TIEl.4/93-88
`
`Standards Project:
`
`T1E1.4: ADSL
`
`Title:
`
`Source:
`
`Recommended Procedures for Exchange of
`DMT Loading Information in ADSL
`
`Amati Communications Corporation
`
`J.A.C. Bingham and P.S. Chow
`101 University Ave., Suite 100
`Palo Alto, CA 94301
`Ph:
`415-617-8982
`fax: 415-327-ADSL (2375)
`
`Date:
`
`April 15, 1993
`
`Distribution to:
`
`T1E1.4
`
`Abstract:
`
`This paper is a continuation and an expansion of TIE 1.4/93-022. It describes, in more,
`though not yet complete, detail, that part of the initialization procedure for an ADSL system
`implemented by DMT that is concerned with feedback to each transmitter of the calculated bits and
`powers.
`
`NOTICE
`
`This contribution has been prepared to assist Standards Committee TI-Telecommunications;
`it is offered to the committee as a basis for discussion and is not a binding proposal on Amati
`Communications Corp. The requirements and recommendations may be changed after further
`study. Amati specifically reserves the right to add to, amend, or withdraw the statements contained
`herein.
`
`

`
`Recommended Procedure for Exchange
`of DMT Loading Information in ADSL
`
`I.Introduction
`
`In order to achieve the best possible performance from a DMT ADSL system on a wide
`variety of subscriber lines, three essential parts of the initialization procedure are
`(a) Measurement, in the two receivers, of the S"N"R * in all used sub-bands. The
`recommended procedure for this has been outlined in TIEl.4193-022, and is described in detail in
`TIE1.4/93-087.
`
`(b) Calculation of the optimum distribution of bits, and available transmit power among the
`sub-carriers (hi and gi, for ;=1 to Ne, where Ne = 256 at the CO and 64 at the RT). Two algorithms
`have been described in [1,2], but neither is totally satisfactory. Better algorithms have been, and
`will continue to be, developed; since, however, they are not essential for the design of a compatible
`ADSL unit, they will probably remain proprietary.
`
`(c) Communication of this "bits and gains" information from each receiver to each
`transmitter. This paper recommends a procedure for this communication.
`
`2. Exchange of Bits and Gains Information
`
`As shown in Fig. 5 of 93-083, the RTU transmits the random channel-analysis signal
`immediately after the periodic line-conditioning signal. Thereafter, the basic sequence of signals is
`as shown in Fig. 1 of this paper. It should be noted that the channel analyses are performed
`simultaneously (i.e., in a full-duplex mode); this is because unfiltered or uncancelled echo may be a
`significant part of the total "noise" seen by a receiver. The subsequent exchange of bits and gains,
`however, is most conveniently performed in a half-duplex mode.
`
`Ultra reliability in the exchange of bits and gains information is very desirable, but at this
`stage of the initialization the two transmitters do not, of course, know how many bits each of their
`subcarriers can carry. Since, however, the amount of information to be sent is quite small Gust 16
`bits for each of the Ne carriers), a set of only four low-frequency carriers (those that would be
`expected eventually to carry up to eleven bits) is used with only QPSK on each. Two other
`safeguards are used: the information is repeated on a second set of fourcarriers, and a l6-bit CRC is
`appended to the data.
`
`The bits and gains signal, which therefore consists of 2Ne symbols, is preceded by a
`header symbol and followed by the CRC (both formats to be recommended later) for a total of
`(2Ne +4) symbols.
`
`The format of the symbols is [bi,gi] = [bli,b2i,b3i,b4i,gli,g2i, ........... g12i], and they are
`transmitted in ascending order of subcarrier number: that is, ; = 1 to Ne*.
`
`* The N is enclosed in quotation marks to emphasize that the "noise" is the sum total of all
`impairments listed in Section 4.1 of TIE 1.4/92-098.
`
`

`
`The bl,b2,b3,b4 are encoded in conventional binary (0000 to 1011 for 0 to 11) onto a
`primary set of carriers 8-9 and a secondary set that is to be determined.
`
`The gl, ....... gI2 represent the ratio of the gain to be transmitted to that used for channel
`analysis, with 0010 0000 0000 being normalized to unity. For example 0100 0000 0000 would
`indicate a 6 dB increase in power.
`
`The symbols [hi,gil are transmitted on the primary set of carriers 8 - 11 and a secondary set
`Nsec - Nsec+3 in conventional QPSK as follows:
`
`blb2 b3b4 g1g2 g3g4
`11
`8
`9
`10
`First symbol:
`First symbol: Nsec+ 0
`1
`2
`3
`
`g5g6 g7g8 g9g10 g11g12
`8
`9
`10
`11
`Second symbol:
`Second symbol: Nsec+ 0
`1
`2
`3
`
`This method of communicating the bits and gains information is extremely reliable, but there
`is, of course, a non-zero probability of error. If the CRC indicates an error then the "basic"
`procedure of Fig. 1 must be refmed to allow for retransmission. Details of this have not been
`fmalized. In the the present implementation a detected error would initiate a complete re(cid:173)
`initialization: this is not necessary.
`
`References.
`
`[1] J.A.C. Bingham, "The Theory and Practice of Modem Design", John Wiley and Sons, 1988.
`
`[2] D. Hughes-Hartogs, "Ensemble Modem Structure for Imperfect Transmission Media", U.S.
`Patent No. 4.679.227, July 1987.
`
`co
`
`Medley
`
`RT
`
`Medley
`
`S = Segue; H = Header (see 93-083)
`
`Figu re 1. Bits and Powers exchange.
`
`* Subcarrier 0 can never be used in a transformer-coupled system. Subcarrier Nc is at the Nyquist
`frequency,fsl2; if used, which would be very rarely, it could be modulated only as BPSK: that is
`bNc may be 0 to 5.