FFT-based CDMA RAKE Receiver System and Method

The present invention relates generally to Code Division Multiple Access (CDMA) communications and, more particularly, to a RAKE receiver for interference cancellation of CDMA signals.

Wireless cellular communications around the world are moving toward CDMA systems in the third generation approaches. A CDMA system has the most efficient usage of the limited radio spectrum. It has been proved that a CDMA system ideally provides a large channel capacity gain over other accesses methods such as Frequency Division Multiple Access (FDMA) and Time Division Multiple Access (TDMA). In general, a CDMA system has to use a RAKE receiver to combine the received signal energy form different paths to combat the effect of multipath fading.

A RAKE receiver implements a form of path diversity by gathering the signal energy from different paths and by optimally combing all the multipath signals together. The path diversity function provides a robust communication channel so that when one path fades, communication is still possible through a non-fading path. A CDMA RAKE receiver detects the multipath signals using either a matched filter method or a correlation method.

Conventionally, a RAKE receiver in a CDMA system uses a spreading code matched filter to despread a multiple access code and a transversal filter for channel impulse response matching. The spreading code matched filter can be implemented at IF band using a SAW filter or at baseband using a digital matched filter. After code despreading, a transversal filter which is implemented at baseband is used to combine the received signal energy form different paths. The drawback of the SAW filter approach is that a SAW filter cannot be easily integrated with the baseband transversal filter in an IC. As IC technology progresses rapidly, the digital spreading code matched filter approach is a preferred choice. Although the current IC technology can provide large computational power, it is still hard to implement a transversal filter based RAKE receiver in IC, especially when the length of the multiple access code is large.

An alternative method to implement a RAKE receiver is to use a bank of correlators. Each correlator is used to detect a received signal path separately. The number of the correlators in the correlator bank is typically three or four. Therefore, this RAKE receiver structure needs to search for three or four stronger paths in the received signal.

Both the above two RAKE receiver implementations need a sounding receiver to estimate the multipath channel impulse response. A RAKE receiver needs to know the delay time, the carrier phase shift, and the strength of the main paths. Moreover, the correlator bank implementation needs extra computations to select the main paths.

One method to increase the CDMA system capacity is to use a parallel interference cancellation (PIC), for example, as described in Dariush Divsalar and Marvin K. Simon, "Improved CDMA performance using parallel interference cancellation," IEEE MILCOM, pp. 911-917, October 1994. Although the CDMA system capacity as this paper described can be increased, their receiver is applicable to a Additive White Gaussian Noise (AWGN) channel only and is unsuitable for a multipath fading channel. Other PIC methods for a multipath fading channel have also been proposed, such as in Matti Latva-aho, Markku Juntti and Markku Heikkila, "Parallel Interference Cancellation Receiver for DS-CDMA Systems in Fading Channels," IEEE PIMRC, pp. 559-564, September 1997. However, these methods adapt a time domain signal processing approach to cancel the multiple access interference and therefore, become complicated.

The present invention overcomes the foregoing limitations by providing a CDMA RAKE receiver that computes a detection in the frequency domain by using Fast Fourier Transform (FFT) matched filters. The receiver is implemented in both downlink and uplink communications when the received signal contains both data and pilot signals. In both downlink and uplink receivers the signal detection process operates in the frequency domain. The received signal is expediently processed in frequency domain by spreading code matched filters, and a channel matched filter prior to making a decision.

Spreading code matched filters are used to despread the spreading code of a pilot signal and a data signal. A channel matched filter is used to combine the received signal from different paths before a decision is made. The channel matched filter contains a channel frequency response estimation unit which is used to compute the coefficient of the channel matched filter.

After code despreading the pilot signal is used to compute the channel frequency response estimation. There are two types of methods for channel frequency response estimation. One uses the pilot signal after code despreading for the channel frequency response estimation directly. The other one transforms the pilot signal after code despreading to the channel impulse response estimation in time domain and then reserves the main paths of the channel impulse response estimation. After the main paths are reserved, the channel impulse response estimation is transformed to the channel frequency response estimation.

