Network Working Group H. Schulzrinne Request for Comments: 4734 Columbia U. Obsoletes: 2833 T. Taylor Updates: 4733 Nortel Category: Standards Track December 2006 Definition of Events for Modem, Fax, and Text Telephony Signals Status of This Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited. Copyright Notice Copyright (C) The IETF Trust (2006).
AbstractThis memo updates RFC 4733 to add event codes for modem, fax, and text telephony signals when carried in the telephony event RTP payload. It supersedes the assignment of event codes for this purpose in RFC 2833, and therefore obsoletes that part of RFC 2833.
1. Introduction ....................................................3 1.1. Terminology ................................................3 1.2. Overview ...................................................3 2. Definitions of Events for Control of Data, Fax, and Text Telephony Sessions ..............................................5 2.1. V.8 bis Events .............................................5 2.1.1. Handling of Congestion ..............................9 2.2. V.21 Events ...............................................10 2.2.1. Handling of Congestion .............................11 2.3. V.8 Events ................................................12 2.3.1. Handling of Congestion .............................15 2.4. V.25 Events ...............................................15 2.4.1. Handling of Congestion .............................17 2.5. V.32/V.32bis Events .......................................18 2.5.1. Handling of Congestion .............................19 2.6. T.30 Events ...............................................19 2.6.1. Handling of Congestion .............................23 2.7. Events for Text Telephony .................................23 2.7.1. Signal Format Indicators for Text Telephony ........23 2.7.2. Use of Events with V.18 Modems .....................27 2.8. A Generic Indicator .......................................28 3. Strategies for Handling Fax and Modem Signals ..................29 4. Example of V.8 Negotiation .....................................30 4.1. Simultaneous Transmission of Events and Retransmitted Events Using RFC 2198 Redundancy ............35 4.2. Simultaneous Transmission of Events and Voice-Band Data Using RFC 2198 Redundancy ............................37 5. Security Considerations ........................................39 6. IANA Considerations ............................................40 7. Acknowledgements ...............................................42 8. References .....................................................43 8.1. Normative References ......................................43 8.2. Informative References ....................................44
RFC 2119 . In addition to those defined for specific events, this document uses the following abbreviations: Fax facsimile HDLC High-level Data Link Control PSTN Public Switched (circuit) Telephone Network RFC 4733  to include the control events and tones that can appear on a subscriber line serving a fax machine, a modem, or a text telephony device. The events are organized into several groups, corresponding to the ITU-T Recommendation in which they are defined. Their purpose is to support negotiation, start-up and takedown of fax, modem, or text telephony sessions and transitions between operating modes. The actual fax, modem, and text payload is typically carried by other payload types (e.g., V.150.1  modem relay, voice-band data as formalized in ITU-T Rec. V.152 , Clearmode  for digital data, T.38  for fax, or RFC 4103  for character-mode text). NOTE: implementers SHOULD NOT rely on the descriptions of the various modem protocols described below without consulting the original references (generally ITU-T Recommendations). The descriptions are provided in this document to give a context for the use of the events defined here. They frequently omit important details needed for implementation. The typical application of these events is to allow the Internet to serve as a bridge between terminals operating on the PSTN. This application is characterized as follows: o each gateway will act both as sender and as receiver; o time constraints apply to the exchange of signals, making the early identification and reporting of events desirable so that receiver playout can proceed in a timely fashion;
o the receiver must play out events in their proper order; o transfer of the events must be reliable. Applications will vary in their ability to recover from missing events. In some cases, an implementation may simply ignore certain events, such as fax tones, that do not make sense in a particular environment. Section 2.4.1 of RFC 4733  specifies how an implementation can use the Session Description Protocol (SDP) "fmtp" parameter within an SDP description  to indicate which events it is prepared to handle. Regardless of which events they support, implementations MUST be prepared to send and receive data signals using payload types other than telephone-event, simultaneously with the use of the latter. This is discussed further in Section 3. In many cases, continuity of playout is critical. In principle, this is achieved through buffering at the receiving end. It is generally desirable to minimize such buffering to reduce round-trip response times. Maintenance of a constant packetization interval at the sending end while reporting events is helpful for this purpose. A further word on time constraints is in order. Time constraints governing the duration of tones do not pose a problem when using the telephone-event payload type: the payload specifies the duration and the receiving gateway can play out the tones accordingly. Problems occur when time constraints are specified for the duration of silence between tones. A silent period of "at least x ms" is not a problem -- event notifications can be received late, but they can still be played out at their specified durations. The problem occurs if silence must last for a specific duration or at most some specific period. The most general constraint of the latter type has to do with the operation of echo suppressors (ITU-T Rec. G.164 ) and echo cancellers (ITU-T Rec. G.165 ). These devices may re-activate after as little as 100 ms of no signal on the line. As a result, in any situation where echo suppressors or cancellers must be disabled for signalling to work, tone events must be reported quickly enough to ensure that these devices do not become re-enabled.
