Phase perturbation compensation system.

0001This invention relates to a phase perturbation compen­sation system for compensating a phase-perturbed signal in a data modem.

0002The invention has a particular application to high-speed modems in switched carrier configurations, such as multi­point networks and half-duplex point-to-point configurations.

0003It is desirable in such networks and configurations for training sequences, which are applied at the commencement of transmissions, to be as short as possible.

0004One problem which arises in modem communication systems is that the communication channel introduces various kinds of deterioration, including carrier phase jitter and carrier frequency offset. For instance, phase jitter introduced on transmission lines such as telephone lines used for data communication is caused by AC ripple in the DC power supply for oscillators used in carrier systems, by incomplete band filtering, and by imperfections in quantization in digital systems. Frequency offset on transmission lines such as telephone lines used for data communication is caused in FDM (Frequency Division Multiplex) systems, since on most long haul lines there is a slight difference in frequency between the modulat­ing and the demodulating carriers. High values for phase jitter and frequency offset can occur, particularly on long haul telephone lines such as are used for half-­duplex point-to-point communication.

0005In both multipoint and half-duplex point-to-point config­urations successive transmissions generally start at random instants. Therefore, it is not possible to start in a phase jitter compensation circuit from the state with previously acquired values and to continue tracking.

0006U.S. Patent No. 4,320,526 discloses an adaptive phase jitter compensator which includes a filter section which receives a derived phase error signal and its own weighted sum output to provide the difference between them as noisy estimates of the actual phase via taps of a delay line. A multiplier receives each tap signal to form a product used to update the tap weights used in the filter section. An up-modulator modulates a carrier sig­nal with the weighted sum output, the up-modulator output being applied to a look-up table to form components com­patible with the received data signal. A complex multi­plier multiplies these components and the equalized data signal to provide both demodulation and phase jitter compensation.

0007It is an object of the present invention to provide a phase perturbation compensation system which achieves fast and reliable phase jitter compensation.

0008Therefore, according to the invention, there is provided a phase perturbation compensation system for compensating a phase-perturbed signal in a data modem, characterized by: phase jitter compensation means adapted to receive a phase error signal and including a tapped delay line having a plurality of output taps, weighting means inc­luding a corresponding plurality of weighting coefficient generating means adapted to adaptively generate and to apply respective weighting coefficients to said plurality of output taps, thereby forming a plurality of weighted tap signals, summing means adapted to sum said weighted tap signals thereby providing a phase error prediction signal and feedback means adapted to feed back said error prediction to input means of said tapped delay line; reference signal determination means responsive to said phase error prediction means to provide a reference signal; error correction means adapted to correct said phase-perturbed signal in dependence on said reference signal; and storage means adapted to store a plurality of parameters developed in said weighting coefficient gen­erating means during an initial modem training sequence and to restore the stored parameters to said weighting coefficient generating means during a subsequent modem training sequence.

0009It will be appreciated that in a system according to the invention, the storage and retrieval of the weighting coefficient parameters enables the achievement of a very rapid adjustment of the phase jitter compensation means to achieve proper tracking.

0010An embodiment of the present invention will now be des­cribed by way of example with reference to the accom­panying drawings, in which:- <ul id="ul0001" list-style="none"><li>Fig. 1 is a block diagram of a multipoint data modem communication system including a master modem and a plur­ality of remote modems;</li><li>Fig. 2 is a block diagram of a remote modem trans­mitter;</li><li>Fig. 3 is a block diagram of the master modem receiver;</li><li>Fig. 4 is a more detailed block diagram of the automatic phase control circuit shown in Fig. 3;</li><li>Fig. 5 is a diagram of the phase and amplitude cor- rection circuit shown in Fig. 4;</li><li>Fig. 6 is a diagram of the amplitude error cor­rection determination circuit shown in Fig. 4;</li><li>Fig. 7 is a diagram of the reference vector deter­mination circuit shown in Fig. 4;</li><li>Figs. 8A and 8B are a diagram of the phase jitter compensation determination circuit shown in Fig. 4;</li><li>Figs. 9A to 9C are diagrams of coefficient generat­ing circuits used in generating certain coefficients shown in Fig. 8;</li><li>Fig. 10 is a diagram of the vector generator circuit shown in Fig. 4;</li><li>Fig. 11 is a diagram of the decision circuit shown in Fig. 4; and</li><li>Fig. 12 is a diagram of a half-duplex point-to-point data modem communication system.</li></ul>

0011Referring first to Fig 1, there is shown a multipoint data modem communication system 10 including a master modem 12 and three remote modems 14, 16, 18. In prac­tice, a greater of lesser number of remote modems may be employed. The master modem 12 is connected to a master data terminal equipment (DTE) unit 20 and the remote modems 14, 16, 18 are connected to remote DTE units 22, 24, 26. The master modem 12 is connected over a 4-wire telephone transmission line 28 to a branch point 30 which is connected over 4-wire telephone transmission lines 32, 34, 36 to the respective modems 14, 16, 18. As is con­ventional, each of the 4-wire telephone transmission lines 28, 32, 34, 36 includes a transmit pair of lines and a receive pair of lines.

