Tackling Delay in Today's Global Voice
There are three very important components of voice quality in todayï¿½s global
telecommunications network: clarity, echo, and delay.
These attributes must be understood and managed in order to maintain voice
quality in the Public Switched Telephone Network (PSTN). While many
consumers take voice quality for granted even in telephone conversations
halfway around the globe, the network has evolved considerably over time to
ensure this level of service.
The understanding of the mechanics of delay and the innovations built into
the PSTN that has evolved to make the voice network delay-free is a
tremendous engineering feat that makes long-distance telecommunications a
monumental achievement. The global community developed strict standards to
help overcome these technology hurdles and todayï¿½s networks are as efficient
and as free of delay as they have ever been.
The latest technology introductions such as voice over Internet Protocol (VoIP),
Wireless Local Loop, and cellular phone traffic are offering new challenges
for todayï¿½s voice networks. While these new technologies offer compelling
capabilities that will ultimately usher in a new era in telecommunications,
the long established standards governing minimum delays in communications
shouldnï¿½t be ignored if voice carriers are to keep customer satisfaction as
There were several good reasons why specifications governing network delay
were put in place; however technology is changing so rapidly that there is a
distinct possibility that unregulated technologies will undermine these
specifications. As in the past though, this rapid integration of new
products is completely manageable. Letï¿½s look at some solutions that can
avoid these delays.
EARLY NETWORKS: A HISTORICAL
In early telephone networks, it was a challenge to carry the voice far
enough to experience delay and any noticeable echo. After all, a fraction of
the speed of light was incomprehensively fast. When the standards in use
today were developed, the challenges for preserving voice clarity had
already been resolved using digitally sampled voice.
Digital processing of the voice requires sampling, storage, and switching,
which leads to delays. To manage this potential for delays, strict
specifications were introduced to avoid eroding the delay quality of the
voice network. From this viewpoint, the shape and nature of the voice
network has evolved. But the underlying specifications have not changed
much, and these specs can be tough to meet as equipment densities increase.
VoIP, Mobile Wireless, and Wireless Local Loop networks can add significant
amounts of delay, so itï¿½s important to monitor the delay efficiency of
circuit-switched TDM networks if it is to support these new services.
In the global voice networks, voice quality is key to retaining customers.
In the case of wireless networks, itï¿½s not just key to keeping customers but
also to keeping them on the phone talking and generating revenue.
EFFECTS OF DELAY
The effects of delay on voice traffic can be severe, which is why the
governing requirements are so strict.
Having good voice quality means that you can have a telephone conversation
without any distracting or annoying effects ï¿½ it really has to be just as
clear and spontaneous as if the other person is right there in front of you.
This way you know that any pauses or silences have meaning.
Delay and echo can be more frustrating than noise and may be present even on
the clearest of lines. This is because delay and echo interrupt the rhythm
of speech, which interrupts the flow of thought and makes it difficult to
maintain the conversation.
Poor voice quality caused by echo and delay can lead to frustration,
stressful conversations, and create customer churn. With pay-by-the-minute
services like mobile and long distance, the business model usually requires
frequent use to cover the fixed costs. A good quality line can mean much
longer conversations and a more profitable service.
NUMBERS TELL THE STORY
Delay is the time it takes for the listener to hear the talkerï¿½s voice. Any
one-way delay less than 100 milliseconds (ms) is not noticeable for most
conversations and does not affect the flow of conversation. Though, highly
interactive business tasks may not tolerate more than 50 ms of one-way
If delay times are longer than 150 ms, it becomes difficult to know whose
turn it is to talk. This adds a level of discomfort or even stress to the
conversation. People are more likely to interrupt each other and certainly
more likely to cut their call short.
After about 200ï¿½250 ms of delay, any interactive conversation needs
perseverance; people are forced into an awkward, half-duplex conversation,
much like talking on a one-way radio.
The negative effects of delay and echo are regulated in the PSTN to ensure
high-quality voice connections. An accepted standard is 100 ms round-trip on
any national network.
For the international leg of a long distance call, another 100 ms is allowed
between any two countries. That provides a total delay budget for
international calls of 300 ms round-trip, 100 ms for each country and 100 ms
for the international trunk. A round trip total of 300 ms or 150 ms one-way
delay is an acceptable limit for a high-quality link. Beyond this, voice
quality and customer satisfaction drops off sharply.
An efficient transport network able to carry voice with minimal delay will
be increasingly valuable when new technologies such as VoIP, mobile
wireless, or wireless local loop become more common.
