WO2007010331A1

WO2007010331A1 - Preamble detection

Info

Publication number: WO2007010331A1
Application number: PCT/IB2005/052399
Authority: WO
Inventors: Henrik Sahlin; Tobjörn KARLSSON
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2005-07-19
Filing date: 2005-07-19
Publication date: 2007-01-25

Abstract

Method of detecting the offset position (k <= 1280) of a repetitive code stream sequence of a first string length (Npre), the offset position laying within a maximum delay value (1280), such as a WCDMA (Wideband Code Division Multiplex Access) uplink preamble. the method comprising the steps of a) continuously receiving analog to digital converting and sampling a stream of incoming signals (r(i)) at a sampling rate equal to at least the chip rate, b) splitting into segments and de-spreading the incoming signals, calculating a sum function value (S(k)) for the given window displacement (k) involving multiplying the incoming signal r(i) with a conjugate chip sequence c*(i),c) forming a sum function (Sos (k)I2 ), where (at periodic intervals) a fictitious value is interpolated between consecutive sum function values, and d) resolving the window displacement (k) corresponding to the maximum value of the sum function (ISos (k)I2 ). A method for a double dwell detector is also disclosed.

Description

Preamble Detection

Field of the invention

The present invention relates to detections of chip sequences in telecommunication systems and in the field of instrumentation and measurements. More specifically, the invention may be used for random access preamble detection for WCDMA uplink traffic. The invention may be applied to a single dwell detector architecture and is particular advantageous when applied to a double dwell detector architecture.

Background of the invention

In code division multiplex access systems (CDMA), a scrambling code is assigned to a given radio cell in order to identify that particular cell from other cells. A set of orthogonal signature codes, e.g. Hadamard codes, is used to distinguish the data streams of individual mobile stations. These signatures are orthogonal to each other and serve as a temporary mobile station identity.

One task of the base station is firstly to identify the mobile station by the given signature code, transmit an acknowledge message to it, and finally decode a payload message transmitted by the mobile station. This payload message is transmitted by the mobile station a short interval after receiving the acknowledge message from the base station.

A common pilot channel is transmitted from the base station to all mobile stations. The pilot channel serves as phase or timing reference for all mobile stations. When the base station receives signals from the mobile stations, these signals will be delayed relative to the pilot channel due to the distance between mobile station and base station. The delay can be measured as a round trip time, i.e. the time for a signal to travel from base station to the mobile station and back again to the base station.

An exact timing is necessary for correctly decoding the Hadamard coded payload signals from the individual base station entity. However, since neither the location nor the timing of the signals from a given mobile station entity can be known in advance, the base station must resolve the prevalent timing properties of the signals relating to a given mobile station. The preamble is used for this purpose. In WCDMA, the preamble is a signal, which comprises a sequence of 4096 chips. The preamble consists of a signature code as mentioned above, and a scrambling code. These preambles are constructed as

where

• N_pre = 4096 is the number of chips in the preamble,

• S_r__{pre n} (k) is a scrambling code

• C _s (k) is a signature (out of 16 possible) and

• is a PAPR (PEAK TO AVERAGE POWER RATIO REDUCTION) (Peak

to Average Power Ratio) modulation.

Here, the preamble signature is constructed as 256 repetitions of a 16 symbols long Ha- damard code. The notations in equation (I) above are according to 3GPP TS 25.213 "Spreading and Modulation (FDD)" Technical Specification Group Access Network, 3rd Generation Partnership Project.

In fig. 1 , the preamble and the possible unknown offset or delay at the base station has been indicated.

In WCDMA, a subset of signatures is available and each mobile station selects one signature randomly from this subset. There are 16 unique signatures available.

The scrambling codes constitutes Gold sequences, see [3GPP TS 25.213 "Spreading and Modulation (FDD)" Technical Specification Group Access Network, 3rd Generation Partnership Project], which has the unique property of being discernable by a sum function independently of the timing properties of the sequence and interference form other Gold sequences. The workings of the decoding of the Gold coding shall be briefly explained in the following. Before the preamble sequence is transmitted over the air interface by a mobile station, it is filtered though a transmitter filter. Also on the receiver side, in the base station, a receiver filter is applied on the received signal. The product of these two filters has a so- called raised cosine impulse response, see figure 3.

