 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a modulator and demodulator (MODEM)
communication system. More particularly, it relates to a multipoint type
MODEM communication system having training means for adapting a receiving
MODEM to a characteristic of a MODEM communication line, and a method for
training the same.
2. Description of the Related Art
MODEM communication systems using telephone lines as communication lines
are extensively known. The MODEM communication systems are categorized
into two types; a point-to-point system in which a pair of MODEMs are
connected through the telephone line, and a multipoint system in which a
MODEM is connected to a plurality of MODEMs through a common telephone
line. The point-to-point system has the advantages of a simple circuit
construction for the MODEM and a simple communication protocol between the
MODEMs. However, if a system is required in which a large number of MODEMs
are provided and connected to each other, the point-to-point system has
the disadvantages of many lines in use, with corresponding high line-use
fees, etc. The multipoint system solves these problems because a single
telephone line is commonly used for the plurality of MODEMs, reducing the
line-use fees and the construction costs. The present invention
essentially relates to the multipoint MODEM communication system.
In the multipoint MODEM communication system, however, a process known in
the art as "TRAINING", which will be described later in detail, is
indispensable and must be carried out prior to the reception of data on
the line. Briefly, the training process is that wherein a receiving MODEM
adjusts control parameters therein in response to a training signal from a
sending MODEM to ensure the reception of data. In the multipoint MODEM
communication system, distances between a center station (CS) and the
plurality of local stations (LS) are not constant, and accordingly, a
characteristic of the line, for example, phase jitter, between the CS and
one LS is not equal to that of a line between the CS and another LS, and
thus, for example, the receiving MODEM in the CS, per se, must adjust
reception parameters therein to match the line connected to the sending
MODEM in the LS, prior to the reception of data from the sending MODEM.
In general, a high speed transfer on the telephone line may result in a
short transfer time for the same amount of data transfer. In the
multipoint MODEM communication system, however, this is not always true
because a considerable length of time is required for the training.
Namely, the high speed transfer will require, for example, a fine
adjustment of the parameters in the receiving MODEM, and this fine
adjustment may require a large amount of training information and the
training may take a long time. For example, according to CCITT
recommendations, the training time should be within 50 ms for a 4800 bit
per second (bps) transfer line (CCITT V. 27 bis), 253 ms for a 9600 bps
transfer line (CCITT V. 29), and 1393 ms for 14.4 Kbps transfer line
(CCITT U. 33). If only a small amount of data is to be transferred, the
time needed for the training may be longer than the time needed for the
data transfer. As a result, the transfer time containing the training time
and the data transfer time in the high speed transfer line may be longer
than the transfer time in the low speed transfer line. To eliminate this
paradox, the training time must be reduced. Shortening the training time
will bring a high transfer efficiency on the line, and accordingly, reduce
the line-use fees.
A variety of training methods for shortening the training time have been
proposed. The prior art training methods, however, still suffer from the
disadvantage of a long training time.
SUMMARY OF THE INVENTION
An object of the present invention is to provide an improved multipoint
type MODEM communication system having a training means and a reduced
training time.
Another object of the present invention is to provide an improved
multipoint type MODEM communication system having the above features and
applicable to a high speed data transfer line.
Still another object of the present invention is to provide an improved
training method applicable to the above improved multipoint type MODEM
communication system.
Yet another object of the present invention is to realize the above objects
without increasing the cost of the system, particularly the MODEM.
According to the present invention, there is provided a modulator and
demodulator (MODEM) communication system including: a telephone line
device having a telephone line; a central station including a MODEM and
operatively connected to the telephone line through the MODEM; and, one or
more local stations each including another MODEM, and operatively
connected to the telephone line through the other MODEM and to the central
station through the other MODEM, the telephone line, and the central
station MODEM. The central station MODEM carries out a polling of the
other MODEMs. The corresponding other MODEM undergoing the polling sends a
signal including a training signal and data to the central station MODEM.
