I. RELATED APPLICATIONS

This application relates to an application by the same inventors entitled “A Low Complexity Correlator for Multicode CDMA” filed on the same date as this application, and incorporated herein by reference. [0001]
II BACKGROUND OF THE INVENTION

This invention relates generally to despreaders, and more particularly to a dualcode despreader that can despread pseudonoise (PN) sequences used for Code Division Multiple Access (CDMA) systems. [0002]

The growing importance of wireless communications has increased the demand for data transmission over mobile radio channels. Although GSM standards have become very popular and data service specifications are developing, most current mobile communications still use CDMA technologies. Future multimedia transmission, however, will require wide bandwidths and high data rates, which in turn will require complex and expensive hardware. [0003]

CDMA systems use a PN sequence to “spread” input data to resist data loss in a noisy wireless environment. The transmitted baseband signal is expressed as
[0004] $\begin{array}{cc}{S}_{T}=\sum _{i}\ue89e\mathrm{bi}\ue8a0\left[j\right]\ue89e\mathrm{Ci},& \left(1\right)\end{array}$

where b[0005] _{i}[j], a scalar value representing the jth bit of user i, is +1 or −1, and C_{i}, a column vector representing the PN code sequence, also has entries of either +1 or −1. The baseband signals are then summed and transmitted.

Receivers must despread the received signal back into an original input symbol by correlating the received signal with the same PN code C[0006] _{k }used to spread the signal as follows:

b _{k} [k]=sign(C _{k} ^{T} S _{R}) (2)

The Sign function outputs a 1 if the input is positive and a −1 if the input is negative. [0007]

Generally, [0008]

S _{R} =S _{T} +n (3)

where n represents the noise from the environment. Combining equations (1), (2), and (3) yields:
[0009] $\begin{array}{cc}\begin{array}{c}{b}_{k}\ue8a0\left[k\right]=\mathrm{sign}\ue8a0\left[{C}_{k}^{T}\ue89e{S}_{T}+n\right],\\ =\mathrm{sign}\ue8a0\left[{C}_{k}^{T}\ue89e\sum _{i}\ue89e\left(\mathrm{bi}\ue8a0\left[j\right]\ue89e{C}_{i}+n\right)\right]\\ =\mathrm{sign}\ue8a0\left[{b}_{k}\ue8a0\left[j\right]\ue89e{C}_{k}^{T}\ue89e{C}_{k}+\sum _{i\ne k}\ue89e\left({\mathrm{biC}}_{k}^{T}\ue89e{C}_{i}+{C}_{k}^{T}\ue89en\right)\right]\end{array}& \left(4\right)\end{array}$

For orthogonal codes, [0010]

C _{k} ^{T} C _{i}≈0, if i≠k (5)

Also, noise n is small compared with an original signal, and [0011]

C _{k} ^{T} C _{k} =m (6)

where m is the spreading factor. [0012]

Because b[0013] _{k} [j] is either 1 or −1,

b _{k} [j]≈sign(b _{k} [j]m)=b _{k} [j] (7)

In addition to the noise signal, other user data can be regarded as another source of interference for CDMA systems. [0014]

Modern CDMA systems use either different orthogonal codes or the same code with different delays. This requires hardware that can despread several codes concurrently. Some researchers have tried to increase throughput by adding two PN sequences as the code input, with the two sequences coming from the same code with different delays. [0015]

Another system to reduce power consumption uses signmagnitude data format and two accumulators for positive and negative partial sums, respectively, with a specialized architecture. None of these systems, however, provides a complete solution to future problems facing CDMA systems. [0016]
III. SUMMARY OF THE INVENTION

A method, consistent with this invention, of despreading an input signal spread with two original codes, comprises forming two new multielement codes from the two original codes such that only one of the corresponding elements of each code is zero, combining the new codes with the input signal, and combining the partial results to form correlation values. Combining the new codes involves forming partial results for each of the new codes, and updating, for each element, the partial result corresponding to the new code for which the corresponding element is not zero. [0017]

A system, consistent with this invention, for despreading an input signal spread with two original codes involves forming two new multielement codes from the two original codes such that only one of the corresponding elements of each code is zero. The apparatus comprises first adder means for combining the new codes with the input signal, and adder subtractor means for combining the partial results to form correlation values. The first adder means includes means for forming partial results for each of the new codes, and means for updating, for each element, the partial result corresponding to the new code for which the corresponding element is not zero. [0018]

