Combined Equalizer and Roll-Off Filter

TECHNICAL FIELD

This invention relates to a communication system employing a digital roll-off filter.

BACKGROUND ART

A quadrature amplitude modulation, referred to hereinafter as QAM, has been increasingly employed in a mul t i-channe1 digital communication system due to its high efficiency capability. Concept of a system configuration of a typical QAM communication system is schematically illustrated in Fig. 1. Explanation will be given hereinafter representatively with the I channel because the circuit con igurations of I and Q channels are symmetrical with each other. In a transmitting station, an internal frequency band (referred to hereinafter as IF) or a base band (both referred hereinafter to representatively as IF) of an I channel and a Q channel are input to input terminals of first digital roll-off filters 102ai and 102aq, respecti ely. Circuit configuration of roll-off filters will be described later in detail. Frequency

characteristics of the first roll-off filters are determined by the tap rating ratios stored in a first read-only memory, referred to hereinafter as a ROM, 113.

Output of first roll-off filter 102ai is input to a digital-to-analog (referred to hereinafter as D/A) converter 131. Unnecessary higher frequency spectrum generated in the output of D/A converter 1311 is eliminated by a low-pass filter 137 i . Outputs from low-pass filters 1321 and 132q respectively of the I channel and Q channel are input to a QAM modulator 134, to which a carrier frequency signal is also input from an carrier generator 133. A QAM modulated radio frequency signal is transmitted via a radio frequency amplifier (not shown in the figure) to a receiving station.

In the receiving station, a preamplifier (not shown in the figure) amplifies and converts the received radio frequency signal to an IF signal, which is then input to a QAM demodulator 154, to which a local frequency signal is input from a local frequency oscillator 156. I channel and Q channel signals output from QAM demodulator 154 is input via a low-pass filter 1521 to an analog-to-d igi ta 1 (referred to hereinafter as A/D) converter 151. Each bit line of a parallel digital signal output from A/D converter 1511 is input to a second roll-off filter 102bi. Frequency transmission characteristic of second roll-off filter 102b i is determined by tap rating ratios stored in, and output from, a ROM 142.

Frequency characteristics of the first and second roll- off filters are chosen such that overall transmission characteristics, i.e. a total of the frequency characteristics of the two filters in each channel, allow the second roll-off filter to output signal pulses in an adequately good shape which causes no i ntermodulat ion in the QAM-modulated signals, as well as to eliminate unnecessary upper frequency spectrum generated from the circuits, such as D/A converter, etc.

Even though the overall frequency characteristics that is the sum of both the first and second roll-off filters are set so that the pulse form at the output at second roll-off filter is in a good shape, fading or some other factors in the transmission system always varies the transmission characteristics, such as frequency vs amplitude, or frequency vs phase-delay, which accordingly deteriorate the pulse forms output from the second roll-off filter.

In order to remedy this deterioration an automatic equalizer 160 i is provided at the output of the second roll-off filter 102bi . Output of second roll-off filter 102bi is input via a pulse divider 146 i , which returns the signal to have the symbol rate, to a first input terminal of automatic equalizer 180 i . Moreover, an output of roll-off filter 102bq of Q channel is input to a second input terminal of automatic equalizer 160 i of the I channel. Symmetrically the same cross-connection is done in the Q

channe1.

Automatic equalizers are formed of a kind of roll-off filter, frequency characteristics of which are variably determined by tap rating ratios given from a calculating circuit 107. Calculating circuit 107 is formed of a micro computer system which monitors the pulse forms of the signal output from the automatic equalizer 160 i and calculates optimum values of the factors so that pulse form of the signal output therefrom is satisfactorily in a good shape.

The output signal is also input to a carrier regeneration circuit 155, an output of which is fed back to control local oscillator 156.

A problem of this circuit configuration is in that the provision of the automatic equalizer causes a cost increase in manu acturing the receiving station.

More important problem is in that, when the frequency characteristics is deteriorated by some cause located after the first roll-off filter and the cause must be urgently removed, it is impossible to locate the cause by checking eye patterns of the waveforms at check points Al , A2 , A3, B3, B2, Bl and BO, each located after the first roll-off filter, without time-consuming manual operations. This is because, even in a normal state where the second roll-off filter is outputting satisfactory waveforms the waveforms in the stages between the two roll-off filters are not in a good shape viewed in the eye diagram.

DISCLOSURE OF THE INVENTION It is a general abject of the invention to provide a digital roll-off filter additionally having a function of automatic equalizer, and a method to find a cause to deteriorate frequency characteristic of the transmission line employing a digital roll-off filter.

