This invention relates to a method of designing a passive RC complex filter employed in a Hartley Image-rejection radio receiver for widening an Image-rejection frequency range.
Although heterodyne receivers typically constitute a receiver architecture in currently available radio appliances, those heterodyne receivers have an essential problem of image interference. Image interference is a phenomenon in which two RF signals having a symmetrical relationship between a high frequency side and a low frequency side with respect to a local oscillator frequency “ω0” are converted into the same intermediate frequency (IF) range by a down conversion.
In
Equation 1
Normally, in order to prevent the image interference, the bandpass filter 100 which passes only the frequency band (ω0+Δω) is required at a front stage of the mixer. However, since the filter can be hardly integrated and a passband thereof is fixed, it is considerably difficult to apply the bandpass filter to a plurality of radio systems having different bands.
Image-rejection receivers are effective for removing the bandpass filter 100 from the receivers. As one of those effective receiver systems, a Hartley receiver shown in
In an actual circuit, as shown in
It should be noted that when differential mixers 101b and 104b are employed, as shown in
The bands of the Hartley receivers have been described in a qualitative manner. Quantitatively, as described in “Explicit Transfer Function of RC Polyphase Filter for Wireless Transceiver Analog Front-End” by H. Kobayashi, J. Kang, T. Kitahara, S. Takigami, and H. Sakamura, 2002 IEEE Asia-Pacific Conference on ASICs, pp. 137-140, Taipei, Taiwan (August 2002), it is only necessary that a complex transfer function “H(s)”, in which one of a highpass characteristic and a lowpass characteristic is set as a real part “Hr(s)” and the other is set as an imaginary part “Hi(s)”, be defined, and a frequency response be observed. At this time, a negative frequency becomes a response with respect to an image, whereas a positive frequency becomes a response with respect to a desirable wave. The complex transfer function H(s) using a first-order filter is expressed in Equation 3, and a frequency response obtained by normalizing “ωc” is shown in
Expansion of bandwidth of the Hartley receivers result in designing problems of passive RC complex filters, and higher-order complex transfer functions having wide band frequency responses in both a passband and a stopband must be designed. One of the conventional techniques for designing the higher-order complex transfer functions is described in “Low-IF topologies for high-performance analog front ends of fully integrated receivers” by J. Crols and M. S. Steyaert, IEEE Trans Circuits Syst.-II, vol. 45, pp. 269-282, March 1998.
In the conventional technique, first of all, a proper prototype lowpass characteristic is designed. As to the prototype lowpass characteristic, various sorts of characteristics are known, for instance, a Butterworth filter, and various higher-order characteristics can be readily designed.
Next, a variable transformation is performed with respect to the prototype lowpass filter so as to shift the frequency response to the side of the positive frequency on the frequency axis. Because of the shift operation, a bandpass type complex transfer function with which a positive frequency band becomes a passband and a negative frequency becomes a stopband is obtained. This transfer function succeeds to the shape of the prototype lowpass characteristic. However, with this method, since the transfer function has a complex pole, such a restriction that a passive RC complex filter can only have a negative real pole in a simple root cannot be satisfied. As a result, there is no choice but to realize the prototype lowpass characteristic by an active filter, resulting in demerits in terms of noise and power consumption of active elements.
Another method is described in “RC Polyphase Filter with Flat Gain Characteristic” by Kazuyuki Wada and Yoshiaki Tadokoro, Proceedings of the 2003 IEEE International Symposium on Circuits and Systems, Vol. I, pp. 537-540, May 2003. In this conventional technique, because the element values are directly designed based on the assumption of the structure of the higher-order RC polyphase filter, an active element is not required. However, although the frequency of the transfer zero point, namely, the resistance/capacitance products at the respective stages are clearly given based upon the equi-ripple model, the resistance values are, properly determined based upon the arbitrary constant “α.” In other words, although the numerator of the transfer function is perfectly designed, the denominator thereof is imperfect. Accordingly, the flatness of the passband cannot be completely guaranteed.
This invention has been made to overcome drawbacks in the conventional designing methods. That is, an object of this invention is to provide a method of designing a complex transfer function which can be realized in a passive RC complex filter at the same time while perfectly succeeding to the feature of the prototype lowpass characteristic.