Examples of this traffic are now being delivered into the PSTN, but they are
not always capable of meeting the strict PSTN delay standards. These
services add enough value to offset some small reduction in overall voice
quality, but problems may yet occur when too many sources of delay are added
to the same voice call.
VoIP COMES INTO PLAY
VoIP is an exploding technology with attractive features driving its
Unlike the PSTN circuit-switched network, VoIP relies on bypassing the
regulated PSTN and sending voice over the Internet. VoIP conventions for
delay are a little bit more relaxed than the PSTN standards. An upper limit
of 250 ms one-way, that some would argue is acceptable, will seem like a
poor quality connection to a PSTN user.
The delay is primarily caused by the encoding, packetizing, queuing,
switching, routing, and buffering in the VoIP equipment. A one-way delay of
150ï¿½200 ms is generally considered to be a reasonable target. In this range,
the overall quality should be acceptable to most people.
Delay occurs when the voice signal is encoded. This can typically range from
eightï¿½20 ms. As the processing gets more complex with higher compression
rates, this delay time grows. When encoding, thereï¿½s a tradeoff between
bandwidth and delay. Historically, VoIP network equipment has been optimized
for bandwidth but now bandwidth is cheaper. With VoIP, delay is likely to
have the biggest affect on the user. Most VoIP equipment can minimize delay
by increasing the bandwidth allocated to each voice channel.
Packetization delay is the time it takes to fill a packet with compressed
speech and usually falls in the range of fiveï¿½50 ms. There is a tradeoff
between packetization delay and the load on the processor. Using short
packets over high-speed trunks can easily shorten the delay, but because
each packet requires overhead, more packets mean lower network efficiency.
With the recent activity from Independent Local Exchange Carriers (ILECs)
and International Exchange Carrier (IXCs), the presumption is that bandwidth
preservation has become a key motivator; they may not be motivated to
minimize delay at the expense of bandwidth.
Queuing delay occurs as packets wait in router queues and typically range
from sevenï¿½10 ms. Many modern routers can be configured to give priority to
voice traffic, but this delay is still a concern on heavily congested
networks that mix voice and data traffic. Not all routers will be able to
differentiate voice from data, especially in the Internet.
Because of the nature of the packet network, not all packets take the same
time or even the same path to the destination. This means the time of their
arrival varies, so data needs to be stored in a First In, First Out device
(FIFO) prior to decoding. The maximum delay in a jitter buffer is
configurable but is typically set between 40 and 60 ms.
Many of the delays in a VoIP call are a product of the codecs, bandwidth,
packet size and quality and can be dynamically adjusted to suit quality of
service (QoS) requirements. Some delays are the product of the components
and traffic flowing in the Internet itself. These delays can be queuing
delays in routers, head of line blocking when mixed with big data packets,
One way to avoid these delays is not to use a shared network. Ethernet over
Synchronous Optical Network (SONET) provides another option that is
currently becoming more popular with network managers. This bypasses the
shared ï¿½Internet cloudï¿½ and can remove ambiguity in delay caused by other
traffic by using the circuit switched network. It cannot eliminate
propogation delay but may assist in keeping a delay budget down to
The global voice network is becoming more and more complex. Today, a variety
of transport and coding technologies may be carrying your voice across the
network and each time the format is changed, delay will be introduced. We
must keep this network efficient if we are to preserve the high voice
quality levels we have today. When designing new equipment and new networks,
delay should be one of the more important considerations.
Voice today generates a huge amount of revenue for our industry. It is going
to continue to attract investment and new entrants. Recent moves flirting
with deregulation of the telephone networks highlight the possibilities for
rapid changes in technologies and methods. Some caution must be applied if
the voice networks are to remain the high-quality service they are today.
These new services will no doubt be deployed, but the underlying quality of
voice connection will determine the adoption rate and long-term success of
these new services. Managing network delay along with the other aspects of
voice quality will go a long way to securing these new revenue streams.
Gordon Oliver is currently responsible for SONET/ SDH PDH interworking
solutions including the TEMAP/OCTLIU (T1/E1 Transport, Test Access and
Transmultiplexer) product families at
PMC-Sierra. Prior to joining PMC-Sierra, he was a Hardware Team Leader
at NEC Australiaï¿½s Multimedia Research Laboratory developing ATM PON-based
If you are interested in purchasing reprints of this article (in either
print or HTML format), please visit Reprint Management Services online at
www.reprintbuyer.com or contact a representative via e-mail at
or by phone at 800-290-5460.
To The July 2004 Table Of Contents ]