0012Data is normally transmitted at a data bit rate of 14,400 bits per second, using a carrier frequency of 1800 Hz and a modulation frequency of 2400 Hz (symbol rate of 2400 baud). The modems operate with a sample clock frequency having a nominal value of 9600 Hz, that is, four times the modulation frequency.

0013Referring now to Fig. 2, there is shown a block diagram of the modem transmitter section of one of the modems 14, 16, 18 shown in Fig. 1. Data signals from the associated data terminal equipment (DTE) are applied over an input line 50 to a scrambler 52. The output of the scrambler 52 is connected over a line 54 to an input of an encoder 56. An initialization control circuit 57 is connected over a line 58 to the encoder 56. The initialization control circuit 57 enables the encoder 56 to provide training signals as will be described hereinafter. The output of the encoder 56 is connected over a line 59 to a low-pass filter 60 the ouput of which is connected over a line 62 to a modulator 64. The output of the modulator 64 is connected over a line 66 to a digital-to-analog converter 68 which has an output 70 connected to the associated telephone transmission line.

0014Referring to Fig. 3, there is shown a block diagram of the master modem receiver 80. The signal received from the associated transmission line is applied over an input line 82 to an analog-to-digital converter 84 forming digital sampling means. The analog-to-digital converter 84 is connected over a line 86 to a band-pass filter 88, an energy on/off control circuit 90 and a timing recovery circuit 92. The timing recovery circuit 92 is connected over a line 94 to the analog-to-digital converter 84 to control the sampling times. The energy on/off control circuit 90 is connected over a line 96 to an initial­ization control circuit 98. The initialization control circuit 98 is connected over a line 100 to the timing recovery circuit 92, over a line 102 to a gain control circuit 104, which is also connected to an output of the band-pass filter 88 over a line 106. The line 106 trans­mits complex-valued signals, as is indicated by the double line symbol. It should be understood that when the energy on/off control circuit 90 detects energy on the line 86, it provides a signal over the line 96 to initiate operation of the initialization control circuit 98. The initialization control circuit 98 provides con­trol signals over the control lines 100, 102, 103 to the timing recovery circuit 92, the gain control circuit 104, and an automatic phase control circuit 118.

0015The gain control circuit 104 is connected over a line 108 to a demodulator 110 which is connected over a line 112 to an equalizer 114. The equalizer 114 is connected over a bidirectional bus line 116 to the automatic phase con­trol circuit 118 which is connected over a bidirectional bus line 120 to a decision circuit 122. The automatic phase control circuit 118 receives a RESET input over the line 103 from the initialization control circuit 98. The decision circuit 122 is connected over a line 124 to a descrambler 126 which has an output line 128 on which DATA signals are provided representing data received by the modem receiver 80.

0016Also included in the modem receiver 80 is a receiver parameter storage unit 130 which is connected over bidir­ectional bus lines 132, 134, 136 to the gain control circuit 104, the equalizer 114 and the automatic phase control circuit 118. The receiver parameter storage unit 130 is used to store equalizer coefficients and receiver parameters at locations asssociated with the respective transmitting modems during initial training sequences and to retrieve the stored coefficients and parameters during subsequent training sequences. The acquired parameters are stored at locations corresponding to the particular remote modem which generated these parameters.

0017It should be understood that the present invention is concerned essentially with the automatic phase control circuit 118, which will be described in detail herein­after, and that a detailed understanding of the remaining circuits shown in Fig. 3 is not essential for an under­standing of the present invention. Consequently such circuits will not be described in detail herein.

