US20060277240A1 - Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices - Google Patents
Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices Download PDFInfo
- Publication number
- US20060277240A1 US20060277240A1 US11/504,849 US50484906A US2006277240A1 US 20060277240 A1 US20060277240 A1 US 20060277240A1 US 50484906 A US50484906 A US 50484906A US 2006277240 A1 US2006277240 A1 US 2006277240A1
- Authority
- US
- United States
- Prior art keywords
- values
- look
- law
- value
- calculated
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/02—Digital function generators
- G06F1/03—Digital function generators working, at least partly, by table look-up
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03K—PULSE TECHNIQUE
- H03K19/00—Logic circuits, i.e. having at least two inputs acting on one output; Inverting circuits
- H03K19/02—Logic circuits, i.e. having at least two inputs acting on one output; Inverting circuits using specified components
- H03K19/173—Logic circuits, i.e. having at least two inputs acting on one output; Inverting circuits using specified components using elementary logic circuits as components
- H03K19/1733—Controllable logic circuits
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2101/00—Indexing scheme relating to the type of digital function generated
- G06F2101/12—Reciprocal functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2101/00—Indexing scheme relating to the type of digital function generated
- G06F2101/16—PCM companding functions
Definitions
- the present invention relates generally to programmable logic devices (PLDs). More specifically, the present invention relates to techniques for efficiently implementing complex circuits in PLDs.
- FIR filters In order to cancel echoes that occur as a result of hybrid coupling within a telephone network, adaptive Finite Impulse Response (FIR) filters have been used.
- FIR filters typically include complicated arithmetic circuits for performing reciprocal calculations, and complicated ⁇ -Law/A-Law expander circuits for expanding ⁇ -Law/A-Law encoded signals.
- PLDs are the devices of choice in implementing these complicated digital signal processors.
- the basic building blocks are the Logic Elements (LEs) or Logic Array Blocks (LABs) that include logic circuits that may be programmed to perform specific logic operations.
- Ls Logic Elements
- LABs Logic Array Blocks
- Each PLD has a limited number of LEs. Therefore, after these arithmetic circuits are implemented, few LEs are left for performing other functions. As a result, it is often difficult to implement an entire digital signal processing system with complicated arithmetic circuits on a single PLD.
- LUTs Look-Up Tables
- LMS Least-Mean-Squared
- a circuit for calculating reciprocal values is implemented with LUT(s) having stored therein pre-calculated reciprocal values.
- the pre-calculated reciprocal values are determined based on the address locations at which they are stored.
- an input value to the circuit is partitioned into a number of segments. Then, one of the segment is selected for indexing the LUT(s) and for retrieving an associated pre-calculated reciprocal value therefrom.
- the retrieved pre-calculated reciprocal value is then shifted, according to a position of the selected segment within the input value, to obtain an output value.
- the selected segment is the first segment, starting from the decimal point, that contains at least one digital “high” value (e.g., a logic “1”).
- the input value is a fraction
- the output of the circuit which is an approximate reciprocal of the input value, is an integer.
- a circuit implementing a normalized LMS algorithm has LUTs storing pre-calculated ⁇ -Law or A-law expansion values.
- An input to the circuit which is either ⁇ -Law or A-Law encoded, is provided to the LUT as an index for retrieving an associated expansion value.
- the retrieved expansion value is then provided to circuits for performing other steps of the normalized LMS algorithm. Accordingly, by using LUTs to store pre-calculated ⁇ -Law or A-Law expansion values, such complicated arithmetic circuits can be implemented with a significantly smaller number of LEs.
- FIG. 1 is a block diagram illustrating a programmable logic device on which embodiments of the present invention may be practiced.
- FIG. 2A is a block diagram illustrating a circuit for calculating reciprocal values in accordance with one embodiment of the present invention.
- FIG. 2B is a block diagram illustrating a circuit for calculating reciprocal values in accordance with another embodiment of the present invention.
- FIG. 3 is a flow chart diagram illustrating a method of calculating reciprocal values using an LUT-based reciprocal calculation circuit in accordance with an embodiment of the present invention.
- FIG. 4A is a block diagram illustrating a Look-Up Table (LUT) for implementing a ⁇ -Law expander in accordance with one embodiment of the present invention.
- LUT Look-Up Table
- FIG. 4B is a block diagram illustrating a Look-Up Table (LUT) for implementing an A-law expander in accordance with one embodiment of the present invention.
- LUT Look-Up Table
- FIG. 5 is a block diagram illustrating a Look-Up Table (LUT) in accordance with an embodiment of the present invention.
- FIG. 1 is a block diagram illustrating a programmable logic device 100 (e.g., Altera FLEXTM 10K device).
- the device 100 includes logic arrays blocks 120 , embedded array blocks 140 , an interconnect grid 160 , and input/output elements 180 .
- Interconnect grid 160 includes sets of horizontal conductors 162 and sets of vertical conductors 164 that are configured in a grid pattern.
- Logic array blocks 120 and embedded array blocks 140 are programmably connectable to horizontal conductors 162 and vertical conductors 164 of interconnect grid 160 .
- Interconnect grid 160 is also programmably connectable to input/output elements 180 .
- PLD 100 may be programmed to implement complex arithmetic circuits of the present invention.
- LUTs Look-Up Tables
- values of the function fn(x) over a range of input values (x) are first determined.
- the pre-determined or pre-calculated values are then stored within a LUT of a programmable logic device.
- the pre-determined values are stored at addresses corresponding to the input values (x). For example, the pre-determined value of fn( 1 ) is stored at address 1 , and the predetermined value of fa( 255 ) is stored at address 255 .
- the LUT can then be used for calculating the value of fn(x) for an input value (x). For example, when an input value k is presented to the look-up table, the value stored at address k, which is fn(k), will then be provided as an output.
- a block diagram of a LUT 500 and its contents according to this embodiment of the present invention is illustrated in FIG. 5 .
- the circuit for matching the input value to the appropriate address of the LUT 500 is well known in the art, and is not illustrated for purposes of clarity.