10] is a general procedure for two endpoints to establish each other's capabilities and to transition between different operating modes, both at call startup and after the call has been established. It supports many of the same terminals as V.8  (Section 2.3 below), but allows more detailed parameter negotiation. It lacks support for some of the older V-series modems defined in V.8, but adds capabilities for simultaneous or alternating voice and data, H.324  multilink, and T.120  conferencing. Following V.8 bis capability negotiations, if the terminals have negotiated a modem-based operating mode, they initiate the actual modem session using either V.8, a truncated version of V.8 (preferred), or V.25 start-up. V.25 is described in Section 2.4. V.8 bis distinguishes between "signals" and "messages". The V.8 bis signals -- ESi/ESr, MRe/MRd, and CRe/CRd -- consist of tones, as described in the next few paragraphs. The V.8 bis messages -- MS, CL, CLR, ACK(1), ACK(2), NAK(1), NAK(2), NACK(3), and NACK(4) -- consist of sequences of bits transported over V.21  modulation. Signals are intended to be comprehensible at the receiver even in the presence of voice content. They consist of two tone segments. The first segment consists of a dual-frequency tone held for 400 ms, and has the function of preparing the receiver and any in-line echo suppressor or canceller for what follows. The specific frequencies depend only on whether the signal is from the initiator or the responder in a transaction. When using the telephone-event payload, the V8bISeg and V8bRSeg events in Table 1 represent the first segment of any V.8 bis signal in the initiating and responding case, respectively. The complete V.8 bis strategy for dealing with echo suppressors or cancellers is described in Rec. V.8 bis Appendix III. The only silent period constraints imposed are of the "at least" type, posing no difficulties for the use of the telephone-event payload. The second segment follows immediately after the first, and is a single tone held for 100 ms. The frequency used indicates the specific signal of the six signals defined. When using the telephone-event payload, the second segment of a V.8 bis signal is represented by the applicable event: CRdSeg, CReSeg, MRdSeg, MReSeg,
ESiSeg, or ESrSeg, as defined in Table 1. ESiSeg and ESrSeg use the same frequencies as V.21 low and high channel '1' bits, respectively (see Table 2), and are therefore assigned the same event codes. V.8 bis messages use V.21  frequency-shift signalling to transfer message content. V.21 is described in the next section. V.8 bis uses V.21 in half-duplex mode at 300 bits/s, with the lower channel assigned to the initiator and the upper channel to the responder. Each V.8 bis message is preceded by a 100-ms preamble of continuous V.21 marking frequency except if it was immediately preceded by an ESi or ESr signal (the second segment of which is that same V.21 marking frequency). The sender SHALL NOT report this preamble tone using the ESiSeg or ESrSeg events; these are to be used only for the V.8 bis signals to which they pertain. Spelling this out, continuous V.21 marking tone immediately following V8bISeg and V8bRSeg is reported as ESiSeg or ESrSeg, respectively. Continuous V.21 marking tone occurring in any other context, and particularly after CRdSeg, CReSeg, MRdSeg, or MReSeg, is reported by other means such as a different payload type or using the V.21 '1' bit events defined in Section 2.2. No events are defined for V.8 bis messages, but a brief description follows. o the V.8 bis CL message describes the sending terminal's capabilities; o the CLR message also describes capabilities, but indicates that the sender wants to receive a CL in return; o the MS establishes a particular operating mode; o the ACK and NAK messages are used to terminate the message transactions. The V.8 bis messages are organized as a sequence of octets. The first two to five octets are HDLC flags (0x7E). Then comes a message type identifier (four bits), a V.8 bis version identifier (four bits), zero to two more octets of identifying information, followed by zero or more information field parameters in the form of bit maps. An individual bit map is one to five octets in length. Up to 64 octets of non-standard information may also be present. The information fields are followed by a checksum and one to three HDLC flags. Because of limits on the size of any one information field, V.8 bis defines segmentation procedures. Excess data is sent in an additional message, but only after prompting from the receiving end.
Applications supporting V.8 bis signalling using the telephone-event payload MAY transfer V.8 bis messages in the form of sequences of bits, using the V.21 bit events defined in the next section. If they do so, the transmitted information MUST include the complete contents of the message: the initial HDLC flags, the information field, the checksum, and the terminating HDLC flags. Transmission MUST also include the extra '0' bits added according to the procedures of Rec. V.8 bis, clause 7.2.8, to prevent false recognition of HDLC flags at the receiver. Implementers should note that these extra '0' bits mean that in general V.8 bis messages as transmitted on the wire will not come out to an even multiple of octets. Sending implementations MAY choose to vary the packetization interval to include exactly one octet of information plus any extra '0' bits inserted into that octet; the resulting variation will be insignificant compared with the amount of buffering required to guard against network delays in delivery of packets to the receiver (see below). One reason for reporting the V.21 bits exactly as presented on the wire is to match the corresponding content if it is also carried by other means, such as voice-band data. The power levels of the V.8 bis and V.21 signals are subject to national regulation. Thus, it seems suitable to model V.8 bis events as tones for which the volumes SHOULD be specified by the sender. If the receiver is rendering the V.8 bis tones as audio content for onward transmission, the receiver MAY use the volumes contained in the event reports, or MAY modify the volumes to match downstream national requirements. Table 1 summarizes the event codes defined for V.8 bis signalling in this document. The individual events are described following the table. Each event begins when the beginning of the tone segment is detected and ends when the tone is no longer detected.
