Multi-channel sample rate conversion method

• US 20060212503A1
• Filed: 03/14/2006
• Published: 09/21/2006
• Est. Priority Date: 03/15/2005
• Status: Active Grant

• Current Assignee: Analog Devices, Inc.
• Original Assignee: Analog Devices, Inc.
• Inventors: Beckmann, Paul E., Stilson, Timothy S.
• Granted Patent: US 7,890,563 B2
• Time in Patent Office: Days
• US Class Current: 708/300
• CPC Class Codes: H03H 17/0628 the input and output signals being derived from two separate clocks, i.e. asynchronous sample rate conversion
• CPC Class Codes: H03H 2218/06 Multiple-input, multiple-output [MIMO]; Multiple-input, single-output [MISO]

Abstract

A sampling-rate conversion method receives N input channels which have been digitized at an input sampling rate, and converts the sampling rate of each input channel to produce N output channels at an output sampling rate. The method comprises computing a common FIR interpolating function which depends upon the input and output sample clocks and the instantaneous output time, and applying the common FIR interpolating function to all N input channels to compute all N output channels.

Claims

1. A method of converting the sampling rate of multiple input channels, comprising:

digitizing N input channels at a first sampling rate; and

converting the sampling rate of each of said N input channels to produce N output channels at a second output sampling rate;

said sampling rate conversion comprising;

computing a common FIR interpolating function which depends upon the input and output sample clocks and the instantaneous output time; and

applying said common FIR interpolating function to all N input channels to compute all N output channels.

2. The method of claim 1, wherein said sampling rate conversion is asynchronous.

3. The method of claim 1, wherein said sampling rate conversion is synchronous.

4. The method of claim 1, wherein said computing of a common FIR interpolating function comprises:

decomposing instantaneous output time (T_out) into integer and fractional portions where T_out=N+t_out, where N=floor(T_out) and t_out=T_out−

N;

providing a set of M precomputed FIR filter coefficient sets, each of said filter sets denoted as h_k[n] and arranged to calculate a fractionally-delayed output at time k/M, with k=0,1, . . . , M−

1;

choosing the x filter coefficient sets nearest T_out;

h_k[n], h_k+1[n], h_k+2[n] , . . . , h<sub>k+(x−

1)</sub>[n] where k=floor(t_out*M)−

2;

decomposing t_out into integer and fractional portions where t_out=T_M+tt,

$T_M=\frac{1}{M}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{floor}<br><br>\left(t_{\mathrm{out}}<br><br>M\right)=\frac{k}{M},$

and tt=t_out−

T_M;

scaling tt to obtain the quantity t_c which lies in the range 0≦

t_c≦

1, where t_c=M tt; and

using polynomial interpolation to compute the desired interpolating function h_net[i] from a linear weighted sum of the filter coefficient sets h_k[n], h_k+1[n], h_k+2[n] , . . . , h<sub>k+(x−

1)</sub>[n] in which the weighting coefficients depend solely on t_c;

h_net[i]=w_a(t_c)h_k[i]+w_b(t_c)h_k+1[i]+w_c(t_c)h_k+2[i]+ . . . +w_x(t_c)h<sub>k+(x−

1)</sub>[i], h_net[i] being said common FIR interpolating function.

5. The method of claim 4, wherein applying said FIR interpolating function to all N input channels to compute all N output channels comprises:

calculating h_net[i]; and

applying h_net[i] separately to each input channel to calculate one sample of each output channel.

6. The method of claim 4, wherein x=4 and said polynomial interpolation employs a cubic polynomial.

7. The method of claim 6, wherein x=4 and said polynomial interpolation employs a cubic polynomial, such that h_net[i]=w_a(t_c)h_k[i]w_b(t_c)h_k+1[i]+w_c(t_c)h_k+2[i]+w_d(t_c)h_k+3[i], where:

8. The method of claim 4, further comprising introducing a latency of 2/M samples such that k=floor(t_out*M).

9. The method of claim 1, wherein said N input channels are respective audio channels.

10. A method of converting the sampling rate of multiple input channels, comprising:

digitizing N input channels at a first sampling rate; and

converting the sampling rate of each of said N input channels to produce N output channels at a second output sampling rate;

said sampling rate conversion comprising;

computing a common FIR interpolating function which depends upon the input and output sample clocks and the instantaneous output time; and

applying said common FIR interpolating function to all N input channels to compute all N output channels, wherein said computing of a common FIR interpolating function comprises;

decomposing instantaneous output time (T_out) into integer and fractional portions where T_out=N+t_out, where N=floor(T_out) and t_out=T_out−

N;

providing a set of M precomputed FIR filter coefficient sets, each of said filter sets denoted as h_k[n] and arranged to calculate a fractionally-delayed output at time k/M, with k=0,1, . . . , M−

1;

choosing the four filter coefficient sets nearest T_out;

h_k[n], h_k+1[n], h_k+2[n] , . . . , h₊₃[n] where k=floor(t_out*M);

decomposing t_out into integer and fractional portions where t_out=T_M+tt,

$T_M=\frac{1}{M}<br><br>\mathrm{floor}<br><br>\left(t_{\mathrm{out}}<br><br>M\right)=\frac{k}{M},$

and tt=t_out−

T_M;

scaling tt to obtain the quantity t_c which lies in the range 0≦

t_c<

1, where t_c=M tt;

using polynomial interpolation to compute the desired interpolating function h_net[i] from a linear weighted sum of the filter coefficient sets h_k[n], h_k+1[n], h_k+2[n], . . . , h_k+3[n] in which the weighting coefficients depend solely on t_c;

h_net[i]=w_a(t_c)h_k[i]+w_b(t_c)h_k+1[i]+w_c(t_c)h_k+2[i]+w_d(t_c)h_k+3[i], where;

$w_a=-\frac{1}{6}<br><br>t_c^3+\frac{1}{2}<br><br>t_c^2-\frac{1}{3}<br><br>t_c$$w_b=\frac{1}{2}<br><br>t_c^3+-t_c^2-\frac{1}{2}<br><br>t_c+1$$w_c=-\frac{1}{2}<br><br>t_c^3+\frac{1}{2}<br><br>t_c^2+t_c$$w_d=\frac{1}{6}<br><br>t_c^3-\frac{1}{6}<br><br>t_c;<br><br>$

h_net[i] being said common FIR interpolating function; and

applying h_net[i] separately to each input channel to calculate one sample of each output channel.

11. The method of claim 10, wherein said N input channels are respective audio channels.