A dualcode correlator, consistent with this invention, is coupled to an input signal and a pair of multielement codes designed such that only one of the corresponding elements of each code is zero, and comprises a decoder element receiving the input signal and the codes, an adder coupled to the output of the decoder element and a register bank containing partial results, and an adder/subtractor circuit to form correlation results from the partial results. [0019]

Another dual code correlator, consistent with this invention, is coupled to an input signal and a pair of multielement codes designed such that only one of the corresponding elements of each code is zero, and comprises an adder circuit receiving the input signal and a partial result, a register bank containing a plurality of partial results, steering circuits for directing the output of the adder circuit to the register circuit and for directing the appropriate one of the partial results to the adder circuit, and an adder/subtractor circuit to form correlation results from the partial results. [0020]

A matched filter, consistent with this invention, comprises a plurality of filter stages connected sequentially, and an adder/subtractor circuit connected to the outputs of the last of the filter stages in the sequence for forming the filter output. Each of the stages includes an adder receiving an input signal and an output from a previous stage, the first of the stages in the sequence receiving a fixed input, and a register bank to hold the output of the adder as an output of that stage. [0021]

Both the foregoing general description and the following detailed description are exemplary and do not restrict the invention claimed. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate systems and methods consistent with the invention and, together with the description, explain the principles of the invention.[0022]
IV. BRIEF DESCRIPTION OF THE DRAWINGS

In the drawings, [0023]

FIG. 1 is a flow diagram of the operations for despreading consistent with this invention; [0024]

FIG. 2 is a block diagram of a system, consistent with this invention, for a two'scomplement representation of input signals; [0025]

FIGS. 3A and 3B are block diagrams of systems consistent with the invention for signals in signmagnitude representation of input signals; [0026]

FIG. 4 is a block diagram of dualcode, chipmatched filter consistent with this invention; [0027]

FIG. 5 is a block diagram of standard design of a correlator for purposes of comparing the complexity of conventional designs with systems consistent with this invention; and [0028]

FIG. 6 is a block diagram of a standard design of a matched filter for purposes of comparing the complexity of conventional designs with systems consistent with this invention.[0029]
V. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

A. Principles of Operation [0030]

1. Dual Code Despreading [0031]

Systems and methods consistent with the present invention generate two new codes from two original PN codes that can be orthogonal or the same with different delays. Proper use of the new codes allows the correlator hardware to be designed with fewer elements and greater speed. [0032]

The two new codes are generated from the original codes C[0033] _{a }and C_{b }as follows:

C _{1}=(C _{a} +C _{b})/2

C _{2}=(C _{a} −C _{b})/2 (8)

The elements of C[0034] _{a }and C_{b }are either 1 or −1, so the elements of C_{1 }and C_{2 }are 1, −1, or 0. Moreover, if C_{1,i }represents the ith entry of vector C_{1}, then

C _{1,i}=0 if C _{2,i≠}0

and [0035]

C _{1,i}≠0 if C _{2,i}=0 (9)

If the transmitted signals include data from two users then: [0036]

S=b _{a} [j]C _{a} +b _{b} [j]C _{b } (10)

To simplify the process, assume noise is zero or negligible. As a result, correlating this signal with the redesigned new codes yields: [0037]

b _{1} =C _{1} ^{T} S=(C _{a} ^{T} +C _{b} ^{T})S/2=(b _{a} [j]+b _{b} [j])m/2 (11)

Similarly, [0038]

b _{2} C _{2} ^{T} S=(C _{a} ^{T} −C _{b} ^{T})S/2=(b _{a} [j]−b _{b} [j])m/2 (12)

Therefore, the following results can be obtained by addition: [0039]

b _{a} [j]=sign(b _{1} +b _{2})=sign(b_{a} [j]m)=b _{a} [j]

b _{b} [j]=sign(b _{1} −b _{2})=sign(b _{a} [j]m)=b _{a} [j] (13)

The total number of computations for C[0040] _{1 }and C_{2 }equals the spreading factor. As will become apparent below, systems and methods consistent with this application require only two extra addition/subtraction operations at the end to performa correlation using the two new codes.