A transversal type digital roll-off filter receiving a single bit line of a parallel digital signal sampled n times from an analog signal carrying a pulse train having a pulse spacing T, comprises: (1) a transversal type delay line comprising an even number of delay elements each having a delay time T/n ; (2) nods between adjacent two of the delay elements; (3) a memory device for providing first tap rating ratios to control signals of the nods, respectively; and (4) a calculation circuit for monitoring pulse forms of an output signal of the roll-off filter, and calculating second tap rating ratios to additionally control central one of the nods and every n-th nods counted from the central nod, where the second tap rating ratios is calculated so as to be optimum to make the output pulse forms good in shape. Thus, the delay line act as a roll-off filter whose frequency characteristic is determined by the first tap rating ratios as well as acts as an automatic equalizer whose frequency characteristic is variably controlled by the second tap rat ing rat ios .

In a method of diagnosing electronic circuits used in the above communication system, the first memory device further comprises a second kind of tap rating ratios to be used to diagnose the system, where the first and second tap rating ratios are switchably output to the first roll-off filter. The second memory device further comprises a third kind of tap rating ratios to be used for diagnosing the system. The second kind tap rating ratios allows the first roll-off filter itself alone to output a satisfactorily good pulse shape at its output terminal, and the third kind of tap rating ratios allow the second roll-off filter to have a flat frequency characteris ic. The method to diagnose the system comprises the steps of (1) switching the first memory device so as to output the second kind tap rating ratios; and (2) comparing a width or a height of an eye in an eye diagram with a pred termined reference level, at a check points provided at a circuit after the first roll-off filter, so as to detect a location of a cause deteriorating the pulse forms.

The above-mentioned features and advantages of the present invention, together with other objects and advantages, which will become apparent, will be more fully described hereinafter, with re being made to the accompanying drawings which form a part hereof, wherein l ike numerals refer to like parts throughout.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 schematically illustrates a circuit configuration of a prior art communication system employing digital roll- off filters and an automatic equalizer;

Fig. 2 schematically illustrates a transmitting station of a first preferred embodiment of the present invention;

Fig. 3 illustrates a. digital roll-off filter of a first preferred embodiment of the present invention;

Fig. 4 schematically illustrates a circuit con iguration of a calculation circuit;

Fig. 5 schematically illustrates tap rating ratios output from a memory deviee and. from a calculation circuit of the first preferred embodiment;

Fig. 6 schematically illustrates a circuit configuration of an accumulator of a second preferred embodiment;

Fig. 7 schematically illustrates tap rating ratios output from a memory device and from a calculation circuit of the second preferred embodiment;

Fig. 8 illustrates a digital roll-off filter of a third preferred embodiment of the present invention;

Fig. 9 schematically illustrates tap rating ratios output from a memory deviee and from a calculation circuit of the third preferred embodiment;

Fig. 10 illustrates a digital roll-off filter of a fourth preferred embodiment of the present invention;

Fi . 11 schematically illustrates tap rating ratios

output from a memory device and from a calculation circuit of the fourth preferred embodiment;

Fig. 12 illustrates a digital roll-off filter of a fifth preferred embodiment of the present invention;

Fig. 13 illustrates a digital roll-off filter of a sixth preferred embodiment of the present invention;

Fig. 14 illustrates a digital roll-off filter of a seventh preferred embodiment of the present invention;

Fig. 15 illustrates a digital rol l-off filter of a eighth preferred embodiment of the present invention;

Fig. 16 schem tically illustrates a circuit configuration of a system where the method for diagnosing a. defect of the circuit applicable according to the present invent ion ;

Fig. 17(a) shows frequency characteristics of the first digital roll-off filter and an input signal thereto;

Fig. 17(b) shows frequency characteristics of the first digital roll-off filter in the ninth preferred embodiment;

Fig. 17(c) shows frequency characteristics of the output signal from the first digital roll-off filter of the ninth preferred embodiment;

Fig. 18(a) shows a ideal eye diagram;

Fig. 18(b) shows a practical eye diagram;

Fig. 19(a) shows a. circuit to detect a deterioration of the pulse forms by an oscilloscope and micro processor;

Fig. 19(b) shows a wired logic circuits to detect a

deterioration of the pulse forms on an analog circuit; and

Fig. 19(c) shows a wired logic circuits to detect a deterioration of the pulse forms on a digital circuit.