0018At the commencement of the first transmission by each of the remote modems 14, 16, 18 there is transmitted an initial training sequence consisting of six segments SG₁ - SG₆, as shown in Table 1: <tables id="tabl0001" num="0001"><table frame="all"><title>Table 1:</title><tgroup cols="7" colsep="1" rowsep="1"><colspec colnum="1" colname="col1" colwidth="22.50mm" /><colspec colnum="2" colname="col2" colwidth="22.50mm" /><colspec colnum="3" colname="col3" colwidth="22.50mm" /><colspec colnum="4" colname="col4" colwidth="22.50mm" /><colspec colnum="5" colname="col5" colwidth="22.50mm" /><colspec colnum="6" colname="col6" colwidth="22.50mm" /><colspec colnum="7" colname="col7" colwidth="22.50mm" /><thead valign="top"><row><entry namest="col1" nameend="col7" align="center"><u style="single">Initial Training Sequence</u></entry></row><row><entry namest="col1" nameend="col1" /><entry namest="col2" nameend="col2" align="center">SG₁</entry><entry namest="col3" nameend="col3" align="center">SG₂</entry><entry namest="col4" nameend="col4" align="center">SG₃</entry><entry namest="col5" nameend="col5" align="center">SG₄</entry><entry namest="col6" nameend="col6" align="center">SG₅</entry><entry namest="col7" nameend="col7" align="center">SG₆</entry></row></thead><tbody valign="top"><row><entry namest="col1" nameend="col1" align="right">(1)</entry><entry namest="col2" nameend="col2" align="right">256</entry><entry namest="col3" nameend="col3" align="right">2976</entry><entry namest="col4" nameend="col4" align="right">128</entry><entry namest="col5" nameend="col5" align="right">78</entry><entry namest="col6" nameend="col6" align="right">48</entry><entry namest="col7" nameend="col7" align="right">48</entry></row><row rowsep="1"><entry namest="col1" nameend="col1" align="right">(2)</entry><entry namest="col2" nameend="col2" align="right">107</entry><entry namest="col3" nameend="col3" align="right">1240</entry><entry namest="col4" nameend="col4" align="right">53</entry><entry namest="col5" nameend="col5" align="right">32</entry><entry namest="col6" nameend="col6" align="right">20</entry><entry namest="col7" nameend="col7" align="right">20</entry></row></tbody></tgroup></table></tables>

0019In Table 1 the first line (1) denotes the number of symbol intervals in the respective segments SG₁ to SG₆ and the second line (2) denotes the corresponding appro­ximate time in milliseconds. The segments are designated as follows: <tables id="tabl0002" num="0002"><table frame="all"><tgroup cols="3" colsep="1" rowsep="0"><colspec colnum="1" colname="col1" colwidth="52.50mm" /><colspec colnum="2" colname="col2" colwidth="52.50mm" /><colspec colnum="3" colname="col3" colwidth="52.50mm" /><tbody valign="top"><row><entry namest="col1" nameend="col1" align="left">SG₁:</entry><entry namest="col2" nameend="col2" align="left">Segment 1:</entry><entry namest="col3" nameend="col3" align="left">Alternations (180° phase alternations)</entry></row><row><entry namest="col1" nameend="col1" align="left">SG₂:</entry><entry namest="col2" nameend="col2" align="left">Segment 2:</entry><entry namest="col3" nameend="col3" align="left">Equalizer Conditioning Pattern</entry></row><row><entry namest="col1" nameend="col1" align="left">SG₃:</entry><entry namest="col2" nameend="col2" align="left">Segment 3:</entry><entry namest="col3" nameend="col3" align="left">Configuration Sequence</entry></row><row><entry namest="col1" nameend="col1" align="left">SG₄:</entry><entry namest="col2" nameend="col2" align="left">Segment 4:</entry><entry namest="col3" nameend="col3" align="left">Alternations (180° phase alternations)</entry></row><row><entry namest="col1" nameend="col1" align="left">SG₅:</entry><entry namest="col2" nameend="col2" align="left">Segment 5:</entry><entry namest="col3" nameend="col3" align="left">Equalizer Conditioning Pattern</entry></row><row rowsep="1"><entry namest="col1" nameend="col1" align="left">SG₆:</entry><entry namest="col2" nameend="col2" align="left">Segment 6:</entry><entry namest="col3" nameend="col3" align="left">Scrambled All Binary 1's.</entry></row></tbody></tgroup></table></tables>

0020It will be appreciated that the total number of symbol intervals in the initial training sequence is 3534, corresponding to a total time of 1472 milliseconds appro­ximately. Segments SG₁, SG₂ and SG₆ are conventional training signal segments, corresponding with CCITT Recommendation V.33. Segment SG₃ contains information relating to set-up conditions such as the data bit rate, the nature of the modulation, and other transmission-­related parameters. The segment SG₄ is utilized in cal­culations related to timing adjustment. The segment SG₅ provides a slight readjustment of the equalizer coeffi­cients which is needed as a result of transmitting the segment SG₃.