The training signal sent from the other MODEM contains a carrier, a timing
signal, and at least two impulses, a time between the two impulses
defining a characteristic of the telephone line between the corresponding
other MODEM and the central station MODEM. The central station MODEM
includes a first unit for receiving and demodulating the signal including
the training signal, a second unit for detecting the carrier and the
timing signal, and performing a gain control and a pulling-in
synchronization to the corresponding other MODEM, and a third unit for
recovering the impulses and for performing an equalization and a carrier
phase control in response to the recovered impulses. The data contained in
the signal sent from the corresponding other MODEM is adjusted in response
to the adjusted gain, synchronization, equalization, and carrier phase.
The training signal sent from the other MODEM may be superimposed on the
carrier, the timing signal, and the impulses. The training signal may
contain a forward impulse prior to the impulses, for guarding the
impulses. The training signal may also contain a scramble signal having a
random pattern data, after the impulses. The training signal may further
contain a reverse impulse between the last impulse and the scramble
signal.
The second unit of the central station MODEM includes a circuit for
detecting the carrier, a circuit for controlling a gain, a circuit for
extracting the timing signal, and a circuit for rotating a phase of the
extracted timing signal.
The third unit of the central station MODEM includes a circuit for
recovering the received impulses, a circuit for equalizing a
characteristic to the other MODEM, and a circuit for controlling a carrier
phase to the other MODEM.
Conversely, the training signal sent from the other MODEM may contain the
carrier, the timing signal, and the impulses in series. The training
signal may further contain a forward impulse between the timing signal and
the first impulse, for guarding the impulses. The training signal may also
contain a scramble signal having a random pattern data after the impulses.
The training signal may further contain a reverse impulse between the last
impulse and the scramble signal.
The second unit of the other MODEM includes a circuit for detecting the
carrier, a circuit for controlling a gain, and a circuit for extracting
the timing signal and pulling-in the synchronization. The third unit of
the other MODEM includes a circuit for recovering the received impulses, a
circuit for equalizing a characteristic to the other MODEM, and a circuit
for controlling a carrier phase to the other MODEM.
The signal sent from the other MODEM may be a quadrature-amplitude-modified
signal.
According to another aspect of the present invention, there is provided a
method for training a MODEM in a MODEM communication system including a
telephone line device having a telephone line, a central station including
a MODEM operatively connected to the telephone line and one or more local
stations each including another MODEM operatively connected to the
telephone line, including the steps of: carrying out polling from the
MODEM in the central station to the other MODEMs in the local station
through the telephone line; sending a signal having a training signal
containing a carrier, a timing signal and at least two impulses and data,
from the polled MODEM to the central MODEM through the telephone line, a
time between the impulses defining a characteristic of the telephone line
between the polled MODEM and the central MODEM; receiving and demodulating
the training signal in the central MODEM; detecting the carrier and the
timing signal in the central MODEM to adjust a gain and to pull-in
synchronization to the polled MODEM; and recovering the impulses and
performing an equalization and carrier phase control in response to the
time of the recovered impulses.
The training signal may be formed as any one of the signals set forth
above.
The method for training a MODEM may further include the steps of
discriminating the scramble signal, and adjusting the equalization and
carrier phase control.
BRIEF DESCRIPTION OF THE DRAWINGS
Other objects and features of the present invention will be described below
in detail with reference to the accompanying drawings, in which:
FIG. 1 is a block diagram of a multipoint type MODE communication system to
which the present invention is applied;
FIGS. 2a to 2c are timing charts of signals from the local stations (LSs)
in FIG. 1 to the central station (CS) in FIG. 1;
FIGS. 3a and 3b are views of a signal interface between the host computer
or the data terminal and the MODEM;
FIGS. 4a to 4e are timing charts of the data transfer from the LS to the
CS;
FIG. 5 is a graph representing the transfer time between the MODEMS through
the telephone line;
FIG. 6 is a waveform of the training signal of the prior art;
FIGS. 7a to 7c are waveforms of the training signals of another prior art;
FIG. 8 is a conceptual view of the training signal of an embodiment of the
present invention;
FIGS. 9a to 9d are waveforms of the training signals shown in FIG. 8;
FIG. 10 is a block diagram of MODEMs of an embodiment of the present
invention;
FIG. 11 is a graph representing 16 QAM;
FIGS. 12a and 12b are views showing the training signal pattern of the
embodiment;
FIG. 13 is a block diagram illustrating the recovery of the impulse;
FIG. 14 is a table explaining the impulse recovery of FIG. 13;
FIGS. 15a to 15c are views of the training signal patterns of the
embodiment;
FIG. 16 is a circuit diagram of the MODEM 101 shown in FIG. 10;
FIGS. 17a to 17f are waveforms of the signals in FIG. 16;
FIG. 18 is a graph explaining the phase rotation in FIG. 16;
FIG. 19 is a graph illustrating the eye distortion of the prior art;
FIG. 20 is a graph illustrating the eye distortion of the embodiment;
FIG. 21 is a circuit diagram of the timing extraction unit shown in FIG.
