\ \(em\ When comparing these maximum values of LRs with LRs determined for
existing networks some discrepancies may be found. If the actual LRs are
greater by 2 or even 3\ dB this is no cause for concern. On the other hand, if a margin of 2 or 3\ dB seems to appear, the permissible attenuation for subscriber lines should not automatically be increased. The first step should instead be to use the margin to improve the traffic\(hyweighted mean values referred to
in\ \(sc\ 1.2.
_
.nr PS 9
.RT
.ad r
\fBTable 1/G.121 [T1.121], p. \fR
.ad b
.bp
.sp 1P
.LP
2.2
\fIDifference in transmission loss between the two directions of\fR
\fItransmission in national systems\fR
.sp 9p
.RT
.PP
It has been found pratical to introduce a certain difference in
loss between the directions 4\(hywire\(hyto\(hy2\(hywire and 2\(hywire\(hyto\(hy4\(hywire.
As can be
seen from Figure\ 1/G.121 this difference is equal to \fID\fR\d\fIo\fR\u\
=\ (\fIR\fR \ \(em\ \fIT\fR )\ dB referred to the 0\ dBr 4\(hywire reference
points. Referred to the VASPs as in
Figure\ 1/G.122 the difference is \fID\fR\d\fIv\fR\u\ =\ (\fIR\fR \ \(em\
\fIT\fR \ \(em\ 7)\ dB. For
international
transmission compatibility it is desirable that Administrations choose
approximately the same value of these differences. Table\ C\(hy1/G.121
indicates
that \fIR\fR \ =\ 7, \fIT\fR \ =\ 0\ dB are the most common pad values,
giving \fID\fR\d\fIo\fR\u\ =\ 7,
\fID\fR\d\fIv\fR\u\ =\ 0 on the average. For planning of new networks,
these are the
preferred values. Thus, the difference in loss between the two directions of
transmission on an international connection should not exceed 8\ dB, preferably
not 6\ dB.
.PP
The following points should be noted:
.RT
.LP
1)
Bearing in mind that most Administrations allocate the
losses of their national extension circuits in much the same sort of way
connections set up in practice should not exhibit differences much in excess
of 3\ dB.
.LP
2)
As far as speech transmission is concerned, from the studies carried
out by several Administrations in\ 1968\(hy1972, it is clear that for
connections with overall LRs falling within the range found in practice, no
great disadvantage attaches to any reasonable difference in LR between
the two directions of transmission.
.LP
3)
When devising national transmission plans, Administrations should take
into account the needs of data transmission between modems
complying with the pertinent Recommendations.
.sp 2P
.LP
\fB3\fR \fBMinimum SLR\fR
.sp 1P
.RT
.PP
Administrations must take care not to overload the international
transmission systems if they reduce the attenuation in their national trunk
network.
.PP
Provisionally a nominal minimum value of \fISLR\fR \ =\ \(em1.5\ dB referred
to a 0\ dBr point or 2\ dB referred to the send virtual analogue switching
point of
the
international circuit is recommended in order to control the peak value
of the speech power applied to international transmission systems. It should
be noted that the imposition of such a limit does not serve to control
the long\(hyterm
mean power offered to the system.
.PP
In some countries a very low sending loudness rating value may occur if
unregulated telephone sets are used. Furthermore, the speech power applied
to the international circuits by operators' sets must be controlled so
that it does not become excessive.
.RT
.sp 2P
.LP
\fB4\fR \fBDetermination of nominal Loudness Ratings\fR
.sp 1P
.RT
.PP
Loudness Ratings and their properties and uses are explained in
Annex\ A to Recommendation\ G.111. There it is described how a particular
LR of a national system may be determined as a sum of the individual LRs
of its parts. Also, rules are given for how to obtain the individual LRs
of these parts, i.e. for telephone sets, subscriber lines, junctions, channel
equipment, etc.
.PP
Note that Send and Receive Loudness Ratings of \fIanalogue telephone\fR
\fIsets\fR | are measured under specified conditions which do not exactly
correspond to those valid for a national system which is part of an international
connection. The measurements are done with a terminating impedance of\
600\ ohms resistive and over a much wider bandwidth (100\(hy8000\ Hz or
200\(hy4000\ Hz) than
the assured bandwidth of the international connection (300\(hy3400\ Hz).
