Note: Descriptions are shown in the official language in which they were submitted.
CA 02248480 1998-09-28
DIGITAL SIGNAL PROCESSING APPARATUS
This invention relates to a digital signal processing app~al~ls. Generally,
tr~ncmicsion of signals requires (a) a sçndin~e port to modulate signals and load the modulated
signals onto a carrier wave, (b) a tr~nsmic.sion medium to carry the modulated carrier wave,
5 and (c) a receiving port to detect the signals from the carrier wave and demodulate them.
Analog signal tran~mission uses analog modulation and demodulation, while digital signal
trancmicsion requires digital modulation and demodulation. Digital signals are sometimes
modulated onto carrier waves for tr~n.cmi.csion as analog signals. A method is known in the
art for performing digital signal tr~n.cmission by mod~ ting digital signals onto orthogonal
10 carrier waves. The sending port modulates digital I- and Q-signals onto orthogonal carrier
waves. The orthogonal carriers have the same frequency but differ in phase by 90 degrees.
The orthogonal carriers are here called "conjugate waves", as each. Carrier is conjugate to
the other. The receiving port demodulates the I- and Q-signals by multiplying the received
signals by the orthogonal carriers. This type of modulation is called "orthogonal
15 modulation" because it uses two orthogonal carrier waves having a phase di~rence of 90
degrees.
A number of prior art docllments have proposed digital orthogonal modulation. For
example, Japanese Patent Laying Open No. 6-14074, entitled "digital signal processing and
mod~ ting apparatus," pointed out the distortion of modulated signals by conventional
20 orthogonal mod~ tion and suggested a new way to reduce this distortion. In this method,
two digital signals are tr~n.cmitted, one being the I-signal (I-channel signal or I-ch) and the
other being the Q-signal (Q-channel signal or Q-ch). The sending port multiplies the digital
I-signal by a carrier, multiplies the digital Q-signal by the conjugate carrier having a 90-
degree phase difference, sums the two products, D/A-converts the digital signal to an analog
CA 02248480 1998-09-28
signal, and transmits the analog signal across a medium.
The receiving port receives the analog signal from the medium, A/D-converts it to a
digital signal, extracts the I-signal by multiplying the signal by the first carrier, extracts the Q-
signal by multiplying the signal by the conjugate carrier, and thus obtains the I- and Q-signals.
5 This is a simple explanation of orthogonal modulation. The frequency of the carrier is
denoted by fc and the angular frequency of the carrier is denoted by Q, where Q=2 ~ fc
One carrier is sin Q t, and the conjugate carrier having a phase di~elel-l by 90 degrees is cos Q
t. The I- and Q-signals will now be represented in brief by I and Q. The sçn~ling signal is
thus denoted by Icos Qt+QsinQt. The receiving port can extract the I-signal by multiplying
10 IcosQt+QsinQt by cosQt, and the Q-signal by multiplying IcosQt+QsinQt by sinQt. The
orthogonality of sinQt and cosQt thus enables the use of orthogonal demodulation to
complete the tran.cmi~sion of two dirrere,.l signals with subst~nti~lly a single carrier of
frequency Q.
In principle, a wave of arbil,a-y frequency can be used as a carrier for digital signal
15 tran.~mission. In practice, a wave of 44MHz or 57 MHz has been chosen as a carrier. The
values of sinQt and cosQt as a function of t are prestored in a ROM, because the speed
limitations of existing hardw~e prevent the real-time calculation of sinQt and cosQt during
modulation. J~p~nese Patent Laying Open No. 6-14074 explained that the modulation of the
baseband signal by harmonics of the carrier folds over to appear as noise in the frequency
20 band of the carrier. This reference solved the problem by specifying the carrier frequency to
be a half-integer multiple of the frequency of the input signal.
Japanese Patent Laying Open No. 6-14074 told that the half-integer multiplier of the
carrier elimin~tes distortion of the modulated signal caused by overlapping of the harmonic-
modulated baseband. For the case of sampling the signal once every period, the carrier
25 frequency fc should be determined by fc=n/2T, where T is a period (cycle) of an input signal
,
CA 02248480 1998-09-28
and n is an integer. For the case of sampling the signal x times in a period, the carrier
frequency fc should be fc=xn/2T, where x is the sampling rate. The difficulty imposed by this
method, however, is the rigorous determination of the carrier frequency by the input signal
and the impossibility of using carriers of arl~ y frequency.
Japanese Patent Laying Open No. 5-153182, entitled "digit~1i7ed orthogonal
modulator", pointed out a problem associated with a mixing type of orthogonal modulation
which modulates digital signals by sinQt and by cosQt, sums the two modulated signals,
D/A-converts the sum to an analog signal, and mixes the analog signal with a local oscillation
frequency ( ~ ). Mixing Icos Q t+Qsin Q t with cl) produces cos ~ t(Icos Q t+Qsin Q t)
10 (I/2){cos( ~ - Q )t+cos( cl) + Q )t}+(Q/2){-sin( c~ - Q )t+sin( cl) + Q )t}. Here cl) > Q . The
mixing type of modulation creates two wave packets having central frequencies of (~ -Q)
and (~+Q) and the same bandwidth as the input signal. The centers of the two wave
packets are distanced by 2 Q from each other. Selecting a Q higher than the bandwidth of
the input signal prevents overlapping and enables a filter to separate the two packets.
15 However, the carrier frequency fc(=Q/2~) is restricted in practice by the speed of the
multiplier, the speed of the adder, the access time of the ROMs, and the settling time of the
AID converter. These restrictions compel the use of a carrier frequency fc only slightly
higher than the input signal frequency. Thus the two wave packets will appear just below
and just above the local oscillation frequency cl~ /2 ~ . Since the distance between the two
20 packets is reduced, their separation by a filter is difficult. Only an excellent filter having an
extremely steep selectivity could separate two wave packets with such a narrow separation.
A filter of such high performance would be expensive, owing to the difficulty of its
m~mlf~ctllre. This reference therefore describes an additional technique, for solving the
shortcoming of mixing type orthogonal modulation. Using two equivalent modulation
25 circuits, this technique modulates two digital signals by two sets of carriers ((sinQt, cosQt)
CA 02248480 1998-09-28
and (-cosQt, sinQt)) having the same frequency but differing in phase by 90 degrees, obtains
two modulated packets of IsinQt+QcosQt and -Icos Qt+QsinQt, and D/A converts and
mixes the two packets in parallel with the local oscillation frequency ~/2 ~ . The mixture
that results is an analog signal of (Isin Qt+Qcos Qt)cos c~) t+(-Icos Qt+QsinQ t)sin cl~ t. The I-
5 signal component is I(sin Q tcos ~ t-cos Q tsin ~ t), which reduces to Isin( Q - ~ )t. In other
words, the counterpart with the central angular frequency (Q+~) disappears in the I-
component. Similarly, the Q-signal component Q(cos Q tcos c ) t+sin Q tsin ~ t) reduces to
Qcos(Q-~)t. Thus a single signal having a central angular frequency (Q-cl)) remains
while its counterpart of frequency ( Q + ~ ) vanishes. Both the I- and Q-signals have only a
10 single wave packet with a central angular frequency of ( Q - ~ ). The absence of the ( Q + ~ )
component removes the need for a filter of steep selectivity.
Japanese Patent Laying Open No. 6-97969, entitled "digital signal orthogonal
modulator", aimed to increase the data tran~mission capacity by enhancing the carrier
frequency of a digital signal orthogonal modulator capable of processing digital signals by an
15 amplitude-phase modulation (e.g., QAM) or a phase-shift keying (PSK). Orthogonal
modulation is a method for transmitting two signals by multiplying one signal (the I-signal)
by sinQt and the other signal (the Q-signal) by cosQt. The I- and Q-signals are digital.
The product Isin Q t is not continually calc~ te~, but is instead only periodically resolved by
multiplying the digital I-signal by selected and prestored values of sin Q t several times in one
20 cycle of the carrier. For example, for an interval of T/n, calculations must be repeated n
times in every period T. Thus the values of sinQt and cosQt are read out from the ROMs
and are multiplied by the I-signal and Q-signal n times in each period.