- the present invention is particularly applicable to the implementation of complicated arithmetic circuits for adaptive digital voice filtering in programmable logic devices.
- exemplary implementations of circuits for carrying out normalized LMS (Least-Mean-Squared) algorithms are presented. It should be appreciated that, however, the present invention is not limited to such circuit implementations.
- P n .(a n ⁇ 1 a n ⁇ 2 a n ⁇ 3 . . . a 1 a 0 ) is an n-bit binary unsigned fraction.
- the reciprocal value of 2 ⁇ L (a n ⁇ (L+1) *2 ⁇ 1 +a n ⁇ (L+2) *2 ⁇ 2 +a n ⁇ (L+3) *2 ⁇ 3 + . . . +a n ⁇ 2L *2 ⁇ L ) can be determined by retrieving the pre-calculated reciprocal value from the look-up table using the coefficients (a n ⁇ (L+1) a n ⁇ (L+2) a n ⁇ (L+3) . . . a n ⁇ 2L ) as the address, and by multiplying the retrieved value by 2 L .
- the reciprocal value of 2 ⁇ mL (a n ⁇ (mL+1) *2 ⁇ 1 +a n ⁇ (mL+2) *2 ⁇ 2 +a n ⁇ (mL+3) *2 ⁇ 3 + . . . +a 0 *2 ⁇ L ) can be determined by retrieving the pre-calculated reciprocal value from the look-up table using the coefficients (a n ⁇ (mL+1) a n ⁇ (mL+2) a n ⁇ (mL+3) . . . a 0 ) as the address, and by multiplying the retrieved value by 2 mL .
- its reciprocal value can be determined by looking up Table 1 to retrieve the value stored at address 1011, which is 16/11, or approximately 0.23.
- its reciprocal value can be calculated by looking up Table 1 to retrieve the value stored at address 1001, which is 16/9, and multiplying 2 4 to the retrieved value.
- the reciprocal value of 0.00001001 (bin) can be easily determined to be 2 4 *16/9, or approximately 28.44.
- Reciprocal calculation can be sped up even further by storing the reciprocal values in the look-up table as binary integer values. Multiplication of binary integer values by 2 n is equivalent to shifting the values to the left by n bits. Naturally, accuracy of the reciprocal calculation may be somewhat compromised. Nevertheless, in some applications where speed or simplicity is desired, such a trade-off may be preferred.
- the least significant segment may include fewer than L-bits (e.g., n ⁇ 1 mod L bits). In that case, the look-up table with fewer entries will be used.
- the present invention may also be applied to signed values. In that embodiment, the magnitude of a signed value is first determined. Then, the reciprocal of the magnitude is determined using techniques discussed above. Thereafter, if the original number is negative, the reciprocal value is two-complemented.
- FIG. 2A is a block diagram illustrating a circuit 200 for calculating reciprocal values in accordance with one embodiment of the present invention.
- Circuit 200 may be implemented using a programmable logic device (e.g., device 100 ).
- circuit 200 includes 2's complement circuits 210 a - 210 b , a plurality of identical look up tables 220 a - 220 n , a plurality of shifters 230 a - 230 n , and an output selector circuit 240 .
- the 2's complement circuit 210 a determines whether the number is negative and determines the magnitude
- is then partitioned into m segments each being L-bit in width.
- 1225 containing a plurality of segments 235 a - 235 n , is also illustrated in FIG. 2A .
- is the sign bit 234 , which is not taken into account when the number is partitioned. Note that the least significant segment 235 n may contain only (n ⁇ 1 mod L) bits.
- each LUT 220 a - 220 n has stored therein a plurality of pre-determined reciprocal values.
- the contents of the LUTs 220 a - 220 n may be identical.
- the reciprocal values are pre-calculated based on the addresses of the LUT at which the reciprocal values are stored. For example, address 1011 stores the reciprocal value of 0.1011, and address 1110 stores the reciprocal value of 0.1110, etc.
- the LUT 220 a retrieves the reciprocal value stored at the address corresponding to the segment 235 a and provides the retrieved reicprocal value to the shifter 230 a .
- the reciprocal value stored at the address corresponding to the segment 235 b will be retrieved and provided to the shifter 230 b .
- address 0000 does not store the reciprocal value of 0.0000. Rather, address 0000 of the look-up table stores a value of zero, or a value that indicates a potential error.
- Each of the shifters 230 a - 230 n shifts the retrieved reciprocal values according to the position of the segment within
- the shifter 230 b is configured to shift the reciprocal value L bits to the left
- the shifter 230 n is configured to shift the reciprocal value (n ⁇ 1)*L bits to the left.
- the output selector circuit 240 select a first non-zero output from the shifters 230 a - 230 n and ignores the rest of their outputs.
- the shifter 230 a corresponding to the most significant bits of the input value, has the highest priority
- the shifter 230 b corresponding to the next most significant bits of the input value, has the second highest priority, etc. For example, if the output of shifter 230 a is non-zero, then the outputs of the shifters 230 b - 230 n are ignored.
- the selective adder 240 selects the output of the shifter 230 b .
- the selected output is then provided to 2's complement circuit 210 b to be converted into 2's complement form if the original number is negative.
- FIG. 2B illustrates a circuit 205 for calculating reciprocal values in accordance with another embodiment of the present invention.
- circuit 205 is suitable for implementation by a PLD.
- circuit 205 includes 2's complement circuits 210 a - 210 b , an input segment selector circuit 250 , a look-up table 260 , and a shifter 270 .
- the LUT 260 has stored therein a plurality of pre-determined reciprocal values that are pre-calculated based on the addresses of the LUT at which the reciprocal values are stored. For example, address 1011 stores the reciprocal value of 0.1011, and address 1110 stores the reciprocal value of 0.1110, etc.
- the 2's complement circuit 210 a determines whether the number is negative and determines the magnitude
- is then partitioned into m segments each being L-bit in width.
- 225 containing a plurality of segments 235 a - 235 n , is also illustrated in FIG. 2A .