BEST MODE FOR CARRYING OUT THE INVENTION

A first preferred embodiment of the present invention is hereinafter described referring to Fig. 2 showing a principal part of a transmitter and Fig. 3 showing a principal part of a receiver. Same or like parts are denoted with corresponding reference numerals throughout the f igures .

The input to the transmitting station are a digital signal , for example a base band, of an I channel and a Q channel , orthogonal with each other, and are respectively of parallel digital signal, for example of eight bits, having a symbol rate T. Detailed description of the circuit confi ur ion will be hereinafter given representati ely for I channe] only.

Each bit line of the input signal is connected to an input terminal of each of delay lines 42i of eight, in this case, first digital roll-off filters 102ai. Only one of the delay l ines is representati ly drawn for each channel in Fig. 3. Circuit configuration of the roll-off filters will be described later in detail.

Output of first roll-off filter 102ai is input to a D/A converter 131 i. Unnecessary higher frequency spectrum

included in the output of D/A converter 131 i is eliminated by a low-pass filter 1371- Outputs from low-pass filters 132 i and 132q respectively of the I channel and Q channel are input to a QAM modulator 134, to which a carrier frequency sign l is also input from an carrier frequency oscillator 133. A QAM modulated radio frequency signal is transmitted via a radio frequency amplifier (not shown in the figure ^' * to a receiving station.

Each of roll-off filter 102ai comprises a finite impulse response (referred to hereinafter as FIR) type transversal delay line 42i . FIR delay circuit 42i is formed of a series of as many as 2m, an even number for example fourteen, delay elements 42-D. Each delay element 42-D is formed of, for example, a flip-flop having a delay time T/n, where T also shows pulse spacing of the pulse train input thereto and "n" is the sampling number, in this example, two. A branch line branches out from each of 2m+ 1 nods connecting adjacent two delay elements. Each branch line is serially provided with a multiplier 43. Signals output to branch l ines are multiplied at respective multipliers 43 by first tap rating ratios o^. —OJQ— q_ _m output from a first memory device 113, typically formed of a ROM (read-only memory). All of thus multiplied signals are summed by a first adder 145. An output of first adder 45, i.e. the sum, is the output of the first roll-off filter 102ai. Thus, FIR-type T/n delay line 42, multipliers 43, first memory device 113 and adder

circuit 45 constitute the first digit l roll-off filter 102ai . Frequency characteristic of first roll-off filter is determined by tap rating ratios q -- O(Q -- oi .

The frequency characteristics of the first roll-off filter is such that the unnecessary upper frequency spectrum input thereto shown in Fig. 17(a) is eliminated, and moreover provides, together with the output of the second roll-off filter in the receiver station, a good output pulse shape at as will be explained later in detail.

In the receiving station, a QAM modulated radio frequency signal received from the transmitter is pre- amplified as well as converted by a pre-ampl ifer and a frequency converter, each of which is not shown in the figure, to an IF signal. The IF signal is input to a demodulator 154, to which a local frequency signal is also input from local frequency oscillator 156. An I channel signal output from demodulator 154 is input to an A/D converter 15 i J ^' whe r e the analog input signal is sampled n times, in the figures n = 2, so as to output a parallel digital signal of 8-bit in this example. Unnecessary high

_^» frequency spectrum in eliminated by a. low-pass-filter 1511 -

Each bit line signal output from low-pass-filter 151 i is input to a second FIR type transversal delay line 1 of second digital roll-off filter llOi . In the first preferred embodiment of the present invention, the second roll-off filter 1101 acts as a roll-off filter plus an automatic

equal i zer .

Delay line 1 is formed of a series of as many as 2m, an even number for example fourteen, delay elements 1-D. Each delay element 1-D is formed of, for example, a flip-flop having a delay T/n, where T is the pulse spacing of the original pulse train and "n" is sampling number of A/D converter 1521. There are as many as 2m+ 1 nods each between adjacent two delay elements 1-D and at both of the input and output terminals of the delay line. A branch line, which may be called a tap, branches out of each nod. A second group 2-3 of branch l ines and a third group 2-1 of branch lines branch out of the center nod of delay line I i and every n-th nods counted from the center nod, totally as many as 2a+l, an odd number symmetric about the center nod. Thus, the signals on the second and third branch lines are delayed by the period T from the adjacent branch lines. A fourth group 2-2 of branch lines branch out from all other nods than those the second group and third group tap lines 2-3 and 2-1. In the figures of the present and the subsequent preferred embodiments n is 2; accordingly, every two nods are connected with the third group braneh l ines in the figure. If n=4, it is needless to say that every four nods are connected with the second and third group branch 1 ines.