16;
FIG. 22 is a circuit diagram of the phase rotation unit shown in FIG. 16;
FIG. 23 is a circuit diagram of the judging unit shown in FIG. 16;
FIG. 24 is a waveform of the impulse recovered in FIG. 16;
FIGS. 25 and 26 are waveforms explaining the phase hold in FIG. 16;
FIG. 27 is a flow chart explaining the operation of the circuit shown in
FIG. 16;
FIG. 28 is a table explaining the variety of trainings;
FIGS. 29 and 30 are block diagrams of the training portions in the MODEM
101 in FIG. 16;
FIG. 31 is another conceptual view of the training signal of another
embodiment of the present invention, corresponding to FIG. 8;
FIGS. 32a to 32d are waveforms of the training signals shown in FIG. 31,
corresponding to FIGS. 9a to 9d;
FIG. 33 is a circuit diagram of another embodiment, corresponding to FIG.
16; and
FIGS. 34a to 34c are views of the training signal patterns of the another
embodiment of FIG. 33, corresponding to FIGS. 15a to 15c.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Before describing preferred embodiments of the present invention, an
explanation will be given of the prior art for reference.
Referring to FIG. 1, a multipoint type MODEM communication system is
represented in a general form. The system includes a center station (CS)
containing a host computer (HOST) 100 and a central MODEM 101 connected
thereto, a telephone line (LINE) 300 and a plurality of local stations
LS.sub.1 to LS.sub.3. The local station, for example, LS.sub.1, contains a
data terminal (DTE) 200 and a local MODEM 201 connected thereto. The local
MODEMs 201, 211, and 221 are operatively connectable to the center MODEM
101 through the LINE 300. The LINE 300 includes an upstream line and a
downstream line and is commonly used for the MODEMs 201, 211, and 221,
thereby reducing the line-use fees. The CS sequentially carries out a
"POLLING" of the LS.sub.1 to LS.sub.3 through the downstream line of the
LINE 300. Upon receipt of the polling signal, the corresponding LS sends a
signal indicating whether or not a data transfer is necessary to the CS
through the upstream line of the LINE 300. When a data transfer is
necessary, the corresponding LS consecutively transmits signals to the CS
through the upstream line. In each local MODEM, as the characteristics of
the line from the local MODEM to the central MODEM were measured upon
installation, and the reception characteristics of each local MODEM to the
central MODEM 101 through the LINE 300 were previously adjusted, the
training in each local MODEM may be easily effected in response to the
polling carrier signal sent by the CS. Accordingly, the training in the
local MODEMs is not the subject of the present invention.
Referring to FIGS. 2a to 2c, signals SIG. 1 to SIG. 3 from the local MODEMs
201 to 221 to the central MODEM 101 are shown. The signals SIG. 1 to SIG.
3 are time-divided. The signal; typically SIG. 1, consists of series of
time-divided "SENTENCE" signals each containing the training signal TR and
the data DATA. The TR contains information to be used for the training,
i.e., adjusting the reception parameters, in the central MODEM 101. The
DATA contains standard signals for a MODEM protocol and/or data to be
actually transferred from the DTE.sub.1 200 to the HOST 100.
FIGS. 3a and 3b are graphs of a standard CCITT V. 24 interface between the
HOST and the central MODEM or the DTE and the corresponding local MODEM.