- is the sign bit 234 , which is not taken into account when the number is partitioned. Note that the least significant segment 235 n may contain only (n ⁇ 1 mod L) bits.
- the segments 235 a - 235 n are then provided to the input segment selector circuit 250 .
- the input segment selector circuit 250 selects the first non-zero segment of the magnitude
- the LUT 260 retrieves the reciprocal value stored at the address corresponding to the first non-zero segment and provides the retrieved value to the shifter circuit 270 .
- the shifter 270 then shifts the retrieved reciprocal value according to a shift control signal provided by the input segment selector circuit 250 .
- the output of the shifter 270 is then provided to 2's complement circuit 210 b to be converted into 2's complement form if the original number is negative.
- FIG. 3 is a flow diagram illustrating a method of determining reciprocal values for an input data using a LUT-based reciprocal calculation circuit (e.g., circuit 205 ) in accordance with one embodiment of the present invention.
- a LUT-based reciprocal calculation circuit e.g., circuit 205
- reciprocal values are stored in a look-up table.
- the reciprocal values may be stored at addresses of the look-up table that correspond to the reciprocal values. For instance, the reciprocal value of 0.0010 maybe stored at address 0010 of the LUT. It should be noted that, in that embodiment, an ERROR value or a zero value may be stored at address 0000 of the LUT.
- an input data is received. Then, at step 330 , the input data is partitioned into a number of segments each having a bit-width corresponding to the bit-width of the addresses of the look-up table. For instance, if the look-up table has 8-bit wide addresses and the input data is 32-bit wide, the input data is partitioned into four 8-bit wide segments.
- one of the segments is selected and provided to the look-up table to determine if there is a match and to retrieve a reciprocal value whose address matches the selected segment.
- the first segment containing a non-zero value may be selected. Note that other selection criteria may also be used.
- the retrieved reciprocal value is shifted according to a position of the selected segment in relation to the input data to generate an approximate reciprocal value for the input data. For instance, if the second 8-bit segment of a 32-bit input fraction is selected, then the retrieved reciprocal value is shifted eight bits to the left. If the third 8-bit segment of the 32-bit input fraction is selected, then the retrieved reciprocal value is shifted sixteen bits to the left. The result of the shifting will be an approximate value of the reciprocal of the 32-bit input fraction.
- the normalized LMS algorithm is a well-known algorithm in the field of telecommunications. Specifically, the normalized LMS algorithm can be used for canceling echoes that occur as a result of hybrid coupling within a telephone network.
- the linear input data x(j) can be represented by the expression: ⁇ -Law(Rin(j)), where Rin(j) is the ⁇ -Law 8-bit compressed value of the input data x(j).
- the linear input data x(j) can be represented by the expression: A-Law(Rin(j)), where Rin(j) is the A-Law 8-bit compressed value of the input data x(j).
- the first step in the normalized LMS algorithm is to square the far end linear input data x(j) for calculation of the input signal power P(j).
- the pre-calculated values are stored in the look-up table using the values (Rin(j)) as storage addresses.
- pre-calculated value for [ ⁇ -Law( 255 )*p-Law( 255 )] may be stored at address 255 of the look-up table.
- the input value Rin(j) is then used as an index to look-up the corresponding [ ⁇ -Law(Rin(j))* ⁇ -Law(Rin(j))] or [A-Law(Rin(j))*A-Law(Rin(j))] value.
- a significant number of logic elements can be saved at the expense of only one embedded array block, which is used for implementing the look-up table.
- FIG. 4A is a block diagram illustrating a look-up table 410 for implementing a ⁇ -Law expander in accordance with one embodiment of the present invention.
- the look-up table 410 includes pre-calculated values of [ ⁇ -Law(Rin(j))* ⁇ -Law(Rin(j))] for all possible input values Rin(j). It should be noted that Rin(j) is an 8-bit value.
- Rin(j) is an 8-bit value.
- 256 pre-calculated ⁇ -Law expansion values are stored within the look-up table 410 .
- FIG. 4B is a block diagram illustrating another look-up table 420 for implementing an A-law expander in accordance with one embodiment of the present invention.
- the look-up table 420 includes pre-calculated values of [A-Law(Rin(j))*A-Law(Rin(j))] for all possible input values Rin(j). Similar to table 410 , the table 420 includes 256 pre-calculated A-Law expansion values.
Abstract
Efficient implementation of arithmetic circuits in programmable logic devices by using Look-Up Tables (LUTs) to store pre-calculated values. A table look-up operation is performed in place of complex arithmetic operations. In this way, at the expense of a few LUTs, many logic elements can be saved. This approach is particularly applicable to circuits for calculating reciprocal values and circuits for performing normalized LMS algorithm.
Description
- This application claims priority to the provisional patent application entitled, “Apparatus and Method for Implementing Complex Arithmetic Circuits in Programmable Logic Devices,” Ser. No. 60/236,244, filed Sep. 28, 2000.
- The present invention relates generally to programmable logic devices (PLDs). More specifically, the present invention relates to techniques for efficiently implementing complex circuits in PLDs.
- Many applications in digital signal processing require the use of complicated circuits for performing complex arithmetic calculations in real time. For example, in order to cancel echoes that occur as a result of hybrid coupling within a telephone network, adaptive Finite Impulse Response (FIR) filters have been used. Such FIR filters typically include complicated arithmetic circuits for performing reciprocal calculations, and complicated μ-Law/A-Law expander circuits for expanding μ-Law/A-Law encoded signals.
- PLDs are the devices of choice in implementing these complicated digital signal processors. For many PLDs, the basic building blocks are the Logic Elements (LEs) or Logic Array Blocks (LABs) that include logic circuits that may be programmed to perform specific logic operations. Due to their complexity, digital signal processors typically require the use of many LEs. Each PLD has a limited number of LEs. Therefore, after these arithmetic circuits are implemented, few LEs are left for performing other functions. As a result, it is often difficult to implement an entire digital signal processing system with complicated arithmetic circuits on a single PLD.
- Accordingly, it would be desirable to provide improved techniques for efficiently implementing complex arithmetic circuits in programmable logic devices.