.PP
Therefore, to avoid confusion, measured values of Send and Receive
Loudness Ratings of \fIanalogue\fR | telephone sets are designated by
the index \*Qw\*U (for wideband). To get the proper values of SLR and RLR
for \fIplanning\fR |
international connections, 1\ dB should be added to the measured values
in order to compensate for bandwidth and impedance mismatch effects. Thus,
.RT
.LP
\fISLR\fR =
\fISLR\fI\d\fIw\fR\u\ +\ 1
.LP
\fIRLR\fR =
\fIRLR\fI\d\fIw\fR\u\ +\ 1
.PP
A \fIdigital\fR | telephone set, however, does not need these
corrections because the codec and filters in the set limit the band anyhow.
.bp
.PP
In general, the loudness loss between \fItwo electrical interfaces\fR , |
the Circuit Loudness Rating CLR, is equal to the corresponding difference in
relative levels. (Unless an interface with a \*Qjump\*U in relative level is
included in the path. See\ \(sc\ 6.3.)
.PP
\*QNominal value\*U here signifies a \*Qreasonable engineering average\*U
for typical conditions as exemplified in what follows, excluding \*Qworst
cases\*U.
.PP
With regard to circuits and other items of equipment, variations with time,
temperature etc. are not included in the nominal CLRs, Circuit Loudness
Ratings.
.PP
For telephone sets, most Administrations today have to accept a large variety
of types which comply with some national specification having rather
wide limits. The requirements for SLR and RLR usually refer to a measuring
setup with a variable artificial line terminated by a feeding bridge and a
nominal impedance which may be complex or, most often, 600\ ohms.
.PP
The specification is often drawn up in the form of upper and lower
limits for SLR\fI\fI\d\fIw\fR\u | and RLR\fI\fI\d\fIw\fR\u | as functions
of line length (or possibly line current). The \*Qnominal\*U SLR\fI\fI\d\fIw\fR\u |
and RLR\fI\fI\d\fIw\fR\u | of telephone set plus
subscriber line may then be interpreted as the arithmetic mean between the
upper and lower limit curves.
.PP
In practice, the subjective quality impression of the overall loudness
changes rather insignificantly for fairly large variations of OLR around
the
optimum value and it is unlikely that sets with worst possible LRs are
associated with limiting line lengths. Therefore, rather wide manufacturing
tolerances, commonly about\ \(+-3\ dB, can be accepted for the individual
set SLR
(set) and RLR (set). (SLR (set) and RLR (set) refer to set measurements
without the subscriber line but as function of line current, including
the\ 1\ dB
bandwidth correction.)
.PP
Note however, that the \fIsum\fR | of SLR (set) + RLR (set) for an
individual 2\(hywire
telephone set must be controlled more carefully so that is does not decrease
below a certain minimum value. The reason is that, under certain circumstances,
subscribers react very unfavourably to strong sidetone and talker echo.
Both
effects depend directly on this LR sum in addition to the unavoidable network
impedance variations. This minimum limit is often translated into a minimum
limit for STMR as measured against a specified impedance. See\ \(sc\ 5 for a
discussion.
.RT
.sp 2P
.LP
\fB5\fR \fBSidetone\fR
.sp 1P
.RT
.sp 1P
.LP
5.1
\fIGeneral\fR
.sp 9p
.RT
.PP
Especially for those connections approaching the limits for high
Loudness Ratings and/or noise, further transmission impairments should be
avoided. One important precaution is to ensure that an adequate \fIsidetone\fR
|
performance is maintained for the various circuit combinations occurring
in the telephone system. (\*QAdequate\*U is in most cases to be interpreted
as a
sufficiently high sidetone loss.)
.PP
For 2\(hywire telephone sets, the sidetone performance is basically
dependent on set sensitivity and impedance variation limits as explained in
Annex\ A to Recommendation\ G.111. Thus, a national transmission plan should
not only give rules for allocation of losses in the network but also provide
an
appropriate impedance strategy to follow. (An example is given in
Supplement\ No.\ 10 of\ Vol.\ VI.)