FIG. 3a shows the interface at the sending side, and FIG. 3b the interface
of the receiving side. FIGS. 4a to 4e are timing charts of the signals in
FIGS. 3a and 3b. Referring to FIGS. 3a, and 3b, and 4a to 4e, the
operation in the system of FIG. 1 will be briefly described. The HOST 100
outputs a "REQUEST TO SEND (RS)" signal to the MODEM 101 at a time
t.sub.11. The MODEM 101 outputs the same signal to the corresponding LS
through the downstream line of the LINE 300. The corresponding local
MODEM, typically MODEM 201, receives the RS signal and transmits the
received RS signal to the DTE.sub.1 200. Upon receipt of the RS signal,
the DTE.sub.1 200 outputs the TR signal of SIG. 1 to the MODEM 101 through
the MODEM 201 and the LINE 300, specially the upstream line. The MODEM 101
starts the training for the LS.sub.1 in response to the reception of the
TR signal. The MODEM 101 detects a carrier included in the TR signal and
outputs a "CARRIER DETECTION (CD)" signal at a time t.sub.12. Anticipating
the completion of the training in the MODEM 101, the MODEM 201 may output
a "CLEAR TO SEND (CS)" signal indicating a permission to send to the MODEM
101 at a time t.sub.13. Upon receipt of the CS signal through the MODEM
101, the HOST 100 outputs a "SEND DATA (SD,)" signal to the MODEM 201 and
the DTE.sub.1 200 through the MODEM 101. The MODEM 201 outputs a "SENDING
TIMING 2 (ST2)" to the DTE.sub.1 200. Subsequently, the DTE.sub.1 200
outputs the DATA containing the MODEM standard signals and actual data of
the SIG. 1 to the MODEM 101 through the MODEM 201. The MODEM 101 outputs a
"RECEIVING DATA (RD)" signal and a "RECEIVING TIMING (RT)" signal to the
HOST 100, signifying that the DATA from the DTE.sub.1 200 is being
received at the HOST 100. After completion of the transfer of the DATA,
the HOST 100 makes the RS signal low level. The MODEM 101 also makes the
CD signal and the RD signal low level. In response to the change in level
of the signals, the MODEM 201 makes the CS signal low level. Accordingly,
one "SENTENCE" transfer is completed through the above operation.
In FIGS. 4a and 4b, the time between times t.sub.11 and t.sub.13, i.e.,
from the RS signal to the CS signal, is known as the "RS-CS" time and
designates the training time. As described above, the RS-CS time should be
within 50 ms for a 4800 bps transfer line and 253 ms for a 9600 bps
transfer line, etc.
FIG. 5 is a graph representing the data transfer characteristics. Curve C1
represents the characteristics for a 9600 bps transfer line and curve C2
the characteristics for 4800 bps transfer line. On curve C1, the
permissible training time is 253 ms and the transfer time is defined along
curve C1 in response to the number of DATA. Note that the transfer time of
the 9600 bps line is longer than that of the 4800 bps line when the DATA
is less than 256 bytes. This is the paradox mentioned before. In general,
the DATA may be N.sub.1, for example, several bytes, smaller than 256
bytes. When the training time of the 9600 bps line is still long, as
mentioned above, the high speed transfer characteristic is not exhibited.
If the training time is shortened to a time t.sub.10 in FIG. 5, the curve
C1 may be shifted to a dotted curve C3 while keeping the same slope as for
9600 bps. In this case, the transfer time of the 9600 bps line for the
DATA N.sub.1 will become shorter than the time of the 4800 bps line.
Accordingly, shortening the training time, i.e., the RS-CS time, is
strongly required.
The training and the training signal will now be described in more detail.
The receiving MODEM 101 performs a pull-in or set-up operation for the
training. The pull-in operation includes a synchronization pull-in,
equalization, and carrier phase control, etc. The sending MODEM 201
provides the TR signal including information used for the above operation.