Each period T(=2 ~ / Q ) thus requires n times of reading in values for sin Q t and cos
Q t and n times of multiplying the I- and Q-signals by sin Q t and cos Q t. The number "n" is
25 called the sampling rate. Digital adders and multipliers must therefore run at a speed n times
CA 02248480 1998-09-28
faster than the modulation frequency fc ( Q /2 ~ ). In general, digital devices have an upper
limit to the operating speed. The reciprocal of the device upper limit speed, divided by n, is
the upper limit of the modulation frequency.
As explained above, a higher modulation frequency can carry a larger amount of
5 information.
Conventionally the modulation frequency is settled to be 44MHz, which is
sufficiently high to provide an adequate tr~n.cmi.csion capacity. A 44MHz modulation
frequency gives rise to rigorous dem~n-ls for electronic devices, however. The digital adders,
digital multipliers, and sin-ROM or cos-ROM must then operate at a speed of 44 x n MHz,
10 which is the sampling number times faster than 44 MHz. Digital devices of such high speed
have not yet been produced. The difficulty of making high speed devices is further
exaggerated when a larger sampling number is used. This effect prevents the use of a large
sampling number n. In practice, therefore, the sampling number n is limited by the current
technology of producing digital ICs including the adders and multipliers.
However, if the sampling number "n" is small, then the orthogonality between sinQt
and cos Q t decreases, degrading the fidelity on the receiving side. The loss of fidelity due to
the decrease in orthogonality is called the "aperture effect", and high fidelity cannot be
m~int~ined at the receiving port if n is too small. Thus this reference suggested taking only
8 phases for coupling I- and Q- signals with different weights instead of multiplying the
20 signals by sinQt or cosQt:
PHASE 0:
PHASE 1: 2-l'2 I + 2.ll2 Q
PHASE 2: Q
PHASE 3: -2-l'2 I + 2-ll2 Q
PHASE 4: -I
CA 02248480 1998-09-28
PHASE 5: 2-"2 I 2-ll2 Q
PHASE 6: -Q
PHASE 7: 2-"2 I 2-"2 Q
Thus Japanese Patent Laying Open No. 6-97969 tran~mitted 8 phases in series
5 repeatedly from a sending port to a receiving port, but it described nothing about the
demodulation on the receiving side. The reference taught that simple calculation of the 8
phases enables the sending port to send a larger amount of information by excluding the time-
wasting calculation of sin Q t and cos Q t and culling out the times of the calculation.
However, the eight phases are only the values of Icos Qt + QsinQt at eight points of
27~k/8 (k=0, -,7) in a cycle T (=27~tQ), which merely corresponds to n=8 in the prior
methods. Thus, this method cannot choose the modulation frequency fc any more freely or
irrespective of T than the method of Japanese Patent Laying Open No. 6-14074.
Several proposals of orthogonal conversion type digital modulation have been
explained. At least five to eight sampling points (n) are necess~ry in a cycle to allow the
15 NCO (numeral controlling oscillator) to generate high C/N (count/noise) carrier waves. The
NCO, which is composed of adders, multipliers, sin-ROMs and cos-ROMs, can digitally
produce any carrier wave with arbitrary frequency. Therefore, the adders, multipliers, sin-
ROMs and cos-ROMs must operate at a high speed which is n-times as fast as the modulation
frequency fc However, if the modulation frequency is 44MHz or 57MHz, the system would
20 be restricted by the speed of digital ICs, as there are at present no digital devices which can
run at a speed of five times to eight times faster than 44MHz or 57MHz. Then the prior
methods selected a carrier frequency to be n/T (n=4,6 or 8) in order to obtain high C/N
carriers. However, if fc = n/T, then the carrier frequency must be uniquely determined by the
modulation frequency. On the contrary, if the modulation frequency is determined apart
25 from the carrier frequency, then some frequency conversion circuit will be needed to adjust
CA 02248480 1998-09-28
the modulation frequency to the carrier frequency.
[Conventional orthogonal modulation circuit: type one]
Fig.6 exhibits a schematic view of a prior orthogonal modulation system of type one.
Figs. 7(a) to 7(f) are frequency spectra of the signals of the parts in Fig. 6.
Multivalued digital signals inputted to the I-channel and Q-channel are briefly called
I-signal and Q-signal, respectively.
The I-signal is interpolated by an interpolation device (IP) 21, processed by a
baseband filter 22, and again interpolated by IP 23. Then higher frequency components are
cut by a low-pass filter 24. The Q-signal is interpolated by IP 25, processed by a baseband
10 filter 26, and interpolated by IP 27. A lowpass filter 28 allows only a lower frequency to
pass. The digital I-signal is multiplied by a multiplier 29 with cos cll t of a local oscillator 31
for producing Icos~t. Similarly, the digital Q-signal is multiplied by another multiplier 30
with sin Cl) t, which signal is produced by delaying the phase of cos cl) t of the local oscillator
31 by ~/2 .
The carrier frequency fif is fjf=cl)/2 . For sending TV signals, the carrier frequency
is determined to be 44 MHz in the USA and 57 MHz in Japan. The outputs Icos cl) t and Qsin
c~) t of the multipliers 29 and 30 are summed to Icos ~ t + Qsin ~ t by an adder 33 in a digital
manner. The digital sum is converted to an analog signal by a D/A converter 34. After
passing through a low-pass filter LPF 35, the analog signal is sent via a tr~ncmi.csion medium
20 to receiving ports.
For giving a concrete, intuitive explanation, we assume that the I-signal and Q-signal
bandwidth f5y~b is equal to 5MHz. The I-signal and Q-signal are interpolated once by IP21
and IP25 and are then restricted by baseband filters 22 and 26, which are lOMHz (=2fSylllb)
wide and operate from 0 to 10 MHz.
We also assume the m~imllm operating frequency of a digital device to be 120 MHz.
CA 02248480 1998-09-28
In other words, the sampling frequency is 120 MHz (= cl) /2 ~ ). The frequency of the output
of the baseband filter is lOMHz (2 f5ymb), but this signal is interpolated by a factor of 12 to
become a signal with a 120 MHz bandwidth (24 f5ymb), and imaged frequency components are
excluded by a low-pass filter LPF running at 120 MHz (24 f5ymb).
The local oscillator generates cos c ~ t and sin ~ t which are carrier wave signals of 44
MHz sampled at 120 M~Iz. In the case of sampling the carrier waves of 44 MHz at 120
~Iz, the sampling number is not an integer but a fractional decimal, that is, 120/44 = 2.73.
When the sampling number is small such as 2.73, rli~it~li7ecl carrier waves contain a large
amount of spurious noise that cannot be removed by a filter and so on in the vicinity of an
10 oscillation frequency, whereby the carrier waves having high quality (high C/N) cannot be
generated. This results in the deterioration of the quality of the modulated signals. In
general, it is said that the sampling carrier waves requires at least five to eight samples per
cycle.
The LPF output signal of bandwidth 120 MHz is multiplied by cos c~) t and sin ~ t that
15 are carrier wave signals of 44 MHz from the digital local oscillator, which brings about the
generation of modulation signals in a range from about 42 to 46 ~Iz and the appearance of
image components of the modulation signals in a range from about 74 to 78 MHz, as shown
in Fig.7(d).
The I- and Q-signals thus multiplied by the carrier waves are added by an adder to
20 create an orthogonal modulation wave of Icos cl) t + Qsin ~ t. A D/A converter converts the
digital signal to an analog signal, and the result is a full wave range 47 centered at 44 ~Iz
and a full wave range 48 centered at 76 MHz, which is an image component of the modulated
signal 47, as shown in Fig.7(e). Another LPF elimin~tes the image component 48 or 50 to
produce the IF (intermediate frequency) signal 51 in a band ranging from about 42 to 46 MHz
25 and centered at 44 MHz. Here, f5ymb, the interpolation rate, and the maximum operating
CA 02248480 1998-09-28
frequency of digital circuits as mentioned above are shown just as examples, but these
albill~ly values are within the bounds of possibility.