.PP
Note that for sidetone evaluations one has to consider the line
impedance \*Qseen\*U by the 2\(hywire telephone set in the actual, \fIcomplete\fR
|
connection. In modern system configurations this impedance cannot always be
simulated by an artificial line terminated by a simple R\(hyC network.
Either one has to use a more elaborate measuring setup or resort to computations
from
known data of the circuits involved. (A number of computer programs exists
which can be employed for such purposes.)
.PP
Of special interest is the fact that a 4\(hywire link inserted in a
2\(hywire connection may cause large impedance variations. As this is a common
network practice\ \(em\ for instance digital exchanges\ \(em\ a simplified
calculation
method is discussed in Annex\ B.
.PP
Ideally, a 2\(hywire telephone set could be designed to have an adaptive
sidetone balancing function, thus widening the acceptable range of line
impedances. Such costly techniques are very exceptional, however, and should
not be prescribed for the \*Qstandard\*U sets to
.bp
.PP
be used in the network. A possible, cheaper alternative is to design a
set with a \fIZ\fR\d\fIs\fR\\d\fIo\fR\u |
varying in a predetermined manner with the line feeding current.
\fINote\ 1\fR \ \(em\ The above method replaces an earlier method in which
the echo loss of the path\ \fIa\fR \(hy\fIt\fR \(hy\fIb\fR was provisionally
defined as the
expression in transmission units of the unweighted mean of the power ratios
in the band\ 500\(hy2500\ Hz. The new method has been found to give better
agreement
with subjective opinion for individual connections. However, study has shown
that echo path loss distributions for large samples of actual connections
calculated by the two methods have almost identical means and standard
deviations. Therefore, data gathered by the older method is still considered
useful in planning studies.
.PP
\fINote\ 2\fR \ \(em\ Evidence was presented which showed that a white noise
signal is not necessarily optimum to measure the residual echo level after
cancellation, because an echo cancellor does not converge to quite the same
condition as it does with a real speech signal. It may be better to use the
conventional telephone signal (Recommendation\ G.227\ [5]) or better still, an
artificial speech signal (see\ [6]). A good compromise is the weighted noise
signal based on the principle recommended above.
.PP
\fINote\ 3\fR \ \(em\ Improved balance return losses at \fIt\fR can be
obtained when the local exchange uses 4\(hywire switching and the local
line is permanently
associated with the 2\(hywire/4\(hywire conversion unit and its balance
network (see Recommendation\ Q.552 for examples). When there is 2\(hywire
switching a compromise balance network must be used.
.bp
.PP
\fINote\ 4\fR \ \(em\ There is evidence that a 4\(hywire handset telephone
in normal use can contribute significant acoustic echo to the communication.
Hence in
some circumstances (low transmission loss, long delay times) echo control
devices may be needed.
.RT
.PP
4.3
An example of how the recommendation quoted in \(sc 4.1 above can be achieved
would be for the mean value of the sum of the transmission
losses\ \fIa\fR \(hy\fIt\fR and\ \fIt\fR \(hy\fIb\fR not to be less than
(4\ +\ \fIn\fR )\ dB with a standard
deviation from the mean not exceeding\ 2
@ sqrt { fIn\fR } @ \ dB, accompanied by an echo balance return loss at the terminating set\ \fIt\fR , of not less than\ 11\ dB with a standard deviation not exceeding\ 3\ dB.
.sp 9p
.RT
.sp 2P
.LP
\fB5\fR \fBEffects of listener echo\fR \fB(receive end echo)\fR |
.FS
The use of \*Qlistener echo\*U in this context might be misleading. It could
be substituted by a more appropriate term. The term \*Qreceived end echo\*U
is a
term preferred by some Administrations.