Each of second group branch l ines 2-3 is serially provided with a multipl ier of second group 3i. Each of

third and fourth group branch lines 2-1 and 2-2 is serially provided with a multiplier circuit of a third and fourth group 4i or 5i. Each of second group branch lines 2-3 is serially provided with a multiplier of second group 3i. Signals output to second group branch lines 2-3 are multiplied at respective second multiplier 3i by respective second group tap rating ratio C , that are C _Q 4'~~ ^C _Q O' — _ _q 4_ , output from calculation circuit 7. Calculation circuit 7 will be explained later in detail. Signals output to third and fourth branch lines 2-1 and 2-2 are multiplied at respective multiplier circuits 4i and 5i by respective third and fourth group tap rating ratios^' and d (o(iς — CHi ₀ —cK-i^). Third tap rating ratios α. ' are of the below- described modifications of the outputs a 1 from a second memory deviee 6 typically formed of a ROM. Between each of the third group multipliers 4i and second memory deviee 6 is provided a second adder 8i, where a sixth group tap rating ratio a _j output from second memory device 6 is added with a fifth group tap rating ratio C output from a calculation circuit 7 so as to output a modified second group tap rating ratio ' • Thus multiplied signals output from third and fourth group multipliers 41 and 5i are input to, and summed by , a th i rd adder 9 i .

Thus multiplied signals output from second group multipl iers 3i are input to a co responding third adder circuit 9q of the Q channel . Symmetrically in the same way,

third adder 9i of the I channel is input with the outputs of corresponding second group multipl iers 3q of Q channe] . Accordingly, the sum of the multiplied signals from thp third and fourth group multipliers 4i and 5i of the I channe] and the multiplied signals from the co responding second group multipliers 3q of the Q channel are output from the third adder 91 of the I channel. A digital output of third adder 9i is divided by n by a divider 101 so as to return to have the symbol rate. An output of the pulse divider IOi is the output on a single bit line of the roll- off filter of the invention. The output is input to calculation circuit 7' . The output is also input to a carrier regeneration (CR) circuit 155, an output of which is fed back to control local oscillator 156.

When the automatic equal izer is not operated, the second and fifth group tap rating ratio C and C, each output from calculation circuit 7, are kept zero. Then, frequency characte istics o f the second roll-off filter is determined by fourth and sixth group tap rating ratios or 1 and < , which are all of fixed values.

Frequency characteristics of the first roll-off filter in the transmitter station and the second roll-off filter in the receiving station is chosen such that the overall frequency characteristics of the two roll-off filters allows the signal pulse output from the second roll-off filter without the use of automatic equalizer function, to be in a

satisfacto ily good shape so that no 1 termodula t j on takes place between the QAM-modulated signals. With this condition the tap rating ratios are fixed and stored in both of first and second memory devices 113 and 6, respectively in the transmitter and in the receiver.

When the automatic equalizer is in operation calculation circuit 7 monitors the pulse form output from divider 10 and calculates most optimum value of each of the second and fifth tap rating ratios C and C.

Internal circuit confi uration of calculation circuit 7 is schematically illustrated in Fig. 4. Calculation circuit 7 comprises a micro processor (CPU) 21, a program ROM 22, a first interface circuit 23, a second interface circuit 24, a RAM (random access memory) 25, and an accumulator 26. Processing program for the CPU 21. is stored in ROM 22. The procedure and an algorithm, i.e. the process programs, stored in ROM 22 are transferred to RAM 25; the data of pulse form input via first interface circuit 23 is compared with a predete ined reference level data; the optimum tap rating ratios to shape the pulse form are calculated according to the comparison result; and the optimum tap rating ratios C and C are output via second interface circuit 24 to second adders 8 and second group multipliers 3. The algorithm is similar to those have been genera]]y employed in the transversal filter type automatic equal izer.

The tap rating ratio to be input to an i-th branch

multiplier is given as follows:

Ci = JD(t)©E(t) dt where m is the nod number counted from the center nod; D(t) is a polarity value of the I and Q channel, respect i vely ;

E(t) is an error bit (at one bit lower than data bit); t is time in each symbol time; and © is an exclusive OR.