Since this method uses filn~mental digital circuits, the LPF and the baseband filter
differ from those used in analog circuits. A problem arises in that the basic system of the
5 orthogonal modulation must be driven at a modulation frequency of 44 MHz, but it is very
difficult to achieve such operation. The oscillator can oscillate easily at 44 ~Iz, but it is
difficult to drive the rest of the modulation system at this rate. In order to create an
oscillation signal of good quality, at least several sampling points must be taken in every cycle,
which requires very high-speed sampling ICs.
10 [Conventional orthogonal modulation circuit: type two]
Another ordinary orthogonal modulation circuit is explained with reference to Figs.8
and 9. The I-signal passes through IP 52 and a roll off filter 53. A multiplier 54 multiplies
the signal by any one of 1, 0 and -l. Hence, the I-signal becomes any one of I, 0 and -I.
The Q-signal is treated in the same fashion, i.e. the Q-signal becomes any one of Q, 0 and -Q
throughthe help of a multiplier 58, IP 56, roll offfilter 57, and so on.
The multipliers 1, 0 and -1 are selected by a multiplication controlling signal. The
I- and Q-signals are added by an adder 60, and are then changed to analog signals by a D/A
converter 61. High-frequency components are excluded by an LPF 62. Further, a mixer 63
multiplies the analog signals by a sin ~t wave of a local oscillator 64. A tran~mi~sion signal,
which lies within in the frequency band centered at 44 MHz, is sent via a b~n-lp~s filter
(BPF) 65 into atr~n~mi~sion medium.
The frequency range of the signal inputted to the roll off filter 53 is the same as in
the former example, that is, 0 to about 2 MHz, (i.e., fs~mb= S MHz). The limit of the
p~sban~ of the filter 53 is 8 fs~ b = 40 ~Iz. Thus outputs of this baseband filter appear
near 0 MHz as a half wave range 67 (0~2 MHz) and near 40 MHz as a half wave range 68
CA 02248480 1998-09-28
being an image component. The band ranges after carrier-wave multiplication are shown in
Fig.9(b), where the " carrier wave" is simplified, because this operation is actually performed
by multiplying the signal by 1, 0, -l and 0 at four points in order instead of multiplying the
signal by sin~t. The multipliers l, 0 and -l, as read out from memories 55 and 59, are
multiplied against the signal. Here, since the frequency of the multiplication controlling
signal is l0 MHz, input signal 67 becomes a modulation wave 69 centered at l0 MHz and an
image component 70 centered at 30 MHz.
The I- and Q-signals digitally modulated at l0 ~Iz are added by an adder to create
an orthogonal modulation wave. This modulated signal is then converted into an analog signal
10 by a D/A converter. Only the first IF modulation signal 75 centered at 10 MHz remains after
e~ccl~ ing the image components and higher harmonics.
The final output should be in a frequency range centered at 44 MHz. Therefore, the
frequency should be increased by mixing. The local oscillator 64 and the adder 63 (mixer)
play this role. The local oscillation frequency may be 54 MHz or 34 MHz. Here, 54 MHz
15 is used. The mixer output is shown in Fig. 9(e). The first IF signal becomes the two full
wave ranges 76 and 78 after multiplication by the 54 MHz signal. The narrow peak 77 at 54
MHz is a leakage of the local oscillation signal. The b~n-~p~cs filter BPF 65 passes only the
frequency range 76 having a bandwidth of 10 MHz and centered at 44 MHz. As a result,
only the full wave range 76 is selected to be the final IF signal 79. As shown in Fig. 9(f),
20 this signal lies in the frequency band of from 44-2 to 44+2 MHz (44 + 2 MHz).Since this method needs no sin-ROM or cos-ROM to hold the values of sin c~ t andcos~t, but rather uses only simple multiplication and addition, a high-quality modulation
signal can be generated. However, this method uses only a particular value of modulation
wave frequency depending on fsyl~b Therefore, to obtain a output on a carrier, another analog
25 frequency conversion circuit is required.
CA 02248480 1998-09-28
In the present invention, the I-signal is multiplied by the complex modulation wave
exp(j 2 7~ fct) of a frequency fc as determined by the expression fc = fjf - n ~ f"",b where n is an
integer, and f5""b is the symbol rate (frequency width) of the baseband input signal. The Q-
signal is multiplied by a complex modulation wave exp(j2 7~ fct+ ~ /2), which is obtained by
5 shifting the phase of exp(j2 7~ fct) by 7~ /2. The results of the two multiplications are
summed. The frequency of the s~1mm~tion signal is then raised by multiples of fs~ b by an
interpolation circuit, in order to generate a plurality of frequency ranges having center
frequencies which increase by multiples of fs~ b A complex bantlpa~s filter selects only the
frequency range that includes a carrier wave frequency fif . The real im~gin~ry component is
10 selected and D/A-converted into an analog signal and then an analog low-pass filter selects
only the analog signal at the carrier wave frequency.
The invention will be explained with reference to Figs. 1-4. The I- and Q-signals
are introduced from I-ch and Q-ch, respectively. Both are digital signals. The I-signal
passes through an interpolation device l and a baseband filter 2, and is multiplied by a
15 complex signal exp(j c~) ct) in a multiplier 3. Here, the modulation angular frequency ~ c( ~
c=27~ fc) is 2 ~times the modulation frequency fc that is obtained by subtracting integer
multiples of the symbol rate fs~ b of the I- and the Q-signals from the carrier wave frequency
fjf. (i.e.fc=fjf - n- f5yl1lb)
The modulation frequency fc is much smaller than the carrier wave frequency fjf. In
20 other words, fc< <fjf. This implies that, the modulation angular frequency ~c is also far
smaller than the carrier angular frequency ~, or ~c< < c~ . In general, oscillator 7
generates a slow complex vibration exp(j c~) ct) instead of quick oscillations such as cos ~ t of
the conventional oscillator 3 l shown in Fig. 6. This invention has two primary
characteristics, one being that the modulation frequency fc (= Cl) J2 ~ ) is small and the other
25 being that the I- and Q-signals are multiplied by the complex signal exp(jc)ct). A phase
CA 02248480 1998-09-28
shifter 8 shifts the phase of the complex signal by 90~ to produce, another complex signal
exp(j ~ ct + j ~ /2).
Similarly, the Q-signal passes through an interpolation device 4 and a baseband filter
5, and is multiplied by the complex signal exp(j ~ct + j ~/2) in a multiplier 6. In this
5 invention, the I- and Q-signals do not m~int~in a real form such as I cosc~)t + Q sin~t, but
rather adopt a complex-form such as I exp(j ~ ct) + Q exp(j c,~ t + j ~ /2) after passing through
adder 9. This complex signal I exp(jcl)ct)+Q exp(j~ct + j~/2) then passes through an
interpolation device 10 and a b~n-lpass filter BPF 11 of complex coefficients which çlimin~tec
lower and higher frequency components. Every calculation to this point is complex, and this
10 feature is the most distinctive of this invention.
A real part computing circuit 12 accepts only the real part. Alternatively, the
im~gin~ry part may be chosen by an im~gin~ry part computing circuit. Here, only the real
part is converted into an analog signal by a D/A converter 13. High frequency components
are further excluded by lowpass filter LPF 14. As a result, analog signals in a frequency
15 range centered at fjf are obtained. The signal spectrum is shown in the bottom right-hand of
Fig. 1. fif could be albill~ily selected but in practice, 44 MHz should be selected as
mentioned above.
It is desirable to select 44 MHz for the carrier frequency in this invention. If so,
however, fractions will appear in fc or f5"~,b, which will complicate the explanation.
