Non-recursive half-band filter

• US 5,177,700 A
• Filed: 12/24/1991
• Issued: 01/05/1993
• Est. Priority Date: 02/19/1987
• Status: Expired due to Fees

• Current Assignee: Ant Nachrichtentechnik GmbH
• Original Assignee: Ant Nachrichtentechnik GmbH
• Inventors: Gockler, Heinz
• Primary Examiner(s): Malzahn, David H.
• Application Number:
  US07/814,265
• Time in Patent Office: 378 Days
• Field of Search: 364/724.1, 364/724.13, 364/724.16
• US Class Current: 708/313
• CPC Class Codes: H03H 17/0657 where the output-delivery frequency is higher than the input sampling frequency, i.e. interpolation
• CPC Class Codes: H03H 17/0664 where the output-delivery frequency is lower than the input sampling frequency, i.e. decimation

Abstract

A non-recursive half-band filter having a filter length N and complex coefficients for receiving either a real input signal s(kT) or a complex input signal s(2kT) and for processing and converting the received input signals into either a complex output signal s(2kT) or a real output signal s(kT) wherein the complex coefficients operate at a function of h(l) where l=-(N-1)/2 to (N-1)/2 to (N-1)/2 and the filter length N is odd. As a result it is possible to convert a real input signal into a complex output signal, by modulating its pulse response to a complex carrier of the frequency equal to 1/4 or 3/4 of the sampling frequency, where the null phase of this frequency is an integer multiple of π/2. It is also possible to convert a complex input signal into a real output signal, by modulating its pulse frequency to the complex carrier of a frequency signal to the input sampling frequency or half thereof, where the null phase of this frequency is an integer multiple of π/2.

Claims

1. A non-recursive half-band filter having a filter length N for processing a real input signal s(kT) and converting the real input signal into a complex output signal s(2kT), said half-band filter comprising:

input means for receiving samples of a real input signal s(kT) at a sampling frequency fA, where fA=1/T, and for providing samples at one-half of the sampling frequency fA; and

means, responsive to the provided samples, for processing and converting the real input signal into a complex output signal s(2kT), said means having complex coefficients h(l), where l=-(N-1)/2 to (N-1)/2 and the filter length N is odd, with alternating purely real and purely imaginary values, said processing and converting means including means for modulating the pulse response of a half-band filter h(l) with exclusively real values and h(l) =h(-l) for all |l|≦

(N-1)/2 and h(l)=0 for l=±

2, ±

4, . . . , onto a complex carrier of a frequency of ±

1/4 of the input sampling frequency to yield

space="preserve" listing-type="equation">h(l)=h(l)·

e.sup.j(±

2π

lfA/4fA+φ

0) =j.sup.±

l ·

e.sup.jφ

0 ·

h(l),

wherein the null phase φ

0 of the complex carrier is an integer multiple m of π

/2 (φ

0=m·

π

/2 where m=0, 1, 2, 3 . . .), and wherein said modulating means is divided into a first branch and a second branch, said first branch comprising a chain of (N-1)/2 delay members each have a delay time of 2T, means for forming a plurality of difference signals and means for weighting said difference signals, with every second sample of the input signal s(kT) being routed into the chain of (N-1)/2 delay members;

said difference signal forming means subtracting from the output signal of the last delay member in the chain the input signal of the first delay member of the chain to form a first difference signal, subtracting from the output signal of the penultimate delay member in the chain the input signal of the second delay member of the chain to form a second difference signal, subtracting from the output signal of the third to last delay member in the chain the input signal of the third delay member of the chain to form a third difference signal, and so on until the outputs of each delay member are processed;

said weighting means including an adder and weighting the respective said difference signals by a function of h(l) of the pulse response and summing the weighted signals via the adder to yield either the real or the imaginary component of the filter output signal s(2kT);

said second branch including a second branch delay member which has a time delay of T·

(N-3)/2, into which is routed every other sample of the input signal, the output signal of said second branch delay member being weighted with a value equal to a function of h(0) to yield the other of the imaginary component and the real component of the filter output signal s(2kT).

2. A non-recursive half-band filter according to claim 1, where N=11 and m=1, and said weighting means weights the first difference signal with a value equal to the function -h(5), the second difference signal with a value equal to the function h(3) and the third difference signal a value equal to the function -h(1);

and the sum of the difference signals yields the real component s_r (2kT) of the filter output signal while the signal weighted with the function h(0)=1/2 yields the imaginary component s_i (2kT) of the filter output signal.

3. A non-reactive half-band filter according to claim 1, where N=11 and m=3, and said weighting means weights the first difference signal with a value equal to the function h(5), the second difference signal with a value equal to the function -h(3), and the third difference signal with a value equal to the function h(1);

and the sum of the difference signals yields the real component s_r (2kT) while the signal weighted with the function h(0)=-1/2 yields the imaginary component s_i (2kT) of the filter output signal.