.FE
.sp 1P
.RT
.sp 1P
.LP
5.1
\fIGeneral\fR
.sp 9p
.RT
.PP
It has been assumed in \(sc\(sc 1 to 4 that only one 2\(hywire\(hy4\(hywire\(hy2\(hywire
loop (further referred to as loop) occurs in a connection. Consequently,
the
requirements in \(sc\(sc\ 1 to 4 are valid for that case, i.e.\ they refer
to the
\*Qsemi\(hyloop\*U seen directly from the VASPs (virtual analogue switching
points). However, in mixed analogue/digital connections several loops may
occur when
4\(hywire digital exchanges (including PABXs) are connected 2\(hywire to other
exchanges. Such loops have typically low loss and short delay times (at
most a few milliseconds). Signals reflected twice, i.e.\ at both hybrids
that terminate a loop, would therefore contribute to listener echo. These
listener echo
signals:
.RT
.LP
\(em
can lead to objectionable \*Qhollowness\*U in voice
communications, and
.LP
\(em
can impair the bit error ratio of received voice\(hyband data
signals.
.PP
In general it has been found that for satisfactory transmission, data modem
receivers require higher values of listener echo loss (in the
band\ 500\(hy2500\ Hz) than speech (in the band\ 300\(hy3400\ Hz).
.PP
The effect of listener echo is characterized by the difference in
level between the direct signal and the multiple reflected signal. In
Figure\ 2/G.122 the loss of the direct path is assumed to be \fIS\fR \
dB, whereas the loss along the path of the reflected signal is L\ dB. The
listener echo loss
(LE) then is \fIL\fR \ \(em\ \fIS\fR \ dB. It can be seen from Figure\
2/G.122 that if only two reflections occur (only double\(hyreflected signals),
the listener echo loss\ LE
equals the loss around the loop (open\(hyloop loss, OLL), as all other losses
are incurred equally by the direct and the reflected signals.
.RT
.LP
.rs
.sp 8P
.ad r
\fBFIGURE\ 2/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.PP
It should be noted that usually the listener echo will consist of a series
of signals being reflected two times, four times,\ etc. and hence LE
and OLL are in principle not equal. In practice however LE and OLL may
be taken as equal when OLL exceeds about\ 8\ dB.
.PP
The loss around the loop can be measured by breaking the loop at some point,
injecting a signal and measuring the loss incurred in traversing the
open loop. All impedance conditions of the closed loop and at the 2\(hywire
ends should be preserved whilst making the measurement. The measured quantity
is the open\(hyloop loss\ (OLL).
.PP
For practical purposes,
semi\(hyloop measurements
may be more
easily carried out, especially in the case of 4\(hywire exchanges with
2\(hywire
circuit
terminations, since it is sometimes difficult to maintain a connection
through a 4\(hywire exchange and interrupt one direction of transmission.
Figure\ 3/G.122 explains the notion of the semi\(hyloop loss (SLL).
.bp
.PP
The sum of the two semi\(hyloop losses of a 2\(hywire/4\(hywire/2\(hywire
device is equal to its open\(hyloop loss (and hence very nearly to its
listener echo
loss)\ \(em\ again assuming that impedance conditions at the 2\(hywire ends are
preserved whilst making the measurements.
.RT
.LP
.rs
.sp 11P
.ad r
\fBFIGURE\ 3/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.sp 2P
.LP
5.2
\fILimitation of listener echo\fR
.sp 1P
.RT
.sp 1P
.LP
5.2.1
\fIVoice\(hyband data transmission\fR
.sp 9p
.RT
.PP
The minimum values for the listener echo loss are under study.
However, the following consideration provides an example and may serve as an
indication of what values of\ OLL might be required by existing types of
modems with a bit rate of up to 2.4\ kbit/s, in order to obtain high quality
data
transmission:
.RT
.LP
\(em
a complete connection should not contain more than five
(exceptionally seven) physical loops;
.LP
\(em
loops with very high OLL (exceeding, e.g. 45 dB) need not be included
in the number of loops in the connection;
.LP
\(em
the OLL of each loop at any frequency in the
band\ 500\(hy2500\ Hz, should not be less than the values indicated in
Table\ 1/G.122 (based on \fIOLL\fR \ =\ 18\ +\ 10\ log\ \fIm\fR , where
\fIm\fR \ =\ total
number of loops).
.ce
\fBH.T. [T1.122]\fR
.ce
TABLE\ 1/G.122
.ps 9
.vs 11
.nr VS 11
.nr PS 9
.TS
center box;
cw(48p) sw(48p) | cw(48p) , c | c | ^ .