20 Therefore, the embodiments of this invention select a carrier frequency of 42 MHz, that is, fjf
= 42 MHz. Understand, however that only this selection is only for the sake of convenience
of explanation, and that it is allowable for this invention to be used with any one of 44 MHz,
54 MHz, and 42 ~ffIz.
This invention is similar to the conventional system shown in Fig. 6 in the
25 modulation of the I- and Q-signals, but differs in the modulation frequency ~ . The
CA 02248480 1998-09-28
conventional example shown in Fig. 6 takes a modulation frequency at 44 MHz (carrier
frequency), because the D/A converted analog signal is tran.~mitted at 44 MHz. However,
using 44 MHz for analog signals is completely different from using 44 MHz for digital signals.
If the signals are modulated digitally at 44 MHz, as shown in Fig. 6, an operating rate of
5 about ten times MHz is required for digital devices such multipliers and the sin-cos ROM.
Further, such a ten times higher speed (i.e., 440 MHz) is also required for generating carrier
waves with high quality (i.e., high C/N ratio). But no digital CMOS operating at such a high
speed has ever appeared on the market. Even if such high speed ICs were available they
would dissipate huge amounts of electric power.
This invention allows an oscillator 7 to produce a modulation wave with an angular
frequency ~ c, which is far lower than 44 MHz. Therefore, the modulation frequency fc =
~ J2 ~ differs from the carrier wave frequency fjf = ~ /2 ~, which is based on an extremely
ingenious idea. The lowering of the operating frequency gives this invention a significant
advantage in reali~ine a digital modulation system. This invention further brings about
another benefit of reduced electric power consumption by lowering the modulation frequency,
which allows the digital integrated circuits to operate at a moderate to low frequency.
Therefore, reduction of electric power dissipation is another one of the merits of this
invention. Here, attention should be paid to the relationship between the frequency f and the
angular frequency c~). The angular frequency cl) is derived by multiplying the frequency f
by 2 7~ and is used for simplifying the arguments of sin and cos. 44MHz multiplied by 2 7~
is cl~, which is 276 megaradian/sec, but ~ is not referred to as 276 megaradian/sec.
Rather ~ is ordinarily referred to as "44 MHz", i.e. by the unit of frequency.
In the conventional example shown in Fig. 6, a 44 MHz carrier wave and a 44 MHz
modulation wave were used. Since there was no way to change the modulation wave
frequency, the I- and Q-signals should be modulated at 44 MHz from the beeinning
CA 02248480 1998-09-28
Therefore, it was impossible to obtain a high-quality modulated signal. In the conventional
example shown in Fig. 8, the modulation signal of 10 MHz was generated by a facile method
of obtaining high-quality carrier waves. Since the modulation frequency was less than 44
MHz, the frequency must then be increased by using a local oscillation frequency of 54 MHz
(=44~ffIz+10MHz).
This invention is completely different from the conventional examples described
above. A low-frequency modulation wave fc may be used because fc = fif - n ~ fs~ b It is
therefore possible to lower the modulation frequency fc by integer multiples of f5~ b below the
tr~ncmiccion (carrier) frequency fjf. As the integer n is freely selected, the modulation
10 frequency fc may even be smaller than fs~ b The modulation frequency fc may therefore be
low enough that digital multipliers and digital phase shifters are permitted to operate at speed
below 440 MHz or so. CMOS LSIs currently on the market are capable of operating in
conjunction with such a low modulation speed.
Furthermore, another outst~nrling characteristic of this invention is to treat signals
15 not as real numbers but as complex ones. Therefore, the oscillator 7 outputs a complex
signal exp(j CL) ct)~ and the BPF 11 is a complex filter. The carrier wave fjf is also made to be
at a high frequency (42 MHz), even while the modulation frequency fc is low. As explained
above, while 44 ~Iz is the officially determined carrier frequency, we are ~Csuming the
carrier frequency to be 42 MHz in order to çlimin~te fractions from our explanation.
The reason why it is possible for this invention to operate so advantageously will be
explained with reference to Figs. 1-3 showing spectra of frequency bands. It is assumed that
f5,""b = 5 MHz for the sake of simplifying the explanation, and that the interpolation factor of
the devices 1 and 4 is 4, fc = 2 MHz, the interpolation factor of the device 10 is 6, and fif = 42
MHz.
IP1 and IP4 interpolate the input and I- and Q-signals raising their frequency from
14
CA 02248480 1998-09-28
f5""b of 5 MHz by multiples of 20 MHz. After that, I- and Q-signals are limited to 20
MHz by baseband filters 2 and 5 to become signals having frequency ranges 8 l and 82 in Fig.
3(a).
The signal fc = exp(j ~ ct), which is a modulation wave of frequency fc = 2 ~Iz, is
produced in the oscillator 7. The modulation wave fc = exp(j c~) ct) is multiplied by the signals
outputted from the baseband filter 2 by the multiplier 3. The modulation wave frequency fc
= 2 MHz is about one-twelfth as large as the conventional modulation wave frequency fif = 42
~Iz, which produces a high-quality oscillation output.
Fig. 3(b) shows the frequency ranges of the signals after multiplication by the
10 modulation wave exp(j c~) ct), in which the signal shown by Fig. 3(a) is shifted by only 2 MHz.
If the modulation wave of frequency fc = 2 MHz were a real component such as cosc~) t, the multiplying output would generate upper and lower bands centered at +2MHz and -
2MHz which would overlap and interfere with each other. Such interference would cause
signal distortion. This effect is easily conr"ll,ed from the fact that the cosine function has
15 two frequency components: i.e., cos ~ t = { exp(j ~ t) + exp(j c~) t) } /2.
After the I-signal is multiplied by exp(j ~ ct), and the Q-signal is multiplied by exp(j
cl) ct + j ~ /2), the I- and Q-signals are added to create an orthogonal modulation signal of I
exp(j ~ ct) + Q exp(j c~) ct + j 7~ /2).
A second interpolation device 10 interpolates the orthogonal modulation output.
20 This device raises the frequency of the signal by multiples of 20 MHz (e.g.; 20, 40, 60, 120
MHzj. By this operation, components ranging from 0 to 20 MHz thus appear repeated in the
ranges of from 20 to 40 MHz, from 40 to 60 MHz, from 60 to 80 MHz, from 80 to lO0 MHz
and from lO0 to 120 MHz. The complex BPF l 1 is used to select only the desired frequency
range of the carrier wave frequency fif from among the repeated components.
Here, the BPF l l for extracting only the band of a desired carrier wave frequency fif
CA 02248480 1998-09-28
is a complex one. A real BPF would pass the unwanted higher harmonics 89 or 100 as
shown in Figs.4G) and 4(k), because a real BPF is composed of two windows 98 and 99
which are symmetric about 60 MHz = fS/2 (f5=120 MHz). However, there exists no
symmetry between the frequency band 93 (87) and the frequency band 100 (89) at fS/2
(60M Hz), because the signals were frequency-shifted by 2 MHz at the beginning This
mi.cm~tçh would creak a serious distortion problem. In the case of the complex BPF,
passage of the unwanted higher harmonics does not occur, as any component outside the
desired frequency range is excluded by the complex filter having an albiLlaly window within
the range of from 0 to fs
10By using interpolation to generate higher harmonics of the signal, this invention
allows modulation of input signals to be performed at a low modulation frequency fc which
will then be raised up to higher frequency ranges, one of which includes fif (fif=fc+n fs~ b)
Up to this point, the signals are digital. A real component extraction circuit 12
extracts the real component in order to change the digital signal, to an analog one. As shown
15in Fig 3(f), if only the real component is kept, the component 94 centered at 42 MHz
remains, while a component 96is newly generated by folding the range 94 at 60 M Hz(95).
The extra image component 96 appears at 78 MHz because of the extraction of the real
component. If the complex signal is used as it is, such an image component never appears.
A D/A converter 13 then converts the digital signal into an analog one. The analog
signal contains components centered at 42MHz and at 78 M:HZ.. An analog low-pass filter
14 absorbs the 78 MHz component and allows the 42 MHz component to pass through, as
shown in Fig. 3(g). The 42MHz analog signal is then tran.cmitted out as a sending signal
from the station.