For a cell with a radius 50 km, which should resemble the maximum coverage of a large cell, the maximum round trip delay is L = 2- 50 km- Fj c m 1280 chips where c=3-10⁸m/s is the speed of light and F_s = 3.84-10⁶ chips/second denotes a chip spaced sampling rate.

A given received signal r(J) (corresponding to a length of 4096 chips plus 1280 chips for maximum round trip delay) from a mobile station entity is therefore sampled in the receiver of the base station. Hence, this received signal is then assumed to comprise the preamble transmitted by the mobile station.

In order to detect the offset k, the received signals are made subject to a sum function whereby a window of 4096 chips is selected from the received signal r(ϊ) ; each sample is multiplied with the known preamble sequence and finally summed. This summation may be performed over the whole sequence of 4096 chips, or over shorter segments e.g. 2048 or 1024 chips. For the case of a summation over all 4096 samples, the result can be expressed as

(H)

where c * (4096 - 1 - /) is the sequence in equation (I) but time reversed and conjugated and where the received signal is assumed to be chip spaced sampled, i.e. one sample for each chip.

In fig. 2, the prior art detection principle employing the equation Il above has been visu- alized: For each window displacement k, the sum function Il is calculated over a preamble length, which in the present context is 4096 chips. This may be done continuously in the detector. In order to detect the transmitted radio symbols, more than one sample per chip may be read and stored. The window of 4096 chips selected from the received signal, is shifted in such a way that different round trip delays are selected. In equation (II) this shift of window is given by the parameter k . In order to cover a cell of 50 km, equation (II) should be evaluated for values of k between 0 and 1280-1 .

The summation in equation (II) is a coherent accumulation. Before detecting preambles, the result from equation (II) is absolute squared, i.e.

is calculated.

The sum function of Gold coded scrambling codes as a function of window displacement k has the specific distribution as indicated in fig. 3, where the impact of the raised cosine filter is included and where the channel from transmitter to receiver only consists of one path.

The maximum S_1113x corresponds unambiguously to the window displacement k corresponding to a matching position of the preamble in the sampled stream. In this manner, detecting the maximum of the sum function provides the reference point - or offset value k - of the sampled stream. This displacement corresponds to the given round trip delay and thus also to the distance between the mobile station and the base station.

In a multi-path fading environment, it should be noted that establishing the maximum value is more difficult. Due to the multi-path fading, more maxima will appear corresponding to the individual delays of the signal paths.

If the number of samples per chips in the received signal r_os{i) is larger than one, the sum function can be expressed as:

(III)

where d_os is the number of samples per chip of the received signal r_os(k) . In order to cover a window of 1280 chips in the received signal, equation (III) must be evaluated for values of k in the interval from 0 to d_os -1280-1. In fig. 4, the result of the sum function has been shown in more detail for a d_os = 2.

By selecting a value of the over sampling factor d_os , the number of samples per chip in S_os(k) is given. As illustrated in figure 4, the chance of detecting a peak in S_os(k) in- creases when the over-sampling factor d_os increases. However, the computational complexity also increases when increasing the over-sampling factor d .

Summary of the invention

It is a first object of the invention to reduce the computational complexity of detecting coded streams for a single dwell detector.

This object has been accomplished by claim 1.

It is a secondary object of the invention to reduce the computational complexity of detecting coded streams for a single dwell detector.

This object has been accomplished by claim 2.

It is a third object to set forth a high performance computation reduction method.

This object has been accomplished by claim 3.

It is a further object to increases the detection probability while coping with slow fading channels.

This object has been accomplished by the subject matter of claim 9.

It is a further object to cope with fast fading channels.