+---------+-------------+-----------+------------+------+---------+ | Event | Freq. (Hz) | Dur. (ms) | Event Code | Type | Volume? | +---------+-------------+-----------+------------+------+---------+ | ESiSeg | 980 | 100 | 38 | tone | yes | | | | | | | | | ESrSeg | 1650 | 100 | 40 | tone | yes | | | | | | | | | CRdSeg | 1900 | 100 | 23 | tone | yes | | | | | | | | | CReSeg | 400 | 100 | 24 | tone | yes | | | | | | | | | MRdSeg | 1150 | 100 | 25 | tone | yes | | | | | | | | | MReSeg | 650 | 100 | 26 | tone | yes | | | | | | | | | V8bISeg | 1375 + 2002 | 400 | 28 | tone | yes | | | | | | | | | V8bRSeg | 1529 + 2225 | 400 | 29 | tone | yes | +---------+-------------+-----------+------------+------+---------+ Table 1: Events for V.8 bis Signals ESiSeg: The second segment of a V.8 bis initiating Escape Signal (ESi). The complete ESi signal is represented by events V8bISeg followed by ESiSeg. ESi will be followed by an MS, CL, or CLR message from the same terminal. A 1.5-s silent interval may come between the ESi signal and the transmission of the MS, CL, or CLR message to accommodate network echo suppressors. ESrSeg: The second segment of a V.8 bis responding Escape Signal (ESr). The complete ESr signal is represented by events V8bRSeg followed by ESrSeg. ESr is always sent by the calling terminal in response to an MRe or CRe from an automatic answering station. It will be followed by an MS, CL, or CLR message. The ESr signal turns off any announcement being generated by the automatic answering station. CRdSeg: The second segment of a V.8 bis Capabilities Request signal (CRd). The first segment of the CRd signal is represented either by V8bISeg or V8bRSeg, depending on context. The other end will return a capabilities list (CL or CLR message).
CReSeg: The second segment of a V.8 bis Capabilities Request signal (CRe) initiated by an automatic answering terminal. The complete CRe signal is represented by events V8bISeg followed by CReSeg. The calling terminal will respond with a CRd signal or a CL or CLR message. MRdSeg: The second segment of a V.8 bis Mode Request signal (MRd). The first segment of the MRd signal is represented either by V8bISeg or V8bRSeg, depending on context. The other end will return a CRd signal or an MS message. MReSeg: The second segment of a V.8 bis Mode Request signal (MRe) initiated by an automatic answering terminal. The complete MRe signal is represented by events V8bISeg followed by MReSeg. The calling terminal will respond with an MRd or CRd signal or an MS message. V8bISeg: The first segment of an initiating V.8 bis signal, which may be one of ESi, CRd, CRe, MRd, or MRe. V8bRSeg: The first segment of a responding V.8 bis signal, which may be one of ESr, CRd, or MRd.
12] is a modem protocol offering data transmission at a maximum rate of 300 bits/s. Two channels are defined, supporting full duplex data transmission if required. The low channel uses frequencies 980 Hz for '1' (mark) and 1180 Hz for '0' (space); the high channel uses frequencies 1650 Hz for '1' and 1850 Hz for '0'. The modem can operate synchronously or asynchronously. V.21 is used by other protocols (e.g., V.8 bis, V.18, T.30) for transmission of control data, and is also used in its own right between text terminals. The V.21 events are summarized in Table 2. Sending implementations SHOULD report a completed event for every bit transmitted (i.e., rather than at transitions between '0' and '1'). Bit events are assumed to begin and end with the clock interval for the event, neglecting the rise and fall times between bit transitions. Thus, it is important for a gateway to determine the actual bit rate in use before beginning to report V.21 events. Sometimes determination of the bit rate is not immediately possible, as in the case of the 100-ms training signal at V.21 mark frequency used before V.8 bis messages. Transmission of a single longer-duration V.21 event is reasonable under these circumstances and should not cause any difficulties at the receiving end. Implementations SHOULD pack multiple events into one packet, using the procedures of Section 188.8.131.52 of RFC 4733 . Eight to ten bits is a reasonable packetization interval. Reliable transmission of V.21 events is important, to prevent data corruption. Reporting an event per bit rather than per transition increases reporting redundancy and thus reporting reliability, since each event completion is transmitted three times as described in Section 184.108.40.206 of RFC 4733 . To reduce the number of packets required for reporting, implementations SHOULD carry the retransmitted events using RFC 2198  redundancy encoding. This is illustrated in the example in Section 4.1. The time to transmit one V.21 bit at the nominal rate of 300 bits/s is 3.33 ms, or 26.67 timestamp units at the default 8000-Hz sampling rate for the telephone-event payload type. Because this duration is not an integral number of timestamp units, accurate reporting of the beginning of the event and the event duration is impossible. Sending gateways SHOULD round V.21 event starting times to the nearest whole timestamp unit.
When sending multiple consecutive V.21 events in a succession of packets, the sending gateway MUST ensure that individual event durations reported do not cause the last event of one packet to overlap with the first event of the next, taking into account the respective initial event timestamps. To accomplish this, the sending gateway MUST derive the individual event durations as the succession of differences between the event starting times (so that, at 8000 Hz, every third event has reported duration 26 units, the remainder 27 units). Where a receiving gateway recognizes that a packet reports a consecutive series of V.21 bit events, it SHOULD play them out at a uniform rate despite the possible one-timestamp-unit discrepancies in their reported spacing and duration. +--------------------+----------------+------------+------+---------+ | Event | Frequency (Hz) | Event Code | Type | Volume? | +--------------------+----------------+------------+------+---------+ | V.21 channel 1, | 1180 | 37 | tone | yes | | '0' bit | | | | | | | | | | | | V.21 channel 1, | 980 | 38 | tone | yes | | '1' bit | | | | | | | | | | | | V.21 channel 2, | 1850 | 39 | tone | yes | | '0' bit | | | | | | | | | | | | V.21 channel 2, | 1650 | 40 | tone | yes | | '1' bit | | | | | +--------------------+----------------+------------+------+---------+ Table 2: Events for V.21 Signals Implementations that choose to transmit V.21 content using a different payload type may wish to use one of the indicator events defined in Table 7 to alert the receiver to the nature of the content. It is not expected that an implementation will send both one of these indicator events and the V.21 bit events defined above for the same content.