4. A non-recursive half-band filter according to claim 1, wherein N=11 and m=0, and said weighting means weights the first difference signal with a value equal to the function h(5), the second difference with a value equal to the function -h(3), and the third difference signal with a value equal to the function h(l);

and the sum of the difference signals yields the imaginary component s_i (2kT) of the filter output signal while the signal weighted with the function h(0)=1/2 yields the real component s_r (2kT) of the filter output signal.

5. A non-recursive half-band filter according to claim 1, where N=11 m=2, and said weighting means weights the first difference signal with a value equal to the function -h(5), the second difference signal with a value equal to the function h(3), and the third difference signal with a value equal to the function -h(l);

and the sum of the difference signals yields the imaginary component s_i (2kT) while the signal weighted with the function h(0)=1/2 yields the real component s_r (2kT) of the filter output signal.

6. A non-recursive half-band filter having a filter length N for processing a complex input signal s(2kT) and for converting the complex input signal for into a real output signal s(kT) at double the input sampling frequency, said half-band filter comprising:

input means for receiving samples of a complex input signal s(2kT) at a sampling frequency fA'"'"'=1/2T;

means for processing and converting the received complex input signal into a real output signal s(kT) at an output sampling frequency fA=2fA'"'"' said means having complex coefficients h(l), where l=-(N-1)/2 to (N-1)/2 and the filter length N is odd, with alternating purely real and purely imaginary values, said processing and converting means including means for modulating the pulse response of a half-band filter h(l) with exclusively real values and h(l)=h(-l) for all |l|≦

(N-1)/2 and h(l)=0 for l=±

2, ±

4, . . . , onto a complex carrier of a frequency of ±

1/4 of the output sampling frequency fA=2fA'"'"' to yield

space="preserve" listing-type="equation">h(l)=h(l)·

e.sup.j(±

2π

lfA/4fA+φ

0) =j.sup.±

l ·

e.sup.jφ

0 ·

h(l),

with the null phase φ

0 of the complex carrier being an integer multiple m of π

/2 (φ

0=m·

π

/2 where m=0, 1, 2, 3, . . .); and

wherein said complex input signal includes a real component and an imaginary component, and said processing and converting means is divided into first and second branches, said first branch comprising a chain (N-1)/2 delay members each having a time delay of 2T, means for weighting said real and imaginary components of the complex input signal and means for forming a plurality of difference signals, the weighting means weights the real component s_r (2kT) with a value equal to a function of h(l) of the pulses response and said forming means feeds the weighted component to the first delay member of the chain of delay members and subtracts the weighted component from the output signal of the last delay member of the chain to produce a difference signal, said difference signal furnishing every second sample of the real filter output signal s(kT);

said weighting means weights additional momentary values of the real component s_r (2kT) of the filter input signal with a value equal to other functions of h(l) of the pulse response and said forming means adds these additional weighted values to a transversal signal of the chain of delay members at further points; and

said second branch has a further delay member with a time delay of T·

(N-3)/2 whose input receives the imaginary component s_i (2kT) of the complex input signal which has been weighted with a value equal to a function of h(0) and whose output yields every second time-shifted sample of the real filter output signal s(kT).

7. A non-reactive half-band filter according to claim 6, where m=0 and N=11, and the weighting means weights the additional momentary values of the real component s_r (2kT) of the filter input signal and the forming means adds the weighted values to the transversal signal to the points of the chain as follows:

at the input of the first delay member of the chain, the real component is weighted with a value equal to the function h(5);

at the input of the second delay member of the chain, the real component is weighted with a value equal to the function -h(3);

at the input of the third delay member of the chain, the real component to weighted with a value equal to the function h(1);

at the input of the fifth delay member of the chain, the real component is weighted with a value equal to the function h(3); and

at the output of the fifth delay member of the chain, the real component is weighted with a value equal to the function -h(5);

where h(0)=1/2.

8. A non-recursive half-band filter having a pulse response h(l), where l=-(N-1)/2 to (N-1)/2 and N is an odd filter length, said filter comprising means for receiving a real input signal s(kT) and means for converting the real input signal s(kT) into a complex output signal s(kT), where k is a running index, while maintaining a sampling frequency fA=1T, said converting means modulating the pulse response h(l) onto a complex carrier with a frequency of ±

1/4 of the sampling frequency fA=1/T, to yield

space="preserve" listing-type="equation">h(l)=h(l)·

e.sup.j(±

2π

lfA/4fA'"'"'φ

0) =j.sup.±

l ·

e.sup.jφ

0 ·

h(l),

with the null phase φ

0 of this frequency being an integer multiple m of π

/2 (φ

0=m·

π

/2 where m=1, 2, 3, . . .), and wherein said converting means includes a chain of (N-1)/2 delay members having a center delay member, means for forming a plurality of difference signals and means for weighting the formed difference signals, where each sample of the input signal s(kT) is routed into the chain of (N-1)/2 delay members each having a delay time of 2T, and the center delay member is divided into two members each having a delay time of T;

said difference signal forming means subtracting from the output signal of the last delay member of the chain the input signal of the first delay member of the chain to form a first difference signal, subtracting from the output signal of the penultimate delay member of the chain the input signal of the second delay member of the chain to form a second difference signal, subtracting from the output signal of the third to last delay member of the chain the input signal of the third delay member of the chain to form a third difference signal, and so on until the output signal of each delay member is processed;

said weighting means includes an adder and weighting each said difference signal by a value equal to a respective function of h(l) of the pulse response and summing the weighted difference signals via the adder to yield either the real or the imaginary component of the filter output signal s(kT);

said forming means forms a fourth difference signal from the center delay member of the chain where the input signal is delayed by a delay time T·

(N-1)/2 and is weighted with a value equal to the function j(0), which results in the other of the imaginary component and the real component of the filter output signal s(kT).

