![](bg5.png)
in this scenario is in the results of the application
packet loss parameter.
In this scenario, CVRV shows a bigger
application packet loss than CRV in the range of
0-50µs (delay added by the play-out buffer). In this
range, test cases utilizing CVRV lose more packets
than the corresponding test cases using CRV.
It turns out that this phenomenon is due to the
length of the packets. Longer packets travel for a
longer time to the destination in comparison to
shorter ones. This network delay adds to the overall
delay of the packet. For delays longer than 50µs,
there is no difference between the two vocoders.
This is explained by the non-significant contribution
of the network delay when added up to the actual
delay, on top of the play-out buffer.
For demonstration: the length of an average
packet is 82B, the length of the longest packet is
104B, therefore the overhead in travel time for
packets longer than the average would be maximum
17.6µs, as per the following calculation:
When these 17.6µs of delay are added on top of
the 0-50µs delay of the play-out buffer, the addition
of the packet length is significant and influences the
results of the total application packet loss parameter.
The results for the application packet loss
parameter show that for the test case of 500
terminals, the difference in the performance of CRV
and CVRV is less significant compared to the other
test cases (100-400). This is explained by the fact
that the test case with 500 terminals suffered from a
much longer delay and jitter relatively to 100-400
terminals. This causes the effect of the network
delay (caused by packet length) to be less significant
(Queue Delay >> Network Delay), affecting less the
difference in packet loss.
According to the above, when the play-out
buffer for CVRV is designed for the receiver’s end,
the delay added will be according to the maximum’s
packet length, rather than the average packet length,
as in CRV. In order to achieve the same packet loss
rate in CVRV and CRV, the delay added in the play-
out buffer of CVRV will need to be longer,
increasing the total delay time of the packets.
6.2 Voice+Data Traffic over LAN
The scenario of only voice traffic is an isolated case.
The more common situation is a LAN where both
voice and data are transmitted. Scenario 2 that is
investigated here is designed to support both voice
and data traffic. The ratio of voice vs. data terminals
is 1:1. Every data terminal has a ready packet to be
transmitted. The voice terminals perform according
to the pre-designation of CRV and CVRV vocoder,
respectively. The measurements and statistical
studies are applied to the voice packets only. The
scenario is carried out repeatedly with an increasing
number of terminals.
Following we detail the results of the QoS
parameters for this test case.
6.2.1 Delay, Jitter, Packet Loss and
Application Packet Loss
The results for delay, jitter, packet loss and
application packet loss show that when both voice
and data are supplied to CRV and CVRV vocoders,
no significant difference is manifested by the two
vocoder types. These parameters: delay, jitter and
packet loss, increase correspondingly to the number
of added terminals.
6.2.2 Summary and Discussion
In this scenario, of transmission of voice and data,
we have seen no significant difference between the
two vocoders. This is explained below.
The comparison between the vocoders was
designed in such a way that the surrounding
environment and its features are as close as possible.
In this scenario that mixes voice and data, the only
difference between CRV/CVRV test cases is the
length of the voice packets. The data packets are
always ready to be transmitted and a voice packet is
ready every 64ms. Consequently, the packets are
transmitted at the exact same times in both vocoders.
According to the behaviour of the LAN, when
the medium is free, a packet is transmitted.
Collisions occur only when more than one terminal
senses the medium as free, and transmits a packet.
The packets transmitted from multiple terminals
simultaneously collide and will need to be
retransmitted. It takes the transmitting terminal a
constant period of time to notice that the packet it
has sent is corrupted due to collision (twice the
propagation time). According to the test case setup,
the packets in CRV and CVRV scenarios are sent at
the same time, and the collisions occur at the same
times, respectively. The identification, in a constant
time, of a collision eliminates the difference between
the voice packet lengths, controlling the scenario
EFFECTS OF VARIABLE BIT RATE VOCODER ON VOIP QOS
97