0021As mentioned above, during the initial training sequence the master modem receiver 80 acquires operating para­meters and equalizer coefficients and stores these para­meters and coefficients in the receiver parameter storage unit 130 at a location associated with the transmitting remote modem. Each subsequent transmission by the remote modem commences with a short training signal, referred to as a subsequent training sequence consisting of two seg­ments, as shown in Table 2: <tables id="tabl0003" num="0003"><img file="EP0332302A2_D0001.tif" /></tables>

0022Thus, the subsequent training sequence consists of a first segment formed by 25 symbol intervals of 180° phase alternations followed by a second segment of 48 symbol intervals of predefined symbols.

0023The remote modem may be identified by transmitting a set of identifying frequencies in synchronism with segment 1 of the subsequent training sequence. Thus, the previ­ously stored receiver parameters and coefficients can be retrieved from the identified location in the receiver parameter storage unit 130.

0024Referring now to Fig. 4, there is shown a block diagram of the automatic phase control circuit 118 and its con­nections with the equalizer 114 and the decision circuit 122. The output of the equalizer 114 is connected over a line 116a forming part of the bus line 116 (Fig. 3) to a phase and amplitude correction circuit 150 having an out­put connected over a line 120a forming part of the bus line 120 (Fig. 3) to the decision circuit 122. In addi­tion to the output 124 (Fig. 3), the decision circuit 122 has an output line 120b which carries a signal represent­ing the amplitude error (or in-phase error), an output line 120c which carries a signal representing quadrature error, and an output line 120d which carries a signal representing phase error. The lines 120b, 120c and 120d form part of the bus line 120 (Fig. 3), and are connected respectively to an amplitude error correction determina­tion circuit 152, a reference vector determination cir­cuit 154 and a phase jitter compensation determination circuit 156. Connections are made at nodes 158, 160 between the lines 120b, 120c and a line 116b which thereby carries a complex-valued signal. The line 116b is connected as a first input line of a complex conjugate multiplier 162, which has a second input line 116c from the reference vector determination circuit 154, and an output connected to a multiplier 164 to which is also applied a constant value c. The complex multiplier 162 multiplies the complex-valued input on the line 116b by the complex conjugate of the complex-valued input on the line 116c. Thus, the imaginary part of the complex-­valued signal on the line 116c is inverted by an inverter (not shown) prior to multiplication. The output of the multiplier 164 is connected over a line 116e to the equalizer 114.

0025The amplitude error correction determination circuit 152 is connected over a line 170 to the phase and amplitude correction circuit 150. The phase jitter compensation determination circuit 156 has an output connected over a line 172 to a vector generator circuit 174, which has a complex-valued output connected over a line 176 to the reference vector determination circuit 154. The complex-­valued output of the reference vector determination cir­cuit 154 is connected over a line 178 to the phase and amplitude correction circuit 150 and to the line 116c forming the second input to the complex multiplier 162. With this arrangement, it will be appreciated that the error feedback for the equalizer 114 which is effected over the lines 116b, 116d, 116e is provided with a proper backwards rotation.

0026Referring now to Fig. 5, there is shown a detailed dia­gram of the phase and amplitude correction circuit 150. The complex-valued output of the equalizer 114, on the line 116a, is applied to a complex multiplier 190, which also receives the complex output of the reference vector determination circuit 154, on the line 178. The multi­plier 190 effects a phase correction, using the output from the reference vector determination circuit 154. The complex-valued output of the multiplier 190 is applied over a line 192 to a multiplier 194 which also receives the output of the amplitude error correction determina­tion circuit 152 over the line 170. Thus, the multiplier 194 effects an amplitude error correction using the out­put of the amplitude error correction determination cir­cuit 152.

0027Referring now to Fig. 6, there is shown a detailed dia­gram of the amplitude error correction determination circuit 152. The input line 120b is connected to a multiplier 200 which also receives a coefficient value d₁. The output of the multiplier 200 is connected over a line 202 to an adder 204, the output of which is applied over a line 206 to a multiplier 208 which also receives a coefficient value d₂. The output of the multiplier 208 is applied over a line 210 to an adder 212 to which is applied a coefficient value d₃. The output of the adder 212 is applied over a line 214 to the output line 170 of the circuit 152, and via a feedback line 216 to a delay 218, the output of which is connected over a line 220 to an input of the adder 204.