Referring to FIG. 6, the training signal as an example of the prior art
disclosed in, for example, U.S. Pat. No. 3,962,637, includes a "TONE"
signal containing a carrier for performing the automatic gain control and
the pull-in of a phase of the carrier, a "TIMING" signal used for
adjusting the carrier phase and a pull-in of the sending timing, and an
"IMPULSE" signal containing two train impulses IPS.sub.1 and IPS.sub.2.
The first impulse IPS.sub.1 is used for controlling a carrier phase, and
the second impulse IPS.sub.2 is used for the equalization. The distances
between the MODEM 101 and the local MODEMs 201 to 221 are different, and,
responses of an impulse depend on these distances. Accordingly, the MODEM
101 measures a response of an impulse sent from the local MODEM and judges
the line characteristics. That is, the MODEM 101 computes distortion of
the first impulse IPS.sub.1 and performs a carrier automatic phase
control, and similarly, computes distortion of the second impulse
IPS.sub.2 and also performs the equalization, and thus, carries out a
fine-adjusting of a carrier automatic phase controller (CAPC) therein,
initial-setting of equalization coefficients to an equalizer (EQL)
therein, and resetting of a code-converter also therein. Subsequently, the
DATA is received and adjusted by the CAPC, the EQL, etc., ensuring
reception of the DATA. In the prior art, a time for training is defined by
a maximum distance between the central MODEM 101 and the most remote local
MODEM, for example, MODEM 221, and is 30 ms, for a 9600 bps line.
Referring back to FIG. 5, the training time T.sub.TR1 of 30 ms for a 9600
bps line is shorter than the time for a 4800 bps line, and thus is greatly
improved. Still, a further shortening of the training time was attempted.
Referring to FIGS. 7a to 7c, other waveforms of the training signals of
another prior art are shown. In FIGS. 7a to 7c, the training signal of the
signal sent from the local MODEM consists of the tone and the timing,
respectively, corresponding to those in FIG. 6, and a single impulse. The
impulse is positioned at a time from the beginning of the tone to the
impulse, for example, T.sub.11 in FIG. 7a, defining the distance between
the central MODEM 101 and the local MODEM 201. The MODEM 101 measures the
time from the beginning of the tone to the impulse and adjusts the
parameters therein. The impulses in FIGS. 7a to 7c are respectively used
for both the CAPC and th EQL set forth above. In FIG. 6, times between the
impulses IPS.sub.1 and IPS.sub.2 and between the impulse IPS.sub.2 and the
DATA are defined on the basis of the maximum distance between the central
MODEM and the local MODEM, taking into consideration the measurement of
the impulse response. Although the times T.sub.11 to T.sub.13 in FIGS. 7a
to 7c are defined as arbitrary values, a discriminating between them can
be carried out. As a result, in the second prior art with reference to
FIGS. 7a to 7c, the training time is reduced to 15 ms for the 9600 bps
line, a half of the above. The prior art in question, however, suffers
from a disadvantage in that an accurate timing adjustment is difficult.
This is because the tone and the timing, are not adjusted when they are
received, and the received impulse is adjusted to the line characteristic
by receiving the tone and the timing and adjusting the parameters using
the same, consequently, the detection of the time from the beginning of
the tone to the impulse may become vague. Accordingly, the training time
can not be made shorter than a predetermined time, i.e., 15 ms for the
9600 bps line. In addition, the prior arts suffer from another
disadvantage in that the carrier detection and the pull-in of the AGC are
difficult because the tone is a single spectrum of .pi./.pi. signals
consisting of a pair of signals shifted by 180 degrees relative to each
other in a 16 quadrature amplitude modification (QAM) plane, which will be
described later.
The present invention will now be described.
FIG. 8 shows a basic scheme of the TR signal of a first embodiment of the
present invention. In FIGS. 6 to 7c, the TR signal consists of the
series-arranged tone, timing signal, and impulse(s). In FIG. 8, the tone,
the timing signal, and the impulses are superimposed upon each other.
