How Do I Minimize Delay Induced
Echo?
Delay in a network originates from a number of different sources and
phenomena.
A primary source of delay is the process of converting speech to VoIP
traffic. The IP Office supports a number of standards based encoding
methods to allow the optimum trade off between quality and bandwidth
to be made
IP Office incorporates integral echo cancellation to minimize the effect
of echo introduced in the VoIP conversion process.
Another source of delay comes from data and voice traffic queuing at
the ports of switches, routers, gateways and or bridges that make up
the network. It is possible that the traffic queuing at a port is minimal
and no action needs to be taken. This would be the case if the available
bandwidth far exceeded the demand. To overcome queuing bottlenecks in
the network, IP Office prioritizes voice traffic using a standard known
as DiffServ. This marks each IP packet carrying voice with a flag so
that routers, etc. can force packets containing voice to the front of
the transmission queue. An alternative method of prioritization that
can be used by switches and routers, with an equally satisfactory result,
is to look at what protocol is being used and prioritize this. All voice
traffic is carried using two easily identifiable protocols, RTP and
RTCP. Both methods are equally good, choose whichever method is the
most cost effective and easiest to implement and manage.
A similar source of delay
can be attributed to specific network nodes that convert from one network
medium to another. For example T1 trunk lines may be carried across
a high speed DSL like connection and converting from the high speed
link back to T1 in the access gateway takes time to perform. Any VoIP
traffic being carried through this link is therefore subject to the
delay introduced by this conversion step. The delay may be minimized
by ensuring that an appropriate QoS mechanism is enabled in the gateway
to prioritize the VoIP traffic. IP Office incorporates integral echo
cancellation to help minimize the effect of this kind of delay introduced
by the network.
Delay can also be introduced as a consequent of collisions occurring
on particular segments of the LAN. Collisions result when two devices
on a shared switch port or segment try to transmit simultaneously. This
causes all devices to stop transmitting for a period of time. This is
a normal characteristic of many older Ethernet networks and, if occasional,
may pass unnoticed. The more devices sharing a switch port, and the
busier they are, the greater the opportunity for collisions. This is
simply resolved by reducing the number of devices on each port, or by
dedicating a port to each VoIP device. If you are just using VoIP to
link two IP Offices together, it’s well worth dedicating a port to each
IP Office and router at either end of the link as the cost implications
are likely to be minimal. In this regard it is important to dimension
a network to cope with existing traffic demands as well as any future
increases in traffic carrying capability.
How Do I Minimize Warble and
Clipping?
Warble, clipping and some distortion quality problems are symptoms
of variable delay and or packet loss.
Variability in the delays of traffic is called jitter. Jitter and packet
loss may be the result of switches and routers that are either faulty
or working outside their design intentions.
IP Office provides jitter buffers to compensate for a moderate amount
of jitter found in networks. Voice traffic is quite tolerant of small
amounts of packet loss so in most cases this may be ignored. Where packet
loss is excessive (greater than 2% say) the cause should be established
and fixed. This could be due to a fault or simply an over worked device
discarding packets. Significant packet loss can cause perceptible losses
in speech, to the extent that no speech may be heard either in one or
both directions.
How Do I Minimize Distortion?
Each time speech is converted into a digital signal and back again,
tiny difference from the original creeps in. The more times this happens
on a single call, the bigger those differences can become. These differences
can become perceptible as distortion. Ideally, the path speech takes
should only require one ‘analog to digital to analog’ conversion and
this will be the case in many instances. Exceptions to this occur when
making calls to mobile telephones or voice mail systems where the analog
to digital to analog conversion may occur twice (once on IP Office and
once on the mobile network, etc).
Different encoding methods will have different effects. IP Office supports
a range of encoding methods to allow you to choose the one with the
right quality versus bandwidth for your network. In general multiple
conversions should be minimized wherever possible.
What Delay
is Acceptable?
Try to keep the overall end-to-end delay to 150 milliseconds or below.
An idea of the delay inherent in the network can be measured by carrying
out a ping test and dividing the result by two. IP Office has built
in echo cancellation to maximize speech quality.
|