By separating the modulation frequency from the carrier frequency, this invention
can use a sufficiently low frequency of, e.g., 2MHz, for modulation while m~int~ining the
16
CA 02248480 1998-09-28
carrier frequency at 44 MHz (again, we are a~sllming that the carrier frequency is 42MHz in
order to simplify the relations among the various frequencies). This invention is already
quite dirrerenl from the typical orthogonal modulations of Fig. 6 or Fig. 7 for its wide
separation of the modulation frequency from the carrier frequency.
The prior art method of Figs.6 and 7 suffered from the difficulty of requiring high-
speed modulation, since the modulation frequency is equal to the carrier frequency. Another
prior art method shown in Figs.8 and 9 uses a sufficiently low modulation frequency, then
requires an extra local oscillator in order to raise the signal frequency from the low
modulation frequency up to the 44MHz of the carrier frequency. However, the present
10 invention can suppress the modulation frequency to be far lower than the carrier frequency
(e.g. 44MHz). Such a low modulation frequency never requires high-speed operation by the
digital ICs, and slow modulation can easily be realized by popular, inexpensive CMOS
devices already available on the market.
Fig. 4(h) shows the spectrum after interpolating the modulated signal (Fig. 3(b)) by
15 multiples of 20 MHz by the device 10 to create frequency ranges 85 to 9l, and is the same as
the spectrum of Fig.3(c). Since the interpolated signal goes through a complex filter 92
having a transparent window between 30 MHz and 54 MHz, only the signal in range 87 is
passed and the images in the other ranges 85, 86, 88, 89, 90 and 91 are rejected. Fig. 4(i)
exhibits the spectrum having only the frequency range 93(87) having a central frequency of
44MHz, which is the same as Fig. 3(e).
The complex filter 92 is capable of only selecting a single frequency range, if a real
filter were used instead of the complex one, an extra 66-90 MHz window 99 would appear in
addition to the 30-54 MHz window 98. In other words, the real filter would select an extra
frequency range 89 centered about 82 MHz in addition to the frequency range 87 centered
about 42 MHz. Fig. 4(j) shows the spectrum having the 42 MHz frequency range 93 (87)
CA 02248480 1998-09-28
and the extra 82 MHz frequency range 100 (89) passed by the real filter. The windows 98 and
99 of the real filter are symmetric to one another with regard to the central frequency 60 MHz,
which is the center of the spectrum expanded from 0 MHz to 120 MHz by the interpolation
device 10. But the signals in the frequency ranges 93 and 100 chosen by the real filter are
not symmetric, because the original baseband signal 81 has been shifted by the modulation
frequency by 2MHz in the step of Fig. 3(b). This 2MHz shift by the modulation breaks the
symmetry of the interpolated frequency ranges 93 and 100 with respect to 60 MHz. When
the real component extractor 12 passes only the real component of the signal, the final
spectrum has extra frequency ranges 101 and 103 which are mirror images of the ranges 93
10 and 100 being symmetric to ranges 93 and 100 with regard to 60 MHz. Since D/A conversion
does not change the frequency ranges, the spectrum after the D/A conversion, has extra ranges
101 and 103 in addition to the intrinsic ranges 102 (93,87) and 104 (100,89), as shown in Fig.
4(1).
In the vicinity of 40MHz, for example, the frequency range is a sum of the inherent
15 range 102 (93, 87) and the superfluous range 101. The range 102 is equal to the intrinsic
frequency range 93 having the central frequency 42 MHz. However the range 101 is only a
mirror image of the range 100 of 82MHz. This superfluous frequency range 101 creates a
serious problem. The upper frequency range near 80 MHz is composed of a similar sum of
the superfluous range 103 and intrinsic range 104. However, the higher counterpart 103 &
20 104 creates no problem, since an analog low pass filter LPF 14 rejects the upper frequency
range.
The lower unifled frequency range 101 & 102 causes signal distortion. The reasonfor this distortion will now be analyzed. Here "g" denotes the frequency range included in
the original baseband frequency range 81 in Fig. 3(a). The baseband frequency range g
25 which includes the original input signal does not extend beyond 2MHz. The modulation by
18
CA 02248480 1998-09-28
the oscillator 7 and the multiplier 3 raises the baseband range 81 of g up to a modulated
frequency range 83 of (g+2). After interpolation, the modulated signal is found in a
frequency range 93 and another frequency range 100 at the output of the real number filter
BPF. The range 93 extends from (42-g) MHz to (42+g) MHz, and the range 100 extends
5 from (82-g) MHz to (82+g) MHz, which ranges are shown in Fig. 4(k). When the upper
range 100 is folded at the middle frequency 60MHz by the real component operator 12, a
mirror range 101 results. The mirror range produced by the real component operator 12
extends from (38-g) MHz to (38+g) MHz. In other words, by the action of the real
component operator, the baseband range g now appears in two frequency ranges, being the
10 range of (38+g) MHz, and the other being the range of (42+g) MHz. If g is smaller than
2MHz, then the two ranges do not overlap. If g is larger than 2MHz, however, the two
ranges overlap. Such overlapping results in the distortion of the signals. Even if
overlapping can be avoided, a practical filter could not fully separate the ranges of (38+g)
MHz and (42+g) MHz. In any case, therefore, two ranges of (38+g) MHz and (42+g)
15 MHz will cause signal distortion, which invites cross-talk between di~elenl channels.
Instead of the real bandpass filter REAL BPF, this invention makes use of a complex
b~n~1pass filter BPF. Therefore, it is entirely immune from such signal distortion, even if the
original baseband has a wide frequency range.
Thus a conspicuous advantage of the present invention is the possibility of adopting a
20 very low modulation frequency such as 2MHz. Prior high-speed modulation at 44MHz
required ultra-high speed digital ICs operating at a speed of 500~Iz or so which is five times
or ten times as high as the 44MHz modulation frequency, this factor being necessary for
m~int~ining a high C/N (count/noise) ratio. Digital CMOS LSIs currently on the market
cannot operate at such a high speed as 500MHz. But the low frequency modulation of e.g.,
25 2MHz of the present invention requires an operating speed of at most 10 MHz. Such a low
19
CA 02248480 1998-09-28
speed modulation can be carried out by popular, inexpensive, and widely available CMOS
LSIs.
How can this invention enjoy such a benefit of low speed modulation over prior art
orthogonal modulation systems? This matter will now be clarified. This invention takes
full advantage of the frequency enhancement of interpolation. In the present system, the
interpolation device makes a plurality of new signal images by adding multiples (n x fsymb) of
f5ymb to the original signal frequency. When the original signal has been expressed in a
frequency h, the interpolation easily produces images at frequencies of h + n x f,ymb where n is
an integer. The carrier frequency fjf is predetermined by government standard to be 44 ~Iz
10 or 57 MHz. The parameters f5ymb and n should be selected so that some of the interpolated
range coincides with the government-determined carrier frequency.
fjf = h + n x f"ymb (where n is an integer) (l)
Once the input signal frequency fsymb of the I-and Q-signals is determined, h can be reduced to
a value smaller than f5ymb by choosing some pertinent integer n (h< fsymb) The original signal
15 resides in a baseband beginning from 0 MHz. Raising the frequency by h lifts the original
baseband up to a suitable signal for introducing the interpolation. In other words, the
modulation only raises the 0MHz starting baseband of the original signal up to the h
frequency range. Then the modulation frequency fc is given by
fc = h= fjf - n ~ fsymb (n:integer) (2)
Since f can be determined to be smaller than fs~ b~ the modulation frequency fc which
is equal to h can also be a value smaller than f5ymb. The example above uses fsymb=SMHz~ n=8,
fjf=42MHz, and fC=2MHz. This example simplifies the frequency relations by using a
dummy modulation frequency of 42MHz to çlimin~te fractions.
The actual carrier frequency fif should be either fjf=44MHz or fjf=57MHz. If fjf25 should be 44MHz, the modulation frequency fc can set at 2MHz (fC=2MHz) by choosing
CA 02248480 1998-09-28
fsylllb=5.25MHz and n=8, because 5.25 x 8 + 2 = 44. Otherwise, an alternate selection of
fs,,l"b=5MHz, n=8 and fC=4MHz gives a 44MHz carrier, since 5 x 8 + 4 = 44.
