1. Field of the Invention
The invention relates generally to the signal processing, and more particularly to signal processing in wireless communications.
2. Discussion of the Related Art
Wireless Local Area Network (WLAN) technology based upon Orthogonal Frequency Division Multiplexing (OFDM) is increasingly gaining in popularity due to very high spectral efficiency and extremely high data rates (e.g. 54 Mbits/second for IEEE 802.11a). The physical layer for OFDM-WLAN systems requires the implementation of FFTs (Fast Fourier Transform) for demodulation, complex division of the OFDM symbol for nominal channel estimation, and complex multiplication for channel equalization and pilot phase correction, all of which must be performed at high speed. The implementation of FFTs and vector-based complex operations necessitates very high computational throughput for the modem signal processing. However, the desire to implement systems with minimal cost, size, and power for VLSI implementation is constraining the performance capabilities.
One example of this is shown in a current system (
There is a need for the following embodiments. Of course, the invention is not limited to these embodiments.
According to an aspect of the invention, a method comprises: providing a reconfigurable circuit having multiple distinct circuit configurations with respective distinct operating modes, configuring the reconfigurable circuit in a first configuration to perform a fast Fourier transform function, configuring the reconfigurable circuit in a second configuration to perform a multiplier function, and configuring the reconfigurable circuit in a third configuration to perform a divider function. The fast Fourier transform function, multiplier function, and divider function may be used for signal demodulation, channel equalization and channel estimation for a WLAN IEEE 802.11 system.
According to another aspect of the invention, an apparatus comprises a reconfigurable mathematical operation circuit having a plurality of reconfigurable components for performing a respective plurality of distinct mathematical functions; and a controller for controlling a configuration of the circuit so that, in a first configuration the circuit performs a fast Fourier transform function, in a second configuration the circuit performs a multiplier function, and in a third configuration said circuit performs a divider function. The individual reconfigurable components may be a plurality of CORDIC circuits.
In an exemplary embodiment, the reconfigurable mathematical operation circuit forms a reconfigurable WLAN IEEE 802.11 receiver circuit including WLAN signal demodulation, channel estimation and channel equalization functions.
These, and other, embodiments of the invention will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following description, while indicating various embodiments of the invention and numerous specific details thereof, is given by way of illustration and not of limitation. Many substitutions, modifications, additions and/or rearrangements may be made within the scope of the invention without departing from the spirit thereof, and the invention includes all such substitutions, modifications, additions and/or rearrangements.
The drawings accompanying and forming part of this specification are included to depict certain aspects of the invention. A clearer conception of the invention, and of the components and operation of systems provided with the invention, will become more readily apparent by referring to the exemplary, and therefore nonlimiting, embodiments illustrated in the drawings, wherein like reference numerals (if they occur in more than one view) designate the same elements. The invention may be better understood by reference to one or more of these drawings in combination with the description presented herein. It should be noted that the features illustrated in the drawings are not necessarily drawn to scale.
The invention permits a small footprint (i.e. small VLSI size) system that can reconfigure its underlying hardware structure in a way that optimally implements a complex N-point parallel FFT butterfly stage, a complex division vector operation, or a complex multiplication vector operation. By doing so, the system may be configured in 802.11a mode for optimum OFDM-FFT processing and may be reconfigured in 802.11a or 802.11b mode for channel estimation or time domain filtering. Moreover, the system circumvents the need for traditional two's complement multiplication modules anywhere in the computation or data path stages through the incorporation of flexible CORDIC hardware module. This makes the implementation much more amenable to VLSI implementation. For IEEE 802.11a, the system may implement a 64-point complex FFT in 38 clock cycles; it also may perform point-wise multiplication or division of a complex vector by a complex vector in 51 clock cycles.
The FFT refers to the computationally efficient implementation of the DFT (Discrete Fourier Transform) by exploiting the following properties of WN, a multiplying factor:
A direct computation of the DFT involves N2 complex multiplications and N*(N−1) complex additions. The DFT is defined as
where
WN=exp(−j2π/N)
j=√{square root over (−1)}
The multiplying factors WN are known as “phase factors” or “twiddle factors.” The Inverse Discrete Fourier Transform (IDFT) is defined as
As the IDFT only differs from the DFT in sign of phase of WN and a scaling factor, for the purposes of discussion, only the DFT is used. However, all derivations below apply to IDFT with simple sign manipulation and scaling factor application.
Based on the equations for the DFT, the equation for a 64-point FFT, as used in the IEEE 802.11 protocol, may be written as
while the 64-point IFFT may be expressed as
where both xk and Xr are, in general, complex vectors. The point-wise complex vector multiplication and division may be described as
{overscore (Z)}={overscore (X)}*{overscore (Y)}
{overscore (Z)}={overscore (X)}/{overscore (Y)}
where {overscore (X)}, {overscore (Y)}, and {overscore (Z)}, are complex vectors of equal length. The point-wise vector multiplication and division performs Z[i]=X[i]×Y[i] and Z[i]=X[i]/Y[i] for each element i of input vectors {overscore (X)} and {overscore (Y)}.
