3.10G.722
The G.722 wideband coder was also used for testing the IP phones in sending and receiving direction in both modes, handset and hands-free. The TMOS results with minimum and maximum scores together with the one-way delays are given in table 6.5 and figure 6.5 respectively. One implementation did not provide hands-free functionality. These tests were therefore skipped. The measured quality scores point out the following:
-
Very high quality differences can be measured for the three phones in wideband handset mode in sending and receiving direction. The TMOS scores differ by approximately 2 points.
-
The two implementations tested in hands-free mode also show high quality differences in sending direction whereas the receiving quality is surprisingly similar.
G.722 – TMOS (TOSQA2001), VAD off, 20 ms, 8 N (except where otherwise stated)
|
|
Delay [ms] / PL / jitter [ms]
|
TMOS
|
Av. one way delay (ms)
|
min
|
av.
|
max
|
min
|
av.
|
max
|
Sending
|
0 / 0% / 0
|
1.8
|
2.9
|
3.8
|
75.3
|
77.7
|
82.0
|
Receiving
|
0 / 0% / 0
|
2.0
|
2.8
|
4.1
|
70.0
|
79.3
|
93.6
|
HFT, sending
|
0 / 0% / 0
|
1.5
|
2.5
|
3.5
|
49.3
|
60.9
|
72.5
|
HFT, receiving
|
0 / 0% / 0
|
2.9
|
3.1
|
3.2
|
72.2
|
73.2
|
74.2
|
Table 6.5: TMOS (TOSQA2001) results, G.722 speech coder, 20 ms packet length (PL)
Fig. 6.5: TMOS results calculated using TOSQA2001; G.722, 20 ms PL
4IP Phone Conversational Aspects
The following analyses summarize parameters relevant for one-way transmission quality (frequency responses, …) and conversational aspects. These tests are carried out using the G.711 speech codec (a-law).
The analyses below are subdivided as follows:
-
Performance for handset use like one-way transmission parameter, echo performance tests, double talk and background noise transmission tests,
-
Tests covering hands-free capability (frequency responses, echo performance tests, double talk capability and background noise transmission quality),
-
Wideband tests in handset and hands-free mode using the G.722 speech coder.
4.1.1Phone No.1
The phone was tested only in handset mode. It provides a sufficient loudness in both directions (recommended SLR 7 ± 3.5 dB, RLR 3 ± 3.5 dB). The sending TMOS of 4.0 is high. The frequency response is flat (fig. 7.1.). TMOS is receiving direction is relatively low for the three application forces.
|
|
|
|
LR
|
TMOS
|
This is caused by the strong high pass characteristic in receiving direction as shown in fig. 7.2. Background noise is completely transmitted (see fig 7.3).
|
SND
|
|
5.4 dB
|
4.0
|
RCV
|
2N
|
---
|
2.2
|
8N
|
3.4 dB
|
2.4
|
13N
|
---
|
2.8
|
|
|
|
|
Fig. 7.1: Sending
|
Fig. 7.2: Receiving
|
Fig. 7.3: Café (VAD on)
|
|
|
|
The ETSI EG 202 396-3 analyses lead to a relatively balanced performance although some requirements are violated (see quality pie). The near end signal is completely transmitted during double talk but overlapped by some echo components (fig. 7.4, type 2a characterization).
|
|
Fig. 7.4: Near end speech (DT)
|
|
The echo attenuation of 29 dB is too low. The echo is detected as strong, unexpected component in the RA analysis (fig. 7.5). The echo level exceeds the background noise level (fig. 7.6, 7.7).
|
|
|
|
|
|
Fig. 7.5: Echo, Rel. Approach
|
Fig. 7.6: Noise transmission with far end signal
|
Fig. 7.7: RA. Noise transm. with far end speech
|
| 4.1.2Phone No.2 4.1.3Phone No.3
The SLR of 18 dB is significantly too high. However, the listening speech quality is still high, TMOS was determined to 4.0 in sending direction. The sending frequency response shows a strong resonance at 3 kHz (fig. 7.15). Listening speech quality in receiving direction is mainly limited by the strong high pass characteristic (fig. 7.16).
|
|
|
|
LR
|
TMOS
|
Background noise is completely transmitted but quality is impaired by audible comfort noise. The dynamic of the original noise can not be anymore seen in fig. 7.17.
|
SND
|
|
17.7 dB
|
4.0
|
RCV
|
2N
|
---
|
2.3
|
8N
|
2.4 dB
|
2.6
|
13N
|
---
|
2.7
|
|
|
|
|
|
|
Fig. 7.15: Sending
|
Fig. 7.16: Receiving
|
Fig. 7.17: Café (VAD on)
|
|
|
|
The listening speech quality in the presence of background noise (ETSI EG 202 396-3) is limited by the audible comfort noise. This leads to an offset for all scores (see quality pie). The near end signal is completely transmitted during double talk but overlapped by a slight echo (type 2a).
|
|
Fig. 7.18: Near end speech (DT)
|
|
The echo attenuation is sufficiently high (TCLw 50 dB). Echo disturbances can not be expected. The Relative Approach analysis in fig. 7.19 detects a slightly varying noise level. The near end background noise is nearly completely transmitted in fig. 7.20 and 7.21.
|
|
|
|
|
|
Fig. 7.19: Echo, Rel Approach
|
Fig. 7.20: Noise transmission with far end signal
|
Fig. 7.21: RA. Noise transm. with far end speech
|
|
Share with your friends: |