In one national system {
Maximum total number of loops in international
connection
}
Number of national loops OLL of each loop
_
.T&
cw(48p) | cw(48p) | cw(48p) .
1 22 dB 3
.T&
cw(48p) | cw(48p) | cw(48p) .
2 25 dB 5
.T&
cw(48p) | cw(48p) | cw(48p) .
3 26.5 dB 7
_
.TE
.nr PS 9
.RT
.ad r
\fBTABLE\ 1/G.122 [T1.122], p.\fR
.sp 1P
.RT
.ad b
.RT
.sp 1P
.LP
5.2.2
\fIVoice transmission\fR
.sp 9p
.RT
.PP
Voice performance in the presence of listener echo can be
quantified in terms of a weighted value of OLL over the voice\(hyfrequency band
of\ 300\(hy3400\ Hz. Two weighting functions have been defined in Supplement\
No.\ 3 in Volume\ V.
.PP
Using the information given in Recommendation P.11 appropriate values of\
OLL may be derived as a function of loop round\(hytrip delay for satisfactory
voice transmission. These values are presently under study.
.bp
.RT
.ce 1000
ANNEX\ A
.ce 0
.ce 1000
(to Recommendation G.122)
.sp 9p
.RT
.ce 0
.ce 1000
\fBMeasurement of stability loss\fR (\fIa\fR \(hy\fIb\fR ) \fBand\fR \fBecho
loss\fR (\fIa\fR \(hy\fIb\fR )
.sp 1P
.RT
.ce 0
.PP
The stability loss
(\fIa\fR \(hy\fIb\fR ) and the
echo
loss
(\fIa\fR \(hy\fIb\fR ) as defined in \(sc\(sc\ 3.1 and\ 4.1 respectively
may be measured by
apparatus at the international centre in accordance with the principle of
Figure\ A\(hy1/G.122.
.sp 1P
.RT
.PP
In respect of the echo measurement, the combined response of the send and
receive filters must be such that the definition given in\ \(sc\ 4.2 of
the text is effectively implemented, e.g.\ such that the difference between a
measured echo loss and one calculated from the loss/frequency characteristic
does not exceed\ 0.25\ dB.
.PP
The allocation of the total response between send and receive is not critical
and any reasonable division may be used provided that:
.RT
.LP
\(em
excessive interchannel interference is avoided in national
transmission systems due to an unrestricted spectrum of the
transmitted signal;
.LP
\(em
unwanted signals that may give rise to errors, e.g. hum,
circuit noise, carrier leak signals, are prevented from entering the
receiver.
.PP
Appropriate arrangements (not shown) are needed for automatic or manual
access to the 4\(hywire switches at the international centre and also to
ensure that due account is taken of the transmission levels at the actual
switching points.
.PP
As far as the stability measurement is concerned, if a sweep
oscillator is used, attention must be paid to the risks of engendering false
operation of national signalling systems.
.PP
For both measurements anomalous results may be obtained if echo
suppressors are encountered in the national extension.
.PP
To measure the echo loss (\fIa\fR \(hy\fIb\fR ), the output of the send filter
is first connected to the input of the receive filter and the appropriate
level set and noted. The apparatus is then connected as in Figure\ A\(hy1/G.122
and the new reading on the meter noted. The loss so indicated is the echo
loss
(\fIa\fR \(hy\fIb\fR ).
.RT
.LP
.rs
.sp 23P
.ad r
\fBFigure\ A\(hy1/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.LP
.bp
.ce 1000
ANNEX\ B
.ce 0
.ce 1000
(to Recommendation G.122)
.sp 9p
.RT
.ce 0
.ce 1000
\fBExplanation of terms associated with the path \fR \fIa\fR \(hy\fIt\fR
\(hy\fIb\fR
.sp 1P
.RT
.ce 0
.ce 1000
(Contribution of British Telecom and Australia)
.sp 9p
.RT
.ce 0
.LP
B.1
\fIReturn loss\fR
.sp 1P
.RT
.PP
This is a quantity associated with the degree of match between two impedances
and is given by the expression:
\v'6p'
.RT
.sp 1P
.ce 1000
Return loss of \fIZ\fR \d1\u versus \fIZ\fR \d2\u =
20 log
\d10
\u
@ left | { fIZ\fR~\d1\u~+~\fIZ\fR~\d2\u } over { fIZ\fR~\d1\u~\(em~\fIZ\fR~\d2\u } right | @ dB
.ce 0
.sp 1P
.LP
.sp 1
.PP
The use of the expression \*Qreturn loss\*U should be confined to
2\(hywire paths supporting signals in the two directions simultaneously.