The COrdinate Rotation DIgital Computer (CORDIC) algorithm is an iterative procedure to compute various elementary functions. The CORDIC algorithm uses a single core routine to evaluate sines, cosines, multiplications, divisions, exponentials, logarithms, and transcendental functions. The CORDIC algorithm computes these functions with n-bits of accuracy in n iterations, where each iteration requires only a small number of shifts and additions. The basic CORDIC equations are as follows:
xi+1=xi−mσi2−S(m,t)yi
yi+1=yi+σi2−S(m,t)xi
zi+1=zi−σiαm,t
where m identifies circular (m=1), linear (m=0), or hyperbolic (m=−1) co-ordinate systems, and for each iteration i=0, 1, . . . , n.
The scale factor is given by
It should be noted that this scale factor is fixed for each mode m, and thus can be pre-calculated. Furthermore, this scale factor may be approximated as sum-of-powers-of-2, thus simplifying its implementation to few adders and multiplexers instead of a multiplier.
Table 1 shows the different elementary functions that can be evaluated by the CORDIC algorithm. The multiplication and division operations using CORDIC have a restriction in that their results must be bounded by the input word length. If fractional fixed-point format is assumed, the multiplication output is always fractional, and thus satisfies the criterion. For division operation, the two fractional inputs must be scaled such that the division result is guaranteed to be fractional.
Referring to
As shown in
This radix-k kernel may be implemented using two radix-k/2 kernels and a twiddle factor 401 of Wk1=−j. For example, as shown in
The n-point FFT may use h stages of radix-k kernels. The same interconnect geometry is utilized for each radix-k FFT stage, thus allowing the sharing of hardware among all the stages. To determine the number of iterations or stages needed when implementing a n-point FFT using a radix-k kernel, the following equation is used:
Number of Iterations=logk(n)
The twiddle factors required between the two radix-k stages are computed using the CORDIC algorithm using Rotation mode in the Circular co-ordinate system, as shown in Table 1. The twiddle factors for FFT and IFFT differ only in sign of their respective phases.
This n-point FFT/IFFT structure of the present invention may be modified to incorporate n/2-element complex vector point-wise multiplication or division, which is defined as follows
{overscore (Z)}={overscore (X)}.*{overscore (Y)}
{overscore (Z)}={overscore (X)}./{overscore (Y)}
or
Z[i]=X[i]×Y[i] for i=0, 1, 2, . . . , n−1
Z[i]=X[i]/Y[i] for i=0, 1, 2, . . . , n−1
This is possible because the same CORDIC core engine for twiddle factors may also be used to compute multiplication and division of two real numbers, as shown in Table 1 (above) using the Rotation/Vectoring mode in the Linear co-ordinate system. To calculate complex number multiplication/division, complex inputs (real and imaginary) are first converted into their polar co-ordinates (magnitude and phase) using the CORDIC in Vectoring mode in Circular co-ordinate system. The multiplication and/or division of input magnitudes is performed using Rotation/Vectoring mode in Linear co-ordinate system of CORDIC. The input phases are added or subtracted for multiplication and division respectively by using CORDIC adders/subtractors. Finally, the resultant magnitude and phase are converted into real and imaginary components of output.
In the alternative, the complex multiplication may also be carried out directly in cartesian co-ordinates by
(I1+jQ1)×(I2+jQ2)=(I1I2−Q1Q2)+j(I1Q2Q1)
which involves 4 real multiplications (using CORDIC in Rotation mode in Linear co-ordinate system) and 2 real adders. However, the multiplication in polar co-ordinates is used here as it's very similar to division operation, thus permitting the reuse of the same control logic.
The terms a or an, as used herein, are defined as one or more than one. The term plurality, as used herein, is defined as two or more than two. The term another, as used herein, is defined as at least a second or more. The phrase any integer derivable therein, as used herein, is defined as an integer between the corresponding numbers recited in the specification, and the phrase any range derivable therein is defined as any range within such corresponding numbers. The terms n and k are any positive integer.
Specific embodiments of the invention will now be further described by the following, nonlimiting examples which will serve to illustrate in some detail various features and advantages of the present invention. The following examples are included to facilitate an understanding of ways in which the invention may be practiced. It should be appreciated that the examples which follow represent embodiments discovered to function well in the practice of the invention, and thus can be considered to constitute preferred modes for the practice of the invention.
For IEEE 802.11a, a system can implement a 64-point complex FFT in 38 clock cycles; it also can perform point-wise multiplication of a complex vector by a complex vector in 51 clock cycles.