2. Despreading Process [0041]

FIG. 1 is a flow diagram [0042] 100 of the operations for despreading consistent with this invention. First, the new PN codes C_{1}, C_{2 }are calculated from the original codes C_{a }and C_{b }(step 110). Then a loop begins (step 120) in which the ith element of PN code C_{1 }is compared to zero (step 130). If they are not equal, then a value, Result_{1}, is incremented by the product of the ith element of code C_{1 }and the ith element of the input signal (step 140). Otherwise, Result_{2 }is incremented by the product of the ith element of code C_{2 }and the ith element of the input signal (step 150). If there are additional elements of the input signal and of the PN codes (step 160), then i is incremented by 1 (step 165) and the loop repeats. Otherwise, the correlation factors are calculated as a sum and difference between Result_{1 }and Result_{2 }(step 170).

3. Examples [0043]

To demonstrate the equivalence of the new codes to the original codes, assume the two original codes are Walsh codes, [1−1 1−1][0044] ^{T }and [1 1−1−1]^{T}. The two new codes, from Eq. 8 become [1 0 0−1]^{T }and [0−1 1 0]^{T}. Correlating an input, [4−2 3−3]^{T}, with these two codes yields:

[1 0 0−1][4−2 3−3]^{T}=7

[0−1 1 0][4−2 3−3]^{T}=5

The operations for multiply and accumulate are the same as for single code despreading, so the correct correlation values can be obtained as [0045]

(7+5)=12 and (7−5)=2

To verify the accuracies of these results, the conventional approach yields the same results: [0046]

[1−1 1−1][4−2 3−3]^{T}=12, and [1 1−1−1][4−2 3−3]^{T}=2

B. Hardware Implementations [0047]

1. Two'sComplement Representation [0048]

FIG. 2 shows a system [0049] 200, consistent with this invention, for a two'scomplement representation of input signals. System 200 includes a decoder or multiplier (e.g., XOR gate) 210 to multiply the codes by the input signal. A first adder 220 adds the output of the multiplier to the partial correlation results in either register 230 or 240, the register to be selected by multiplexer 250. The output of the first adder goes to either registers 230 or 240 according to demultiplexer 260. Both multiplexer 250 and demultiplexer 260 are controlled by a “select” signal derived from the codes. The “select” signal selects the proper registers according to the one of the codes that is not zero.

System [0050] 200 also includes buffer registers 235 and 245. The inputs of buffer registers 235 and 245 connect to registers 230 or 240, respectively, and their outputs connect to second adder 270 and subtractor 280. Adder 270 and subtractor 280 form the correct correlation values. Because one of the two new codes is always zero, the system needs to add the output of multiplier 210 to only one partial result, thus the system needs only one first adder 210.

System [0051] 200 can also be implemented in a bitserial manner if the spreading factor is larger than the wordlength. This also allows the second adder to operate slower than the first adder.

In addition, the dynamic range of the adder can be reduced. This is because the nonzero portions of the sum or difference of orthogonal PN sequences are about half their length due to the nature of PN sequences and orthogonality. [0052]

2. SignMagnitude Representation [0053]

A system consistent with this invention that can provide a lower power design uses signmagnitude representation. Such a representation generally has fewer transitions between states, which will consume less power in most CMOS implementations. One known architecture reduces the power consumption of the receiver using two correlators, one for positive values and the other for negative values. That architecture can be adapted for use with the two new codes to provide improved receivers. [0054]

FIG. 3A shows a system [0055] 300 consistent with the invention for signals in signmagnitude representation. System 300 includes an adder 310 for positive values and an adder 320 for negative values. Each adder has an associated demultiplexer, demultiplexer 312 for adder 310 and demultiplexer 322 for adder 320. The demultiplexers route partial sums to one of two associated registers, depending upon which code is zero. Registers 314 and 317 store the partial sums from adder 310 and provide an input to adder 310 through multiplexer 319, and registers 324 and 327 store partial sums from adder 320 and provide an input to adder 320 through multiplexer 329. A “select” signal for multiplexers 319 and 329 and for demultiplexers 312, 322 works as described above for system 200 in FIG. 2. In addition, registers 334 and 337 store the outputs of registers 314 and 317, respectively, and registers 344 and 347 store the outputs from registers 324 and 327, respectively.