In this invention, as a first step, a prototype lowpass transfer function F(p) having a pole on a unit circle is designed. For the F(p), a Butterworth filter (Butterworth characteristic) or a elliptic filter (simultaneous Chebyshev characteristic) in which a maximum passband loss “α” and a minimum stopband loss “A” satisfy a condition equation of α=A/√(A2−1) is employed.
Next, as a second step, a bilinear variable transformation expressed by p=j(s−j)/(s+j) is performed with respect to the prototype transfer function F(p) so as to derive a complex coefficient transfer function G(s). The complex coefficient transfer function G(s) has a pole in a simple root on a negative real axis, and a passband is present in a positive frequency and a stopband is present in a negative frequency.
As a third step, a passive RC complex filter H(s) having a transfer function identical to the complex coefficient transfer function G(s) is designed, and thereafter, an impedance scaling is performed, whereby the designing method is accomplished. The above-mentioned designing flow is shown in
According to the above-mentioned designing method, while perfectly taking the feature of the prototype lowpass transfer function F(p), a higher-order complex transfer function G(s) which can be realized by a passive RC complex filter can be simultaneously designed. A Hartley receiver using the higher-order complex transfer function G(s) can reject images over a broadband and can maintain flatness with respect to a desirable wave at the same time. In addition, the Hartley receiver is excellent in terms of circuit noise and power consumption, as compared with those using an active filter.
The present invention can be appreciated by the description which follows in conjunction with the following figures, wherein:
First, a description is made of a prototype lowpass transfer function “F(p)” having a pole on a unit circumference. A first characteristic corresponds to a Butterworth characteristic. A transfer function of an n-th order Butterworth filter in which a power half value (3-dB loss) frequency is normalized is expressed by the following equation:
A second-order Butterworth transfer function to a fifth-order Butterworth transfer function are designed, and coefficients of respective terms are expressed in the following table when the transfer function F(p) is represented by a rational function form as shown in the following equation:
A pole location of Butterworth filters is shown in
As the second, an elliptic filter is listed up. In order to describe the elliptic filter, an elliptic function is required, so that this elliptic function will now be briefly described. First, the Jacobian elliptic function is defined by the following equation:
Subsequently, first kind complete elliptic integrals are defined by the following equation:
While various sorts of specific values of elliptic functions are known, the specific values required in this case are given as follows:
While the above-mentioned items are prepared, various sorts of parameters will now be defined which are required so as to describe the pole location of the elliptic filter. First of all, L1 and L2 are defined from both a maximum passband loss “α” and a minimum stopband loss “A” in accordance with the following equation:
Next, as solutions of the following equation, parameters “k1” and “k2” are defined. In this equation, “n” indicates an order of the transfer function. It should be noted that the parameter “k1” is called a “selectivity”, while there is such a relationship that a passband edge frequency is √k1, and a stopband edge frequency is 1/√k1.
Also, as a solution of the following equation, a parameter “ξ0” is determined.
sn(ξ0K(L2), L2)=1/α Equation 11
The transfer function F(p) of the elliptic filter is given by the following equation by employing the above-mentioned parameters:
A detailed content as to the above-mentioned method of designing the elliptic filter is described in “APPROXIMATION AND STRUCTURE” written by Masamitsu Kawakami and Hiroshi Shibayama, published by KYORITSU publisher in 1960.
In a general-purpose design of the elliptic filter, since the maximum passband loss αand the minimum stopband loss A are independently selected, no pole is arranged on a unit circle. However, in the case where a restriction condition of Equation 13 is given between α and A, a pole is arranged on the unit circle.
α=A/√{square root over (A2−1)} Equation 13
At this time, since the following equation can be established based upon Equation 11, it is determined that ξ0=½ from the specific value of the elliptic function of Equation 8:
Furthermore, due to the specific value relationship of the elliptic function, α0=1, and since α0=1, Tk=1.
As described above, also in the elliptic filter, since the restriction condition of Equation 13 is added, all of the poles are represented as being arranged on the unit circle of the complex plane.