9. A non-reactive half-band filter according to claim 8, where N=11 and m=1, and the weighted means weights the first difference signal with a value signal equal to the function -h(5), the second difference signal with a value equal to the function h(3) and third difference signal with a value equal to the function -h(1);

and the sum of the difference signals yields the real component s_r (kT) of the filter output signal while the signal weighted with the function of h(0)=1/2 yields the imaginary component s_i (kT) of the filter output signal.

10. A non-recursive half-band filter according to claim 8, where N=11 and m=3, and the weighting means weights the first difference signal with a value equal to the function h(5), the second difference signal with a value equal to the function -h(3), the third difference signal with a value equal to the function h(1);

and the sum of the difference signals yields the real component s_r (kT) of the filter output signal while the signal weighted with the function of h(0)=1/2 yields the imaginary component s_i (kT) of the filter output signal.

11. A non-reactive half-band filter according to claim 8, wherein N=11 m=0, and the weighting means weights the first difference signal with a value equal to the function h(5), the second difference signal with a value equal to the function -h(3), and the third difference signal with a value equal to the function h(1);

and the sum of the difference signals yields the imaginary component s_i (kT) of the filter output signal while the signal weighted with the function of h(0)=1/2 yields the real component s_r (kT) of the filter output signal.

12. A non-recursive half-band filter according to claim 8, where N=11 and m=2, and the weighting means weights the first difference signal with a value equal to the function -h(5), the second difference signal with a value equal to the function h(3), the third difference signal with a value equal to the function -h(1);

and the sum of the difference signals yields the imaginary component s_i (kT) of the filter output signal while the signal weighted with the function h(0)=-1/2 yields the real component s_r (kT) of the filter output signal.

13. A non-recursive half-band filter having a filter length N and a pulse response h(l) where l=-(N-1)/2 to (N-1)/2 and the filter length N is odd, said half-band filter comprising:

means for receiving a complex input signal, and means for converting said complex input signal s(kT), where k is a running index, into a real output signal s(kT), while maintaining a sampling frequency fA=1/T, said converting means modulating the pulse response h(l) with reference to the sampling frequency fA, onto a complex carrier at a frequency of ±

fA/4 to yield

space="preserve" listing-type="equation">h(l)=h(l)·

e.sup.j(±

2π

lfA/4fA+φ

0) =j.sup.±

l ·

e.sup.jφ

0 ·

h(l)

with the null phase φ

0 of this frequency being an integer multiples m of π

/2 (φ

0=m·

π

/2 where m=0, 1, 2, 3, . . .), and wherein said complex input signal has a real component and an imaginary component; and

said converting means includes a chain of (N-1)/2 delay members each having a time delay of 2T and a center delay member divided into two members, with each divided member having a delay time of T, means for weighting the real and imaginary components with a value equal to a function of h(l), and means for forming a plurality of difference signals; and

the weighting means weights the imaginary component s_i (kT) with values equal to functions of h(l) of the pulse response and said forming means feeds the weighted component to the first delay member of the chain and subtracts the weighted component from the output signal of the last delay member of the chain to form a difference signal, said difference signal furnishing the real filter output signal s(kT);

said weighting means weights additional momentary values of the imaginary component s_i (kT) of the filter input signal with a value equal to other functions of h(l) of the pulse response and said forming means adds the additional weighted values to a transversal signal of the chain of delay members at further points; and

the weighting means further weights the real component s_r (kT) of the complex filter input signal with the function h(0), and additively feeds the weighted signal to the transversal signal of the chain at the center of said center delay member.

14. Non-recursive half-band filter according to claim 13, where m=0 and 2, respectively, and N=11, and the weighting means weights the additional momentary values of the real component s_r (kT) and of the imaginary component s_i (kT), respectively, of the filter input signal and the forming means adds the weighted values to the transversal signal to the points of the chain as follows:

at the input of the first delay member of the chain, with a value equal to the function ±

h(5);

at the input of the second delay member of the chain with a value equal to the function ±

h(3);

at the input of the third delay member of the chain with a value equal to the function ±

h(1);

at the input of the penultimate delay member of the chain with a value equal to the function ±

h(1);

at the input of the last delay member of the chain with a value equal to the function ±

h(3); and

at the output of the last delay member of the chain with a value equal to the function ±

h(5);

where h(0)=±

1/2.