Other congestion-related considerations depend on the specific application for which the V.21 bit events are being used. 9] is an older general negotiation and control protocol, supporting startup for the following terminals: H.324  multimedia, V.18  text, T.101  videotext, T.30  send or receive fax, and a long list of V-series modems including V.34 , V.90 , V.91 , and V.92 . In contrast to V.8 bis , in V.8 only the calling terminal can determine the operating mode. V.8 does not use the same terminology as V.8 bis. Rather, it defines four signals that consist of bits transferred by V.21  at 300 bits/s: the call indicator signal (CI), the call menu signal (CM), the CM terminator (CJ), and the joint menu signal (JM). In addition, it uses tones defined in V.25  and T.30  (described below), and one tone (ANSam) defined in V.8 itself. The calling terminal sends using the V.21 low channel; the answering terminal uses the high channel. The basic protocol sequence is subject to a number of variations to accommodate different terminal types. A pure V.8 sequence is as follows: 1. After an initial period of silence, the calling terminal transmits the V.8 CI signal. It repeats CI at least three times, continuing with occasional pauses until it detects ANSam tone. The CI indicates whether the calling terminal wants to function as H.324, V.18, T.30 send, T.30 receive, or a V-series modem. 2. The answering terminal transmits ANSam after detecting CI. ANSam will disable any G.164  echo suppressors on the circuit after 400 ms and any G.165  echo cancellers after one second of ANSam playout. 3. On detecting ANSam, the calling terminal pauses at least half a second, then begins transmitting CM to indicate detailed capabilities within the chosen mode. 4. After detecting at least two identical sequences of CM, the answering terminal begins to transmit JM, indicating its own capabilities (or offering an alternative terminal type if it cannot support the one requested).
5. After detecting at least two identical sequences of JM, the calling terminal completes the current octet of CM, then transmits CJ to acknowledge the JM signal. It pauses exactly 75 ms, then starts operating in the selected mode. 6. The answering terminal transmits JM until it has detected CJ. At that point, it stops transmitting JM immediately, pauses exactly 75 ms, then starts operating in the selected mode. The CI, CM, and JM signals all consist of a fixed sequence of ten '1' bits followed by a signal-dependent pattern of ten synchronization bits, followed by one or more octets of variable information. Each octet is preceded by a '0' start bit and followed by a '1' stop bit. The combination of the synchronization pattern and V.21 channel uniquely identifies the message type. The CJ signal consists of three successive octets of all zeros with stop and start bits but without the preceding '1's and synchronizing pattern of the other signals. Applications MAY report each instance of a CM, JM, and CJ signal, respectively, as a series of V.21 bit events (Section 2.2), or may use another payload type to carry this information. Applications supporting V.8 signalling using the telephone-event payload MAY report the synchronization part of the CI signal (ten '1's followed by '00000 00001') both as a series of V.21 bit events and, when it has been recognized, as a single CI event. Note that the CI event covers only the synchronization part of the CI signal. The remaining call function octet and its start and stop bits need to be transmitted also, either as a series of V.21 bit events or in some other payload format. Presumably, the calling end gateway will use the same format for the CM and CJ signals. The overlapping nature of V.8 signalling means that there is no risk of silence exceeding 100 ms once ANSam has disabled any echo control circuitry. However, the 75-ms pause before entering operation in the selected data mode will require both the calling and the answering gateways to recognize the completion of CJ, so they can change from playout of telephone-event to playout of the data-bearing payload after the 75-ms period.
