Download 118 - Telecommunications Industry Association

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Broadcast television systems wikipedia , lookup

UniPro protocol stack wikipedia , lookup

Serial digital interface wikipedia , lookup

Last mile wikipedia , lookup

Public address system wikipedia , lookup

Telecommunications engineering wikipedia , lookup

Telecommunications relay service wikipedia , lookup

Telecommunication wikipedia , lookup

History of telecommunication wikipedia , lookup

Transcript
Telecommunications Industry Association
(TIA)
McLean, VA
TR-30.1/02-09-118
September 10 – 12, 2001
COMMITTEE CONTRIBUTION
Technical Committee TR-30 Meetings
SOURCE:
TR-30 Chair
CONTACT:
Fred Lucas
TITLE:
Correspondence on TTY over VoIP
PROJECT:
PN-4628
DISTRIBUTION:
Members of TR-30.1
____________________
ABSTRACT
This document is correspondence from the Interactive Voice Response Forum to the TTY Forum
describing a problem they have found with the transport of TTY over VoIP. It is provided for
information purposes
COPYRIGHT STATEMENT:
The contributor grants a free, irrevocable license to the Telecommunications Industry Association (TIA) to incorporate text or
other copyrightable material contained in this contribution and any modifications thereof in the creation of a TIA Publication; to
copyright and sell in TIA's name any TIA Publication even though it may include all or portions of this contribution; and at
TIA's sole discretion to permit others to reproduce in whole or in part such contributions or the resulting TIA Publication. This
contributor will also be willing to grant licenses under such copyrights to third parties on reasonable, non-discriminatory terms
and conditions for purpose of practicing a TIA Publication incorporates this contribution.
This document has been prepared by the Source Company(s) to assist the TIA Engineering Committee. It is proposed to the
Committee as a basis for discussion and is not to be construed as a binding proposal on the Source Company(s). The Source
Company(s) specifically reserves the right to amend or modify the material contained herein and nothing herein shall be
construed as conferring or offering licenses or rights with respect to any intellectual property of the Source Company(s) other
than provided in the copyright statement above.
Sponsored by the Alliance for
Telecommunications Industry Solutions (ATIS)
July 15, 2002
1200 G Street, NW, Suite 500
Washington, DC 20005
Tel: +1-202-628-6380
Fax: +1-202-393-5453
www.atis.org
Ed Hall
Chair, TTY Forum
The Alliance for Telecommunications Industry Solutions
1200 G Street, NW
Suite 500
Washington, DC 20005
Subject: Liaison Report from IVR Forum on Voice Over Internet Protocol
Dear Ed,
At the June 5, 2002 meeting of the ATIS-sponsored Interactive Voice Response Accessibility
Forum (IVR Forum), a technical presentation was given regarding Voice Over Internet Protocol
(VoIP) and how it may affect TTY transmissions. Recent advances in this technology have
increased the overall voice quality to the point where some carriers are using it for voice traffic
in the Public Switched Telephone Network, and many other carriers have publicly announced
their plans to do so. The telecommunications industry is moving toward this technology in order
to lower operating costs by minimizing the facilities required to handle a given volume of traffic.
According to the presentation given at the IVR Forum, implementing VoIP may have a negative impact
on TTY traffic over both the wireline and wireless networks. For instance, a VoIP connection with a
packet loss rate of 0.5% (considered excellent) is likely to yield unacceptable TTY error rates. I have
attached an overview of the VoIP issue that explains its affect on TTY transmissions in more detail. This
overview was developed by Paul Michaelis from Avaya. The IVR Forum brings this to your attention
knowing that the TTY Forum members have just completed major network changes to accommodate
TTY. This is sent as information to insure that the TTY Forum is cognizant of this pending issue. The
IVR Forum has been made aware that the Network Reliability and Interoperability Council (NRIC) Focus
Group 3 will be addressing this issue.
I am available at you convenience if you have questions or are in need of further discussion on
this topic.
Respectfully,
Ken Evans
Vice Chairman – IVR Forum
Cingular Wireless
404-713-8888
cc:
Megan Hayes, TTY Forum Secretariat
Jim Tobias, IVR Forum Chair
Developed by Paul Michaelis, Avaya
Transmission of TTY Tones on Voice Over Internet Protocol
(VoIP) Systems
The purpose of this document is to summarize for non-engineers some of the technical
challenges associated with using Voice over Internet Protocol (VoIP) mechanisms to transport
the “Baudot tones” that are generated by TTYs. But first, an important disclaimer: Industry
standards, such as T.140, describe mechanisms for providing reliable real-time text conversations
on Internet Protocol systems. This document does not address those standards. Again, the issue
being addressed is the use of voice channels to transport TTY tones.
In the United States, the most commonly used TTY communication standard is described by
ANSI/TIA/EIA 825 (“A 45.45 Baud FSK Modem"). With regard to VoIP-related issues, the
important aspects of this standard are:

TTYs are silent when not transmitting.
 Unlike fax machines and computer modems, TTYs have no “handshake” procedure at the
start of a call, nor do they have a carrier tone during the call.
 Although this approach tends to limit the speed of transmission, it has the advantage of
permitting TTY tones, DTMF (Dual Tone Multi-Frequency signals, also known as “touch
tones”), and voice to be intermixed on the same call.

Operation is “half duplex.”
 TTY users must take turns transmitting, and typically cannot interrupt each other. If both
people try to type at the same time, their TTYs will show no text at all, or will show text
that is gibberish.
 There is no automatic mechanism that lets TTY users know when a character they have
typed correctly has been received incorrectly.