With regard to these elements, the sign magnitude implementation in FIG. 3A resembles the two'scomplement circuits from FIG. 2 if one circuit is used for positive values and one circuit is used for negative values. Again, only two adders are needed because the value of one of the codes will always be 0. [0056]

The final elements of system [0057] 300 are subtractor 350 to subtract the negative values from register 344 from the positive values in register 334, and subtractor 355 to subtract the negative values from register 347 from the positive values in register 337. Adder 360 and subtractor 365 then determine the correct correlation values.

FIG. 3B shows a similar system [0058] 370 with one core adder 380, one demultiplexer 390, and one multiplexer 395. Such an architecture is possible because only one of adders 310 and 320 is in use at any one time. In such a system, the “select” signal indicates which code is not zero and the sign of the code.

Alternatively, a gated clock can also be applied to control the data flow and reduce power consumption. The dynamic range of elements [0059] 230, 235, 240, 245 can be reduced by 1 bit (and the dynamic range of elements 314, 317, 324, 327, 334, 337, 344, 347 can be reduced by 2 bits) because the probability of a “1” or −“1” is about equal. Bitserial adder/subtractors, which have fewer elements, can also be used to save hardware costs.

3. ChipMatched Filter Architecture [0060]

The use of two codes can also be applied to a chipmatched filter design. FIG. 4 shows one possible architecture for dualcode, chipmatched filter [0061] 400. Because the coefficients of the matched filter are set at +1 or −1, most of the structure can be hardwired, eliminating the need for a multiplexer.

Filter [0062] 400 contains several stages 410, 420 , . . . 460. Each stage includes adder 412, 422 , . . . 462, two registers to hold the partial results, 414/416, 424/426 , . . . 464/466, some stages have a register 418, 428 to hold the input signal. Only an additional adder 470 and subtractor 475 are needed to generate the correct outputs, thus realizing significant hardware savings for very long matched codes. This is because the adder at each stage just accumulates one of the new codes to correlate the partial results. Thus, hardwiring the new codes in the adders allows the adders just to decide whether to add the positive or negative inputs with one of the partialresult registers.

Both two'scomplement and signmagnitude representations are possible, but only the two'scomplement result is shown. Bitserial techniques cannot be applied to the final adder/subtractor due to the pipelined constraint. [0063]

C. Complexity Analysis [0064]

Table 1 contains a hardware complexity analysis. Each heading, except the first and last, identifies the type of component used in the associated circuit, and the number in parentheses represents the number of transistor gates for the component. The numbers in the columns represent the number of gates needed for 1bit implementation. The design of the multiplexer/demultiplexers assumes pass transistors are the wordlength of the input is four bits, and the spreading factor is 16. As a result, an 8bit adder is typically needed for the correlator. [0065]

Only seven bits are used for system
[0066] 300 in FIG. 3, but two bitserial adder/subtractors are appended, so Table 1 lists nine 1bit adders. Similarly, 6bit adders are used for architecture
400 in FIG. 4, but four bitserial adders/subtractors are needed.
TABLE 1 


Complexity comparison of several architectures 
 XOR     
 gates  Adders  Registers  Demux/Mux (1)  Total 
Architecture  (3)  (10)  (6)  (pass transistor)  gates 

Standard  4  8  16  0  288 
2's complement 
2 x standard  8  16  32  0  576 
2's complement 
System 200  4  9  28  14  384 
System 300  0  16  48  48  496 
System 370  0  10  48  48  436 
Standard chip  32  128  192  0  2528 
matchedfilter 
2 x standard chip  64  256  384  0  5056 
matched filter 
System 400  32  128  288  0  3104 


FIG. 5 shows the standard design of a correlator for each code, and FIG. 6 shows a standard design of a matched filter for each code. The proper comparisons are between the 2×standard designs and systems [0067] 200, 300, 370, and 400 because one of the standard designs would be needed for each code.

D. Conclusion [0068]

The specific hardware used to implement the correlators and the chipmatched filters is not critical to this invention. Persons of ordinary skill in the art will know to use whatever technologies or circuit designs are appropriate for their particular needs while still taking advantage of the savings attendant the present invention. Therefore, the scope of the appended claims is not to be limited to those specific examples. [0069]