For IEEE 802.11a, the required length of a FFT/IFFT transform is 64. One embodiment of the invention implements this 64-point FFT/IFFT using a radix-4 FFT kernel and CORDIC core engine for complex rotations of twiddle factors. The radix-4 FFT kernel performs the following operation:
This radix-4 kernel may be implemented using two radix-2 kernels 400 and a trivial twiddle factor 401 of W41=−j, as shown, for example, in
The 64-point FFT may use three identical stages of radix-4 kernels. An embodiment of one such radix-4 kernel is shown in
This 64-point FFT/IFFT structure may be modified to incorporate 32-element complex vector point-wise multiplication or division, which is defined as follows
{overscore (Z)}={overscore (X)}.*{overscore (Y)}
{overscore (Z)}={overscore (X)}./{overscore (Y)}
or
Z[i]=X[i]×Y[i] for i=0, 1, 2, . . . , 31
Z[i]=X[i]/Y[i] for i=0, 1, 2, . . . , 31
Referring now to
The lines in
Only a 4-point vector slice out of 64-point vector is shown in
In FFT/IFFT mode, the architecture of
All of the signals received at the inputs 907–909, 917 proceed through the radix 4 stage 800 shown in
Table 2 shows the number of clock cycles required to compute vector FFT/IFFT, multiplication and division functions, for IC CORDIC iterations in accordance with the present invention. To achieve W-bit accuracy at the output, one needs to perform (W+log2W) iterations of the CORDIC algorithm.
As shown in
Another embodiment of the invention can solve a 32-point FFT/IFFT using a radix-2 FFT kernel and CORDIC core engine for complex rotations of twiddle factors.
The 32-point FFT may use five stages of radix-2 kernels 1000 as shown in
This 32-point FFT/IFFT structure may be modified to incorporate 16-element complex vector point-wise multiplication or division, which is defined as follows
{overscore (Z)}={overscore (X)}.*{overscore (Y)}
{overscore (Z)}={overscore (X)}./{overscore (Y)}
or
Z[i]=X[i]×Y[i] for i=0, 1, 2, . . . , 15
Z[i]=X[i]/Y[i] for i=0, 1, 2, . . . , 15
The architecture for the current embodiment of the invention may be reconfigured to implement a 32-point complex FFT/IFFT, 16-point complex vector multiplication, or 16-point complex vector division. The reconfiguration may be done during the receiving and multiplexing of the incoming signals, where it is controlled by software. This reconfiguration may also take place each time a signal is received as many microprocessor chips have clock rates faster than that of the transmission rate. Reconfiguration may also be implemented in the form of a state machine in a microprocessor chip as variables such as delays and the number of bits transmitted are known.
Another embodiment of the invention can solve a 16-point FFT/IFFT using a radix-4 FFT kernel and CORDIC core engine for complex rotations of twiddle factors.
The 16-point FFT may use 2 stages of radix-4 kernels 1100 as shown in
This 16-point FFT/IFFT structure may be modified to incorporate 8-element complex vector point-wise multiplication or division, which is defined as follows
{overscore (Z)}={overscore (X)}.*{overscore (Y)}
{overscore (Z)}={overscore (X)}./{overscore (Y)}
or
Z[i]=X[i]×Y[i] for i=0, 1, 2, . . . , 7
Z[i]=X[i]/Y[i] for i=0, 1, 2, . . . , 7
The architecture for the current embodiment of the invention may be reconfigured to implement a 16-point complex FFT/IFFT, 8-point complex vector multiplication, or 8-point complex vector division. The reconfiguration may be done during the receiving and multiplexing of the incoming signals, where it is controlled by software. This reconfiguration may also take place each time a signal is received as many microprocessor chips have clock rates faster than that of the transmission rate. Reconfiguration may also be implemented in the form of a state machine in a microprocessor chip as variables such as delays and the number of bits transmitted are known.
A practical application of the invention that has value within the technological arts is that it enables mapping of generic algorithms used in digital communications and wireless modems. One embodiment of the invention, as shown in
A reconfigurable vector-FFT/IFFT and vector-multiplier/divider with a VLSI micro-footprint, representing an embodiment of the invention, is cost effective and advantageous for at least the following reasons. One such embodiment of the invention is reconfigurable so that different operations are based on the same underlying CORDIC kernel. An embodiment of the invention does not utilize multipliers or dividers, thus reducing the area it requires and costing less to make.
The invention enables improved bit-level accuracy for traditionally, computationally intensive functions, such as division and FFT. The invention also allows for WLAN 802.11 as well as other possible forms of FFT/IFFT and complex number operations. The invention improves quality and/or reduces costs compared to previous approaches.
Each of the reference listed are hereby incorporated by reference in their entirety.
Number | Name | Date | Kind |
---|---|---|---|
4231102 | Barr et al. | Oct 1980 | A |
4737930 | Constant | Apr 1988 | A |
4811210 | McAulay | Mar 1989 | A |
4896287 | O'Donnell et al. | Jan 1990 | A |
6003056 | Auslander et al. | Dec 1999 | A |
6608863 | Onizawa et al. | Aug 2003 | B1 |
6643678 | Van Wechel et al. | Nov 2003 | B1 |
6691144 | Becker | Feb 2004 | B1 |
6778591 | Sato | Aug 2004 | B1 |
6836839 | Master et al. | Dec 2004 | B1 |
6874006 | Fu et al. | Mar 2005 | B1 |
20040103265 | Smith | May 2004 | A1 |
Number | Date | Country | |
---|---|---|---|
20040064493 A1 | Apr 2004 | US |