.sp 1P
.LP
B.2
\fIBalance return loss\fR
.sp 9p
.RT
.PP
A clear definition is given in the preamble of Recommendation
G.122. Figure\ B\(hy1/G.122 illustrates the definition.
.RT
.LP
.rs
.sp 14P
.ad r
\fBFigure\ B\(hy1/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.PP
The 2\(hywire portion must be in the condition appropriate to the
study, e.g., if speech echo is being studied the telephone set must be
in the speaking condition.
.PP
In the particular case (which occurs very often) in which the
impedances presented by each of the paths in the 4\(hywire portion is
also\ \fIZ\fR\d\fIB\fR\u\ (e.g.\ 600\ ohms) then the terminating set presents
an impedance of the 2\(hywire point which is substantially equal to\ \fIZ\fR\d\fIB\fR\u.
Figure\ B\(hy2/G.122 illustrates this case.
.RT
.LP
.rs
.sp 11P
.ad r
\fBFigure B\(hy2/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.LP
.bp
.PP
The term \*Qbalance return loss\*U (\fInot\fR | eturn loss) should always
be used for the contribution to the loss of the path\ \fIa\fR \(hy\fIt\fR
\(hy\fIb\fR attributable to the degree of match between\ \fIZ\fR\d\fIB\fR\uand\
\fIZ\fR\d\fIT\fR\\d\fIW\fR\u.
.sp 1P
.LP
B.3
\fITransmission loss of the path a\(hyt\(hyb\fR
.sp 9p
.RT
.PP
There is room for confusion here because the concept can be applied to
arrangements in which there is no physical point\ \*Q\fIt\fR \*U at all,
e.g.\ as in
some laboratory simulations of long connections in which echo is introduced
by a controlled unidirectional path bridging the two 4\(hywire paths. The
point\ \*Q\fIt\fR \*U is necessary in the Recommendation because practical
public switched telephone networks are being dealt with.
.PP
Thus in general two cases arise.
.PP
\fICase\ 1:\fR \ There does exist a point \*Q\fIt\fR \*U (Figure\ B\(hy3/G.122).
.RT
.LP
.rs
.sp 13P
.ad r
\fBFigure\ B\(hy3/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.PP
The transmission loss of the path \fIa\fR \(hy\fIt\fR \(hy\fIb\fR may be
calculated
from
\v'6p'
.sp 1P
.ce 1000
loss (\fIa\fR \(hy\fIt\fR ) + 20 log
\d10
\u
@ left | { fIZ~\dB\u\fR~+~\fIZ~\dTW~\u\fR } over { fIZ~\dB\u\fR~\(em~\fIZ~\dTW~\u\fR } right | @ + loss (\fIt\fR \(hy\fIb\fR )
.ce 0
.sp 1P
.PP
.sp 1
The diagram is drawn in terms of the virtual switching points of the international
circuit with their associated relative levels. The
subscript\ \fIi\fR in the abbreviation\ dBr\fI\fI\d\fIi\fR\usignifies that
these relative
levels are with respect to a 0\ dBr point of the international circuit.
.PP
It is clear that any other convenient pair of relative levels
(differing by 0.5\ dB in the correct sense) can be used in practice, e.g.,\
the actual switching levels used in an international centre.
.PP
\fICase\ 2:\fR \ There does not exist any \*Q\fIt\fR \*U (Figure B\(hy4/G.122).
.PP
This relates particularly to laboratory testing arrangements.