Referring to FIGS. 9a to 9d, specific waveforms of the TR signal in FIG. 8
are shown. In FIGS. 9a to 9d, the tone and the timing signal are omitted
to simplify the drawings. A signal SIG.A in FIG. 9a includes a TR signal
containing a guard impulse GIP.sub.1 and first and second impulses
IPS.sub.11 and IPS.sub.12, and the DATA. A second TR signal in FIG. 9b
contains the forward guard impulse GIP.sub.1 and two impulses IPS.sub.21
and IPS.sub.22 and a reverse guard pulse GIP.sub.2. A third TR signal in
FIG. 9c contains the forward guard impulse GIP.sub.1, two impulses
IPS.sub.31 and IPS.sub.32, the reverse impulse GIP.sub.3, and a scramble
pattern Z (SCRZ) consisting of random pattern data. A fourth TR signal in
FIG. 9d contains the forward guard impulse GIP.sub.1, two impulses
IPS.sub.41 and IPS.sub.42, the reverse impulse GIP.sub.4 and the SCRZ.
The concept of the embodiment is that the local MODEM outputs the
corresponding one of the TR signals in FIGS. 9a to 9d, the central MODEM
receives the TR signal, recovers the two impulses, for example, IPS.sub.11
and IPS.sub.12, and detects a time T.sub.A between the recovered impulses
to determine parameters for adjusting the CAPC and the EQL, etc. The times
T.sub.A, T.sub.B, T.sub.C, and T.sub.D show a line characteristic defined
by the distances between the central MODEM, and the local MODEMS, but do
not correspond to the time between the impulses IPS.sub.1 and IPS.sub.2 in
FIG. 6. The times T.sub.A to T.sub.D may be arbitary values sufficient to
carry out a discrimination between them.
In FIGS. 9a to 9d, the forward guard impulse GIP.sub.1 and the reverse
guard impulses GIP.sub.2, GIP.sub.3 and GIP.sub.4 are not essential to the
present invention. When the distance between the central MODEM and the
local MODEM is short, the guard impulse should not be provided, because
line noises and distortion of the signal may be neglected. In this case,
the first impulse, for example, IPS.sub.11, may function as a guard
impulse. The central MODEM may easily and correctly recover the impulses
and detect the time therebetween. When the distance in question is an
intermediate distance, the guard impulse GIP.sub.1 should be provided to
ensure the reception of the two impulses. When the distance is long, the
reverse guard impulse should be provided in addition to the forward guard
impulse.
A time between the forward guard impulse and the first impulse, for
example, IPS.sub.11, may be constant. Similarly, a time between the last
impulse, for example, IPS.sub.22 and the reverse impulse GIP.sub.2 also
may be constant.
The processing of the guard impulses is not described in the following
text.
In FIGS. 9c and 9d, the SCRZ may be provided to carry out a fine adjustment
of the equalizer.
As set forth above, with reference to FIG. 8, the embodiment will shorten
the training time by using the shortened training signal in which the
tone, the timing and the impulses, and in addition, the SCRZ, are
superimposed. Note, this embodiment will solve the above problems without
the provision of a parallel operation of the EQL, the CAPC, etc., which
may be easily conceived in the art. In other words, the embodiment will
solve the above problems without increasing the cost of the MODEM, as will
be disclosed later.
Below, an explanation will be given of the basic nature of MODEM training
in embodiments of the present invention, with reference to an example of
data transmission between a central station CS and local stations LS shown
in FIG. 1.
FIG. 10 is a general schematic view of MODEM of an embodiment of the
present invention. FIG. 11 is a graph showing the principle of a 16 QAM.
FIGS. 12a and 12b are views for explaining the training pattern in an
embodiment of the present invention, and FIG. 13 is a block diagram
explaining impulse recovery.
FIG. 10 shows just the transmission unit in a MODEM 201 connected to a data
terminal DTE.sub.1 200 and just the reception unit in a MODEM 101
connected to a HOST 100. When the MODEM 201 receives a transmission
request signal RS from the DTE.sub.1 200, a training data generator (TRG)
202 generates training data TR.sub.O and supplies the same to a modulator
(MOD) 203. The modulator MOD 203 sends a carrier signal CR from a carrier
signal generator (CRG) 204 through a transmission route (line) 300 to the
MODEM 101 connected to the HOST 100 under quadrature amplification
modulation (QAM) based on this training data TR.sub.O. The MODEM 101
performs the initialization for data reception based on the received
training signal TR. In anticipation of the completion of the
initialization, the MODEM 201 issues a transmission enable signal CS and
begins to send a signal modulated by the send data SD to the MODEM 101.