Interference cancellation techniques are used to increase the CDMA system capacity. In a downlink receiver, we estimate the pilot interference. Then, the estimated pilot interference is subtracted from the received signal in frequency domain for pilot interference cancellation. In an uplink receiver, the method utilizes multi-stage parallel interference cancellation. Such multi-stage parallel interference cancellation can be used in a fast fading channel because at each stage, interference signal estimation can be obtained from both the channel frequency response estimation and a tentative decision of each user at the previous stage.

The foregoing aspects and many of the attendant advantages of this invention will become more readily appreciated as the same becomes better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein:

FIGS. 1A and 1B are general architectural diagrams of a CDMA RAKE, system for a downlink and uplink receiver, respectively;

FIG. 2A is a block diagram of a downlink RAKE receiver with optional reserving main paths and data signal reconstruction;

FIG. 2B is a block diagram of an alternate downlink RAKE receiver;

FIG. 3A is a block diagram of an uplink RAKE receiver with optional reserving main paths;

FIG. 3B is a data detection flowgraph of the uplink RAKE receiver for the receiver shown in FIG. 3A;

FIGS. 4A and 4B are block diagrams illustrating different pilot signal estimations performed in the downlink RAKE receivers shown in FIGS. 2A and 2B;

FIG. 5 is a block diagram of data signal reconstruction optionally performed in the downlink RAKE receiver shown in FIG. 2A and 2B; and

FIG. 6 is a block diagram of reserve main paths optionally performed in the RAKE receivers shown in FIGS. 2A, 2B and 3A.

The FFT based CDMA RAKE receiver is designed for a direct sequence spread spectrum system (DSSS). The transmitted signal of this DSSS contains a data signal and a pilot signal. A data signal is modulated by using either binary phase shift keying (BPSK) or quadrature phase shift keying (QPSK) and multiplied with a multiple access spreading code (a data signal spreading code). A pilot signal is unmodulated and multiplied with a pilot signal spreading code. Then, the data signal and the pilot signal are combined to be a transmitted signal. Here the multiple access code of this DSSS and the pilot signal spreading code of this DSSS are equal-length short codes. That is, the period of a multiple access code or the period of a pilot signal spreading code is the same as the period of a transmitted symbol.

FIGS. 1A and 1B are architectural diagrams of downlink and uplink RAKE receiver systems 20 and 40 provided by the present invention. The downlink RAKE system 20 includes an antenna 22, a RF front-end 24, an analog-to-digital (AD) converter 26, a fast Fourier transform unit (FFT) 28, and a FFT-based CDMA RAKE receiver 30. The uplink RAKE receiver 40 includes the same components as the downlink RAKE receiver 20 and in addition includes a FFT-based CDMA RAKE receiver 50 for each user. A radio frequency (RF) signal is received by the antenna 22 and 42. The RF signal from the antenna enters the RF front-end 24 and 44 and converts the RF signal to an equivalent baseband complex signal (complex envelope) with a real I(t) and imaginary Q(t) component. The analog-to-digital (AD) converter 26 and 46 samples the I(t) and Q(t) components at a sampling rate of 1/Tc (or its integral multiple) to generate discrete time signals I[n] and Q[n], respectively, where Tc denotes the chip period for a spreading code. The discrete time signals I[n] and Q[n] compose an equivalent baseband discrete time complex signal r[n], where r[n]=I[n]+jQ[n]. The signal r[n] is transmitted to the FFT transform unit 28 and 48 which computes a N points FFT of the r[n], denoted by the symbol R[k], where N is the length of the pilot and data signal spreading code (or its integral multiple). The N points FFT computation is synchronized with received symbol time. The signal processing is then performed in the frequency domain. The signal R[k] is transmitted to the FFT-based CDMA RAKE receiver(s) 30 and 50.

FIG. 2A is a block diagram of a downlink RAKE receiver 30 with optional reserving main paths and data signal reconstruction. The downlink RAKE receiver 30 includes a pilot signal spreading code matched filter 62, a pilot interference cancellation unit 64, a data signal spreading code matched filter 66, a channel matched filter 68, and a decision unit 70.