This object has been accomplished by the subject matter of claim 11. Further objects and advantages will appear from the following detailed description of preferred embodiments of the invention. Brief description of the drawings

Fig. 1 shows an exemplary preamble and the possible unknown offset or delay at the base station,

fig. 2 shows a prior art detection principle for detection offset value k,

fig. 3 shows an impulse response of a raised cosine filter and - when a root raised cosine filter is used at receiver and transmitter - the response also equals a sum function of Gold coded signatures as a function of window displacement k ,

fig. 4, shows the result of the prior art Gold sum function for a sampling rate of twice the chip rate,

fig. 5, shows the result of the Gold sum function according to the invention for a sampling rate of twice the chip rate,

fig. 6 shows a single dwell detector according to the invention,

fig. 7 is a flow diagram relating to a single dwell detector of fig. 6,

fig. 8 shows a double dwell detector according to the invention, and

fig. 9 is a flow diagram relating to the double dwell detector of fig. 8.

Description of preferred embodiments of the invention

Fig 6 shows a single dwell detector according to a first embodiment of the invention, comprising a multiplier 21 , a summation unit 22, a plurality of fast Hadamard transform units 23, an interpolation unit 24, and a dwell detector 25. The dwell detector 25 comprises a plurality of absolute quadratic calculators 26, a set of summation units 27 and a discrimination unit 28.

The functionality of the single dwell detector according to the invention shall now be ex- plained with reference to fig. 7, which is a flow diagram relating to a single dwell detector of fig. 6.

In step a) of fig. 7, the radio signal is sampled, r(i).

In step b), the sampled radio signal is segmented into segments, e.g. into 4 segments (M). This is done in order to combat fading and frequency errors. Then each segment is demultiplexed into 16 streams, 36, corresponding to 16 possible signatures. All samples of each stream are multiplied in multiplier 21 with the cell specific scrambling code S_r-Pre, _n, 35, time reversed and conjugated. A coherent accumulation of all samples in each stream is then carried out, 22, which is a summation over 64 values in the case of 4 segments and 16 streams (i.e. 4096(N_pre) /16(s) /4(M)=64). The results of these coherent accumulations are then fed into respective fast Hadamard transform (FHT) units 23, in which the signals are multiplied with the respective signature codes. From this unit 23, one value is delivered for each signature. Also, this unit 23 is repeated for all seg- ments and all antennas (in the present example 1 antenna is provided).

The above procedure is carried out for K_os =1280 offset delays. In fig. 7, it has been illustrated that in multiplier 21 , a data amount corresponding to K_os = 1280 times s=16 = 20480 data streams/values are processed.

Subsequently, in step c), an interpolation is performed in 16 parallel stages (times 1280).

The results from the FHT unit 23 are interpreted in unit 24 as described in the following.

According to a first aspect of the invention, the incoming signal is sampled at a sample rate in the range of 1 sample per chip. To compensate for the reduced amount of information due to the low sampling rate, fictitious symbol values at intermediate positions of the samples are approximated. From the chip spaced sampled values of S(Jc) in equation (II), higher rate sampled values are achieved according to the following method of interpolation performed in unit 24:

Note that the interpolated values S_os(k) for k = 0,... ,2-1280-1 , are based on chip sampled values S(k) for /c = 0,...,1280-l .

D According to the invention, the absolute square of S_os(k) is used for detecting preambles. The interpolation can then be formulated as

which is subsequently performed in the quadratic calculators 26 in the single dwell detector 25.

The interpolation according to equation (V) at an over sampling rate of 2 samples per chip, corresponding to a sampling interval of V₂ a chip, has been indicated in fig. 5. The interpolated absolutely squared sum function values have been indicated between the sampled values. As appears from comparing with fig. 4, the interpolated peak value differs from the true value as indicated in fig. 5. It can be shown that the effect of more accurately detecting the peak also generally appears for other outcomes than the exemplary displacement shown in fig. 5. Accordingly, in step d), the maximum sum function is resolved (in dwell detector 25) for the given window displacement k in each of the 16 parallel stages. Here, the results from all of the 16 parallel stages are compared in the dwell detector 25 dwell detector, in which the signals are absolutely squared in calculators 26, added over all segments and all antennas (not shown) in unit 27. The signature number, corresponding to one of the parallel stages, which results in the largest value for at least one window displacement k, is regarded as a detected signature by the processing of discrimination unit 28.