The tap rating ratios Ci output from calculation circuit 7, shown in Fig. 4, are typically obtained with an accumulator circuit 26 shown in Fig. 6. The first bit Indicating its polarity is input from the output of the second roll-off filter llOi to a delay line 20 formed of as many as 2a. delay element each having a delay T. The bit indicating an error ( typica1 ] y a second bit for 4PSK or a third bit for 1.6QAM) is input to another delay l ine 21 formed of as many as "a" delay element each having a delay T. An output of delay line 21 and an output from each nod of delay line 20 are input to each exclusive-OR 22, whose correlation values Sl _a , S 1 _a _^ , .. S 1 Q , .. Si_ _a , are input to each of adders 23. Each of outputs of adders 23 is input to a delay circuit 24 having a delay T, referred to hereinafter as a T-delay circuit) whose outputs Ci _Q , i _a _- .. Oi ₀ , .. , Ci_ _a are returned to adder 23, respectively. Thus, adder 23 and the T-delay circuit 24 locally form an accumulator.

The calculation operation of the calculation circuit 7 is continuously repeated so as to meet the always changing transmission condition.

Though calculation circuit 7 is drawn with a single block in Fig. 5 the content includes four subsidiary calculation circuits, where a first subsidiary calculation circuit is input with the I channel output so as to output fifth group tap rating ratios C, i .e. Ci _} -- Ci_^, a second subsidiary calculation circuit is input with both the I and Q channe] outputs so as to output second group tap rating ratios C , i.e. Cl'^ -- Ci' _* , a third subsidiary calculation circuit is input with the Q channel output so as to output fifth group tap rating ratios Cq of the Q channe] , I.e. Cq^ -- Cq_4, and a fourth sub calculation circuit is input with the I and Q channel outputs so as to output second group tap rating ratios Cq of the Q channel , i .e. C' ^ -- C _* . In Fig. 3, the first and second subsidiary calculation circuits are drawn as a single block 7, and the third and fourth sub calculation circuits are drawn as a single block 7Q. This is similarly done in other fjgures.

A second preferred embodiment of the present inven ion is schematically illustrated in Fig. 7, where second adder 8 is combined with the accumulator shown in Fig. 6, accordingly calculation circuit 7 is modified to be denoted wi h numeral 7' . Fifth group tap rating ratios, i.e. outputs Si , Si _a _ι .. Si Q , .. Si_ _a from calculation circuit

7' are respectively input to adders 25i. Outputs of adders 251 are respectively input to a first input terminal of selection circuit 26. A second input terminal of selection circuit 26 is input with each of the sixth group tap rating ratios a (oli 4, .. oti Q, .. o(i _ *) output from memory device 6'. Output of each selection circuit 26 is input to a T- delay circuit 27, whose output js returned to respective adder 251- When selection circuit 26 is reset with its reset terminal , the output of memory 6' is selectively output. When the reset is released, the accumulation operation starts. Then, the outputs α. '(ou' 4, • • oji , .. oU'_ .) of T-delay circuits 27 are the third group tap rating ratios a ' described in the first preferred embodiment. Thus, in the second preferred embodiment the circuit configuration can be simple while the reset function is additionally provided.

A third preferred embodiment of the present invention i schematically illustrated in Fig. 8, where only the portion of the second roll-off filter is drawn. As seen in the above preferred embodiments the tap rating ratios are generally symmetric about the center nods. Moreover, the I channel and the Q channel are symmetric with each other in principle. Accordingly, the fixed values, i.e. the fourth group tap rating ratios a - (di 5, o(i -, oli _j ) are employed commonly for the first half and the second half of the delay l ine about the center tap, as well as commonly for the 1

channel and Q channel. Consequently, the capacity of memory 6' ' can be reduced. Thus reduced state is shown in Fig. 9, where the marks * indicate the deletion of the memory outputs which existed in Fig. 5.

A fourth preferred embodiment of the present invention is schematically illustrated in Fig. 10, where only the portion of the second roll-off filter is drawn. The fourth preferred embodiment is to further simplify the circuit configur tion for the case where the distortions are equally generated in both I and Q channels, for example, the case where the input to the second roll-off filter is an IF signal . The sixth group tap rating ratios o 1 ( o \ ₄ , .. o(iø .. o(i _4) and second group tap rating ratios C (C^, C' ) C' Q) in the first preferred embodiment are used commonly for the I and Q channels. The fourth group tap rating ratios a - < ^i 5 _> •• oli _j .. oti _ ₅ > in the first preferred embodiment are used commonly for both the halves about the center nod as well as both the I and Q channels. Consequently, the outputs of memory 6' ' ' and calculation circuit 7' can be remarkably reduced. Thus reduced state is shown in Fig. 11, where the marks * indicate the deletion of the memory outputs which existed in Fig. 5.

