1. Field of the Invention
The present invention relates generally to digital filters.
2. Description of the Related Art
In the processing of digital signals, it is often advantageous to change from a first data rate suitable for a first network of digital circuits to a different second data rate that is suitable for a subsequent second network. If the second data rate is less than the first, the rate change is typically called decimation which comprises a filtering process configured to reduce aliasing followed by a downsampling process to effect the rate change. If the second data rate is greater than the first, the rate change is typically called interpolation which comprises an upsampling process to effect the rate change followed by a filtering process configured to reduce images.
In systems that process data at high rates, it is extremely important that the structure of the digital filter 24 is configured to reduce computational complexity. For this purpose, it has been found that a particularly advantageous version of the filter 24 is the combination of a comb filter 31 and a subsequent integrator filter 32 which are shown with their transfer functions in the filter system 30. These filters essentially perform a recursive running-sum process on the upsampled data stream out of the switch 23 wherein the comb filter 31 provides a moving sum and the integrator filter 32 provides an average of this sum.
The filter system 30 is significantly simplified by interchanging the comb filter and the switch as shown in the filter system 40. Because this interchange translates the comb filter to the low rate portion of the filter system, the differential delay R of the comb filter 31 in the system 30 reduces to 1 so that it is becomes a simple differentiator 41 in the system 40. That is, a 1-sample delay before upsampling by R is equivalent to an R-sample delay after the upsampling.
The comb filter 41 thus subtracts a delayed data stream element from the current input data stream element whereas the integrator filter 32 is an accumulator that adds the current input data stream element to the previous output data stream element. Accordingly, the comb filter has a feed-forward structure and the integrator filter has a feed-back structure.
The filter system 40 of
When these filters are required, however, to operate at ever-higher data rates it is generally found that integrated circuit implementations of CIC filters run into problems. For example, data timing problems become excessively problematical and filter dissipation rises because, at these higher rates, the filter adders must be realized with pipelined adder structures that substantially increase fabrication costs and system heating.
The present invention is directed to interleaved comb and integrator filter systems that enhance filter performance. The novel features of the invention are set forth with particularity in the appended claims. The invention will be best understood from the following description when read in conjunction with the accompanying drawings.
In particular, the system 60 serially couples an input filter section 61, a resampler 62, and an output filter section 63 between an input port 64 and an output port 65. An input data stream is received at the input port 61 and the system 60 processes this stream into an output data stream at the output port 65. As stated above, the input data stream has an input data rate and the output data stream has an output data rate that is greater than the input data rate. In particular, the output data rate is at a clock (clk) rate and the input data rate is at a rate of clk/R wherein R is an integer.
To effect this interpolation process, the filter system 60 has a CCI arrangement in which the input filter section 61 is formed with comb structures 67 (each denoted with a C) and the output filter section 63 is formed with integrator structures 68 (each denoted with an I). Data storage in the comb structures is significantly simplified by positioning them at the lower data rate portion of the system. In contrast, the integrator structures follow the resampler and, accordingly, process the higher data rates of the system.
As the demands of modern signal processing require higher and higher data rates, it has been found that filter rates approach or exceed those at which digital gates (e.g., complementary metal-oxide-semiconductor (CMOS) gates) can reliably operate. To resolve this problem, the filter system 20 configures the output filter section 63 so that it interleavably processes resampled data streams from the resampler over N integrator processing paths to thereby provide the output data stream. Because the output filter section processing is interleavably carried out over N processing paths, each path can operate at the reduced rate of clk/N.
Although the processing demand is not as severe in the input filter section 61 (because it processes the lower-rate input data stream), it may also be configured to interleavably process the input data streams over M integrator processing paths to thereby provide M intermediate data streams to the resampler 62. Because the input filter processing is interleavably carried out over M processing paths, each path can operate at the reduced rate of clk/RM.
To facilitate the interleaved processing of the input filter section 61, it has a buffer system 70 which provides successively-delayed input data streams at a data rate of clk/R. In addition, the resampler 62 is configured to convert the M intermediate data streams from the input filter section 61 into N successively-delayed resampled data streams which are provided to the output filter section 63. The intermediate data streams have an intermediate data rate of clk/RM and the resampled data streams have a resampled data rate of clk. The integrator structures operate at the rate clk/N. Dividers 69A, 69B and 69C are provided to divide the clk signal and provide the required clock signals clk/R, clk/RM and clk/N.
In the filter system 20, therefore, processing rate in the comb structures 67 is reduced by a factor M and processing rate in the integrator structures 68 is reduced by a factor N. In the exemplary system embodiment of
The buffer system 70 of
A first summer 83 is arranged to difference elements of the data stream at the port 81 with elements of the delayed data stream at the port 82 to thereby provide an output element y(n) at an output port 84 wherein y(n)=x(n)−x(n−1). The first summer 83 thus subtracts a delayed data stream element from the current input data stream element which is the comb process performed by the comb filter 40 of
In order to interleavably process elements of the delayed data stream at the port 82, further delayed elements (e.g., x(n−2)) must be available. Accordingly, the comb structure 80 includes a delay register 85 which is coupled to the port 81 to provide data stream elements x(n−2) that are differenced with delayed data stream elements x(n−1) in a second summer 86. The difference provides an output element y(n−1) at an output port 87 wherein y(n−1)=x(n−1)−x(n−2). Thus, elements x(n) and x(n−1) of a data stream at the input port 81 and a delayed data stream at the input port 82 are interleavably processed into elements y(n) and y(n−1) of intermediate data streams at the output ports 84 and 87.