Second embodiment of the invention - double dwell detector

It is known in the art, to resolve the preamble in a double dwell detector.

Fig 8 shows a double dwell detector according to a second embodiment of the invention, comprising a multiplier 21 , a summation unit 22, an interpolation unit 24, a distribution unit 31 , a first dwell detector 29, a data selection unit 24, a plurality of fast Hadamard transform signature filters 23, and a second dwell detector 25.

The first dwell detector comprises a plurality of absolute quadratic calculators 26, a set of summation units 27 and a discrimination unit 28. The second dwell detector comprises a plurality of absolute quadratic calculators 26, a summation unit 27 and a discrimination unit 28.

The double well approach is based on having a two step preamble detector where a reduced number of possible delays are selected in the first dwell whereby only the latter delays are used in signature filtering and evaluated in a second dwell. In order to further reduce computationally complexity, the despreading with scrambling codes may be done with one sample per chip. Then an interpolation to two samples per chips must be done just before first dwell detector. This interpolation reduces the chip rate processing significantly.

The principle shall now be explained with reference to figs. 8 and 9.

In the double dwell preamble detector, the detection of preambles is done in two steps. In the first step, confer first dwell detector 29, a detection is done in order to select a

(small) subset of delays for which preambles were received out of the whole set of de- lays from 0 to 1280 chips. This set of delays should be at least half chip sampled, i.e. 2*1280 positions should be considered. This detection is done without considering which preamble number, which was transmitted in the mobile station. Consequently, the first dwell 29 only detects delays, not preamble signature numbers.

Before information is transmitted to the second dwell detector 25, signature matched filters 23 are used, but only for delays selected in the first dwell 29. Preamble numbers can now be detected in the second dwell detector.

More specifically according to step a), the radio signal is received and sampled. Up to and including the despreading, all delays are de-spread, without any over sampling, i.e. the despreading is done at a rate equal to the chip rate.

In the preamble detection, the whole received signal should not be added coherently in order to combat frequency errors and fading. Before detection, the signal is thus split into a number of segments, typically 2 or 4. All coherent accumulations are done within these segments individually.

In step b), the signal stream is multiplied with the scrambling code corresponding to the given cell. The above procedure is carried out for K_os =1280 offset delays. In fig. 7, it has been illustrated that in multiplier 21, a data amount corresponding to K_os = 1280 times s=16 = 20480 data streams/values are processed.

In order to improve detection performance, these chip spaced delays are interpolated, step c), to half chip spaced samples whereby the interpolation according to the invention is carried out as explained in the foregoing (Times 1280).

For each segment, step d), the received samples are de-multiplexed into 16 streams, each containing every 16'th sample of the segment. All samples in each stream are now modulated with the same preamble symbol, compare equation (I), such that each of these streams now can be de-spread separately. Thus each of these streams is de- spread with the scrambling code complex conjugated and a summation is done over samples. In order to remove the PAPR modulation, see equation (I), a de-rotation is also done at the same time as multiplication with S_r__{pre n} (k) . In the first dwell detector 29, step e), each of the 16 streams is absolute squared individually, and non-coherently accumulated over all 16 streams, all segments and all antennas.

Next, the delays, which should be passed on to the second dwell, should be chosen. Several approaches might be applied for this selection. For example, a decision variable can be compared against a threshold for each delay, and those delays for which this decision variable exceeds the threshold are passed on to the second dwell. This threshold should then be scaled with a noise variance estimate.

For each of these reduced numbers of delays, 16 filters matched to the signatures are applied. The Fast Hadamard Transform (FHT) implement these 16 signature filters 23.