- Techniques for efficient implementation of complex arithmetic circuits in programmable logic devices are disclosed. In one aspect of the present invention, Look-Up Tables (LUTs) of a programmable logic device are used to store pre-calculated intermediate or final calculation values. A table look-up operation is performed in place of complex arithmetic operations. In this way, at the expense of a few LUTs, many logic elements can be saved. This approach is particularly applicable to circuits for calculating reciprocal values and circuits for performing a normalized Least-Mean-Squared (LMS) algorithm.
- According to one aspect of the invention, a circuit for calculating reciprocal values is implemented with LUT(s) having stored therein pre-calculated reciprocal values. The pre-calculated reciprocal values are determined based on the address locations at which they are stored. In operation, an input value to the circuit is partitioned into a number of segments. Then, one of the segment is selected for indexing the LUT(s) and for retrieving an associated pre-calculated reciprocal value therefrom. The retrieved pre-calculated reciprocal value is then shifted, according to a position of the selected segment within the input value, to obtain an output value. In one embodiment of the invention, the selected segment is the first segment, starting from the decimal point, that contains at least one digital “high” value (e.g., a logic “1”). Further, according to the present invention, the input value is a fraction, and the output of the circuit, which is an approximate reciprocal of the input value, is an integer.
- In another aspect of the present invention, a circuit implementing a normalized LMS algorithm has LUTs storing pre-calculated μ-Law or A-law expansion values. An input to the circuit, which is either μ-Law or A-Law encoded, is provided to the LUT as an index for retrieving an associated expansion value. The retrieved expansion value is then provided to circuits for performing other steps of the normalized LMS algorithm. Accordingly, by using LUTs to store pre-calculated μ-Law or A-Law expansion values, such complicated arithmetic circuits can be implemented with a significantly smaller number of LEs.
- For a better understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings, in which:
-
FIG. 1 is a block diagram illustrating a programmable logic device on which embodiments of the present invention may be practiced. -
FIG. 2A is a block diagram illustrating a circuit for calculating reciprocal values in accordance with one embodiment of the present invention. -
FIG. 2B is a block diagram illustrating a circuit for calculating reciprocal values in accordance with another embodiment of the present invention. -
FIG. 3 is a flow chart diagram illustrating a method of calculating reciprocal values using an LUT-based reciprocal calculation circuit in accordance with an embodiment of the present invention. -
FIG. 4A is a block diagram illustrating a Look-Up Table (LUT) for implementing a μ-Law expander in accordance with one embodiment of the present invention. -
FIG. 4B is a block diagram illustrating a Look-Up Table (LUT) for implementing an A-law expander in accordance with one embodiment of the present invention. -
FIG. 5 is a block diagram illustrating a Look-Up Table (LUT) in accordance with an embodiment of the present invention. - Like reference numerals refer to corresponding parts throughout the drawings.
- In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known structures, devices, methodologies, etc., are not described in detail in order to avoid obscuring aspects of the present invention.
- A. General Description of a Programmable Logic Device
-
FIG. 1 is a block diagram illustrating a programmable logic device 100 (e.g., Altera FLEX™ 10K device). The device 100 includeslogic arrays blocks 120, embeddedarray blocks 140, aninterconnect grid 160, and input/output elements 180. Interconnectgrid 160 includes sets ofhorizontal conductors 162 and sets ofvertical conductors 164 that are configured in a grid pattern.Logic array blocks 120 and embeddedarray blocks 140 are programmably connectable tohorizontal conductors 162 andvertical conductors 164 ofinterconnect grid 160. Interconnectgrid 160 is also programmably connectable to input/output elements 180. By programming a connection between input/output elements 180,logic array blocks 120 and embedded array blocks 140 throughhorizontal conductors 162 andvertical conductors 164, the user may create desired logical functions. Particularly, PLD 100 may be programmed to implement complex arithmetic circuits of the present invention. - B. Implementation of Arithmetic Circuits using Look-Up Tables
- According to the present invention, complicated arithmetic circuits may be efficiently implemented within a programmable logic device (e.g., device 100) by using Look-Up Tables (LUTs) to store pre-calculated values. A table look-up operation can then be performed in place of complex arithmetic operations. In this way, at the expense of a few LUTs, many logic elements can be saved.
- As an example, it is desired to implement a circuit for calculating a complicated arithmetic function fn(x) in response to an input value x. Generally, complicated logic circuits requiring a large number of logic elements are used. However, according to an embodiment of the present invention, values of the function fn(x) over a range of input values (x) are first determined. The pre-determined or pre-calculated values are then stored within a LUT of a programmable logic device. In one embodiment, the pre-determined values are stored at addresses corresponding to the input values (x). For example, the pre-determined value of fn(1) is stored at
address 1, and the predetermined value of fa(255) is stored ataddress 255. The LUT can then be used for calculating the value of fn(x) for an input value (x). For example, when an input value k is presented to the look-up table, the value stored at address k, which is fn(k), will then be provided as an output. A block diagram of aLUT 500 and its contents according to this embodiment of the present invention is illustrated inFIG. 5 . The circuit for matching the input value to the appropriate address of theLUT 500 is well known in the art, and is not illustrated for purposes of clarity. - The present invention is particularly applicable to the implementation of complicated arithmetic circuits for adaptive digital voice filtering in programmable logic devices. In the following, exemplary implementations of circuits for carrying out normalized LMS (Least-Mean-Squared) algorithms are presented. It should be appreciated that, however, the present invention is not limited to such circuit implementations.
- C. Implementation of a Reciprocal Circuit in a Programmable Logic Device
- In various digital signal processing applications, efficient calculations of reciprocal values are often required. A brief description of a fast LUT-based algorithm for computing a reciprocal value given a fraction with arbitrary bit width follows.