In this case, with respect to 3 conditions in which the minimum stopband loss “A” is 30 dB, 40 dB, and 50 dB, when a second-order elliptic filter transfer function F(p) to a fifth-order elliptic filter transfer function F(p) are designed which satisfy Equation 13, and are expressed by a rational function form as shown in the following equation, coefficients of the respective terms are indicated in the following tables. It should be noted that odd-order terms of a numerator are not present, and are therefore omitted. Also, the passband edge frequency “√k1” and the stopband edge frequency “1/√k1” are additionally expressed:
Locations of poles of the above-mentioned elliptic filter are indicated in
Next, a complex transfer function “G(s)” is derived. In general, a transfer function “F(p)” of a prototype lowpass characteristic having a pole on a unit circle is expressed by the following equation:
A bilinear variable transformation is defined based upon Equation 18. If Equation 18 is solved with respect to “ω”, then responses with respect to Ω=0 and Ω=∞ of a prototype lowpass characteristic “H(p)” are mapped to ω=1 and ω=−1, respectively. As a consequence, such a new transfer function G(s) that an area in the vicinity of ω=1 corresponds to a passband and an area in the vicinity of ω=−1 corresponds to a stopband is obtained. Similarly, responses with respect to power half value frequencies Ω=−1 and Ω=1 of the Butterworth filter are mapped to ω=0 and ω=∞, respectively. A passband edge frequency Ω=±√k1 and a stopband edge frequency Ω=±1/√k1 of the elliptic filter are mapped to ω=(1±√k1)/(1∓√k1) and ω=(±1√k1)/(□1+√k1), respectively.
Since the prototype lowpass transfer function F(p) can be decomposed to a product of a first-order transfer function and a second-order transfer function, respective function cases thereof will now be described. First, in case of the first-order transfer function, this function can be modified as a form having a negative real pole and a negative zero on an imaginary axis:
Next, in case of the second-order transfer function, it is conceivable that the second-order function has a complex pole and thus may be defined as Q>½, the second-order function may be modified as a form having two different negative real poles and two zeros on an imaginary axis:
It should be noted that an inverse number relationship of the following equation can be established:
As apparent from the foregoing description, the higher-order function may be expressed by the following form:
Next, a transfer function “Gn(s)” normalized in such a manner that a response in a DC becomes G(0)=1 is expressed by the following equation:
Now, results of deriving the complex transfer functions Gn(s) from the prototype lowpass transfer function F(p) shown in Tables 1 to 4 are represented. Tables 5 to 8 represent “ρk” and “σk” in the order of magnitudes when transfer functions which constitute rational functions are indicated in a factorization form as shown in the following equation. Also, both passband edge frequencies and stopband edge frequencies are additionally expressed.
A chain matrix “mi” of the RC polyphase filter per stage is given by the following equation. When “n” stages of the RC polyphase filters are connected in the cascade connection manner, the resultant RC polyphase filters become a product “M” of the respective chain matrixes, and a transfer function H(s) is obtained by an inverse number of a 1-row 1-column element of the product “M.” A denominator D(s) constitutes an n-order polynomial of “s.” However, there is such a restriction that the denominator D(s) can have only a negative real simple root “−λAi”, while the restriction is caused by a passive RC circuit. It should be noted that when it is considered that all capacitors are opened in a DC in view of a nature of a circuit, the DC gain becomes H(0)=1.
In the beginning, a numerator polynomial is determined. “ω” may be directly adapted with respect to “ρ” which has already been designed. As to a denominator polynomial D(s), “n” sorts of resistance values must be determined.
First, the transfer function H(s) is a non-dimensional amount, and an impedance level is indefinite, so any one of the resistance values is arbitrarily determined. Similar to the case of the numerator polynomial, if “λ” may be directly adapted to “σ”, then the denominator polynomial D(s) becomes preferable. However, normally, the denominator polynomial D(s) has a form which can be hardly factorized. As a consequence, the denominator polynomial is once expanded so as to equally arrange coefficients from a first-order term up to an (n−1)-order term, and (n−1) pieces of simultaneous equations related to the remaining (n-1) sorts of resistance values may be solved.
On the other hand, in the above-described methods, several different sorts of solutions are obtained depending upon corresponding sequences between “ρ” and “ω.” There are some possibilities that a negative resistance value may be obtained depending upon the corresponding sequence. However, a circuit cannot be realized by employing this negative resistance value. As a result, “ρ” should be combined with “ω” by repeating trial and error until a positive resistance value is obtained. If all of element values are positive, any of solutions may be employed. However, if an extent of element values is narrow, an RC polyphase filter is easily manufactured.
