talk, Phone 1 @ HATS				Fig. 5.133: Double Talk, Phone 2 @ HATS

Download 0.63 Mb.

Share with your friends:

1 ... 5 6 7 8 9 10 11 12 13

The database is protected by copyright ©ininet.org 2024
send message