- Suppose Pn=.(an−1an−2an−3 . . . a1a0) is an n-bit binary unsigned fraction. The decimal value of Pn can be expressed as follows:
- A certain symmetry can be seen from the above expression. This symmetry of the expression can be exploited. Particularly, all possible reciprocal values of (an−1*2−1+an−2*2−2+an−3*2−3+ . . . +an−L*2−L) can be pre-calculated and stored within a look-up table, using the coefficients (an−(L+1)an−(L+2)an−(L+3) . . . an−2L) as addresses. The reciprocal value of 2−L(an−(L+1)*2−1+an−(L+2)*2−2+an−(L+3)*2−3+ . . . +an−2L*2−L) can be determined by retrieving the pre-calculated reciprocal value from the look-up table using the coefficients (an−(L+1)an−(L+2)an−(L+3) . . . an−2L) as the address, and by multiplying the retrieved value by 2L. Similarly, the reciprocal value of 2−mL(an−(mL+1)*2−1+an−(mL+2)*2−2+an−(mL+3)*2−3+ . . . +a0*2−L) can be determined by retrieving the pre-calculated reciprocal value from the look-up table using the coefficients (an−(mL+1)an−(mL+2)an−(mL+3) . . . a0) as the address, and by multiplying the retrieved value by 2mL.
- As an example, suppose the fraction Pn is divided into m equally wide fields each having 4 bits. To calculate the reciprocal of Pn, m look-up tables each having 4-bit wide addresses (or indices) and each capable of storing sixteen reciprocal values can be used. The contents of one such look-up table are illustrated below in Table 1.
TABLE 1 Address Content 0000 0 0001 16 0010 8 0011 16/3 0100 4 0101 16/5 0110 8/3 0111 16/7 1000 2 1001 16/9 1010 8/5 1011 16/11 1100 4/3 1101 16/13 1110 8/7 1111 16/15 - In operation, for a fraction 0.1011 (bin), its reciprocal value can be determined by looking up Table 1 to retrieve the value stored at address 1011, which is 16/11, or approximately 0.23. For a fraction 0.00001001 (bin), its reciprocal value can be calculated by looking up Table 1 to retrieve the value stored at address 1001, which is 16/9, and multiplying 24 to the retrieved value. In the present example, the reciprocal value of 0.00001001 (bin) can be easily determined to be 24*16/9, or approximately 28.44.
- Reciprocal calculation can be sped up even further by storing the reciprocal values in the look-up table as binary integer values. Multiplication of binary integer values by 2n is equivalent to shifting the values to the left by n bits. Naturally, accuracy of the reciprocal calculation may be somewhat compromised. Nevertheless, in some applications where speed or simplicity is desired, such a trade-off may be preferred.
- Note that the least significant segment may include fewer than L-bits (e.g., n−1 mod L bits). In that case, the look-up table with fewer entries will be used. It should also be noted that the present invention may also be applied to signed values. In that embodiment, the magnitude of a signed value is first determined. Then, the reciprocal of the magnitude is determined using techniques discussed above. Thereafter, if the original number is negative, the reciprocal value is two-complemented.
-
FIG. 2A is a block diagram illustrating acircuit 200 for calculating reciprocal values in accordance with one embodiment of the present invention.Circuit 200, according to the present invention, may be implemented using a programmable logic device (e.g., device 100). As illustrated,circuit 200 includes 2's complement circuits 210 a-210 b, a plurality of identical look up tables 220 a-220 n, a plurality of shifters 230 a-230 n, and anoutput selector circuit 240. - In operation, when a number Pn is received, the 2's
complement circuit 210 a determines whether the number is negative and determines the magnitude |Pn| of the number. The magnitude |Pn| is then partitioned into m segments each being L-bit in width. The partitioned magnitude |Pn| 1225, containing a plurality of segments 235 a-235 n, is also illustrated inFIG. 2A . The most significant bit of the partitioned magnitude |Pn| is thesign bit 234, which is not taken into account when the number is partitioned. Note that the leastsignificant segment 235 n may contain only (n−1 mod L) bits. - The segments 235 a-235 n are then provided to look-up tables 220 a-220 n . According to the present embodiment, each LUT 220 a-220 n has stored therein a plurality of pre-determined reciprocal values. In the embodiment illustrated in
FIG. 2A , the contents of the LUTs 220 a-220 n may be identical. Further, the reciprocal values are pre-calculated based on the addresses of the LUT at which the reciprocal values are stored. For example, address 1011 stores the reciprocal value of 0.1011, and address 1110 stores the reciprocal value of 0.1110, etc. Thus, when asegment 235 a of the magnitude |Pn| is presented to theLUT 220 a, theLUT 220 a retrieves the reciprocal value stored at the address corresponding to thesegment 235 a and provides the retrieved reicprocal value to theshifter 230 a. Likewise, when asegment 235 b is presented to theLUT 220 b, the reciprocal value stored at the address corresponding to thesegment 235 b will be retrieved and provided to theshifter 230 b. It should be noted, however, that address 0000 does not store the reciprocal value of 0.0000. Rather, address 0000 of the look-up table stores a value of zero, or a value that indicates a potential error. - Each of the shifters 230 a-230 n shifts the retrieved reciprocal values according to the position of the segment within |Pn|. For example, the
shifter 230 b is configured to shift the reciprocal value L bits to the left, and theshifter 230 n is configured to shift the reciprocal value (n−1)*L bits to the left. - The shifted values are then provided to the
output selector circuit 240. In accordance with the present embodiment, theoutput selector circuit 240 select a first non-zero output from the shifters 230 a-230 n and ignores the rest of their outputs. In this embodiment, theshifter 230 a, corresponding to the most significant bits of the input value, has the highest priority, and theshifter 230 b, corresponding to the next most significant bits of the input value, has the second highest priority, etc. For example, if the output ofshifter 230 a is non-zero, then the outputs of theshifters 230 b-230 n are ignored. As another example, if the output ofshifter 230 a is zero (indicating that the first segment 225 a contains 0000), but the output of theshifter 230 b is non-zero, then theselective adder 240 selects the output of theshifter 230 b. The selected output is then provided to 2'scomplement circuit 210 b to be converted into 2's complement form if the original number is negative. - In another embodiment of the present invention, instead of using multiple LUTs in parallel, a single LUT may be used to determine the reciprocal values.