0028The operation of the amplitude error correction deter­mination circuit 152 approximates that of a leaky inte­grator. The circuit 152 adapts quickly to compensate for amplitude errors and adapts slowly by leakage towards a target value 0.5. It should be understood that gain control in the modem receiver 80 is derived from three separate contributions. A first contribution is provided by the gain control circuit 104 (Fig. 3). A second con­tribution is provided by the amplitude correction discus­sed hereinabove with reference to Fig. 5. A third, small contribution is provided in the equalizer 114. In the circuit 152, the following coefficient values are utilized: d₁ = 0.0150 d₂ = 0.9998 d₃ = 0.0001

0029There is a target value of 0.5 for the output of the amplitude error correction determination circuit 152. The circuit 152 has internal feedback through the delay element 218, which also has the target value 0.5. The circuit 152 adapts its output quickly to compensate for an amplitude error such as a small, abrupt change of telephone line attenuation. Next, the circuit 152 adapts slowly by leakage and returns slowly to its target value of 0.5. During this time, the gain control circuit 104 (Fig. 3) adapts slowly to a new gain for level normal­ization following the aforementioned small, abrupt change of telephone line attentuation.

0030Referring now to Fig. 7, there is shown a detailed dia­gram of the reference vector determination circuit 154 (Fig. 3). The input line 120c is connected to multi­pliers 230 and 232 which have coefficients b and a₁ applied thereto, respectively. The output of the multi­plier 232 is connected to an adder 234 which has an out­put connected to an adder 236 and to a delay 238. The output of the delay 238 is connected to an input of a multiplier 240 to which is applied a coefficient a₂, and which has an output connected to an input of the adder 234. The output of the multiplier 230 is connected to an input of the adder 236. It will be appreciated that the components 232, 234, 238 and 240 function as a leaky integrator 242 for the averaged quadrature error feedback and the multiplier 230 functions for proportional quadra­ture error feedback.

0031The output of the adder 236 is connected to an complex-­valued feedback loop 244 which includes a multiplier 246, an adder 248 and a delay 250, connected as shown in Fig. 7. The output of the adder 248 is connected to a complex multiplier 252 which also receives an input over the line 176 from the vector generator, and which provides compen­sation for phase jitter that is present in the output of the reference vector determination circuit 154.

0032Referring now to Figs. 8A and 8B, there is shown a detailed diagram of the phase jitter compensation determination circuit 156. The input line 120d is connected to a multiplier 260 which also receives a coefficient value f₁, and which has an output connected to multipliers 262, 264 which receive coefficient values e₁, e₂, respectively. It will be appreciated that the multitpliers 260, 262, 264 provide gain adaptation. The output of the multiplier 264 is connected to a subtrac­tive input of an adder 266, the output of which is con­nected to the first delay 268-1 of a chain of sixteen delays 268-1 to 268-16 constituting a tapped delay line 269 having tap nodes 270-1 to 270-16 which are connected to respective multipliers 272-1 to 272-16 which receive inputs from respec-tive leaky integrator circuits 274-1 to 274-16 which provide adaptive weighting factors in the form of coefficient values for the multipliers 272-1 to 272-16. It should be understood that a greater or lesser number of delays and leaky integrator circuits may be utilized in alternative arrangements.

0033The outputs of the multipliers 272-1 to 272-16 are commonly connected to an adder 276 which has an output connected to an input of a multiplier 278 which also receives a coefficient value f₂. The output of the multiplier 278 forms a phase error prediction signal which is applied to the output line 172. The output of the multiplier 278 is also connected via a feedback line 280 to an input of a multiplier 282 which receives a coefficient value e₃. The output of the multiplier 282 is connected to an input of the adder 266. It will be appreciated that the output signal from the multiplier 282 corresponds to the phase jitter prediction after a gain adaptation, and that the output of the adder 266 corresponds to the noisy actual phase error.

0034The output of the multiplier 262 is connected to first inputs of multipliers 284-1 to 284-16 which have second inputs connected to the nodes 270-1 to 270-16. The outputs of the multipliers 284-1 to 284-16 are connected respectively to the leaky integrator circuits 274-1 to 274-16 which include respecive adders 286-1 to 286-16, delays 288-1 to 288-16 and multipliers 290-1 to 290-16. The multipliers 290-1 to 290-16 receive a common coefficient value e4, constituting the leakage factor. With this arrangement it will be appreciated that the outputs of the delays 268-1 to 268-16 are correlated with the output of the multiplier 262, with the correlation being effected by multiplication and integration.