.RT
.LP
.rs
.sp 15P
.ad r
\fBFigure\ B\(hy4/G.122, p. \fR
.sp 1P
.RT
.ad b
.RT
.LP
.bp
.PP
In this case the loss of the path \fIa\fR \(hy\*Q\fIt\fR \*U\(hy\fIb\fR may be
calculated from:
(\fIR\fR \ +\ \fIE\fR \ +\ \fIS\fR )\ dB (assuming acoustic feedback at
the 4\(hywire telephone to be negligible).
.PP
In both cases the loss of \*Qthe path \fIa\fR \(hy\fIt\fR \(hy\fIb\fR \*U
can in principle be directly measured by the principles described in Annex\
A, i.e.\ by injecting a signal at\ \fIa\fR and measuring the result at\
\fIb\fR , so that one may properly say for all cases
\v'6p'
.RT
.sp 1P
.ce 1000
@ left { pile { { ttransmission~loss } above { ~of~the~path~\fIa\fR~\(hy\fIt\fR~\(hy\fIb\fR } } ~ right } @ \(==
@ left { pile { { ttransmission~loss } above { ~between~\fIa\fR~and~\fIb\fR } } ~ right } @
.ce 0
.sp 1P
.LP
.sp 1
.LP
or, more shortly
\v'6p'
.sp 1P
.ce 1000
loss (\fIa\fR \(hy\fIt\fR \(hy\fIb\fR )\ \(==\ loss\ (\fIa\fR \(hy\fIb\fR )
.ce 0
.sp 1P
.LP
.sp 1
B.4
\fIStability and echo losses\fR
.sp 9p
.RT
.PP
So far the quantities dealt with are functions of frequency and
yield a graph of attenuation/frequency distortion. When it is required to
characterize such a graph with a single number, additional qualifying
indications are, for example, stability loss\ (\fIa\fR \(hy\fIb\fR ) and echo
loss\ (\fIa\fR \(hy\fIb\fR ).
.PP
The text of this Recommendation gives the definitions of these
single\(hynumber descriptions thus: the stability loss (\fIa\fR \(hy\fIb\fR
) is the least
value (measured or calculated) in the band 0\(hy4\ kHz (see \(sc\(sc\ 2.1
and\ 3.1), and
the echo loss (\fIa\fR \(hy\fIb\fR ) is a weighted integral of the loss/frequency
function over the band 300\(hy3400\ Hz, as defined in \(sc\ 4.2.
.PP
When the echo\(hypath loss/frequency characteristic is available in
graphical or tabular form, alternative techniques for the calculation of
echo loss (\fIa\fR \(hy\fIb\fR ) are preferable to that suggested for the
field measurement
given in Annex\ A.
.PP
\fINote\fR \ \(em\ When evaluating echo loss from graphical or tabulated
data, sufficient frequency points should be taken to ensure that the influence
of the shape of the amplitude/frequency characteristic is adequately preserved.
The
more irregular the shape, the more points should be taken for a given
accuracy.
.RT
.sp 1P
.LP
\fIGraphical data (trapezoidal rule)\fR
.sp 9p
.RT
.PP
If the loss/frequency characteristic of the echo\(hypath is available in
graphical form (or the data were suitably measured) the echo loss may be
calculated by using the trapezoidal rule as follows:
.RT
.LP
1)
Divide the frequency band (300 to 3400 Hz) into \fIN\fR
sub\(hybands of equal width on a log\(hyfrequency scale.
.LP
2)
Read off the echo loss at each of the \fIN\fR \ +\ 1 frequencies at the
edges of the \fIN\fR sub\(hybands, and express it as an output/input
power ratio, \fIA\fR\d\fIi\fR\u.
.LP
3)
Calculate the echo loss using the formula:
\v'6p'
.sp 1P
.ce 1000
\fIL
\de\u\fR = \(em10 log
\d10
\u
[Formula Deleted]
.ce 0
.sp 1P
.LP
.sp 1
\fITabulated data\fR
.sp 9p
.RT
.PP
When the loss/frequency data are only available at \fIN\fR \ +\ 1 discreet
frequencies, which are nonuniformly spaced on a log\(hyfrequency scale,
proceed
as follows:
.bp
.PP
An approximation to the formula for echo loss (\fIa\fR \(hy\fIb\fR ) given in