FIG. 2B illustrates acircuit 205 for calculating reciprocal values in accordance with another embodiment of the present invention. Likecircuit 200,circuit 205 is suitable for implementation by a PLD. - As shown,
circuit 205 includes 2's complement circuits 210 a-210 b, an inputsegment selector circuit 250, a look-up table 260, and ashifter 270. TheLUT 260 has stored therein a plurality of pre-determined reciprocal values that are pre-calculated based on the addresses of the LUT at which the reciprocal values are stored. For example, address 1011 stores the reciprocal value of 0.1011, and address 1110 stores the reciprocal value of 0.1110, etc. - In operation, when a number Pn is received, the 2's
complement circuit 210 a determines whether the number is negative and determines the magnitude |Pn| of the number. The magnitude |Pn| is then partitioned into m segments each being L-bit in width. The partitioned magnitude |Pn| 225, containing a plurality of segments 235 a-235 n, is also illustrated inFIG. 2A . The most significant bit of the partitioned magnitude |Pn| is thesign bit 234, which is not taken into account when the number is partitioned. Note that the least significant segment 235 n may contain only (n−1 mod L) bits. - The segments 235 a-235 n are then provided to the input
segment selector circuit 250. The inputsegment selector circuit 250 selects the first non-zero segment of the magnitude |Pn| 225 (e.g., the non-zero segment that is closest to the sign bit 234) to be provided to theLUT 260. In response, theLUT 260 retrieves the reciprocal value stored at the address corresponding to the first non-zero segment and provides the retrieved value to theshifter circuit 270. Theshifter 270 then shifts the retrieved reciprocal value according to a shift control signal provided by the inputsegment selector circuit 250. The output of theshifter 270 is then provided to 2'scomplement circuit 210 b to be converted into 2's complement form if the original number is negative. -
FIG. 3 is a flow diagram illustrating a method of determining reciprocal values for an input data using a LUT-based reciprocal calculation circuit (e.g., circuit 205) in accordance with one embodiment of the present invention. As shown inFIG. 3 , atstep 310, reciprocal values are stored in a look-up table. According to an embodiment of the invention, the reciprocal values may be stored at addresses of the look-up table that correspond to the reciprocal values. For instance, the reciprocal value of 0.0010 maybe stored at address 0010 of the LUT. It should be noted that, in that embodiment, an ERROR value or a zero value may be stored at address 0000 of the LUT. - At
step 320, an input data is received. Then, atstep 330, the input data is partitioned into a number of segments each having a bit-width corresponding to the bit-width of the addresses of the look-up table. For instance, if the look-up table has 8-bit wide addresses and the input data is 32-bit wide, the input data is partitioned into four 8-bit wide segments. - At
step 340, one of the segments is selected and provided to the look-up table to determine if there is a match and to retrieve a reciprocal value whose address matches the selected segment. According to the present embodiment, the first segment containing a non-zero value may be selected. Note that other selection criteria may also be used. - Then, at
step 350, the retrieved reciprocal value is shifted according to a position of the selected segment in relation to the input data to generate an approximate reciprocal value for the input data. For instance, if the second 8-bit segment of a 32-bit input fraction is selected, then the retrieved reciprocal value is shifted eight bits to the left. If the third 8-bit segment of the 32-bit input fraction is selected, then the retrieved reciprocal value is shifted sixteen bits to the left. The result of the shifting will be an approximate value of the reciprocal of the 32-bit input fraction. - D. Implementation of a μ-Law/A-Law Expander Circuit
- The normalized LMS algorithm is a well-known algorithm in the field of telecommunications. Specifically, the normalized LMS algorithm can be used for canceling echoes that occur as a result of hybrid coupling within a telephone network. Generally, the normalized LMS algorithm can be represented by the following equations:
P(j)=(1−a)*P(j−1)+a*x(j)*x(j)
h i(j+1)=h i(j)+(μ·e(j)/P(j))·x(j−1), where i=0, . . . , N−1. - For μ-Law encoded data Rin(j), the linear input data x(j) can be represented by the expression: μ-Law(Rin(j)), where Rin(j) is the μ-Law 8-bit compressed value of the input data x(j). For A-Law encoded data Rin(j), the linear input data x(j) can be represented by the expression: A-Law(Rin(j)), where Rin(j) is the A-Law 8-bit compressed value of the input data x(j).
- Due to its complexity, the normalized LMS algorithm, if implemented as a circuit within a programmable logic device, would require more than six hundred logic elements. Logic elements, however, are critical resources in a programmable logic device. Thus, it is desirable to implement a circuit for performing LMS algorithms that do not require as many logic elements. The present invention presents a new approach that is significantly more efficient than prior art approaches.
- As shown above, the first step in the normalized LMS algorithm is to square the far end linear input data x(j) for calculation of the input signal power P(j). According to the present invention, instead of first calculating the linear value of x(j)=μ-Law(Rin(j)) or A-Law(Rin(j)), and then calculating the value of x(j)*x (j), pre-calculated values for [μ-Law(Rin(j))*μ-Law(Rin(j))] or [A-Law(Rin(j))*A-Law(Rin(j))] are stored in a single look-up table. In particular, the pre-calculated values are stored in the look-up table using the values (Rin(j)) as storage addresses. For example, pre-calculated value for [μ-Law(255)*p-Law(255)] may be stored at
address 255 of the look-up table. Then, during the LMS calculation, the input value Rin(j) is then used as an index to look-up the corresponding [μ-Law(Rin(j))*μ-Law(Rin(j))] or [A-Law(Rin(j))*A-Law(Rin(j))] value. In this way, a significant number of logic elements can be saved at the expense of only one embedded array block, which is used for implementing the look-up table. -
FIG. 4A is a block diagram illustrating a look-up table 410 for implementing a μ-Law expander in accordance with one embodiment of the present invention. As shown, the look-up table 410 includes pre-calculated values of [μ-Law(Rin(j))*μ-Law(Rin(j))] for all possible input values Rin(j). It should be noted that Rin(j) is an 8-bit value. Thus, in the present embodiment, 256 pre-calculated μ-Law expansion values are stored within the look-up table 410. -
FIG. 4B is a block diagram illustrating another look-up table 420 for implementing an A-law expander in accordance with one embodiment of the present invention. The look-up table 420 includes pre-calculated values of [A-Law(Rin(j))*A-Law(Rin(j))] for all possible input values Rin(j). Similar to table 410, the table 420 includes 256 pre-calculated A-Law expansion values. - The present invention, the use of embedded memory for efficient implementation of complex arithmetic circuits, has thus been disclosed. The foregoing descriptions of specific embodiments of the present invention are presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed, obviously many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents.