The quantity of the delay element 1-d in the delay ] jne 1 is determined depending on the required characteristics as a roll-off filter. The quantity of the every n-th branches 3 from the nods of delay line 1 is determined depending on

the required characteristics as the automatic equal izer. Accordingly, the quantity of the delay elements and the quantity of the every n-th branches can be arbitrarily chosen as a design choice. The delay elements and the branches must be always symmetric about the center nod.

A fifth preferred embodiment of the present invention is hereinafter described referring to Fig. 12, where the cross connection, i.e. the input from the opposite channel, employed in the previous preferred embodiments is deleted. Accordingly, there is none of the second group multipliers 3, and the second group tap rating ratios C output from calculation circuit. This circuit configuration is widely applicable to a system performing no orthogonal modulation there in .

A sixth preferred embodiment of the present invention is hereinafter described referring to Fig. 13, where the third adder circuit 9' ' has been modified. A fourth adder 21i and 21q are additionally provided at each of the outputs of the roll-off filters of I channel and Q channel of the previous preferred embodiments, respectively. Moreover, the signals output from the second group multipl iers 31 of the I channel are input to the fourth adder 21q newly added to the Q channe] , instead to third adder 9q. In the symmetrical way, the signals output from the second group multipliers ?,q of the Q channel are input to the fourth adder 21 i newly added to the I channel, instead to third adder 9q. Accordingly,

third adder 9i sums only the outputs from third, and fourth group adders 4i and 5i of the I channel, i.e. own channel . Advantageous effect of this circuit configuration is in that the load of third adder 9 summing so many of the signals input thereto ean be eased so as to accomplish a high speed opera,t i on .

Hereinafter described are modifications of the structure of the second roll-off filter so as to further accomplish a higher speed operation. A seventh preferred embodiment of the present invention hereinafter described referring to Fig. 14. A third transversal type delay line 11 is formed of as many as 2m (an even number, 2m = 10 in Fig. 14) delay elements 11-D, each having a delay time T/n, typically formed of a flip-flop having the delay time T/n. In Fig. 14 n is chosen two. Between the delay elements 11-D is provided an adder 12 of a fourth group. Between the input terminal of this roll-off filter and each of the fourth group adders 12 are provided multipliers 4-1 and 5-1 of a fifth and sixth group, which are the same as the third and fourth group multipliers 4 and 5 of the previous preferred embodiments. Allocation of the fourth group adders and the fourth and fifth group multipliers are denoted with the suffixes of the previous preferred embodiments. The inputs to the third and fourth group multipliers 4 and 5 from the nods of the previous preferred embodiments are replaced with the input to the roll-off filter; and outputs of the fifth

and sixth multiplier circuits 4-1 and 5-1 are respectively input to the adders 12 of the fourth group. Input to the first delay element is from multiplier o(i_c. At the output of the last delay element is provided an adder 16 to which an output from the last multiplier circuit whose multiplication factor is αi ^ is input. Multipl ication factors of the fifth and sixth multipl iers are input from second memory device 6 and calculation circuit 7 in the same way as the previous preferred embodiments. Another inpnl to each of the fifth group adder is an output of the adjacent delay element 11-D. Output of each of the fifth group adders is input to the next delay element.

A fourth transversal type delay line 13 is formed of an odd numbers of delay elements 13-D each having a delay time T, typically formed of a flip-flop having the delay time T.

Between the T delay elements 13-D is provided an adder 14 of a fifth group. Between the input terminal of the opposite Q channel roll-off filter and each of the fifth group adders 1 1 is provided each of multiplier 3-1 of a seventh group, which are the same as the second group multiplier 3 of the previous preferred embodiments. Allocation of the seventh group multipl iers 3-1 are denoted with the suffixes of the second group multipliers of the previous preferred embodiments. The inputs to the seventh group multipliers 3-1 from the nods of the previous preferred embodiments are replaced by the input to the

O"

roll-off filter of the opposite channel, i.e. of the Q channel and the outputs of the seventh group multipl iers 3-1 are respectively input to the adders 14 of the fifth group.

Multiplication factors to be input to the seventh group multipliers are of the same values (Ci '_ ₄ , .. Oi' _β , • .Ci '_ ₄ ) as the second group multipliers 3 output from the calculation circuit 7 of the previous preferred embodiments. Another input to each of the fifth group adders is an output of the adjacent delay element 13-D. Output of each of the fifth group adders is input to the opposite adjacent delay element. Input to the first delay element of the fourth delay line 13 is from a multiplier to which a multipl ic tion factor C i_5 is input. At the output of the last delay element of the fourth delay line is provided an adder 16 to which a tap rating ratio C' i^ is input from calculation circuit 7. Output of the last adder of the fifth group 14 is input to a delay circuit 15 having a delay time T/2, because n = 2 in this preferred embodiment, so that the delay times of the third and the fourth delay lines can be co-phased.