The modulator (MOD) 203 of the MODEM 201 performs quadrature amplification
to 16 value upon, for example, the carrier signal CR from the carrier
generator CRG 204, as shown in FIG. 11. The signal points shown in FIG. 11
correspond to the amplitude and phase of the modulated signals and can be
expressed by complex numbers.
The MODEM 101 of FIG. 10 has a demodulator (DEM) 102 for receiving and
demodulating a signal transmitted from the MODEM 201, an impulse
extraction unit (IMP) 103 for extracting a signal corresponding to an
impulse signal from training data demodulated in the demodulator (DEM)
102, and a signal recovery unit (SRC) 104 which extracts RS-CS time data
by the time difference of two extracted impulse signals and is subjected
to training by the training data and the impulses.
The training pattern is comprised of three segments SEG1, SEG2, and SEG3,
as shown in FIG. 12a. Below, an explanation will be made of the pattern of
each segment.
The first transmission pattern of the training pattern, i.e., the segment
SEG1, preferably satisfies the following conditions: (1) CD detection is
easy, (2) AGC (automatic gain control) pull-in can be performed at high
speeds (i.e., the data of the line level can be extracted accurately), and
(3) there is a timing component. The pattern which satisfies these three
conditions is one in which the phase of FIG. 11 is shifted 90 degrees and
intersects points A and B of the same amplitude. As the training start
pattern of segment SEG1, an "AB" pattern comprised of the six symbols
"ABABAB" is used. Conventional tones are single spectrum in nature, so
there is no timing component and the .pi./.pi. signal (pair of signals
shifted 180 degrees in FIG. 11) are insufficient for carrier detection or
AGC pull-in.
Next, the segment SEG2 has to be a pattern able to recover the first
impulse together with the segment SEG1. On the other hand, the impulse
recovery algorithmn at the reception side, i.e., the MODEM 101, delays the
input, i.e., received signal, by one symbol's worth of time by the tap T1
152, as shown in FIG. 13. It obtains the sum with the received signal by
the adder A1 153 and further delays this by one symbol's worth of time by
the tap T2 155. It then obtains the difference from the added output by
the adder A2 156 to recover the impulse. Therefore, the segment SEG 2 is
designated as "X.sub.1 X.sub.2, X.sub.3, X.sub.4, X.sub.5, X.sub.6 " and
the optimal symbols for the impulse recovery are determined with reference
to FIG. 14. The input IN, comprised of the segments SEG1 and SEG2, is
"A,B,A,B,A,B, X.sub.1, X.sub.2, X.sub.3 X.sub.4, X.sub.5, X.sub.6 ". If
there is a one-symbol shift at the tap T1 152, it becomes SX.sub.1.
Therefore, the added output at the adder A1 153 becomes SX.sub.2.
Further, if the added output SX.sub.2 is shifted one symbol's worth by the
tap T2 155, it becomes SX.sub.3. When the difference is obtained by the
adder A2 156, the output OUT of the difference becomes "A,B,0,0,0,0
X.sub.1 -A, X.sub.2 -B, X.sub.3 -X.sub.1, X.sub.4 -X.sub.2, X.sub.5
-X.sub.3, X.sub.6 -X.sub.4, -X.sub.5, -X.sub.6 ".
Therefore, X.sub.1 to X.sub.6 for recovery of the first impulse Z of FIG.
12b are determined as follows with reference to the QAM pattern of FIG.