While such a case of a third-order elliptic filter of a minimum stopband loss (=30 dB) is exemplified, a description is made of a specific example of the above-mentioned design sequence. First, “ρ” is read from Table 6, and the read “ρ” is adapted to “ω” in the following manner:
ω1=ρ3=1/ρ1
ω2=ρ2=1
ω3=ρ1=0.39559514072625 Equation 26
Also, “σ” is read from Table 6 so as to form the following polynomials:
σ1=0.21226147136004
σ2=1
σ3=1/σ1
(s+σ1)(s+σ2)(s+σ3)=s3+5.92343s2+5.92343s+1 Equation 27
On the other hand, the denominator polynomial D(s) of the RC polyphase filter transfer function is calculated from the chain matrix M so as to obtain the following equation with respect to unknown variables R1, R2, and R3.
For example, when R2=1, if the above-mentioned two polynomials are calculated by equally arranging coefficients, then the following element values are obtained:
As other examples than the above-mentioned example, even when “ω” and “ρ” are combined with each other in the following manner, the element values are obtained. It should be noted that in this designing example, solutions obtained from combinations except the above-listed combinations are improper:
Tables 9 to 12 represent results obtained by designing one set of respective element values of each of RC polyphase filters which correspond to the complex transfer functions shown in Tables 5 to 8:
While the various sorts of designing examples have been described above, all of those element values have been normalized. As a result, in an actual apparatus, if ratios of the element values of the elements which constitute a circuit are maintained, then the effect achieved by this embodiment may be obtained. Also, a center of the passband and a center of the stopband are 1 rad/s=0.16 Hz, and −1 rad/s=−0.16 Hz, respectively. In order to apply those centers to an actual circuit, impedance scaling must be performed. First, as to a resistance value, since a standard resistance value is approximately 1 kΩ in view of a circuit design, the resistance value may be uniformly multiplied by, for instance, 1,000. A capacitance value may be determined in accordance with a relationship between a desirable wave and an image. If the capacitance value is uniformly divided by, for example, 1 billion, the centers of the passband and the stopband are multiplied by 1 million, so that those centers are moved to 1 Mrad/s=160 kHz, and −1 Mrad/s=−160 kHz, respectively, and thus, those calculated values become element values suitable for such a case where the desirable wave is apart from the image by 320 kHz.
Although Embodiment 1 has described the designing example based upon the RC polyphase filters, there are other passive RC complex filters capable of realizing the complex transfer function Gn(s). As an example, other modes of passive RC complex filters capable of realizing a third-order complex elliptic filter (A=30 dB) are represented in
Frequency responses of
Since the RC polyphase filter of
The passband of the passive RC complex filter according to this invention becomes flat in the light of the maximum flat model if the Butterworth characteristic is selected as the prototype lowpass characteristic, and also becomes flat in the light of the equi-ripple model if the elliptic filter is selected. Also, the stopband of the passive RC complex filter becomes flat in the light of the maximum flat model if the Butterworth characteristic is selected as the prototype lowpass characteristic, and also becomes flat in the light of the equi-ripple model if the simultaneous Chebyshev characteristic is selected. As a consequence, by employing the passive RC complex filter according to this invention, the superior reception characteristic can be obtained.
This invention can be applied to a radio communication receiving apparatus.
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/JP2004/017644 | 11/19/2004 | WO | 00 | 5/11/2007 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2006/054364 | 5/26/2006 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
6212540 | Murakami et al. | Apr 2001 | B1 |
6529100 | Okanobu | Mar 2003 | B1 |
6854005 | Thiele | Feb 2005 | B2 |
20060106906 | Shabra et al. | May 2006 | A1 |
Number | Date | Country |
---|---|---|
08-065083 | Aug 1994 | JP |
2001-045080 | Jul 1999 | JP |
2001-077648 | Sep 1999 | JP |
2004-108792 | Sep 2002 | JP |
Number | Date | Country | |
---|---|---|---|
20080010617 A1 | Jan 2008 | US |