An output from the third delay line 11, i.e. the output from the last adder 16, and an output from the fourth delay line 13, i.e. an output from the T/2 delay element 15, are summed by a sixth adder 17. A pulse train output from sixth adder 17 is divided by n in divider 18 so as to output a pulse train having the symbol rate. An output from the

divider 18 is the output of the I channel roll-off filter also acting as an automatic equalizer, of the present invention. A symmetrical circuit configuration is provided for the Q channel, as well.

A variation of the seventh preferred embodiment is shown in Fig. 15, as an eighth preferred embodiment. The output of the third transversal delay line 13, i.e. the output of the last one 16 of the fourth group adders is input to divider 18' . Output of the last adder of the fifth group 14 is input to a delay element 15' having a delay time T, so that the delay times of the third and the fourth delay lines are co-phased. An output of divider 18' and the output of the T delay element 15' are summed by an eighth adder 17' . An output from the eighth adder 17' is the output of the I channel roll-off filter also acting as an automatic equalizer of the present invention.

Adder circuit 9 in the first to fifth preferred embodiments has to sum so many of the signals input thereto that a considerably large time must be consumed in this summing operation. Advantageous effect of the seventh and eighth preferred embodiments is in that the summing operation by adder .9 is sequentially carried out by the distributed fourth and fifth adders. Accordingly, high speed operation can be accomplished.

If the delay line of the seventh or eighth preferred embodiment is not employed, the transversal type automatic

equalizer circuit llOi has to be placed serially after the second roll-off filter 102b i as was shown in Fig. 1.

In the above preferred embodiment circuit configurations, the delay time spent in passing through the roll-off f i 1 ter/automat i e equalizer of the present invention is much less than the delay time spent in passing through the serial connection of the second roll-off filter and the automatic equalizer. This is because the delay time to be spent is proportional to the quantity of the serially arranged delay elements through which the signal passes. Thus shortened delay time contributes to solve the problem in that the long delay time in the filter and equalizer deteriorates the effect of the feedback of the carrier regeneration circuit 56 via local frequency oscillator 156 to orthogonal demodulator 154.

Moreover, it is apparent that the circuit configur tion in the first to fourth preferred embodiments, having less number of delay element, is much simpler, consequently its production cost is less expensive, than the prior art confi uration shown in Fig. 1.

An ninth preferred embodiment of the present invention is hereinafter described referring to Fig. 16, which is also an abstract sumniary of the above seven preferred embodiments except below-descr ibe additional feature of two ROMs 113-1 and 6-6, each for determining the frequency characteristics of the first and second roll-off filters 102a and 110. The

~ -

same numerals denote the same as Fig. 1 and Fig. 3.

As described earlier the frequency charact risti s of the roll-off filters are set in advance so that overall frequency characteristics excluding the automatic equal izer function provides at the output of the second roll-off filter such satisfactory pulse forms that no intermodula t i on takes place between the signals carried on the QAM modulated wave .

In the ninth preferred embodiment according to the present invention first ROM 11 -1 stores at its different addresses two kjnds of the tap rating ratios; first kind of which is used for the usual operation as described in the previous preferred embodiments, and the second kind of which is to make the first roll-off filter itself have the above mentioned overall frequency characteristics as illustrated in Fig. 17(b), where the overall frequency preferred characteristics have been shared by 50/50 by the first roll-off filter and the second roll-off f i 1 ter/equa1 izer in the previous preferred embodiments. It is seen in Fig. 17(c) the high frequency spectrum within the pass-band of the output signa] from the first roll-off filter is enhanced. The second ROM 6-1 in the receiving station h s the usual tap rating ratios to be used for the υsual operation described in the previous preferred embodiment, and additionally the second kind ones to make the second roll-off filter/equalizer llOi have a flat frequency

characteristic, referred to hereinafter as transparent. The tap rating ratios to make the transparent frequency characteristics in the second roll-off filter/equalizer 1101 are typically such that only the tap rating ratios input to the central multiplier a 0 and CO are 1 while all other tap rating ratios are set zero. The frequency characteristics taken by each of the roll-off filters will be explained later i detail referring to Figs. 17.