If fjf should be 57MHz, then values of f5",l,b=5MHz and n=12 will realize a
modulation frequency fc = -3MHz, because 5 X 12+(-3) = 57. Another combination of
fsyr~b=6MHz~ n=9 and fC=3MHz allows a carrier frequency of fjf=57MHz since 6 x 9+3 = 57.
Note that even if fjf is different from 44~Iz or 57MHz, a low modulation frequency fc can
still be obtained through selection of n and f5,"r,b.
Furthermore, this invention will enjoy an additional advantage by providing the
complex BPF with a function to correct the aperture effect. The aperture effect is caused by
10 the continual outputting of the D/A converter. Fig. 5(b) shows the aperture effect 111, where
the abscissa is frequency and the ordinate is amplitude. Here, the sampling frequency of the
D/A converter is assumed to be 12Q MHz. The D/A converted signal m~int~in~ an
amplitude nearly equal to that of the digital input signals for low frequencies far from the
sampling frequency of 120 MHz, but the output amplitude is reduced for high frequencies
15 close to 120 MHz. The amplitude of the D/A converted signal decreases to 0 at 120 MHz.
The fall of the D/A conversion amplitude with respect to input frequency, is represented by a
sinc function (sin ~
Hp( ~ j) = sin( c(~ T/2)/( c~) T/2), (3)
where T is a sampling period (l/T=120MHz) and ~ is the angular frequency of the digital
signal. The fall of the curve 111 of Fig. 5(b) demonstrates the aperture effect.If a complex bandpass filter BPF having a flat window 110 were adopted, the
aperture effect would cause the amplitude of the output of the D/A converter to exhibit
frequency dependence even within the window between 34 MHz and 48 MHz. In order to
compensate the aperture effect of the D/A converter, it is preferable to adopt a filter having a
window 112 with a rising dependence on frequency as shown in Fig. 5(c). The output of the
CA 02248480 1998-09-28
D/A converter will be fl~ttçned in the range between 34 MHz and 48 MHz by the balance
between the aperture effect of the D/A converter and the rising dependence of the filter on the
input frequency.
Conventional digital orthogonal modulation systems required that the modulation
5 frequency should be equal to the carrier frequency, since the prior systems multiplied the I-
and Q- signals by sinc~)t and cosc~)t, sl-mmed the modulated signals as Icos~t+Qsin~t, and
D/A-converted and tr~nemitted them. The carrier frequency is high enough, for example, at
44 MHz or 57 MHz. Digital ICs must operate at a speed several times to ten times as high
as the carrier frequency, however, and such digital ICs have not been sold in the market. In
10 contrast, this invention can modulate digital signals at a frequency far lower than the carrier
frequency by separating the modulation frequency from the carrier frequency. The low
modulation frequency allows digital ICs to operate at low speeds for modulation. The use of
inexpensive devices cuts the cost of production. There is a large difference between the
modulation frequency and the carrier frequency, but the interpolation step can easily raise the
15 mod~ tion frequency up to the carrier frequency without requiring local oscillation. Also,
the structure of the circuits is simple. The modulation frequency fc is determined to be fc = fif
- n ~ f5yr"b, and thus the modulation frequency can be less than f5,,ll,b.
In the accompanying drawings:
Fig. l is a schematic view of components of a orthogonal digital signal modulation
20 apparatus of the present invention. Here, fi~ b is an intrinsic frequency of digital input
signals of I-ch and Q-ch, fjf is a carrier frequency, and fc is a modulation frequency.
Fig. 2 is half of Fig. 1 having an I-ch branch for showing only significant
components for the orthogonal modulation.
Fig. 3 shows spectra of the signals processed by the components of the orthogonal
25 modulation system from the initial baseband I-ch signals to an IF output. Fig. 3(a) is a
22
CA 02248480 1998-09-28
spectrum of the output of a baseband filter 2. Fig. 3(b) is a spectrum of the signal multiplied
by the modulation wave. Fig. 3(c) is a spectrum of the interpolated signal with a step of 20
MHz up to 120 MHz. Fig. 3(d) is a spectrum of a window of a complex BPF. Fig. 3(e) is a
spectrum of the frequency range of the signal treated by the complex BPF. Fig. 3(f) is a
5 spectrum of an analog output of a D/A converter. Fig. 3(g) is a spectrum of the IF signal as
an output of the analog low pass filter LPF.
Fig. 4 shows spectra of the signals processed by the components of the orthogonal
mod~ tion appa,~Lus of the present invention for showing the merits of reducing signal
distortion by using a complex BPF instead of a real BPF. Fig. 4(h) is a spectrum of the
10 frequency ranges generated by the interpolation device by raising the frequency by multiples
of 20 ~Iz and passed by the complex filter. Fig. 4(i) shows a spectrum of a frequency
range passed by the complex BPF. Fig. 4(j) is a spectrum that shows the selection of two
ranges by the real BPF. Fig. 4(k) is a spectrum of the output of the real BPF having two
windows. Fig. 4(1) is a spectrum of the output of the D/A converter, where mirror images
15 overlap the original frequency ranges due to folding of the spectrum with regard to the center
of 60 MHz.
Fig. S shows spectra for explaining the aperture effect of reduced amplitudes at
higher frequencies which derives from a definite sampling frequency (l/T) of the D/A
converter. Fig. S(a) is a spectrum of a window of an ordinary bandpass filter BPF. Fig.
20 5(b) is a spectrum of the output of the D/A converter operating at a sampling frequency of
120 MHz, which shows an aperture effect. Fig. S(c) is a spectrum of a window of a BPF
designed to cancel the aperture effect of the D/A converter.
Fig. 6 is a schematic view of a typical prior art, orthogonal digital signal modulation
appa~lus having a modulation frequency fc which is equal to the carrier frequency fif. The
25 rigid relation fC=fif requires such a high rate of modulation that digital arithmetic devices
23
CA 02248480 1998-09-28
cannot perform the necessary sin, cosine, addition, or multiplication operations. In addition,
even if m~nllf~ct~lrers could produce excellent digital devices capable of operating at such a
high speed, the devices would consume large amounts of electric power, which would
prohibit wide-spread use of digital orthogonal modulation for tr~n~mi~sion.
Fig. 7 shows spectra of the signals processed by the components of the prior artapparatus of Fig. 6. Fig. 7(a) is a spectrum of the frequency ranges of the I-ch signal treated
by the baseband filter. Fig. 7(b) is a spectrum of the frequency ranges of the I-ch signal
interpolated by multiples of 10 MHz by the interpolation device IP, which has a plurality of
frequency ranges with lO MHz steps. Fig. 7(c) is a spectrum of the frequency ranges of the
10 signal which has passed the digital lowpass filter LPF. Fig. 7(d) is a spectrum of the signal
multiplied by a modulation frequency (=carrier frequency). Fig. 7(e) is a spectrum of the
output signal of the D/A converter. Fig. 7(f) is a spectrum of the IF signal as an output of the
analog LPF.
Fig. 8 is a sçhem~tic view of another prior art simplified orthogonal digital
15 modulation system. Instead of multiplying the signals by sin cl) t and cos ~ t, the signals are
multiplied by a series ofthe numbers of l, 0, -1, 0 in such order.
Fig. 9 shows spectra of the frequency ranges of signals processed by the components
of the prior art app~ s of Fig. 8. Fig. 9(a) is a spectrum of the output of the baseband
filter. Fig. 9(b) is a spectrum of the signal multiplied by the simplified modulation numbers
20 of 1, 0, -1, 0 in turn. Fig. 9(c) is a spectrum of the output signal of the D/A converter. Fig.
9(d) is a spectrum of the first IF signal without the high-frequency image components which
have been rejected by an analog lowpass filter. Fig. 9(e) is a spectrum of the signal which
has been frequency-converted by a mixer and a local oscillator for converting the frequency to
the carrier frequency (44 MHz). Fig. 9(f) is a spectrum of the IF signal treated by a
25 bandpass filter BPF which has a window that includes the 44 MHz range.