The signal R[k] is transmitted from the FFT unit 28 to the pilot signal spreading code matched filter 62. The pilot signal spreading code matched filter 62 removes (despreads) the pilot signal spreading code and includes a multiplier 72, a complex conjugate unit 74, and a unit 76 for storing the FFT of the pilot signal spreading code c.sub.p [n]. The FFT of c.sub.p [n] is C.sub.p [k], which can be stored in a ROM within the unit 76. The complex conjugate unit 74 computes the complex conjugate of the C.sub.p [k] to produce C.sub.p *[k]. The product of R[k] and C.sub.p *[k] produced by the multiplier 72 removes the pilot signal spreading code to produce R[k]C.sub.p *[k].

The R[k] is also transmitted to the pilot interference cancellation unit 64. The pilot interference cancellation unit 64 includes a pilot signal estimation unit 80 and an adder 82. The pilot signal estimation unit 80 estimates the pilot interference signal P[k] and is described in more detail in FIGS. 4A and 4B below. The adder 82 subtracts P[k] from the R[k] signal to cancel the pilot interference to produce R1[k]. A data signal reconstruction component 84 is an optional component that is coupled to the pilot estimation unit 80 for removing data signal effect and is described in more detail below in FIG. 5.

The data signal spreading code matched filter 66 operates similarly as the pilot signal spreading code matched filter 62 but removes (despreads) the data signal spreading code from R1[k] received from the pilot interference cancellation unit 64. The data signal spreading code matched filter 66 includes a multiplier 88, a complex conjugate unit 90, and a storage unit 92. The storage unit 92 stores the FFT of the data signal spreading code c.sub.d [n] which is C.sub.d [k] and it can be stored in ROM. The complex conjugate unit 90 computes the complex conjugate of the C.sub.d [k] to produce C.sub.d *[k]. The multiplier 88 multiplies the R1[k] with C.sub.d *[k] to produce R1[k]C.sub.d *[k].

The channel matched filter 68 combines the received data signal power from different paths before a decision is made. The channel matched filter 68 contains a channel frequency response estimation unit 94, a multiplier 96, and a block sum unit 98. The channel frequency response estimation unit 94 includes a complex conjugate unit 100, a delay unit 102, a block-by-block average unit 104, and an optional reserve main paths unit 106. The signal R[k]C.sub.p *[k] from the pilot signal spreading code matched filter 62 is transmitted to the channel frequency response estimation unit 94. The R[k]C.sub.p [k] is delayed at the delay unit 102 for one spreading code period. Then, average unit 104 computes a weighted block-by-block average of the delayed version of the R[k]C.sub.p *[k] to generate a channel frequency response estimation H[k], where the block period is one spreading code period. The complex conjugate unit 100 computes the complex conjugate of the H[k] to produce H*[k]. The multiplier 96 multiplies the R1[k]C.sub.d *[k] from the data signal spreading code matched filter 66 with H*[k] to produce R1[k]C.sub.d *[k]H*[k]. The block sum unit 98 computes the total sum of the R1[k]C.sub.d *[k]H*[k] within one spreading code period, which is denoted by ##EQU1##

The decision unit 70 receives ##EQU2## from the channel matched filter 68 and makes a data decision. If the data modulation of the data signal is BPSK, the decision unit determines whether the real part of the ##EQU3## is greater or less than 0. If the data modulation of the data signal is QPSK, the decision unit determines whether the real part of the ##EQU4## and the imaginary part of the ##EQU5## are greater or less than 0.

FIG. 2B is a block diagram of an alternate downlink RAKE receiver 110 with optional reserving main paths. In this embodiment, pilot signal estimation unit 80 is also coupled to multiplier 72. When pilot signal estimation unit 80 is coupled in this manner the delay unit 102 and the average unit 104 are no longer necessary in the channel frequency response estimation unit 94, because, as will be described later, the pilot signal estimation unit 80 already includes these components.