+--------+----------------------+------------+------+---------+ | Event | Frequency (Hz) | Event Code | Type | Volume? | +--------+----------------------+------------+------+---------+ | ANSam | 2100 x 15 | 34 | tone | yes | | | | | | | | /ANSam | 2100 x 15 phase rev. | 35 | tone | yes | | | | | | | | CI | (V.21 bits) | 53 | tone | yes | +--------+----------------------+------------+------+---------+ Table 3: Events for V.8 Signals ANSam: The modified answer tone ANSam consists of a sinewave signal at 2100 Hz, amplitude-modulated by a sine wave at 15 Hz. The beginning of the event is at the beginning of the tone. The end of the event is at the sooner of the ending of the tone or the occurrence of a phase reversal (marking the beginning of a /ANSam event). Phase reversals are used to disable echo cancellation; if they are being applied, they occur at 450-ms intervals. An ANSam event packet SHOULD NOT be sent until it is possible to discriminate between an ANSam event and an ANS event (see V.25 events, below). The modulated envelope for the ANSam tone ranges in amplitude between 0.8 and 1.2 times its average amplitude. The average transmitted power is governed by national regulations. Thus, it makes sense to indicate the volume of the signal. /ANSam: /ANSam reports the same physical signal as ANSam, but is reported following the first phase reversal in that signal. It begins with the phase reversal and ends at the end of the tone. The receiver of /ANSam MUST reverse the phase of the tone at the beginning of playout of /ANSam and every 450 ms thereafter until the end of the tone is reached. CI: CI reports the occurrence of the V.21 bit pattern '11111 11111 00000 00001' indicating the beginning of a V.8 CI signal. The event begins at the beginning of the first bit and ends at the end of the last one. This event MUST NOT be reported except in a context where a V.8 CI signal might be expected (i.e., at the calling end during call setup). Note that if the calling modem
sends the CI signal at all, it will typically repeat the signal several times. It is expected that the CI event will be most useful when the modem content is being transmitted primarily using another payload type. The event acts as a commentary on that content, allowing the receiver to recognize that V.8 signalling is in progress. 13] is a start-up protocol predating V.8  and V.8 bis . It specifies the exchange of two tone signals: CT and ANS. CT (calling tone) consists of a series of interrupted bursts of 1300-Hz tone, on for a duration of not less than 0.5 s and not more than 0.7 s and off for a duration of not less than 1.5 s and not more than 2.0 s. . Modems not starting with the V.8 CI signal often use this tone. ANS (Answer tone) is a 2100-Hz tone used to disable echo suppression for data transmission , . For fax machines, Recommendation T.30  refers to this tone as called terminal identification (CED) answer tone. ANS differs from V.8 ANSam in that, unlike the latter, it has constant amplitude. V.25 specifically includes procedures for disabling echo suppressors as defined by ITU-T Rec. G.164 . However, G.164 echo suppressors have now for the most part been replaced by G.165  echo
cancellers, which require phase reversals in the disabling tone (see ANSam above). As a result, Recommendation V.25 was modified in July 2001 to say that phase reversal in the ANS tone is required if echo cancellers are to be disabled. One possible V.25 sequence is as follows: 1. The calling terminal starts generating CT as soon as the call is connected. 2. The called terminal waits in silence for 1.8 to 2.5 s after answer, then begins to transmit ANS continuously. If echo cancellers are on the line, the phase of the ANS signal is reversed every 450 ms. ANS will not reach the calling terminal until the echo control equipment has been disabled. Since this takes about a second, it can only happen in the gap between one burst of CT and the next. 3. Following detection of ANS, the calling terminal may stop generating CT immediately or wait until the end of the current burst to stop. In any event, it must wait at least 400 ms (at least 1 s if phase reversal of ANS is being used to disable echo cancellers) after stopping CT before it can generate the calling station response tone. This tone is modem-specific, not specified in V.25. 4. The called terminal plays out ANS for 2.6 to 4.0 seconds or until it has detected calling station response for 100 ms. It waits 55-95 ms (nominal 75 ms) in silence. (Note that the upper limit of 95 ms is rather close to the point at which echo control may reestablish itself.) If the reason for ANS termination was timeout rather than detection of calling station response, the called terminal begins to play out ANS again to maintain disabling of echo control until the calling station responds. The events defined for V.25 signalling are shown in Table 4. +-------------------+----------------+------------+------+---------+ | Event | Frequency (Hz) | Event Code | Type | Volume? | +-------------------+----------------+------------+------+---------+ | Answer tone (ANS) | 2100 | 32 | tone | yes | | | | | | | | /ANS | 2100 ph. rev. | 33 | tone | yes | | | | | | | | CT | 1300 | 49 | tone | yes | +-------------------+----------------+------------+------+---------+ Table 4: Events for V.25 Signals
ANS: The beginning of the event is at the beginning of the 2100-Hz tone. The end of the event is at the sooner of the ending of the tone or the occurrence of a phase reversal (marking the beginning of a /ANS event). An initial ANS event packet SHOULD NOT be sent until it is possible to discriminate between an ANS event and an ANSam event (see V.8 events, above). /ANS: /ANS reports the same physical signal as ANS, but is reported following the first phase reversal in that signal. It begins with the phase reversal and ends at the end of the tone. The receiver of /ANS MUST reverse the phase of the tone at the beginning of playout of /ANS and every 450 ms thereafter until the end of the tone is reached. CT: The beginning of the CT event is at the beginning of an individual burst of the 1300-Hz tone. The end of the event is at the end of that tone burst. The gateway at the calling end SHOULD use a packetization interval smaller than the nominal duration of a CT burst, to ensure that CT playout at the called end precedes the sending of ANS from that end.