24
CA 02248480 1998-09-28
Fig. 10 is a view of a basic structure of a first embodiment of the present invention.
The first embodiment modulates the input digital signal by a complex oscillation at a
frequency far lower than the carrier frequency fjf, frequency-converts the modulated signal by
an interpolation device IP6, selects the frequency ranges by a complex bandpass filter BPF,
and D/A-converts the digital signal into an analog signal.
Fig. 11 is a view of a detailed structure of the first embodiment which performs the
complex oscillation by a sum of cos and sin oscillations and assembles the complex filter
from a real filter and an im~gin~ry filter.
Fig. 12 is a view of a basic structure of a second embodiment of the present invention
10 for digital orthogonal modulation. The second embodiment modulates I-ch and Q-ch
signals by multiplying with 0, -1, 0, +1 in turn, adds the modulated I-ch and Q-ch signals,
modulates the sum at a frequency fc which is far lower than the fif, frequency-converts the
modulated sum by an interpolation device IP6, selects the frequency ranges with a complex
ban~p~s filter BPF, and D/A-converts the digital signal into an analog signal.
Fig. 13 is a schematic view of a third embodiment applied to the VSB modulation.This embodiment shifts the frequency of the signal with the VSB filter, modulates the
frequency-shifted signal at a frequency fc lower than fif, interpolates, and selects the desired
frequency range with a complex ban-lp~s filter BPF.
Fig. 14 shows spectra of frequency ranges at the processes of the VSB modulation of
20 Fig. 13. Fig. 14(a) is a spectrum ofthe output signal ofthe VSB Nyquist filter. Fig. 14(b)
is a spectrum of the signal multiplied by a modulation frequency lower than the carrier
frequency. Fig. 14(c) is a spectrum of the output of the interpolation device IP6 which
increases the frequency of the signal by multiples of 20 MHz. Fig. 14(d) is a spectrum of the
output of the complex b~n~p~ss filter BPF which has a single window. Fig. 14(e) is a
25 spectrum of the signal processed by a real component operator and a D/A converter. Fig.
CA 02248480 1998-09-28
14(f) is a spectrum of the IF signal selected by a lowpass filter.
Fig. 15 is a detailed view of the third embodiment. The complex filter consists of a
real filter and an im~gin~ry filter. The complex oscillation is composed of cos and sin
oscillations.
Fig. 16 is a component view of a prior art VSB filter by Hilbert's conversion (phase
shift). Fig. 16(a) is a simplified view of a prior art phase-shifted VSB filter. Fig. 16(b) is a
detailed figure of a prior art phase-shifted VSB filter.
Fig. 17 is a schematic view of a prior art digital VSB modulation.
Fig. 18 is a circuit of the modulation signal generator which consists of a phase
10 accllmlll~tor, a phase register, and a sin&cos ROM.
[Embodiment 1 (Fig. 10)]
Fig. 10 shows a basic example of the orthogonal modulation system of the presentinvention. Fig. 10 is quite similar to Fig. 1. The input signals are the I-ch digital signal and
the Q-ch digital signal. Interporation devices IP4 121 and 124 interpolate the input signals.
15 Baseband filters 122 and 125 reject the spectrum components of the signal frequency higher
than 20MHz. The I-ch signal is multiplied by a modulation wave exp(j ~ct) and the Q-ch
signal is multiplied by another modulation wave exp(j c()ct+j ~/2), where ~/2 ~ is the
modulation frequency. These two modulation waves are orthogonal to each other due to the
phase di~erence of 90 degrees. Then the orthogonally modulated signals are added by an
20 adder 129. Interpolation device IP6 130 interpolates the sum by multiples of 20 MHz up to
120 MHz. A complex b~n-lpass filter BPF 131 selects only the frequency range including fif
(44 MHz in the example). Rejecting the im~gin~ry component, a real component operator
132 passed the real component of the signal. A D/A converter 133 converts the digital
signal into an analog signal. An LPF 134 elimin~tes images to extract only the frequency
25 range including fif (44 MHz in the example). Thus the modulation frequency fc which is
CA 02248480 1998-09-28
determined to be fc =f~ - n ~ fs~ b is lower than fs~ b
[DETAILS OF EMBODIMENT 1 (FIG. 11)]
Fig. 11 is a detailed figure of Fig. 10. An oscillator 127 consists of a cosine
generator 147 and a sine generator 148 for producing a complex oscillation. A (-1)(1)
5 selection circuit 149 multiplies the sine or cosine waves by (-1) or (+1) and sends the products
to the multipliers 143, 151, 146 and 150 for making products ofthe I-ch and Q-ch signals by
sine and cosine waves. The multiplier 143 makes Icoscl)ct. The multiplier 151 produces
Isin~ct. The multiplier 146 produces -Qsinc-)ct and the fourth multiplier 150 makes Qcosc~)
ct. The adder 129 of Fig. 10, in practice, consists of a first adder 153 and a second adder 152
10 for performing complex calculation. The first adder 153 calculates a real part (Icos~ct-Qsin
c~)ct), while the second adder 152 calculates an im~gin~ry part (Isin~ct+Qcos~ct). The
complex multiplication requires Iexp(j c~) t)+Qexp(j ~ t+j 7~ /2) in the present invention instead
of the conventional (Icos cL) t + Qsin c~) t), where the subscript c is omitted in ~ c for simplicity:
Iexp(j ~ t)+Qexp(j cl) t+j 7~ /2) = Icos cl) t+jIsin c~) t+Qjcos ~ t-Qsin ~ t
= Icos c~) t-Qsin c~) t+j {Isin ~ t+Qcos ~ t}. (4)
The adder 153 produces the real component and the adder 152 produces the
im~gin~ry component. The BPF 131 of Fig. 10 comprises a BPF 155 for the real component
and another BPF 157 for the im~in~ry component. Both BPFs 155 and 157 have a
window for selecting suitable frequency ranges. An adder 158 sums the real and the
20 im~gin~ry components up to Icos~t-Qsinc~t-Isin~t-Qcos~t. The D/A converter 133 then
converts the digital signal to an analog signal at a sampling frequency of 120MHz.
[Embodiment 2 (simplified type: Fig. 12)]
Fig. 12 shows a second embodiment of a simplified orthogonal modulation circuit.
The simplified method multiplies the input signal by 0,-1,0, +1 in turn instead of multiplying
25 by cos~t or sinc(~t. The operation thus makes a series of I, -Q, -I, Q, I,-Q, . Then a
CA 02248480 1998-09-28
oscillator 170 outputting exp(j2 ~ fct) multiplies the series I, -Q, -I, Q, I,-Q, of signals by
exp(j27~fct)=exp(j ct)ct). The modulation frequency ~c is small enough to permit such
operation. The interpolation device IP6 172 raises the frequency of the modulated wave by
multiples of 20 MHz (20 MHz, 40 MHz, ,120 MHz). An adder 158 sums the real and the
5 im~gin~ry components. A D/A converter (not shown) converts the digital signal into an
analog one. The receiving port extracts the signals by repetition of I, -Q, -I, Q, I, -Q, to
retrieve the I- and Q-signals. Here f5~,,nb is 5 MHz and 24f5",l,b is 120 ~Iz.
[Prior Art: VSB filter by Hilbert conversion (phase shift)]
This invention can be applied to a VSB (vestigial sideband) modulation. Before
10 explaining the embodiment, a conventional VSB modulation circuit is now briefly reviewed
with reference to Fig. 16(a). Here ~ in is an angular frequency of the input signals and ~ c
is a modulation frequency of an oscillator 233. The input signal is modulated by a multiplier
231 on one path. The same input signal is phase-shifted by 7~ /2 (90 degrees) by a phase-
shifter 232, and the phase-shifted signal is also modulated with a 90-degrees advanced (234)
modulation wave by another multiplier 235. An adder 236 sums the modulated signals.
