Method for digital interpolation

• US 5,717,618 A
• Filed: 07/31/1995
• Issued: 02/10/1998
• Est. Priority Date: 08/08/1994
• Status: Expired due to Term

• Current Assignee: Deutsche ITT Industries GmbH (TDK Corporation)
• Original Assignee: Deutsche ITT Industries GmbH (TDK Corporation)
• Inventors: Menkhoff, Andreas, Temerinac, Miodrag, Witte, Franz-Otto, Winterer, Martin
• Primary Examiner(s): Elmore, Reba I.
• Assistant Examiner(s): MOISE, EMMANUEL LIONEL
• Application Number:
  US08/509,517
• Time in Patent Office: 925 Days
• Field of Search: 364/724.1, 364/724.01, 364/724.16, 364/724.17, 364/723, 324/605, 324/607, 324/613-617, 341/50, 341/61, 341/122, 341/123, 375/225, 375/350
• US Class Current: 708/313
• CPC Class Codes: H03H 17/028 Polynomial filters
• CPC Class Codes: H03H 17/0642 the ratio being arbitrary or irrational

Abstract

An improved method for digital interpolation of signals for a second interpolation filter is disclosed which permits a high signal/noise ratio with a minimum amount of circuitry for an overall system comprising first and second interpolation filters. The method for digital interpolation of signals requires multiplying delayed input values locked to a first signal by corresponding weighting factors which are dependent on a time-difference value determined by the interpolating instant and the time grid of the first clock signal. The weighting factors are determined by an impulse response in the time domain. The associated transfer function has an attenuation characteristic in the frequency domain which, with respect to the stop bands, is limited essentially to the alias regions located at the frequency multiples of the first clock signal. Each of these alias regions is assigned at least two adjacent zeros, or in the presence of double-order zeros, at least one of the alias regions and the associated periodic alias regions are assigned at least one further zero of the transfer function.

Claims

1. A method for digital interpolation of signals comprising the steps of:

receiving a digital signal input value sequence x(k) corresponding to an analog signal sampled at a predetermined frequency;

locking delayed input values of said digital sequence x(k) to a first clock signal, said first clock signal having an interpolating instant and a time grid;

determining weighting factors by means of an impulse response in a time domain with a transfer function in a frequency domain having an attenuation characteristic which, with respect to stop bands, is limited to alias regions located at frequency multiples of said first clock signal;

multiplying said locked delayed input values by said corresponding weighting factors;

determining a time-difference value as a function of said interpolating instant and said time grid;

adding said multiplied values wherein said values depend on said time-difference value to produce a second digital signal sequence y(d) at a second sampling rate;

wherein each of said alias regions is assigned at least two zeros of said transfer function in the frequency domain, said at least two zeros in each alias region lying side by side of if two zeros lie on top of one another as a second-order zero, at least one of said alias regions assigned at least one further zero of said transfer function;

wherein said impulse response is dependent on said time difference value.

2. The method as recited in claim 1 further comprising the step of determining a new impulse response whose first further zero of said transfer function in the frequency domain is located at a predetermined frequency, said new impulse response at a time t determined by adding an original impulse response at said time t to an original impulse response at a time t minus half said predetermined frequency.

3. The method as recited in claim 2 wherein additional zeros of said transfer function are arbitrarily positioned by repeating the step of determining a new impulse response.

4. The method as recited in claim 1 wherein said each of said alias regions is assigned at least one further zero of said transfer function in the frequency domain such that all alias regions have at least three zeros.

5. The method as recited in claim 4 wherein at least three zeros are located at centers of respective alias regions at frequency multiples of said first clock signal.

6. The method as recited in claim 5 wherein said impulse response, of a sample-and-hold function with centrally located zeros which are at least of the third order, is defined by:

##EQU5## wherein n is an order of the central zeros of the associated transfer function in each alias region;

k is a running numerical value for addition, which assumes values zero to n;

σ

(t) is a unit step at instant t; and

σ

(t-kT1) is a unit step at instant t-kT1.