14] is a modem using phase-shift keying with quadrature amplitude modification. It operates on a carrier at 1800 Hz, modulated at 2400 symbols/s. The basic data rates for V.32 are 4800 and 9600 bits/s. V.32bis  extends the data rates up to 14,400 bits/s. Most or all existing deployments are V.32bis, typically in support of point-of-sale terminals and the like. One reason V.32bis is still used is because of its relatively rapid start-up sequence, particularly on leased lines. Operating over the public telephone network, the start-up begins as follows: a. the answering end begins with the V.25 answering procedure (1.8 to 2.5 s of silence followed by continuous ANS tone to a maximum of 3.3 s, with possible phase reversals to disable echo cancelling equipment); b. the calling end waits in silence until it has detected ANS for 1 s; c. the calling end begins to transmit a V.32/V.32bis pattern designated AA, i.e., a series of '0000' bit sequences transmitted at 4800 bits/s; d. upon detecting the AA pattern for at least 100 ms, the called modem is silent for 75 +/- 20 ms, then responds with an AC pattern, which is a series of '0011' bit sequences transmitted at 4800 bits/s. The difference in leased line operation is that the calling modem starts the session by sending AA. After that, the called modem responds with AC, and the rest of the sequence is unchanged. In support of V.32/V.32bis operation, Table 5 defines two events, V32AA and V32AC. +----------------+------------------+------------+------+---------+ | Event | Bit Pattern | Event Code | Type | Volume? | +----------------+------------------+------------+------+---------+ | V32AA | b'0000' repeated | 63 | tone | yes | | | | | | | | V32AC | b'0011' repeated | 27 | tone | yes | +----------------+------------------+------------+------+---------+ Table 5: Events for V.32/V.32bis Signals
V32AA: Indicates that the AA calling pattern of a V.32/V.32bis terminal has been detected. V32AC: Indicates that the AC answering pattern of a V.32/V.32bis terminal has been detected. Each of these two events begins at the beginning of its pattern, and ends nominally when the pattern stops being received. Following the sending of either of these events the session may continue using V.150.1 modem relay  or Clearmode  as negotiated or configured in advance. To help make the transition as quickly as possible, the V32AA or V32AC event SHOULD be reported as soon as the corresponding pattern is detected. It seems likely that the implementation will be transmitting the event reports simultaneously with the same data in an alternate form, typically using RFC 2198  redundancy. 8] defines the procedures used by Group III fax terminals. The pre-message procedures for which the events of this section are defined are used to identify terminal capabilities at each end and negotiate operating mode. Post-message procedures are also included, to handle cases such as multiple document transmission. Fax terminals support a wide variety of protocol stacks, so T.30 has a number of options for control protocols and sequences. T.30 defines two tone signals used at the beginning of a call. The CNG signal is sent by the calling terminal. It is a pure 1100-Hz tone played in bursts: 0.5 s on, 3 s off. It continues until timeout
or until the calling terminal detects a response. Its primary purpose is to let human operators at the called end know that a fax terminal has been activated at the calling end. The called terminal waits in silence for at least 200 ms. It then may return CED tone (which is physically identical to V.25 ANS), or else V.8 ANSam if it has V.8 capability. If called and calling terminals both support V.8, the called terminal will detect CI or more likely CM in response to its ANSam and will continue with V.8 negotiation. Otherwise, the called terminal stops transmitting CED after 2.6 to 4 seconds, waits 75 +/- 20 ms in silence, then enters the T.30 negotiation phase. In the T.30 negotiation phase the terminals exchange binary messages using V.21 signals, high channel frequencies only, at 300 bits/s. Each message is preceded by a one-second (nominal) preamble consisting entirely of HDLC flag octets (0x7E). This flag has the function of preparing echo control equipment for the message that follows. The pre-transfer messages exchanged using the V.21 coding are: Digital Identification Signal (DIS): Characterizes the standard ITU-T capabilities of the called terminal. This is always the first message sent. Digital Transmit Command (DTC): A possible response to the DIS signal by the calling terminal. It requests the called terminal to be the transmitter of the fax content. Digital Command Signal (DCS): A command message sent by the transmitting terminal to indicate the options to be used in the transmission and request that the other end prepare to receive fax content. This is sent by the calling end if it will transmit, or by the called end in response to a DTC from the calling end. It is followed by a training signal, also sent by the transmitting terminal. Confirmation To Receive (CFR): A digital response confirming that the entire pre-message procedure including training has been completed and the message transmissions may commence.
Each message may consist of multiple frames bounded by HDLC flags. The messages are organized as a series of octets, but like V.8 bis, T.30 calls for the insertion of extra '0' bits to prevent spurious recognition of HDLC flags. T.30 also provides for the transmission of control messages after document transmission has completed (e.g., to support transmission of multiple documents). The transition to and from the modem used for document transmission (V.17 , V.27ter , V.29 , V.34 ) is preceded by 75 ms (nominal) of silence). Applications supporting T.30 signalling using the telephone-event payload MAY report the preamble preceding each message both as a series of V.21 bit events and, when it has been recognized, as a single V.21 preamble event. The T.30 control message following the preamble MAY be reported in the form of a sequence of V.21 bit events or using some other payload type. If transmitted as bit events, the transmitted information MUST include the complete contents of the message: the initial HDLC flags, the information field, the checksum, the terminating HDLC flags, and the extra '0' bits added to prevent false recognition of HDLC flags at the receiver. Implementers should note that these extra '0' bits mean that in general T.30 messages as transmitted on the wire will not come out to an even multiple of octets. The training signal sent by the transmitting terminal after DCS consists of a steady string of V.21 high channel zeros (1850-Hz tone) for 1.5 s. Since the bit rate (nominally 300 bits/s) should have been clearly established when processing the preceding signalling, it is natural that if the telephony-event payload type is being used, this training signal will also be sent as a series of V.21 bit events at that bit rate. However, if the sending gateway is capable of recognizing the transition from the end of the DCS to the start of training, it MAY report the training signal as a single extended V.21 (high channel) '0' event. The events defined for T.30 signalling are shown in Table 6. The CED and /CED events represent exactly the same tone signals as V.25 ANS and /ANS, and are given the same codepoints; they are reproduced here only for convenience.