Each TTY character consists of a sequence of seven individual tones.
 The first tone is always a “start tone” at 1800 Hz.
 This is followed by a series of five tones, at either 1400 or 1800 Hz, which specify the
character.
 The final tone in the sequence is always a “stop tone” at 1400 Hz. The “stop tone” is a
border that separates this character from the next.
 Each of the first six tones is 22 milliseconds in duration. The final “stop tone” may also
be 22 milliseconds, but is permitted to be as long as 44 milliseconds. This means that the
duration of each TTY character is at least 154 milliseconds, which works out to
approximately six and a half characters per second. (The description of this as a “45.45
Baud” protocol is based on the number of 22-millisecond tones that can be transmitted in
one second, not the number of characters.)
Many of the techniques that are commonly employed by telephone systems to digitize audio
signals (such as voice) are able to digitize TTY tones with perfect accuracy. Unfortunately,
some techniques that are optimized for low-bit-rate encoding of voice signals tend to distort TTY
tones. An example of the former is the G.711 encoding that is available on most VoIP systems.
An example of the latter is the GSM encoding used on many wireless telephones. The
discussion of VoIP below assumes that a TTY-compatible technique, such as G.711, is being
employed. The essential point of the discussion is that it’s the transmission techniques, not the
audio encoding techniques, that are the primary barrier to TTY compatibility in VoIP systems.
Looking now at VoIP telephony, it’s important to understand a key factor that distinguishes
VoIP systems from traditional approaches. Traditional telephone systems, both analog and
digital, are often described as circuit switched. For the purposes of this discussion, this
mechanism may be envisioned as follows: when two people engage in a telephone conversation,
a circuit switched telephone network establishes a direct circuit between them – in essence, their
own dedicated set of wires. By contrast, VoIP systems route calls through a mechanism called
packet switching.
Specifically, VoIP systems transmit audio streams, such as voice and TTY tones, by digitally
encoding the audio and then breaking the streams into individual packets. (A typical packet
contains a 20-millisecond stream of audio, although packets of other lengths may also be
employed.) Each of these audio packets is tagged with header information, such as the audio
encoding scheme that was used, a sequence number, and the destination’s IP address. The
complete packet is then delivered by the originating device to the network, which transports the
packet via shared pathways that often contain packets from many different sources, with many
different destinations.
Here is a very important point: although the destination is specified in the packet header
information, the route to the destination is not specified. The ability for each packet to take what
is, at that instance, the “best” route to its destination is where VoIP derives a lot of its economic
advantage. It’s also the reason why TTY-on-VoIP can be unreliable: because packets are free to
take different pathways, they cannot be relied upon to arrive at the receiving device before it’s
their “turn” to be played. Although these packets often arrive eventually, they are regarded as
lost because they did not arrive in time, and must therefore be discarded. (Although it’s
acceptable for the downloading of an Internet web page to pause momentarily, waiting for a
packet to be delivered, it’s obvious that pauses of this sort would be intolerable in a telephone
conversation.)
Under most circumstances, the loss of occasional packets is not detectable in voice
communication. Although 20-millisecond periods of silence would certainly be noticeable in a
voice stream (sounding a bit like static), VoIP telephones employ packet loss concealment
algorithms that trick the human ear, typically by mimicking the adjacent packets that have been
received. Although these techniques work well with voice, they do not work with TTY tones. If
a packet containing a TTY tone is lost, the current generation of VoIP techniques is unable to
recover it or rebuild it.
With regard to the percentage of packets that one might expect to lose, it is generally the case
that packet loss of 0.2% or less is achievable when the two VoIP endpoints are on the same
campus, using communication pathways that are not congested. By contrast, for VoIP calls that
originate or terminate “off campus” – in other words, for calls in which there is a wider range of
packet routing possibilities – or for VoIP calls that are transported on congested networks, packet
loss of 2.0% or higher is typical.
What is the impact of packet loss on TTY? Consider the following illustrative example: assume
that the VoIP packet size is 20 milliseconds (a typical value) and that the packet loss rate is 0.5%
(a rate generally regarded as excellent for VoIP communication). Keep in mind that an
individual TTY text character is at least 154 milliseconds in length, and therefore spans eight
packets. This means that, if there is a 0.5% likelihood that any one of those packets is missing,
approximately four percent of all TTY characters will lose one of their packets. If any one of the
eight packets within a character is lost, that character will not be displayed properly on the
receiving device.
Even though the simple statistical model above would seem to predict a four percent TTY error
rate under the described conditions (20-millisecond packet size, 0.5% packet loss rate), the actual
error rate would tend to be much higher. This is because, if the lost packet is the one that
contained the “stop tone” for that character, subsequent characters, even if transmitted without
packet loss, might nevertheless be decoded improperly.
As a point of comparison, a TTY character error rate of more than one percent is generally
regarded as unacceptable, chiefly because the transmission of information such as bank balances
and credit card numbers becomes unreliable (see, for example,
http://www.atis.org/pub/ttyforum/tty19-01-09-26-11.pdf or http://www.accessboard.gov/sec508/commrept/eitaacrpt.htm). Using a simple statistical model that is based on a
20-millisecond packet size, and ignoring the additional deleterious effects that result from
dropping a “stop tone,” the one percent character error rate threshold is exceeded when VoIP
packet loss rates exceed approximately 0.12% – a packet loss rate generally regarded as
unachievable in standard VoIP systems.