11:
______________________________________
X.sub.1 - A = z .thrfore.X.sub.1 = -A
X.sub.2 - B = 0 .thrfore.X.sub.2 = B
X.sub.3 - X.sub.1 = 0
.thrfore.X.sub.3 = -A
X.sub.4 - X.sub.2 = 0
.thrfore.X.sub.4 = B
X.sub.5 - X.sub.3 = 0
.thrfore.X.sub.5 = -A
X.sub.6 - X.sub.4 = 0
.thrfore.X.sub.6 = B
______________________________________
The maximum power of X.sub.1 -A is obtained with X.sub.1 =-A, so is with
the point C rotated 180 degrees from the point A in the phase plane of
FIG. 11. Therefore, X.sub.1 =X.sub.3 X.sub.5 -C. On the other hand,
X.sub.2, X.sub.4, X.sub.6 is the point B rotated 90 degrees from the point
C. In this way, the pattern of the segment SEG2 for recovering the first
impulse Z can be found from the "CB" pattern comprised of the six symbols
"CBCBCB".
If the segment SEG1 is determined as an "AB" pattern and the segment SEG2
as a "CB" pattern in this way and the segment SEG3 is similarly sought in
the same way as above as a pattern enabling recovery of a second impulse
with the segment SEG2, then the segment SEG3 becomes a "DC" pattern
comprised of six symbols "DCDCDC".
The training pattern becomes as shown in FIG. 11. The patterns in the
segments SEG1, SEG2, and SEG3 are comprised of points perpendicularly
intersecting in the phase plane in FIG. 11. Among the segments SEG1, SEG2,
and SEG3, it should be noted, one of the component points of a former
segment is included as one of the component points of a latter segment.
On the other hand, if the segment SEG2 is X.sub.2 -B=Z (impulse), the
X.sub.2 =B and X.sub.1 =A, X.sub.3 =A, X.sub.3 =X.sub.2 =-B, X.sub.5
=X.sub.3 =A, X.sub.5 =X.sub.4 =D and the pattern may be a "DA" pattern
comprised the six symbols "DADADA" too. In this case, either the "DC"
pattern or the "BA" pattern may be selected for the segment SEG3.
In this way, as shown in FIG. 12a, the segments SEG1, SEG2, and SEG3 are
used for the recovery of the second impulse.
The time difference T of the recovered impulse can be varied by the symbol
length of the second segment SEG2.
According to this training pattern, a variety of the RS-CS times is
possible, as shown in FIGS. 15a to 15c.
In other words, by making the symbol length of the second segment SEG2 of
the transmission training pattern of a local station LS1 a close distance
away T.sub.1, the symbol length of the second segment SEG2 of the
transmission training pattern of a local station LS2 a medium distance
away T.sub.2, and the symbol length of the second segment SEG2 of the
transmission training pattern of the local station LS3 a long distance
away T.sub.3, the parent station can discriminate between the RS-CS times
by the time difference T.sub.1 to T.sub.3 of the recovered impulses.
Further, by adding to the medium distance and long distance local stations
LS2 and LS3 a scrambler Z (SCRZ) signal or a binary random signal as a
training pattern, it is possible to send a fine adjustment pattern of an
automatic equalizer for fine adjustment of the equalizer.
The segments of such a training pattern perpendicularly intersect each
other in the phase plane in FIG. 11, so there is no interference among
segments and transmission of an impulse is possible in the smallest
duration, thus making possible dense impulse transmission.
Further, use is made of the outermost data points of the phase plane in
FIG. 11, e.g., point A, point B, point C, and point D, so the maximum
signal energy is obtained and the S/N ratio is improved.
Further, since training of an equalizer is possible using the average value
of two recovered impulses, the noise resistance characteristics are
greatly improved.
This applies not only to the 16 value QAM illustrated in FIG. 10, but also
to an 8 value QAM and 4 value QAM.
Next, an explanation will be given of training in the MODEM 101 connected
to the HOST 100 which receives the above-mentioned training signal.
FIG. 16 is a constitutional view showing in detail the reception unit,
illustrated in FIG. 10, of the MODEM 101 which performs the relevant
training. Reference numeral 1 indicates a band pass filter (BPF) which
limits the band of the signal received from the line 300. Reference
numeral 2 is an analog/digital converter (ADC) which converts the analog
signal output from the band pass filter (BPF) 1 at a period of the
sampling clock, mentioned later, to a digital received signal. Reference
numeral 3 is a demodulator (DEMOD) which demodulates the | | |