+--------------------+----------------+------------+------+---------+ | Event | Frequency (Hz) | Event Code | Type | Volume? | +--------------------+----------------+------------+------+---------+ | CED (Called tone) | 2100 | 32 | tone | yes | | | | | | | | /CED | 2100 ph. rev. | 33 | tone | yes | | | | | | | | CNG (Calling tone) | 1100 | 36 | tone | yes | | | | | | | | V.21 preamble flag | (V.21 bits) | 54 | tone | yes | +--------------------+----------------+------------+------+---------+ Table 6: Events for T.30 Signals CED: The beginning of the event is at the beginning of the 2100-Hz tone. The end of the event is at the sooner of the ending of the tone or the occurrence of a phase reversal (marking the beginning of a /CED event). An initial CED event packet SHOULD NOT be sent until it is possible to discriminate between a CED event and an ANSam event (see V.8 events, above). /CED: /CED reports the same physical signal as CED, but is reported following the first phase reversal in that signal. It begins with the phase reversal and ends at the end of the tone. The receiver of /CED MUST reverse the phase of the tone at the beginning of playout of /CED and every 450 ms thereafter until the end of the tone is reached. CNG: The beginning of the CNG event is at the beginning of an individual burst of the 1100-Hz tone. The end of the event is at the end of that tone burst. V.21 preamble flag: This event begins with the first V.21 bits transmitted after a period of silence. It ends when a pattern of V.21 bits other than an HDLC flag is observed. This means that the V.21 preamble event absorbs the initial HDLC flags of the following message.
It is expected that the V.21 preamble flag event will be most useful when the modem content is being transmitted primarily using another payload type. The event acts as a commentary on that content, allowing the receiver to prepare itself to transition to fax mode. Section 2.4.1). Similarly, the receiver may extend playout of the preamble event while waiting for further reports. However, gaps or extended playout of the V.21 sequences are not feasible. This means, as with V.8 bis, that the receiver must manage its playout buffer appropriately to increase robustness in the face of congestion. RFC 4103  packetization of character data. In anticipation of this migration, it is RECOMMENDED that text carried in the PSTN by legacy modem protocols be converted to RFC 4103 packets at the sending gateway. During a transitional period, however, gateways of a lesser capability may be able to recognize the nature of incoming content, but may only be able to encode it as voice-band data on the packet side. In such circumstances, it will help to optimize processing of the signal at the receiving end if that end receives an indication of the nature of the voice-encoded data signals. The events defined in this section provide such indications, and MAY be used in conjunction with ITU-T Recommendation V.152 , as one example, to carry the content as voice-band data. Implementers should take note of an additional class of text terminals not considered in the events below. These terminals use dual tone multi-frequency (DTMF) tones to encode and exchange signals. This application is described in RFC 4733 , Section 3.1, in conjunction with the registration of DTMF events.
The events shown in Table 7 correspond to signals coming from the following modem types: o Baudot , a five bit character encoding nominally operating at 45.45 or 50 bits/s with frequencies 1800 Hz = '0', 1400 Hz = '1'; o EDT, which is V.21  operating at 110 bits/s in half-duplex mode (lower channel only); characters are 7-bit IA5 plus initial start bit, trailing parity bit, and two stop bits; o Bell 103 mode (documented in Recommendation V.18 Annex D), which is structurally similar to V.21, but uses different frequencies: lower channel, 1070 Hz = '0', 1270 Hz = '1'; upper channel, 2025 Hz = '0', 2225 Hz = '1'; characters are US ASCII framed by one start bit, one trailing parity bit, and one stop bit; o V.23  based videotex, in Minitel and Prestel versions. V.23 offers a forward channel operating at 1200 bits/s if possible (2100 Hz = '0', 1300 Hz = '1') or otherwise at 600 bits/s (1700 Hz = '0', 1300 Hz = '1'), and a 75 bits/s backward channel, which is transmitting 390 Hz (continuous '1's) except when '0' is to be transmitted (450 Hz); o a non-V.18 text terminal using V.21  at 300 bits/s. Characters are 7-bit national (e.g., US ASCII) with a start bit, parity, and one stop bit.
+----------+-----------+----------------+---------+-------+---------+ | Event | Bit Rate | Frequency (Hz) | Event | Type | Volume? | | | bits/s | | Code | | | +----------+-----------+----------------+---------+-------+---------+ | ANS2225 | N/A | 2225 | 52 | tone | yes | | | | | | | | | V21L110 | 110 | 980/1180 | 55 | other | no | | | | | | | | | V21L300 | 300 | 980/1180 | 30 | other | no | | | | | | | | | V21H300 | 300 | 1650/1850 | 31 | other | no | | | | | | | | | B103L300 | 300 | 1070/1270 | 56 | other | no | | | | | | | | | V23Main | 600/1200 | 1700-2100/1300 | 57 | other | no | | | | | | | | | V23Back | 75 | 450/390 | 58 | other | no | | | | | | | | | Baud4545 | 45.45 | 1800/1400 | 59 | other | no | | | | | | | | | Baud50 | 50 | 1800/1400 | 60 | other | no | | | | | | | | | XCIMark | 1200 | 2100/1300 | 62 | tone | yes | +----------+-----------+----------------+---------+-------+---------+ Table 7: Indicators for Text Telephony ANS2225: indicates that a 2225-Hz answer tone has been detected. This is a pure tone with no amplitude modulation and no semantics attached to phase reversals, if there are any. The sender SHOULD report the beginning of the event when the tone is detected. The sender MAY send updates as the tone continues, and MUST report the end of the event when the tone ceases. The tone concerned is generated by a Bell 103-type modem in answer mode. This event MUST NOT be reported outside of the startup context (i.e., on the answering side at the beginning of a call). V21L110: indicates that the sender has detected V.21 modulation operating in the lower channel at 110 bits/s. Note that it may take some time to distinguish between 300 bits/s and 110 bits/s operation. It is expected that implementations will not transmit both this event and individual V.21 bit events for the same content.