Claims (14)
1.-10. (canceled)
11. A circuit for performing normalized least-mean-squared calculations, comprising:
an input for receiving a μ-Law encoded data value; and
a look-up table coupled to the input, the look-up table having stored therein a plurality of pre-calculated intermediate data values, the pre-calculated intermediate data values being operands for normalized least-mean-squared calculations, wherein the look-up table is configured to provide a corresponding one of the re-calculated intermediate data values in response to the μ-Law encoded data value.
12. The circuit of claim 11 , wherein the plurality of pre-calculated intermediate data values comprise a plurality of pre-calculated μ-Law expansion values.
13. The circuit of claim 11 , wherein the look-up table stores a plurality of possible input values associated with the plurality of pre-calculated intermediate data values.
14. The circuit of claim 13 , wherein the look-up table uses the plurality of possible input values as indices.
15. A circuit for performing normalized least-mean-squared calculations, comprising:
an input for receiving an A-Law encoded data value; and
a look-up table coupled to the input, the look-up table having stored therein a plurality of pre-calculated intermediate data values, the pre-calculated intermediate data values being operands of normalized least-mean-squared calculations, wherein the look-up table is configured to provide a corresponding one of the pre-calculated intermediate values in response to the A-Law encoded data value.
16. The circuit of claim 15 , wherein the plurality of pre-calculated intermediate data values comprise a plurality of pre-calculated A-Law expansion values.
17. The circuit of claim 15 , wherein the look-up table stores a plurality of possible input values associated with the plurality of pre-calculated intermediate data values.
18. The circuit of claim 17 , wherein the look-up table uses the plurality of possible input values as indices.
19. A method of performing normalized least-mean-squared calculations, the method comprising:
receiving an input data, the input data comprising a μ-Law encoded data value; and
using the input data as an index to retrieve a pre-calculated intermediate data value from a look-up table, the look-up table having stored therein a plurality of pre-calculated intermediate values, wherein the pre-calculated intermediate data values comprise operands for normalized least-mean-squared calculations and correspond to a pre=determined range of μ-Law encoded data values.
20. The method of claim 19 , wherein the plurality of pre-calculated intermediate values comprise a plurality of pre-calculated i-Law expansion values.
21. A method of performing normalized least-mean-squared calculations, the method comprising:
receiving a A-Law encoded data value; and
using the A-Law encoded data value as an index to retrieve a pre-calculated intermediate data value from a look-up table, the look-up table having stored therein a plurality of pre-calculated intermediate values, wherein the pre-calculated intermediate data values comprise operands for normalized least-mean-squared calculations and correspond to a pre-determined range of A-Law encoded data values.
22. The method of claim 21 , wherein the plurality of pre-calculated intermediate values comprise a plurality of pre-calculated A-Law expansion values.
23.-26. (canceled)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/504,849 US20060277240A1 (en) | 2000-09-28 | 2006-08-16 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US23624400P | 2000-09-28 | 2000-09-28 | |
US09/863,650 US7058675B1 (en) | 2000-09-28 | 2001-05-22 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
US11/264,778 US7124161B2 (en) | 2000-09-28 | 2005-10-31 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
US11/504,849 US20060277240A1 (en) | 2000-09-28 | 2006-08-16 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/264,778 Division US7124161B2 (en) | 2000-09-28 | 2005-10-31 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
Publications (1)
Publication Number | Publication Date |
---|---|
US20060277240A1 true US20060277240A1 (en) | 2006-12-07 |
Family
ID=35997468
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/863,650 Expired - Fee Related US7058675B1 (en) | 2000-09-28 | 2001-05-22 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
US11/264,778 Expired - Fee Related US7124161B2 (en) | 2000-09-28 | 2005-10-31 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
US11/504,849 Abandoned US20060277240A1 (en) | 2000-09-28 | 2006-08-16 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
Family Applications Before (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/863,650 Expired - Fee Related US7058675B1 (en) | 2000-09-28 | 2001-05-22 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
US11/264,778 Expired - Fee Related US7124161B2 (en) | 2000-09-28 | 2005-10-31 | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices |
Country Status (1)
Country | Link |
---|---|
US (3) | US7058675B1 (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080201396A1 (en) * | 2007-02-15 | 2008-08-21 | Fujitsu Limited | Signal processing apparatus and the correcting method |
US20090164543A1 (en) * | 2007-12-23 | 2009-06-25 | Vinodh Gopal | Apparatus and method to compute reciprocal approximations |
US20090254759A1 (en) * | 2006-07-12 | 2009-10-08 | Koninklijke Phillips Electronics, N.V. | Tamper resistance of a digital data processing unit |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FI20011610A0 (en) * | 2001-08-07 | 2001-08-07 | Nokia Corp | Method and apparatus for performing a division calculation |
US7346642B1 (en) * | 2003-11-14 | 2008-03-18 | Advanced Micro Devices, Inc. | Arithmetic processor utilizing multi-table look up to obtain reciprocal operands |
JP2008097194A (en) * | 2006-10-10 | 2008-04-24 | Olympus Corp | Inverse calculation apparatus, inverse calculation method and inverse calculation program |