7. A method for digital interpolation of signals comprising the steps of:

receiving a digital signal input value sequence x(k) corresponding to an analog signal sampled at a predetermined frequency;

multiplying delayed input values of said digital sequence (x(k)) locked to a first clock signal (f1) by corresponding weighting factors (Ci;

C_l) and

adding up the multiplied values, which are depending on a time-difference value (Δ

t;

dt) which is determined by an interpolating instant (tP) and a time grid of the first clock signal (f1), to produce a second digital signal sequence y(d) at a second sampling rate;

wherein the weighting factors (Ci;

C_l) are determined by means of an impulse response h(t) in the time domain, with the transfer function H(f) in the frequency domain, having an attenuation characteristic which, with respect to the stop bands, is limited essentially to the alias regions (1S, 2S, . . . pS, . . . ) located at the frequency multiples of the first clock signal (f1);

wherein each of said alias regions is assigned at least two zeros of the transfer function H(f) in the frequency domain, the at least two zeros in each alias region lying side by side or at least one of the alias regions (pS) being assigned at least one further zero of the transfer function H(f) if the two zeros lie on top of one another as a second-order zero (K=2);

wherein the impulse response for calculating the weighting factors is defined in sections with respect to the time-difference value.

8. The method as recited in claim 7, wherein a new impulse response h(t) whose first further zero of the transfer function H(f) in the frequency domain is to be located at a predetermined frequency fn is formed from an original impulse response h_old (t) as follows:

space="preserve" listing-type="equation">h(t)=h.sub.new (t)=h.sub.old (t)+h.sub.old (t-ta) with ta=1/2fn.

9. The method as recited in claim 7, wherein additional further zeros of the transfer function H(f) can be arbitrarily positioned by repeated use of a new impulse response h(t) whose first further zero of the transfer function H(f) in the frequency domain is to be located at a predetermined frequency fn is formed from an original impulse response h_old (t) as follows:

space="preserve" listing-type="equation">h(t)=h.sub.new (t)=h.sub.old (t)+h.sub.old (t-ta) with ta=1/2fn.

10.

10. The method as recited in claim 7, wherein each alias region (1S, 2S, . . . pS, . . . ) is assigned at least one further zero of the transfer function H(f) in the frequency domain, so that all alias regions have at least three zeros.

11. The method as recited in claim 10, wherein at least three zeros are located at the centers of the respective alias regions (1S, . . . pS, . . . ) at the frequency multiples (1f1, 2f1, . . . pf1, . . . ) of the first clock signal (f1).

12. The method claimed in claim 11, wherein the impulse response h(t) of a sample-and-hold function hn(t) or h(K=n) with centrally located zeros which are at least of the order K=n=3 is defined by:

##EQU6## wherein n is the order of the central zeros of the associated transfer function Hn(f) in each alias region 1S, 2S, . . . pS, . . . ;

k is the running numerical value for the addition, which assumes the values zero to n;

σ

(t) is the unit step at instant t; and

σ

(t-kT1) is the unit step at instant t-kT1.

13. The method as recited in claim 12, wherein a new impulse response h(t) whose first further zero of the transfer function H(f) in the frequency domain is to be located at a predetermined frequency fn is formed from an original impulse response h_old (t) as follows:

space="preserve" listing-type="equation">h(t)=h.sub.new (t)=h.sub.old (t)+h.sub.old (t-ta) with ta=1/2fn.

14. The method as recited in claim 12, wherein additional further zeros of the transfer function H(f) can be arbitrarily positioned by repeated use of a new impulse response h(t) whose first further zero of the transfer function H(f) in the frequency domain is to be located at a predetermined frequency fn is formed from an original impulse response h_old (t) as follows:

space="preserve" listing-type="equation">h(t)=h.sub.new (t)=h.sub.old (t)+h.sub.old (t-ta) with ta=1/2fn.

15. The method as recited in claim 7, wherein a new transfer function H_new (f) is formed from an original transfer function H_old (f) by a shift in the frequency domain by the frequency values +Δ

f and -Δ

af according to the following formula;

space="preserve" listing-type="equation">H.sub.new (f)=H.sub.old (f-Δ

f)·

H.sub.old (f+Δ

f),

whereby the respective stop band is widened as compared with a central zero position.

16. The method as recited in claim 15, wherein additional zeros of the transfer function are generated by repeating the step of forming a new transfer function from an original transfer function by a shift in the frequency domain.

17. The method as recited in claim 7, wherein the weighting factors are formed by means of a matrix stage.

18. The method as recited in claim 7, wherein the impulse response is implemented with an FIR filter structure, the weighting factors being formed by means of a matrix stage.

19. The method as recited in claim 19, wherein the weighting factors are formed by means of a switchable matrix stage.