The first and second kinds of the tap rating ratios are swltchable to output from the first ROM 113-1 and second ROM 6-1, respectively, by an address switch signal ADR. With the second kind tap rating ratio input to the first roll-off filter, the pulse forms after the output of the first roll- off filter 102ai, i.e. at the check points Al at the output of the first roll-off filter 102ai , A2 at the output of D/A converter 1311, A3 at the output of first low-pass filter 132i, B3 at the output of demodulator 154i , B2 at the output of second low-pass filter, Bl at the output of A/D converter 1511, and BO at the output of second roll-off filter 1101 , should be in a good shape, unless some cause deteriorates the transmission characteristics.

Pulse forms at the cheek points of the digital signals can be checked by viewing an eye diagram on an oscilloscope. An idea] eye diagram of sequential pulses observed on the oscilloscope is shown with solid lines in Fig. 18(a). If the frequency characteristics is such as to cut high

frequency spectrum, the pulse transition exhibits a slope as shown with the dotted line on the left hand side of Fig. 18(a). If jitter takes place, the transitions are dispersed as shown with dotted line on the right hand side of Fig. 18(a). Accordingly, each of the two causes decreases the area of the blank portion, i.e. the eye, in the eye diagram as seen in Fig. 18(b) showing a practical eye diagram. By checking the eye patterns at the cheek points the cause which has deteriorated the frequency ch racteristics can be located.

The eye diagram can be observed not only with human eyes via oscilloscope, but also with an electronic means, such as a micro processor. As shown in Fig. 19(a ⁾ the micro processor 67 may have in its ROM 62 a reference level with which the value measured on the eye diagram in the oscilloscope 61 is compared so as to detect the cause, i .e. the defect of the circuit . When the comp rison result indicates that the eye diagram does not meet the reference level the micro processor outputs an alarm signal. Figs. 19(b) and (c ⁾ show wired logic for detecting the deterioration. In Fig. 19(b) an analog signal input from the check points A , A3, B3 or B2 is converted to a digital signal by an A/D converter 63. In this digital signal two error bits, for example D3 and D4 , which are located just below data bits DO -- D2 , in the ease of 64QAM, are input to an exelusive-OR gate 64. The two error bits are averaged by

exelusive-OR gate 64 and a low-pass filter 65. Thus averaged output for low-pass filter 65 is compared with a reference level by a comparator 66. When the averaged level exceeds the reference level the comparator outputs an alarm signal. In Fig. 19(c) a digital signal input from the cheek points Al or Bl is averaged and processed in the same way by the exclusive-OR 64 and low-pass filter 65 of Fig. 15(b).

Thus, the cause, such as a defect in the electronic circuit, can be easily and quickly located. Without this method of the present invention it is normal for the eye patterns to be deformed, i.e. not adequately open or wide, at those check points even if no defect is in the electronic circuit, because the first roll-off filter itself alone does not provide in the usual operation the full frequency characteristics which provides the good shape of the pulse forms .

A tenth preferred embodiment of the present invention is hereinafter described, which is a variation of the eighth preferred embodiment. The first and second ROMs 113-1 and 6-1 are additionally provided with a third kind of the tap rating ratios. The third kind of the tap rating ratios are such that provide the first roll-off filter with less roll- off ratio, for example 3090, compared with 50% of the ninth preferred embodiment. At this time, the second roll-off i 1ter/equal i zer 1101 is kept transparent in the same way as the ninth preferred embodiment.

With thus reduced roll-off factor, the eye diagram at the check points completely open in the vertical direction, however, the width in the horizontal direction becomes narrower. Thus narrowed eye diagram emphasizes the phenomena caused from the characteristics deterioration, accordingly allows easier and more accurate detection of the defect .

Though the cheeking method of the present invention explained above with reference to the eighth and ninth preferred embodiments recites the circuit configuration of the first to eighth and ninth preferred embodiments having the second roll-off filter/automatic equalizer 1101 , it is apparent that this cheeking method may be applied 1o the circuit configuration where the automatic equal izer is provided independent ly from, and serially to, the second roll-off filter as shown in Fig. 1.

The many features and advantages of the invention are apparent from the detailed specification and thus, it is intended by the appended claims to cover all such features and advantages of the methods which fall within the true spirit and scope of the invention. Further, since numerous mo ifications and changes will readily occur to those skilled in the art, it is not detailed to limi the invention and accordingly, all suitable modifications are equivalents may be resorted to, falling within the scope of the invention.