FIG. 3A is a block diagram of an uplink RAKE receiver with optional reserve main paths. The uplink RAKE receiver, as shown in FIG. 1B, can detect L number of users at the same time. Each user has the same type of the uplink CDMA RAKE receiver 50. That is, the L users share the same units from the antenna 42 to the FFT units 48, and then R[k] is transmitted to the L uplink CDMA receivers 50 for the L users detection. The uplink receiver 50 is implemented using a multi-stage parallel interference cancellation method to increase the system capacity. The multi-stage interference cancellation method is an iterative process. At each stage, the reconstructed interference signals generated by the uplink receivers 50 of other users at the previous stage of operation is subtracted from R[k] for each user. Then, the signal remaining after interference cancellation is used for data detection. After data detection, the reconstructed interference signal is generated at each users' uplink receiver and provided to the other users' uplink receivers for the next stage detection process.

As shown in FIG. 3A, the uplink receiver 50 includes a multi-user interference cancellation unit 110, a data signal spreading code matched filter 112, a channel matched filter 114, a decision unit 116, an interference signal estimation unit 118, and a pilot signal spreading code matched filter 120. Without loss of generality, the following will describe the operation of the uplink receiver for a first user, user 1 or n=1.

The multi-user interference cancellation unit 110 removes the reconstructed interference signals from other users from the signal R[k] by using a first adder 122 and a second adder 124 to generate the signal remaining after multi-user interference cancellation which is denoted by R1.sub.i [k]. The index i of R1.sub.i [k] is the i'th stage of data detection of the present received symbol. The first adder 122 combines the reconstructed interference signals from other users to generate the total reconstructed interference signal represented by ##EQU6## , where In.sub.j,i-1 [k] is the reconstructed interference signal generated by the uplink receiver of the j'th user at the i-1 stage of data detection. Furthermore, all of the reconstructed interference signals (In.sub.j, 0 [k]=0, j=1 . . . L) are set to 0 at the first stage of data decision (i=1). After gathering the reconstructed interference signals generated from the uplink receivers of other users, the second adder 124 subtracts the signal ##EQU7## from the signal R[k] to produce a signal R1.sub.i [k]. The signal R1.sub.i [k] is represented by the following equation: ##EQU8##

At the i'th stage of data detection of the present received symbol, the R1.sub.i [k] is transmitted through the data signal spreading code matched filter 112, and a channel matched filter 114. The data signal spreading code matched filter 112 removes (despreads) the data signal spreading code to produce ##EQU9## The data signal spreading code matched filter 112 is described in more detail above with respect to FIG. 2A.

The signal ##EQU10## from the data signal spreading code matched filter 112 is transmitted to the channel matched filter 114. When i=M, M is total number of stages, switch 126 closes, thereby allowing the pilot signal spreading code matched filter 120 to produce ##EQU11## and send it to the channel matched filter 114. The channel matched filter 114 includes a channel frequency response estimation unit 128, a multiplier 130, and a block sum unit 132. The channel frequency response estimation unit 128 includes a delay unit 136, an average unit 138, an optional reserve main paths unit 140, and a complex conjugate unit 142. H[k], produced by average unit 138 from the output of the delay unit 136, is sent to the complex conjugate unit 142. When i=M, the output of the pilot signal spreading code matched filter 120 is sent to the delay unit 136. The delay unit 136 delays its received signal until the beginning of the next symbol processing. The complex conjugate unit 142 computes the complex conjugate of the H[k] to produce H*[k]. The multiplier 130 multiplies the ##EQU12## from the data signal spreading code matched filter 112 with H*[k] from the complex conjugate unit 142 to produce the ##EQU13## Again, the block sum unit 132 computes the total sum of the ##EQU14## within one spreading code period, which is denoted by the symbol ##EQU15##

The decision unit 116 receives the block sum from the channel matched filter 114 and makes a data decision. If the data modulation of the data signal is BPSK, the decision unit determines whether the real part of the ##EQU16## is greater or less than 0 and generates a tentative decision d.sub.1,i of the present received symbol. If the data modulation of the data signal is QPSK, the decision unit determines whether the real part of the ##EQU17## and the imaginary part of the ##EQU18## is greater or less than 0 and generates a tentative decision d.sub.1,i. In the QPSK example, the generated tentative decision d.sub.1,i is a complex signal.

At the i'th stage of data detection for the present received symbol, the signal H [k] from the channel frequency response estimation unit 128 and the tentative decision d.sub.1,i of the present symbol from the decision unit 116 are transmitted to an interference signal estimation unit 118 for producing the reconstructed interference signal. The interference signal estimation unit 118 is used to reconstruct the transmitted symbol of the user, which is the multiple access interference of user 1 to be used by the other users' uplink receivers. The reconstructed interference signal of user 1 is transmitted to the other users' uplink receivers for multi-user interference cancellation. The interference signal estimation unit 118 includes a first multiplier 150, an adder 152, a second multiplier 154, and a normalization unit 155. The first multiplier 150 multiplies the FFT of the data signal spreading code C.sub.d [k] with the signal d.sub.1,i from the decision unit 116 to generate the data modulated data signal spreading code C.sub.d [k]d.sub.1,i. The adder 152 combines the FFT of the pilot signal spreading code with the signal C.sub.d [k]d.sub.1,i to generate C.sub.p [k]+C.sub.d [k].multidot.d.sub.1,i. The second multiplier 154 multiplies the C.sub.p [k]+C.sub.d [k].multidot.d.sub.1,i from the adder 152 with the H[k] from the channel frequency response estimation unit 128 to produce (C.sub.p [k]+C.sub.d [k].multidot.d.sub.1,i).about.H[k]. Since the channel frequency response estimation H[k] is weighted by the .vertline.C.sub.p [k].vertline..sup.2 due to the pilot signal spreading code matched filter 120, the normalization unit 155 is used to remove the weighting factor .vertline.C.sub.p [k].vertline..sup.2. The normalization unit 155 normalizes the output of the second multiplier 154 by .vertline.C.sub.p [k].vertline..sup.2 to generate the reconstructed interference signal denoted by the In.sub.1,i [k]. The signal In.sub.1,i [k] is represented by the following equation:

The uplink RAKE receiver 50 does M stages of data detection for each received symbol, that is, after M times of data detection, the uplink receiver will process the next symbol for data detection in a similar way.

At the M'th stage of data detection of the present received symbol (the last time of data detection of the present received symbol), the switch 126 closes and the signal R1.sub.M [k] is transmitted from the multi-user interference cancellation unit 110 to the pilot signal spreading code matched filter 120. The pilot signal spreading code matched filter 120 removes (despreads) the pilot signal spreading code and produces a ##EQU19## signal. The signal ##EQU20## is transmitted to the channel frequency response estimation unit 128 within the channel matched filter 114 for channel frequency response estimation.

FIG. 3B is a non-user specific data detection flowgraph (use user n as an example) of the uplink receiver shown in FIG. 3A. At step 160, for each received symbol, the uplink receiver 50 initializes i=0 and sets all of the reconstructed interference signals from other users to 0 (In.sub.j,0 [k]=0, j=1 . . . L). At step 162, the index i of data detection stage is incremented by one, denoted by the equation i=i+1. At step 164, the receiver 50 gets or receives the reconstructed interference signals In.sub.j,i-1 [k] generated from other users. At the first stage of data detection, the uplink receiver of user n does not receive the reconstructed interference signal (In.sub.j,0 [k]) from other users. The uplink receiver of user n sets the reconstructed interference signal (In.sub.j,0 [k]) to 0 at the first stage of data detection of the present received symbol. After the first stage of data detection of the present received symbol, the uplink receiver of the user n receives the reconstructed interference signals from other users which are generated at the previous stage of data detection. At step 166, the reconstructed interference signals from other users are subtracted from R[k] as stated in the following equation: ##EQU21## At step 168, R1.sub.i [k] is multiplied with the complex conjugate of the FFT of data signal spreading code to generate ##EQU22## At step 170, the output of the step 168 is matched to the channel frequency response estimation to generate ##EQU23## At step 172, a tentative decision d.sub.n,i is made for the present received symbol. If at decision step 174 the i'th data detection stage equals the last stage of data detection M, the tentative decision is d.sub.n,M. Otherwise, at step 176, the reconstructed interference signal of user n at the i'th stage of data detection is computed and then sent to the uplink receivers of the other users, where the reconstructed interference signal of the users n is represented by the equation In.sub.n,i [k]=(C.sub.p [k]+C.sub.d [k].multidot.d.sub.n,i).multidot.H[k]/.vertline.C.sub.p [k].sup.2. The data detection process is then looped back to the step 162 where the index i is incremented by one for the next stage data detection of the present received symbol.