Referring to
Although the comb structure 80 of
The comb structure 90 includes elements of the comb structure 80 with like elements indicated by like reference numbers. In addition, the comb structure 90 adds a third input port 91 and a third output port 92 and moves the output of the delay register 85 down to a third summer 93 which is coupled between the added ports 91 and 92. Thus, the second delay register 85 continues to receive data stream elements x(n) from the input port 81 but, because M is now 3, the delay register 85 provides delayed data stream elements x(n−3).
Therefore, the first summer 83 differences an element x(n) of the data stream at the port 81 with an element x(n−1) of the delayed data stream at the port 82 to provide an output element y(n)=x(n)−x(n−1) at the output port 84 and summer 86 differences an element x(n−1) of the delayed data stream at the port 82 with an element x(n−2) of the further-delayed data stream at the port 91 to provide an output element y(n−1)=x(n−1)−x(n−2) at the output port 87. Finally, the added summer 93 differences the element x(n−2) at the input port 91 with the delayed data stream element x(n−3) from the delay register 85 to provide an output element y(n−2)=x(n−2)−x(n−3) at the added output port 92.
From
Attention is now directed to
The delay register 102 is coupled to the Nth one of the output ports 108 to provide a delayed version of the filtered data stream at that output port. In the system 60 of
The N−1 summers 104 are arranged to provide a sum 109 of the elements of the resampled data streams at the input ports 107 and the summer 103 sums the delayed filtered data stream with elements of the sum 109 to thereby provide an Nth one of the filtered data streams. As shown in
A first one of the N−1 summers 105 at the output port 108 is arranged to difference the element y(n) with the element x(n) at the first one of the input ports 107 to thereby provide elements y(n−1)=y(n)−x(n) of another of the output ports 108. It is apparent from the integrator filter 32 of
From the series (1), it is also apparent that a previous output data stream element y(n−1) is given by subtracting a current input data stream element x(n) from a subsequent output data stream element y(n). Accordingly, the output of the first one of the summers 105 does indeed provide filtered data stream elements y(n−1).
In a similar manner, remaining ones of the summers 105 are arranged to difference the element y(n) with elements at respective ones of the input ports 107 to thereby provide N−2 additional ones of the filtered data streams. In particular, one of the summers 105 differences element y(n) with the sum x(n)+x(n−1) at one of the input summers 104 to thereby provide y(n)−x(n)−x(n−1) which becomes element y(n−2)=y(n−1)−x(n−1) when it is noted that y(n−1)=y(n)−x(n). This equation for y(n−2) is also consistent with the series (1).
Continuing in this manner, the other of the summers 105 differences element y(n) with the sum x(n)+x(n−1)+x(n−2) at one of the input summers 104 to thereby provide y(n−3)=y(n−2)−x(n−2) at another of the output ports 108.
From
More particularly,
An embodiment 120 of the resampler 62 of
In one embodiment, the reclocker 123 is simply a register that receives the data stream 126 at a data rate of clk/R and is clocked at a rate of clk so that each incoming data element is clocked out R times. Accordingly, the reclocker generates a resampled data stream 127 that has a data rate of clk.
The buffer system 124 has a structure similar to the system 70 of
The filter system 60 of
The system embodiments of the invention, however, also include interleaved structures which are arranged to perform a decimation process in which the output data rate is less than the input data rate by a factor R. In these embodiments, the comb structures 67 (each denoted with a C) of the input filter section 61 in
Interleaved filter embodiments have been described which can generate output data streams with rates that exceed those of which the system components are capable. These embodiments also enhance system performance, e.g., they reduce data timing problems and, in addition, they reduce filter dissipation because they reduce the need for pipelined adder structures.
The embodiments of the invention described herein are exemplary and numerous modifications, variations and rearrangements can be readily envisioned to achieve substantially equivalent results, all of which are intended to be embraced within the spirit and scope of the invention as defined in the appended claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 4999798 | McCaslin et al. | Mar 1991 | A |
| 5517529 | Stehlik | May 1996 | A |
| 6501406 | Mecchia | Dec 2002 | B1 |
| 6559781 | Freidhof | May 2003 | B2 |
| 6643675 | Piirainen | Nov 2003 | B2 |
| 6707255 | Coumou et al. | Mar 2004 | B2 |
| 6788233 | Quiquempoix | Sep 2004 | B1 |
| 6831900 | Blake | Dec 2004 | B2 |
| 6871207 | Nanda et al. | Mar 2005 | B1 |
| 7170959 | Abbey | Jan 2007 | B1 |
| 7196648 | Ding et al. | Mar 2007 | B1 |
| 20050160124 | Bisiaux | Jul 2005 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20070230642 A1 | Oct 2007 | US |