0035It should be understood that the line 136 is connected to the delays 288-1 to 288-16 in the leaky integrators 274-1 to 274-16, such that the delay values acquired during initial training can be stored in the receiver parameter storage unit 130 and restored to these delays at the com­mencement of subsequent training. It should also be understood that at the commencement of subsequent train­ing, the delays 268-1 to 268-16 are cleared in response to a RESET signal on the line 103. Thus, during subse­quent training, the leaky integrators 274-1 to 274-16 provide weighting factors which are initially set and then updated synchronously. This is effected by multi­plication of the phase error signal on the line 120d (after gain adaptation by factors f₁, e₁ in the multi­pliers 260, 262) with the noisy estimate of the actual phase error shifted through the delays 268-1 to 268-16. The leaky integrators 274-1 to 274-16 operate as accumu­lators which store and update the weighting factors for the multipliers 272-1 to 272-16. The delays 268-1 to 268-16 of the delay line 269 contain successive estimates for the actual phase. The adder 276 combines the out­puts of the multipliers 272-1 to 272-16 to provide the weighted sum output.

0036The values of the various gain factor coefficients referred to above that are applied during the automatic phase control (APC) operations which take place during the last part of the initial training sequence and during subsequent training sequences are as shown in Table 3. <tables id="tabl0004" num="0004"><table frame="all"><title>Table 3</title><tgroup cols="3" colsep="1" rowsep="1"><colspec colnum="1" colname="col1" colwidth="52.50mm" /><colspec colnum="2" colname="col2" colwidth="52.50mm" /><colspec colnum="3" colname="col3" colwidth="52.50mm" /><thead valign="top"><row><entry namest="col1" nameend="col1" /><entry namest="col2" nameend="col2" align="center">During last part of initial training</entry><entry namest="col3" nameend="col3" align="center">During subsequent training</entry></row></thead><tbody valign="top"><row rowsep="0"><entry namest="col1" nameend="col1" align="left">initial e_1,i</entry><entry namest="col2" nameend="col2" align="char" char=".">0.25</entry><entry namest="col3" nameend="col3" align="char" char=".">0.0</entry></row><row rowsep="0"><entry namest="col1" nameend="col1" align="left">decrease e_1,d</entry><entry namest="col2" nameend="col2" align="char" char=".">0.99976</entry><entry namest="col3" nameend="col3" align="char" char=".">0.99217</entry></row><row><entry namest="col1" nameend="col1" align="left">final e_1,o</entry><entry namest="col2" nameend="col2" align="char" char=".">0.125</entry><entry namest="col3" nameend="col3" align="char" char=".">0.125</entry></row><row rowsep="0"><entry namest="col1" nameend="col1" align="left">initial e_2,i</entry><entry namest="col2" nameend="col2" align="char" char=".">0.21895</entry><entry namest="col3" nameend="col3" align="char" char=".">0.07812</entry></row><row rowsep="0"><entry namest="col1" nameend="col1" align="left">decrease e_2,d</entry><entry namest="col2" nameend="col2" align="char" char=".">0.99951</entry><entry namest="col3" nameend="col3" align="char" char=".">0.9325</entry></row><row><entry namest="col1" nameend="col1" align="left">final e_2,0</entry><entry namest="col2" nameend="col2" align="char" char=".">0.03125</entry><entry namest="col3" nameend="col3" align="char" char=".">0.03125</entry></row><row rowsep="0"><entry namest="col1" nameend="col1" align="left">initial e_3,i</entry><entry namest="col2" nameend="col2" align="char" char=".">0.99976</entry><entry namest="col3" nameend="col3" align="char" char=".">0.0</entry></row><row rowsep="0"><entry namest="col1" nameend="col1" align="left">decrease e_3,d</entry><entry namest="col2" nameend="col2" align="char" char=".">0.99927</entry><entry namest="col3" nameend="col3" align="char" char=".">0.875</entry></row><row><entry namest="col1" nameend="col1" align="left">final e_3,o</entry><entry namest="col2" nameend="col2" align="char" char=".">0.25</entry><entry namest="col3" nameend="col3" align="char" char=".">0.25</entry></row><row><entry namest="col1" nameend="col1" align="left">e₄</entry><entry namest="col2" nameend="col2" align="char" char=".">0.99997</entry><entry namest="col3" nameend="col3" align="char" char=".">0.99997</entry></row><row><entry namest="col1" nameend="col1" align="left">f₁</entry><entry namest="col2" nameend="col2" align="char" char=".">16.0</entry><entry namest="col3" nameend="col3" align="char" char=".">16.0</entry></row><row><entry namest="col1" nameend="col1" align="left">f₂</entry><entry namest="col2" nameend="col2" align="char" char=".">4.0</entry><entry namest="col3" nameend="col3" align="char" char=".">4.0</entry></row><row><entry namest="col1" nameend="col1" align="left">a₁</entry><entry namest="col2" nameend="col2" align="char" char=".">0.002</entry><entry namest="col3" nameend="col3" align="char" char=".">0.002</entry></row><row><entry namest="col1" nameend="col1" align="left">a₂</entry><entry namest="col2" nameend="col2" align="char" char=".">1.0</entry><entry namest="col3" nameend="col3" align="char" char=".">1.0</entry></row><row rowsep="1"><entry namest="col1" nameend="col1" align="left">b</entry><entry namest="col2" nameend="col2" align="char" char=".">0.016</entry><entry namest="col3" nameend="col3" align="char" char=".">0.016</entry></row></tbody></tgroup></table></tables>