US11562217B2 (en) * | 2020-03-30 | 2023-01-24 | Alibaba Group Holding Limited | Apparatuses and methods for approximating nonlinear function |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4357674A (en) * | 1979-09-22 | 1982-11-02 | Kokusai Denshin Denwa Co., Ltd. | PCM Signal calculator |
US4401855A (en) * | 1980-11-28 | 1983-08-30 | The Regents Of The University Of California | Apparatus for the linear predictive coding of human speech |
US5177734A (en) * | 1988-05-02 | 1993-01-05 | Itt Corporation | Multirate wire line modem apparatus |
US5559793A (en) * | 1993-01-08 | 1996-09-24 | Multi-Tech Systems, Inc. | Echo cancellation system and method |
US6522688B1 (en) * | 1999-01-14 | 2003-02-18 | Eric Morgan Dowling | PCM codec and modem for 56K bi-directional transmission |
US6535844B1 (en) * | 1999-05-28 | 2003-03-18 | Mitel Corporation | Method of detecting silence in a packetized voice stream |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4991132A (en) * | 1987-12-17 | 1991-02-05 | Matsushita Electric Industrial Co., Ltd. | Apparatus for executing division by high-speed convergence processing |
JPH0831029B2 (en) * | 1989-07-11 | 1996-03-27 | 日本電気株式会社 | Approximate inverse generator for division |
US5818744A (en) * | 1994-02-02 | 1998-10-06 | National Semiconductor Corp. | Circuit and method for determining multiplicative inverses with a look-up table |
US5825681A (en) * | 1996-01-24 | 1998-10-20 | Alliance Semiconductor Corporation | Divider/multiplier circuit having high precision mode |
US5768171A (en) * | 1996-06-28 | 1998-06-16 | Intel Corporation | Method and apparatus for improving the precision or area of a memory table used in floating-point computations |
US6260054B1 (en) * | 1998-10-29 | 2001-07-10 | Neomagic Corp. | Reciprocal generator using piece-wise-linear segments of varying width with floating-point format |
-
2001
- 2001-05-22 US US09/863,650 patent/US7058675B1/en not_active Expired - Fee Related
-
2005
- 2005-10-31 US US11/264,778 patent/US7124161B2/en not_active Expired - Fee Related
-
2006
- 2006-08-16 US US11/504,849 patent/US20060277240A1/en not_active Abandoned
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4357674A (en) * | 1979-09-22 | 1982-11-02 | Kokusai Denshin Denwa Co., Ltd. | PCM Signal calculator |
US4401855A (en) * | 1980-11-28 | 1983-08-30 | The Regents Of The University Of California | Apparatus for the linear predictive coding of human speech |
US5177734A (en) * | 1988-05-02 | 1993-01-05 | Itt Corporation | Multirate wire line modem apparatus |
US5559793A (en) * | 1993-01-08 | 1996-09-24 | Multi-Tech Systems, Inc. | Echo cancellation system and method |
US6522688B1 (en) * | 1999-01-14 | 2003-02-18 | Eric Morgan Dowling | PCM codec and modem for 56K bi-directional transmission |
US6535844B1 (en) * | 1999-05-28 | 2003-03-18 | Mitel Corporation | Method of detecting silence in a packetized voice stream |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090254759A1 (en) * | 2006-07-12 | 2009-10-08 | Koninklijke Phillips Electronics, N.V. | Tamper resistance of a digital data processing unit |
US8543835B2 (en) * | 2006-07-12 | 2013-09-24 | Irdeto B.V. | Tamper resistance of a digital data processing unit |
US20080201396A1 (en) * | 2007-02-15 | 2008-08-21 | Fujitsu Limited | Signal processing apparatus and the correcting method |
US20090164543A1 (en) * | 2007-12-23 | 2009-06-25 | Vinodh Gopal | Apparatus and method to compute reciprocal approximations |
US8301680B2 (en) * | 2007-12-23 | 2012-10-30 | Intel Corporation | Apparatus and method to compute reciprocal approximations |
Also Published As
Publication number | Publication date |
---|---|
US7124161B2 (en) | 2006-10-17 |
US7058675B1 (en) | 2006-06-06 |
US20060053192A1 (en) | 2006-03-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7124161B2 (en) | Apparatus and method for implementing efficient arithmetic circuits in programmable logic devices | |
US6943579B1 (en) | Variable fixed multipliers using memory blocks | |
US6438570B1 (en) | FPGA implemented bit-serial multiplier and infinite impulse response | |
US5255216A (en) | Reduced hardware look up table multiplier | |
US5367477A (en) | Method and apparatus for performing parallel zero detection in a data processing system | |
EP0693236B1 (en) | Method and arrangement in a transposed digital fir filter for multiplying a binary input signal with tap coefficients and a method for designing a transposed digital filter | |
EP1105967B1 (en) | Multiplierless digital filtering | |
GB2303009A (en) | Finite impulse response filter | |
US5742534A (en) | Electronic circuit for modular computation in a finite field | |
US7356554B1 (en) | Variable fixed multipliers using memory blocks | |
US5177703A (en) | Division circuit using higher radices | |
US4958311A (en) | Composite finite impulse response digital filter | |
KR100798523B1 (en) | Digital filter with state storage | |
EP0795819A1 (en) | Method and apparatus for the processing of digital signals in one-hot RNS | |
KR19980080981A (en) | Multiplication method and multiplication circuit | |
JP3003467B2 (en) | Arithmetic unit | |
Sasao et al. | On LUT cascade realizations of FIR filters | |
Ramirez et al. | High performance, reduced complexity programmable RNS-FPL merged FIR filters | |
KR100712864B1 (en) | System for varying the dynamic range of cofficients in a digital filter | |
US7194723B1 (en) | Techniques for mapping functions to lookup tables on programmable circuits | |
JPH08107366A (en) | Inversion circuit of a finite body origin | |
Naik et al. | An efficient reconfigurable FIR digital filter using modified distribute arithmetic technique | |
Zhang et al. | Parallel distributed arithmetic FIR filter design based on 4: 2 compressors on Xilinx FPGAs | |
KR0146655B1 (en) | Multi-nary and logic device | |
WO2005085990A1 (en) | A high speed shifter circuit |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |