WO2013172808A1

WO2013172808A1 - Decision-directed nlms equalizer by despreading with a parent code of a group of active codes.

Info

Publication number: WO2013172808A1
Application number: PCT/US2007/073401
Authority: WO
Inventors: Ralph Rudolf VESELI; Tamela J. CRAFT-MOLARAHIMA
Original assignee: Nxp B.V.
Priority date: 2006-07-12
Filing date: 2007-07-12
Publication date: 2013-11-21

Abstract

A method for equalization in a code-division multiple access radio receiver adapts a pilot-aided equalizer more than once every primary common pilot channel (PCPICH) period for tracking fast time -varying channels. It then uses partial-despreading over a PCPICH period with a parent code to obtain a plurality of consecutive y- filter outputs during each PCPICH period, wherein an interference subspace region may also include active codes.

Description

DECISION-DIRECTED NLMS EQUALIZER BY DESPREADING WITH A PARENT CODE OF A

GROUP OF ACTIVE CODES

FIELD OF THE INVENTION

This invention relates to radio communications systems, and more particularly to overcoming multi-path effects in direct-sequence code-division multiple access receivers. The received coding chip waveform, distorted by the multi-path channel, is equalized prior to de-spreading, so the orthogonality of the signal from the basestation can be restored at the code chip-level.

BACKGROUND

Radio communications systems can use many types of modulation. One of the earliest kinds invented was amplitude modulation (AM). But AM radio broadcasts were easily interfered with by the many sources of electrical noise present. So frequency modulation (FM) became popular as a way to receive much clearer and interference-free broadcasts. Modern cellular telephones and wireless local area networks now use direct- sequence code-division multiple access (CDMA) modulation techniques that allow many users to share the same frequency channels. More sophisticated schemes use several of the radio channels in a band, e.g., eight frequency channels, in various orthogonal techniques to frequency hop among individual channels to avoid sources of interference. Some of these schemes can survive one or more of the channels becoming useless, and still lose none of the data.

But one of the more difficult problems has been the ill-effects of multipath distortion. Transmitted signals can take the direct path and arrive at the receiver first. But signals that have been reflected and taken a longer path can also arrive, albeit later, at the receiver on the right frequencies and with the right spreading codes. All the channels being used in an orthogonal scheme can be involved in this multipath problem, and the individual channels can vary amongst themselves in strength and phase because the multipath channel time delays can be highly frequency sensitive and changeable. In the forward link of DS-CDMA systems, such multipath upsets the spreading codes orthogonality, therefore causing multi-user interference (MUI). Ideal non-adaptive zero-forcing (ZF) and MMSE chip equalizer receivers can restore the orthogonality and suppress the MUI in DS-CDMA systems employing aperiodic scrambling codes, but these receivers cannot deal with rapidly fading multipath channels. No continuous training chip sequence is available to allow receiver chip equalizers to track fast enough. Some have proposed pilot-aided adaptive fractionally-spaced chip equalizer receivers. These use the continuous pilot signal in third generation cellular and LEO satellite communication systems to continuously update at the symbol rate. E.g., using a simple normalized least mean square (NLMS) or a more advanced recursive least squares (RLS) adaptive scheme. Such a receiver supposedly out-performs RAKE receivers with perfect channel knowledge over a wide range of normalized Doppler spreads, and its complexity is independent of the number of users.

Adaptive interference suppression has been successfully used for base-to-mobile downlinks in direct sequence (DS) based cellular communication systems. Each base station transmits all the signals destined for the different mobiles. Orthogonal spreading sequences are used for different mobiles to avoid intra-cell interference. However, the orthogonality is, in practice, disturbed by having to use channels plagued by multi-path effects.

In related applications, channel-equalizing terminal receivers are being used for wideband code-division multiple-access (WCDMA) downlinks. The channel equalizers provide the multiple access interference (MAI) suppression needed to maintain adequate receiver performance with large numbers of active users. Equalizers are now being universally employed in CDMA mobile receivers to compensate for multipath- effects. Minimum mean square error (MMSE) focused equalizers focus on both suppressing noise and interference at the same time, and so they perform better than signal-to-noise ratio (SNR) maximization- oriented RAKE receivers. The RAKE receiver uses a filter matched to the spreading operations, pulse shape filtering and channel filtering. Such maximizes the signal-to-interference-plus-noise ratio (SINR) at its output if the interference-plus-noise is white, e.g., when user-dependent aperiodic scrambling is used. This could be used for third generation (3G) system uplinks. But it's not very useful in the synchronous downlinks with cell-dependent scrambling, orthogonal codes and a common channel.

Pilot-aided equalizer designs for CDMA systems are more complex than for global system for mobile communications (GSM) and other time-division multiple access (TDMA) systems. TDMA systems have a common pilot signal which is time-multiplexed with the payload data. User data cannot interfere with the pilot signals. Only additive white Gaussian noise (AWGN) interferes.

In universal mobile telecommunications system frequency division duplex (UMTS-FDD) and other CDMA downlinks, the primary common pilot channel (PCPICH) data is code-multiplexed with all the other existing user and control channels. So they are also subject to the same interferences, it cannot be used for equalizer weight training at the chip rate. Some considered despreading the received signal with the pilot tone's channelization code, to suppress most of the interference over the PCPICH signal. Others extended this to fractionally spaced implementations. A basic schematic of this "conventional" scheme is shown in Fig. 3, and is quite effective. The long despreading operation adapts only once every 256-chips, e.g., one PCPICH period. This technique is slow to converge and track the ever-changing channels.

Conventional 3G orthogonal variable spreading factor (OVSF) code generator blocks generate OVSF code from a set of orthogonal codes. Such are primarily used to preserve the orthogonality between different channels in a communication system. OVSF codes are defined as the rows of an N-by-N matrix, CN, which is defined recursively as follows. First, define CI = [1]. Next, assume that CN is defined and let CN(k) denote the kth row of CN. Define C2N by,

C_N (Q) c_N(m

-c_Nm

C_N(D C_N (X)

C_N(N- 1) C_N(N -1)

C_N(N- D -€_N(N- 1)

CN is only defined for N a power of 2. It follows by induction that the rows of CN are orthogonal. The OVSF codes can also be defined recursively by a tree structure, as shown in Figs. 1 and 2.

Conventional wisdom says the symbol level equalization can only be done at PCPICH symbol rate by despreading with the full PCPICH channelization code. Although the PCPICH chip sequence is a constant sequence, re-adapting the equalization at higher rates seems possible by despreading a part of the PCPICH channelization code with a parent code. But the primary common control physical channel (PCCPCH) and all the child codes of that parent code also get despread.

If [C] is a code length 2r at depth r in the tree, where the root has depth 0, the two branches leading out of C are labeled by the sequences [C, C] and [C, -C], which have length 2r+l . The codes at depth r in the tree are the rows of the matrix CN, where N = 2r. Two OVSF codes are orthogonal if and only if neither code lies on the path from the other code to the root. Since codes assigned to different users in the same cell must be orthogonal, this restricts the number of available codes for a given cell. If a code C4, 1 in a tree is assigned to a user, the codes C1,0, C2,0, C8,2, C8,3, and so on, cannot be assigned to any other user in the same cell.

Fig. 3 is typical of a conventional pilot-aided symbol level LMS module 300 for equalization. Such includes a tap delay line 302, a descrambler and despreader 304 with the PCPICH channelization code

Cch,256,0 on each branch originating from the associated tap, and a filter 306 of length equal to the number of taps N in a conventional LMS adaptation scheme. An adaptive filter input xn serves as an input regressor, a filter output yn is the PCPICH symbol estimate, "d" is the desired signal, e.g., the correct PCPICH symbol, and "e" is the error signal. Adaptive filter is the conventional NLMS algorithm that trains via xn, yn and "d". Fig. 3 shows a T-spaced implementation, wherein a fractionally spaced scheme is well- known to artisans.

SUMMARY OF THE INVENTION In an example embodiment, a method for equalization in a code-division multiple access radio receiver adapts a pilot-aided equalizer more than once every primary common pilot channel (PCPICH) symbol period for tracking fast time-varying channels. It then uses partial-despreading over a PCPICH period with a parent code to obtain a plurality of consecutive y-filter outputs during each PCPICH period, wherein an interference subspace region may also include active codes.

An advantage of the present invention is a receiver is provided which avoids most of the interference generated due to the existence of PCCPCH, and other child codes, and enables training the equalizer weights at higher rates.

Another advantage of the present invention is a receiver is provided which is applicable to any WCDMA system where multi-access is realized via orthogonal variable spreading factor (OVSF) codes like frequency division duplex (FDD) downlinks.

A still further advantage of the present invention is a receiver is provided for WCDMA systems, and more particularly to improving high speed downline packet access (HSDPA) service quality with decision directed NLMS equalizers in receivers with OVSF multi-access codes and FDD downlinks.

The above summary of the present invention is not intended to represent each disclosed embodiment, or every aspect, of the present invention. Other aspects and example embodiments are provided in the figures and the detailed description that follows.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention may be more completely understood in consideration of the following detailed description of various embodiments of the invention in connection with the accompanying drawings, in which:

FIG. 1 is a diagram representing an OVSF code hierarch tree;

FIG. 2 is a diagram representing a OVSF subtree rooted from Cch,16,0 branch;

FIG. 3 is a prior art diagram of a conventional symbol level least mean squares (LMS) module in a wideband code-division multiple access (WCDMA) receiver for downlink equalization, SF=256, d= CPICH symbol;

FIG. 4 is a diagram representing a decision directed pseudo-symbol level LMS data filtering and symbol estimation module embodiment of the present invention;

FIG. 5 is a diagram representing a decision directed pseudo-symbol level LMS filter adaptation module embodiment of the present invention;

While the invention is amenable to various modifications and alternative forms, specifics thereof have been shown by way of example in the drawings and will be described in detail. It should be understood, however, that the intention is not to limit the invention to the particular embodiments described. On the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.

DETAILED DESCRIPTION

Fig. 4 represents a decision directed pseudo-symbol level LMS data filtering and symbol estimation module embodiment of the present invention, and is referred to herein by the general reference numeral 400. Such is useful in WCDMA receivers.

Conventional pilot-aided equalizers adapt only once every PCPICH period, e.g., every 256 chips, and so they cannot track fast time-varying channels. Module 400 does a partial-despreading instead of the full despreading over the PCPICH. The SF=256 spreading factor of Fig. 3 is replaced with one of SF=16. The parent code of Cch, 16,0, is used to obtain sixteen consecutive y- filter outputs during each 256-chip long PCPICH period. However, such de-spreading will not be restricted to only the PCPICH. The PCCPCH, and all the other active child codes of the parent code Cch, 16,0, will also de-spread, as represented in Fig. 2. The code subspace shown by interference subspace region 202 might contain the possible active codes. Mathematically, the de-spread signal will be:

S partial ^{~ S} PCPICH _ partial ^ ^S PCCPCH _ partial ^~*^{~ S}i nartial ⁿ partial

where, : partially despread signal;

PCPICH _ partial .

: partially despread PCPICH symbol (the desired response);

PCCPCH _ partial .

: partially despread PCCPCH symbol (the interference);

partial

: partially spread symbols from other active codes in the interference subspace 202; and

n . : sum of interference plus noise due to multipath, intercell interference and thermal noise. _u , , ^SPCPICH partial ^~*^{~ W} · , ■ _A . _e . . _™ ^S PCPICH partial

But only - is desired at the niter output. The - serves as the

^ PCCPCH partial

response "d", and "n" serves as the error signal "e". Unfortunately, - ^y and

_ partial

amount to undesired interference. ς, - _l _ partial

PCCPCH partial

Embodiments of the present invention re-characterize - and ¹ . Instead of being interfering signals "n", these become desired signals "d". This transformation is done by finding the

Ys i _ partial

sum of the estimates PCCPCH _ partial ^ ^j ^ _an(j ^_e p_artj_aj PCPICH symbol

^S PCPICH partial

- as " , as implemented m Fig. 4.

An algorithmic process is run on consecutive packets of data each relevant to one CPICH transmitted symbol. A processing window of 256-chips is moved like a sliding window for the next packet processing. Adapted filter weights from the previous packet are used for data filtering. The filter weights from the conventional RAKE receiver are used for the initial packet filter weight assignments.

Referring now to Fig. 4, a received chip rate signal (rn) is passed through a FIR filter 402 whose weights are obtained in the previous PCPICH symbol period. A mixer 404 and correlator 406 are used to descramble and despread with the sixteen child codes of Cch, 16,0 at level 256, e.g., {Cch,256,0,

Cch,256, 1 , ... , Cch,256, 15} . Multiple despreading operations begin by despreading with the parent code Cch,16,0. Then second-step multiple correlations are done jointly, e.g., using Fast Walsh Hadamard Transformation (FWHT) 410 of size sixteen (FWHT- 16). If despreading is done independently with the sixteen codes, {Cch,256,0, Cch,256,l, Cch,256,15}, then the complexity is 256* 16 = 4096 units. The complexity of FWHT-N is Nlog2(N). Here, the complexity diminishes to 256+16*log2(16) = 320 units. After despreading, a first correlator output corresponds to a PCPICH linear symbol estimate , a second correlator output is a PCCPCH symbol estimate ² (appears at the 9th place in the FWHT output which is just an output order difference), the remaining are the pseudo-symbol estimates represented from any active code in the interference subspace 202. A serial to parallel converter (S/P) 408 follows.

Once sixteen linear (pseudo-) symbol estimates are obtained, their quality is refined by post processing. The symbol value (1+j) is already known for the pilot element PCPICH, but its real/imaginary amplitude

2

(Al) is not known. Such amplitude and the symbol error variance ^σ are estimated in a control block 412. Once A 1 is determined, a final estimate for PCPICH, ^{1 1 J} ' , is out ut. Similarl , A2 can be

estimated by computing from what is already known about and ² . With A2 in hand, a hard decision block 414 uses a linear PCCPCH symbol estimate ² to obtain a refined estimate

² . Each of the remaining codes estimate the received powers as ^ ^ and a linear minimum mean square error (LMMSE) weighting module 418 outputs ^{+ σ} )

Referring now to Fig. 5, a filter adaptation module 500 includes a FWHT 502, a parallel-to-serial converter

(P/S) 504, and a NLMS equalizer 506. The final estimates, e.g., { ^Sl , *² , ^S }, (Fig. 4) are connected to the corresponding FWHT 502 input ports, where they are re-spread to level sixteen. Such produces sixteen consecutive desired-signal estimates d in each PCPICH symbol period. These sixteen values are used to adapt the NLMS equalizer weights sixteen times in the same packet interval. Final equalizer weights are

f

used for the next packet filter ^{J 11+1} .

All decision directed schemes can suffer from misconvergence. Such occurs when the equalizer locks to a rotated constellation (state) and cannot recover. To avoid misconvergence, embodiments of the present invention make use of the PCPICH signal. Such is a 45-degree vector at both chip and symbol level after descrambling. A Super-PCPICH- Symbol is obtained every ten, or carefully selected number of PCPICH symbol periods. This Doppler-spread and noise dependent design parameter, can be taken less or more often. The f filter weights are de -rotated by an angle , which is equal to the difference between the phase of the estimated Super-PCPICH Symbol and 45-degrees.

Equalizer method embodiments of the present invention can be implemented in both software and hardware. In both cases, an input buffering of at least 256-chips is necessary, and the total processing of the two stages 400 and 500 must be completed in one PCPICH symbol period. Therefore, chip- level processing should progress at least twice the chip rate. The hardware clock or software processor should work at 2*3.84Mhz = 7.68Mhz, or faster.

Such treatment used the pseudo-symbol of Cch, 16,0 of the desired signal for the NLMS algorithm for doing decision direction through the (pseudo)-symbols of sixteen codes at SF=256. However it is only an example. It can be easily generalized for parent codes at other spreading factors. Higher spreading factors such as 32, 64, or 128, e.g., despreading with Cch,32,0, Cch,64,0 or Cch, 128,0, would be more suitable for low SNR and low speed conditions. Whereas lower spreading factors such as 2,4,8 would be more suitable for high SNR and high speed conditions. Window sizes smaller than 256-chips can be used in high speed packet data access (HSDPA) service. Knowing the codes can help improve making reliable symbol estimates via hard decisions. It may be appropriate not to exploit the code space under parent code Cch, 16,0. For example, if there are eight existing HSDPA codes occupying codes {Cch,16,8, Cch,16,9, Cch,16, 10, Cch, 16,l 1, Cch,16,12,

Cch, 16, 13, Cch, 16, 14, Cch, 16,15}, they have a common parent code Cch,2, 1. Table I lists the steps in method.

TABLE-I (1) The window size can be set to sixteen;

(2) Despreading is done with the eight HSDPA codes {Cch, 16,8, Cch, 16,9, Cch, 16, 10, Cch, 16, 1 1, Cch, 16, 12, Cch, 16, 13, Cch, 16, 14, Cch, 16, 15} . Despreading is done first with the parent code Cch,2,l as the first common part, and then use FWHT-8 over eight consecutive values, similar to the despreading correlator bank in Fig. 4;

(3) Hard decisions are made over the eight HSDPA soft symbols; and

(4) The eight hard decision values are fed to FWHT-8 input and the desired signal "d" is obtained every two chips. In Fig. 3, the corresponding SF for the adaptive filtering part is two, and the despreading code is Cch,2, l .

Since the window size in such a scheme will be much less than 256, in this case two, it will be much more robust than the original scheme for highly varying channel conditions because the filter weight latency is only 2-chips for the next window.

In another example, four existing HSDPA codes, {Cch, 16,4, Cch, 16,5, Cch, 16,6, Cch, 16,7}, will have a common parent code Cch,4, 1. Table-II lists the steps in a next method. TABLE-II

(1) The window size is set to sixteen;

(2) Despreading is done with the four HSDPA codes;

(3) Hard decisions are made over the four HSDPA soft symbols; and

(4) The hard decision values are fed to a FWHT-4 input. The desired signal "d" is output every four chips. The corresponding SF for the adaptive filtering part in Fig. 3 will be four, and the despreading code will be Cch,4, l . A third example uses three existing HSDPA codes, {Cch, 16,4, Cch, 16,5, Cch,16,6 } . The fourth code in the previous example, Cch,16,7, is not used to carry HSDPA service. With this single difference, LMMSE weighting over the soft estimate of Cch, 16,7 can be employed.

An alternative embodiment of the present invention for HSDPA service uses the PCPICH and PCCPCH aided scheme described by Figs. 4 and 5, and switches to the HSDPA codes aided schemes once a certain equalization quality level is achieved. DD-NLMS and other schemes can be jumped between, depending on the knowledge of codes, the energy over the code domain, Ior/Ioc values, mobile speed, etc.

While the present invention has been described with reference to several particular example embodiments, those skilled in the art will recognize that many changes may be made thereto without departing from the spirit and scope of the present invention, which is set forth in the following claims.

1. A method for equalization in a code-division multiple access radio receiver, comprising:

adapting a pilot-aided equalizer more than once every primary common pilot channel (PCPICH) period for tracking fast time -varying channels;

partial-despreading over a PCPICH period using a parent code to obtain a plurality of consecutive y- filter outputs during each PCPICH period, wherein an interference subspace region may also include active codes, and a de-spread signal will be:

S partial ^{~ S} PCPICH _ partial ^ ^S PCCPCH _ partial ~*~ Y, ^Si nartial ⁿ

s partⁱa . p_artj_ajjy despread signal;

^ PCPICH Dartial

- : partially despread PCPICH symbol (the desired response);

^ PCCPCH nartial

- : partially despread primary common control physical channel (PCCPCH) symbol e);

: sum of interference plus noise due to multipath, intercell interference and thermal noise.

PCPICH partial

2. The method of Claim 1, wherein - ^y is a desired response "d", and "n" is an error g Ys i _ partial

signal "e", and PCCPCH _ partial _an(j ^j amount to undesired interference.

3. The method of Claim 1, further comprising:

Claims

_g Y^ S i _ partial

re-characterizing PCCPCH _ partial _an(j ^j as desired signals "d" by finding ξ Y i _ partial

the sum of the estimates PCCPCH _ partial ^ ^j ^ _an(j _mg p_artj_aj pcpjcpj sym ol

^S PCPICH _ partial ^

4. The method of Claim 1, further comprising:

running an algorithmic process on consecutive packets of data each relevant to one CPICH transmitted symbol, wherein a processing window of N-chips is moved like a sliding window for the next packet processing, and adapted filter weights from the previous packet are used for data filtering, and the filter weights from a conventional RAKE receiver are used for the initial packet filter weight assignments.

5. An equalizer for a CDMA radio, comprising:

a data filter (402) with an input for a received chip rate signal (rn), and for using weights obtained in a previous PCPICH symbol period;

means (404) to descramble and despread with a plurality of child codes for multiple despreading operations that begin by despreading with a parent code;

means for doing multiple correlations jointly with a Fast Walsh Hadamard Transformation (FWHT) (410) of size sixteen (FWHT-N), wherein the complexity of FWHT-N is Nlog2(N);

wherein, a first correlator output corresponds to a PCPICH linear symbol estimate ¹ , a second correlator output is a PCCPCH symbol estimate ^Sl , and the remaining are the pseudo-symbol estimates represented from any active code in an interference subspace.

6. The equalizer of Claim 5, further comprising:

means (400) for further refining the quality of a plurality of linear (pseudo-) symbol estimates, wherein a symbol value (1+j) is already known for a pilot element PCPICH, but its

real/imaginary amplitude (Al) is not known;

2

means for estimating (412) amplitude Al and the symbol error variance ^σ ; means for calculating a final estimate for PCPICH, ^{1 1 J} ' , and estimating A2 by

computing from what is already known about and ² , and then using A2 in a hard decision block (414) in a linear PCCPCH symbol estimate ⁵² to obtain a refined estimate ⁵² ; wherein, each of the remaining codes estimate the received powers as ^ = ( ² -^²)

2 and a linear minimum mean square error (LMMSE) weighting module 418 outputs ^ ' '

7. The equalizer of Claim 5, further comprising:

a filter adaptation module (500) that includes a FWHT (502) and a NLMS equalizer (506), such that final estimates, e.g., { ¹ , ² , ^Λ }, are connected to corresponding FWHT input ports, to be re- spread to produce consecutive desired-signal estimates d _{in ea}ch PCPICH symbol period, and such values are used to adapt NLMS equalizer weights an equal number of times in a same packet interval, and final

f

equalizer weights are used for the next packet filter ^{J 11+1} .