0037It should be understood that the APC operations according to Table 3 commence, for the initial training, at a pre­determined time during the segment SG₂, Table 1, (for example, 500 symbol intervals after the commencement of the initial training sequence), and, for subsequent training sequences at the commencement of segment 2 of a subsequent training sequence (Table 2). Furthermore, it will be appreciated that the values of e₁, e₂ and e₃ vary during these APC periods, whereas the values of e₄, f₁, f₂, a₁, a₂ and b are fixed during such periods.

0038Circuits 298A, 298B and 298C for generating values of e₁, e₂ and e₃ which vary in accordance with Table 3 will now be described with reference to Figs. 9A to 9C. Referring first to Fig. 9A, the circuit 298A includes a delay 300A which is initially set with the value e_1,i. The output of the delay 300A is connected to a multiplier 302A, which also receives as an input the value e_1,d. The output of the multiplier 302A is connected to the input of an adder 304A which also receives the coefficient value e_1,0. The output of the adder 304A provides the output signal representing the coefficient e₁, which signal is also fed back to form the input of the delay 300A. With this arrangement, the value of e₁ will change from its initial value e_1,i to its final value e_1,0. The value of e_1,d is related to the speed at which e₁ con-verges to e_1,0. The values of e_1,i, e_1,d, e_1,0 are indicated in Table 3. The circuits for generating the coefficients e₂ and e₃, shown in Figs. 9B and 9C are identical to the circuit shown in Fig. 9A, with corres­ponding components identified by B and C suffixes, respectively. The appropriate input values are applied as shown in Figs. 9B and 9C to generate the coefficients e₂ and e₃.

0039Referring now to Fig. 10, there is shown a detailed dia­gram of the vector generator circuit 174 (Fig. 3), which receives a real-valued input representing a phase error prediction signal on the line 172 and provides a complex-­valued output vector on the line 176. The vector gener­ator circuit includes adders 310, 312 and multipliers 314, 316, 318, 320 connected as shown in Fig. 10. The adder 310 and the multipliers 314, 316 have coefficients c₀, c₂ and c₃ applied thereto with the following values:- c₀ = 1.0 c₂ = -0.497401 c₃ = -0.166147

0040These values are derived from the known approximation formulae: cos x = 1 - 0.497401 x² sin x = x - 0.166147 x³, which for -0.35 &lt;x&lt; 0.35, corresponding to a phase jitter level below 40 degrees, give a very accurate approxi­mation.

0041Referring now to Fig. 11, there is shown a detailed dia­gram of the decision circuit 122 (Fig. 3). The input line 120a, which carries a complex-valued input signal (p,q) is connected to a slicer 330. The slicer 330 provides, for each new complex-valued input (p,q), an output (p̂,q̂) on a line 332, which output corresponds to the target vector point selected from the set of signal vector points of the QAM signal constellation that is nearest to the input vector point. The slicer also provides a data signal on the line 124 corresponding to the data represented by the target vector point. The line 332 is connected to a table look-up unit 334 which provides on a line 335 an output (p*,q*) of which the phase corresponds to the phase of (p̂,q̂) and of which the vector length is equal to 1. The line 332 is also con­nected to a table look-up unit 336 which provides on a line 338 a real-valued output corresponding to the inverse of the vector length of (p̂,q̂). An adder 340 provides on a line 342 the error signal (p̂,q̂) - (p,q) which is applied via the line 342 as an input to a com­plex conjugate multiplier 344 which multiplies the value on the line 342 by a value (p*, -q*) derived from the line 335. The output of the complex multiplier 344 corresponds to the error signal (p̂,q̂) - (p,q) after a phase rotation to position the error along the real axis. The imaginary part of the multiplier output is provided on a line 346 and corresponds to the quadrature error. The real part is provided on a line 348. The lines 346, 348 are connected to the lines 120c, 120b, respectively. The line 346 is also connected to an input of a multi­plier 350 which has a second input connected to the output line 338 of the table look-up unit 336. The output of the multiplier 350 provides the phase error signal on the line 120d.