V21L300: indicates that the sender has detected V.21 modulation operating in the lower channel at 300 bits/s. Note that it may take some time to distinguish between 300 bits/s and 110 bits/s operation. It is expected that implementations will not transmit both this event and individual V.21 bit events for the same content. V21H300: indicates that the sender has detected V.21 modulation operating in the upper channel at 300 bits/s. It is expected that implementations will not transmit both this event and individual V.21 bit events for the same content. B103L300: indicates that the sending device has detected Bell 103 class modulation operating in the low channel at 300 bits/s. V23Main: indicates that the sending device has detected V.23 modulation operating in the high-speed channel. As described below, this indicator may alternate with the XCIMark indication. V23Back: indicates that the sending device has detected V.23 modulation operating in the 75 bit/s back-channel. Baud4545: indicates that the sending device has detected Baudot modulation operating at 45.45 bits/s. Baud50: indicates that the sending device has detected Baudot modulation operating at 50 bits/s. XCIMark: Indicates that the sending device has detected the specific bit pattern (0) 1111 1111(1)(0)1111 1111(1) sent at 1200 bits/s using V.23 upper-channel modulation, following a period of V.23 main channel "mark" (1300 Hz).
It is assumed in all cases that the event reports described here are being transmitted in addition to another media encoding, typically G.711  voice-band data, reporting the same information. A natural method to do this is to combine the voice-band data with event reports in an RFC 2198  redundancy payload. The handling of ANS2225 has been indicated above. Since it is a specific tone, it can be handled like any other tone event. For all of the other indicators, the sender SHOULD generate an initial event report as soon as the nature of the audio content has been recognized. For reliability, the initial event report SHOULD be retransmitted twice at short intervals. (20 ms is a suggested value, although the packetization period of the associated media may be sufficient.) The sender MAY continue to send additional reports of the same indicator event, although these have little value once the receiver has adjusted itself to the type of content it is receiving. If the nature of the content changes (e.g., because it is coming from a V.18 terminal in the probing stage), the sender MUST send an event report for the new content type as soon as it is recognized. If the sender has been sending updates for the previous indicator, it SHOULD report the end of that previous indicator event along with the beginning of the new one. 11] defines a terminal for text conversation, possibly in combination with voice. V.18 is intended to interoperate with a variety of legacy text terminals, so its start-up sequence can consist of a series of stimuli designed to determine what is at the other end. Two V.18 terminals talking to each other will use V.8 to negotiate startup and continue at the physical level with V.21 at 300 bits/s carrying 7-bit characters bounded by start and stop bits.
The V.18 terminal is also designed to interoperate with the text modems listed in the previous sub-section. The startup sequences for all these different terminal types are naturally quite different. The V.18 initial startup sequence specifically addresses itself to V.8-capable terminals and V.21 terminals and, by the combination of signals, to V.23 videotex terminals. During the initial startup sequence, the V.18 terminal listens for frequency responses characterizing the other terminal types. If it does not make contact in the preliminary step, it probes for each type specifically. By the nature of the application, V.18 has been designed to provide an extremely robust startup capability. The handling of the V.18 XCI signal is a specific case of the procedures described in the previous section. XCI is a signal transmitted in high-band V.23 modulation to stimulate V.23 terminals to respond and to allow detection of V.18 capabilities in a DCE. The 3-second XCI signal uses the V.23 upper channel having periods of "mark" (i.e., 1300 Hz) alternating with the XCIMark pattern. The full definition is found in V.18, Section 3.13. The sender SHOULD indicate V23Main during the transmission of the "mark" portion of XCI, and change the indication to XCIMark when that pattern is detected. 33]. As with the indicators in the previous section, the sender SHOULD generate an initial event report as soon as the nature of the audio content has been recognized. For reliability, the initial event report SHOULD be retransmitted twice at short intervals. (20 ms is a suggested value, although the packetization period of the associated media may be sufficient.) The sender MAY continue to send additional reports of the VBDGen event, although these have little value once the receiver has adjusted itself to the type of content it is receiving.
+--------+---------------+------------+-----------+-------+---------+ | Event | Bit Rate | Frequency | Event | Type | Volume? | | | bits/s | (Hz) | Code | | | +--------+---------------+------------+-----------+-------+---------+ | VBDGen | Variable | Variable | 61 | other | no | +--------+---------------+------------+-----------+-------+---------+ Table 8: Generic Modem Signal Indicator VBDGen: indicates that the sender has detected tone patterns indicating the operation of some form of modem. This indicator SHOULD NOT be sent if a more specific event is available.