The input signal is denoted by Vcos ~ jnt and the modulation wave is denoted as cos c~) ct. The
output signal Vout may thus be expressed as
Vout = Vcos CL) jntcos C() ct + Vsin ~ jntsin ~ ct = Vcos( c~ in~ Cl) c)t . (5)
The frequency of the output signal has been reduced. The shift toward a lower frequency
results from the 90-degree advancement of both the modulation wave and the input signal.
If one of the modulation wave and the input signal is delayed by 90 degrees in~tea(l, then the
modulated signal shifts the output frequency higher, i.e. to Vcos(cl) jn+cl)c)t. Fig. 16(b)
shows details of a VSB modulator making use of the phase shift of 90 degrees at shifters 242
and 245.
Fig. 17 shows a simplified prior art VSB (vestigial sideband) modulation circuit
28
CA 02248480 1998-09-28
cont~inin,e the VSB baseband Nyquist filter. The I-ch signal is interpolated by interpolation
device IP 250 and is treated by a VSB baseband Nyquist filter 251 to produce a signal having
the VSB property, which means a frequency shift from C~) jn to (CI)i~+ ~C) The signal
Vcos( c~ jn- ~ c)t is multiplied by 1, 2-"2,o, -2-"2,- l, -2-"2, 0, 2-"Z, 1, , which are the values of
cos2 7~ fc at 2fC ~ /4) x n, where n is an integer, in turn at a frequency fc The signal is
D/A converted by a D/A converter 254. A lowpass filter 255 elimin~tes high frequency
components. The signal has a low central frequency fc which is far less than 44 MHz.
Then a mixer 256 mixes the fc-centered signal by the oscillation f of a local oscillator 257,
where f= fif-fC or f= fif+fC Mixing raises the frequency up to the carrier frequency fjf (for
10 example, 44 MHz or 54 MHz).
[Embodiment 3 (Application to VSB modulation: Fig. 13)]
This invention can be applied to the VSB modulation. Figs. 13 and 14 demonstratea VSB modulation improved by the idea of this invention. The inherent signal speed f5ymb is
assumed to be f~ymb=lOMHz for convenience of explanation. An interpolation device IP2
15 181 raises the frequency of the signal by multiples of 2fsymb=20~Iz A VSB Nyquist filter
182 which runs at a speed of fsymJ4 makes a signal having the VSB property. Fig. 14(a) shows
the spectrum of the signal processed by the Nyquist filter. The output of the Nyquist filter
has a spectrum 191 having a center frequency of-f5ymb/4 and another spectrum 192 having a
center frequency of 2f5ymb - fsymJ4 The baseband Nyquist filter is a complex filter.
A local oscillator 184 generates a modulation wave exp(j2 7~ fct) for multiplying the
filtered signal at a multiplier 183. The modulation frequency fc should be selected as fc = fjf
+ fsymJ4 - n ~ f5ymb (where n is an integer). The modulation frequency fc can be made
sufficiently low by choosing an appropriate integer n. In the example, fc = 4.5MHz, i.e.
4.5~Iz = 42MHz + (10/4)MHz - 4 x lOMHz. The fc modulation increases the frequency of
25 the signals in ranges 191 and 192 by fc up to ranges 193 and 194 in Fig. 14(b). An
29
CA 02248480 1998-09-28
interpolation IP6 185 raises the signal by 6 multiples of 20 MHz (+20 MHz, +40 MHz, ,
+120 MHz) to produce frequency ranges 195, 196,, 201, as shown in Fig. 14(c). A
complex bandpass filter BPF 186 has a single window that includes the carrier frequency 42
MHz, and does not have a mirror window which would have a center frequency of 78MHz.
Then a complex filter BPF 186 passes only a frequency range 197. Fig. 14(d) shows the
single range 202 (197) in the vicinity of 42 MHz.
A real component operator 187 takes only the real component of the signal. The
real part has two frequency ranges 203 and 204 as shown in Fig. 14(e). Since the real
component operation introduces image spectra, an extra range 204 appears at 78 MHz. The
10 digital signal is converted into an analog signal by a D/A converter 188. An analog lowpass
filter LPF 189 cuts the higher mirror range 204 and keeps only the lower range 205 in Fig.
14(f) which was the range 203 in Fig. 14(e) or 202 in Fig. 14(d). Thus the frequency range
205 at 42 MHz is outputted from the LPF 189 as an IF signal.
[DETAILED EMBODIMENT 3 (VSB MODULATION; FIG. 15)]
Fig. 15 shows details of the third embodiment as applied to VSB modulation. The
VSB modulation has only an I-ch signal, as opposed to the orthogonal modulation which has,
both I-ch and Q-ch signals. The signal frequency f59n,b is 10MHz. An interpolation device
IP2 210 raises the frequency of the signal by multiples of 2fs9n,b=20MEIz. The complex
Nyquist filter 182 consists in practice of a real Nyquist filter 211 and an im~gin~ry Nyquist
20 filter212. TheNyquistfilters211 and212selecttheranges 191 and 192inFig. 14(a).
The local oscillator 184 for multiplying the signal by exp(j2~fct) in Fig. 13 isconstructed in Fig. 15 with a cos oscillator 213 producing cos 2 ~ fct and a sine oscillator 214
producing sin2~fct. A(-l)(+1) selection circuit 215 makes cos, sin, -cos, -sin components
from the outputs of the sine oscillator 214 and the cosine oscillator 213. Instead of using
25 two independent oscillators for sine wave and cosine wave, it is possible to omit one oscillator
CA 02248480 1998-09-28
and replace its output by ~hi~ing the phase of a single oscillator by 90 degrees. Similarly the
complex BPF 186 is constructed with a real part BPF 223 and an im~gin~ry part BPF 225.
[MODULATION WAVE GENERATrNG CIRCUIT]
~ig. 18 shows an example of a modulation wave generating circuit which
corresponds to either the oscillator 127 in Fig. 10, the oscillators 147 and 148 in Fig. 11, the
oscillator 170 in Fig. 12, the oscillator 184 in Fig. 13, or the oscillators 213 and 214 in Fig. 15.
If analog signals were to be carried, the carrier wave would be a continual sine wave which
could be easily generated. However, the modulation wave is not an analog wave but rather a
digital wave. For making digital modulation waves, this invention uses a look-up table
10 ROM 263 for outputting sin ~ and cos ~ for independent values of ~ . A phase
accuml-lator 261 makes saw (ramp) waves of an a l,i~raly period (-l/fc) A saw wave 264,
265, 266, means a wave which rises in subst~nti~lly linear plopollion to 2 ~ at time t, falls
suddenly from 2 ~, then rises again at the same speed. The ramp waves are repeated in the
phase accum~ tor 261. A frequency control input all,iLl~lily determines the modulation
15 frequency fc by ch~nging the slope of the ramp wave 264. A phase register 262 creates
discrete phases c.~t from the ramp wave of the phase accumulator 261. When a discrete
phase c~) t is inputted to the look-up table ROM 263, the ROM 263 supplies the applop,iate
values of sin ~ t and cos ~ t.
Since the modulation circuit is a digital circuit, the multipliers perform multiplication
20 at discrete intervals which are called sampling times. A signal ramp wave 264 includes a
plurality of sampling times (1/f,). The phase interval ~ ~ is a sampling interval in phase
that corresponds to the sampling time 1/fS. Values of sin~t and cos~t are read out at all of
the phases separated by ~ ~ . A narrower sampling phase interval ~ ~ thus realizes a finer
modulation. However, a narrower sampling interval ~ ~ also requires digital devices
25 having a higher operating speed. 2 ~ is the number of sampling intervalsin a period.
CA 02248480 1998-09-28
If this sampling number is ten, the conventional modulation of 44 MHz would require
ultrahigh speed digital ICs which could run at 440 MHz. This invention, however, alleviates
the need for high-speed digital ICs by separating the modulation frequency from the carrier
frequency. When the modulation frequency fc is 2 MHz in this invention, the maximum
S speed imposed upon the digital ICs is only 20 MHz. When the modulation frequency is 4.5
MHz, digital devices which operate at 45 MHz will allow the apparatus to take ten sampling
points in each period.
32