0042In operation, an initial, long training sequence in accordance with Table 1 is applied. Initially, the con­tents of the delays 268-1 to 268-16, and 288-1 to 288-16 in the phase jitter compensation determination circuit 156, and the delay 238 in the reference vector deter­mination circuit 154 are in a zero condition. During the initial training, the contents of these delays are synthesized to provide optimum compensation. At the conclusion of the initial training the master modem receiver 80 acquires receiver parameters and equalizer coefficients which are stored in the receiver parameter storage unit 130. These parameters include the contents of the delays 288-1 to 288-16 in the phase jitter compen­sation determination circuit 156, and the delay 238 in the reference vector determination circuit 154, the con­tents of these delays being transmitted over the bus line 136 to the storage unit 130.

0043During subsequent transmissions, a short training sequence in accordance with Table 2 is applied. After identification of the remote modem the relevant previ­ously stored parameters and coefficients are reloaded from the storage unit 130, including the values for the delays 288-1 to 288-16 in the phase jitter compensation determination circuit 156, and the delay 238 in the reference vector determination circuit 154.

0044Thus, the reference vector determination circuit 154, which effects frequency shift compensation, is loaded immediately with the appropriate value for the delay 238. Also, the phase jitter compensation determination circuit 156 is loaded immediately with the appropriate values for the weighting coefficients in the delays 288-1 to 288-16.

0045The delay line elements 268-1 to 268-16 of the delay line 269 start from zero since there is no estimate of the actual phase because the phase of the phase jitter fluctuation is unknown.

0046Next, the phase jitter compensation determination circuit 156 can train fast during the 48 predefined symbols of segment 2 of the subsequent training sequence, starting from the state with appropriate settings for the delays 288-1 to 288-16 and with an appropriate setting for the delay 238 in the reference vector determination circuit, for effecting frequency shift compensation.

0047Referring briefly to Table 3 above, it will be appre­ciated that the second column of values gives optimum settings for the gain values, including the varying values e₁, e₂ and e₃, the initial settings for which differ from the initial settings in the first column of values since during subsequent training a fast training is effected during the 48 predefined symbols starting from the state with appropriate settings for the weight­ing coefficients.

0048Although the invention has been described hereinabove in relation to the multipoint modem configuration shown in Fig. 1, it should be understood that the invention is also applicable to a half-duplex point-to-point configu­ration. Referring to Fig. 12, such point-to-point con­figuration includes a local station DTE unit 370 connected to a local modem 372. The local modem 372 is connected over a 2-wire transmission line 374 to a remote modem 376 to which is connected a remote DTE unit 378. On the 2-wire transmission line 374 communication is in a half-duplex mode, wherein transmissions may take place in either direction. Each modem 372, 376 includes a recei­ver corresponding to that shown in Fig. 3, except that the receiver parameter storage unit 130 utilizes storage locations for the single remote modem only, instead of for a plurality of remote modems.

0049With this half-duplex point-to-point configuration, during the first local to remote and first remote to local transmission, there is sent a long training sequence as in Table 1, during which receiver parameters and coeffients are acquired and saved, including the contents of the delays 288-1 to 288-16 and the delay 238 at each location.

0050In summary, there has been described a modem receiver wherein reliable and accurate adaptive phase tracking is achieved to compensate for phase jitter and frequency offset. It will be appreciated that in both multipoint systems and point-to-point systems, successive trans­missions may start at random instants, so that it is not possible to start in the phase jitter compensation deter­mination circuit from the state with all previous acquired values and to continue tracking. The present invention enables a very short training to be used wherein fast and reliable compensation for phase jitter and frequency offset is achieved. Thus, a very short training sequence can be used in conjunction with a high data transmission rate.