FIG. 4A illustrates a pilot signal estimation unit 180 for low speed users. The received R[k] signal is delayed at a delay unit 182 for one spreading code period. Since the pilot signal is fixed, that is, it is unmodulated, a simple block-by-block average of the R[k] estimates the pilot interference. An average unit 184 then computes a weighted block-by-block average of the R[k] signal to generate the pilot interference P[k], where the block period is one spreading code period. This setup is effective in a low speed system, because the averaging can be performed over a long period of time.

FIG. 4B illustrates a pilot signal estimation unit 186 for high speed users. A summer 188 is included between the delay unit and the average unit. The summer 188 removes data signal component d.multidot.C.sub.d [k].multidot.H[k]/.vertline.C.sub.p [k].sup.2 from the delayed signal. The data signal component is reconstructed by the data signal reconstruction unit 84 described in FIG. 5 below. Since the average length of the signal is not large enough with respect to high speed users, the reconstructed data signal component must be removed for pilot signal estimation to avoid "data signal cancellation."

As shown in FIG. 5, the data signal reconstruction unit 84 includes a first and second multiplier 190 and 192 and a normalization unit 194 for reconstructing the data signal using a decision feedback method. The first multiplier 190 multiplies the data detection output of the previous received symbol d with the FFT of the data signal spreading code C.sub.d [k] to generate d.multidot.C.sub.d [k]. The second multiplier 192 multiplies d.multidot.C.sub.d [k] with the channel frequency response estimation H[k] that is then normalized by normalization unit 194 to generate the reconstructed data signal component of the previous received symbol, d.multidot.C.sub.d [k].multidot.H[k]/.vertline.C.sub.p [k].vertline..sup.2.

As shown in FIG. 6, the reserve main paths unit 106 includes an Inverse Fast Fourier Transform (IFFT) unit 200, a search peak unit 202, a selecting unit 204, and a FFT unit 206. The IFFT unit 200 receives the signal from the pilot signal spreading code matched filter 62 or the average unit 104 or the average unit 138, see FIGS. 2B, 2A and 3A, respectively, and computes the IFFT, thereby generating the channel impulse response estimation denoted by the symbol h[n]. The search peak unit 202 receives h[n] from the IFFT unit 200 and searches or determines the peak amplitude of h[n]. Then, the select path unit 204 selects signals with high amplitudes from h[n] and discards the remaining signals that have low amplitudes by setting the low amplitude signals to zero. For example, a high amplitude signal is typically a signal with an amplitude which is greater than 10 dB below the peak value of h[n] while a low amplitude signal is typically a signal with a relatively smaller amplitude. The selection of high amplitude signals generates a modified channel impulse response estimation signal h1[n] containing only signals on the main paths. The FFT unit 206 calculates the FFT transformation of the h1[n] from the selecting unit 204 to generate a channel frequency response estimation H[k]. The H[k] provides the coefficient of the channel matched filter 68 and 114. The reserve main path unit 106 is not required, but does provide more reliable channel estimation.

Although specific embodiments have been illustrated and described, it will be obvious to those skilled in the art that various modifications may be made without departing from the spirit which is intended to be limited solely by the appended claims. For example, although FIGS. 3A, 3B, and 3C described the uplink receiver of the user 1, the similar process can be repeated for the data detection of the other user. Therefore, while the preferred embodiment of the invention has been illustrated and described, it will be appreciated that various changes can be made therein without departing from the spirit and scope of the invention.