Before the second dwell detector 25, the result of the FHT 23 is absolute squared in calculators 26 and finally non-coherently added in summation units 27 over all segments and all antennas resulting in a decision variable, one for each signature number.

A threshold is then used in the second dwell detector, step f), which should be scaled according to a noise variance estimate. If the decision variable exceeds this threshold for at least one delay value a corresponding signature number is consequently regarded as detected.

As is known in the art, by using a double dwell detector, the computational complexity can be reduced. This reduction is appreciated, as the number of samples, which can be removed before signature filtering. According to the invention and as appears from the flow diagram of fig. 9, the amount of processing power is considerably reduced when using the approximation method described above in connection with a double dwell detector. As can be understood from fig. 9, the amount of information streams, needing computational processing are considerably reduced, due to the early data reduction for the approximation and the delayed signature branching stemming form the double dwell approach.

An explanation shall now be given which specifies the calculations performed in the double dwell detector, whereby reference is made to Table A showing equations (1) to (13). Selection every 16'th sample

Every 16'th sample is de-multiplexed into 16 separate streams as

(1) where r_t{ή) is a received signal for antenna number i and for

•

• all streams 0 < n < 16 ,

• all delays 0 < k_MDS ≤ L — l and

• all antennas 0 < /< N_α -l .

Segmentation:

The received signal is split into M segments as

(2), where , for all segments 0≤ m ≤ M-l

Despreading and PAPR demodulation: of segment number m and delay k_MDS :

(3) where ( )^* denotes complex conjugate.

Coherent accumulation: (4) for 0 < n ≤ 15 where . The processing gain in this coherent accumulation is

K /16 .

Interpolation:

(5)

Note the division by square root of two in the interpolation. This scaling aims at keeping the same noise variance for interpolated and non-interpolated values. First Dwell non-coherent accumulation: (6)

First Dwell criteria: (7) where σ_n ² is an estimate of the noise variance, here approximated by

(8) and (9) is the processing gain in the coherent accumulation of the despreading in (8) such that equals the signal to noise ratio in the first dwell. The parameter K_noise is used to specify the sampling rate for the noise estimate.

Selecting the delays for the second dwell: is done as the N₀₂ delays, which result in largest, first dwell decision variable V_D2 (k_os) .

Signature filter (or Fast Hadamard Transform): (10) for all signatures 0 < s ≤ 15.

Second dwell non-coherent accumulation: (11).

Second dwell criteria: (12) where

(13) is the processing gain in the coherent accumulation of the despreading in (4) and (10).

Here the signal energy is the energy per chip of the received signal. Table A

Claims

Patent claims

1. Method of detecting the offset position (k <= 1280) of a repetitive code stream se- quence of a first string length (N_pre), the offset position laying within a maximum delay value (1280), such as a WCDMA (Wideband Code Division Multiplex Access) uplink preamble, the method comprising the steps of

wherein for each window displacement (k)

a) continuously receiving analog to digital converting and sampling a stream of incoming signals (r(i)) at a sampling rate ( F_s = 3,84 • 10⁶ chips / second) equal to at least the chip rate,

b) splitting into segments and de-spreading the incoming signals, calculating a sum function value (S(k)) for the given window displacement (k) involving multiplying the incoming signal r(i) with a conjugate chip sequence c*(i),

c) forming a sum function

), where (at periodic intervals) a fictitious value is interpolated between consecutive sum function values, and

d) resolving the window displacement (k) corresponding to the maximum value of the sum function

2. Method of detecting the offset position (k <= 1280) of a repetitive code stream sequence of a first string length (N_pre), the offset position laying within a maximum delay value (1280), such as a WCDMA uplink preamble, the method comprising the steps of

wherein for each window displacement (k)

a) continuously receiving analog to digital converting and sampling a stream of incoming signals (r(i)) at a sampling rate (F_s = 3,84 -10⁶ chips / second) equal to at least the chip rate,

b) de-multiplexing, splitting into segments, despreading, and coherently accumulating the incoming signal,

c) performing an interpolation to an increased number of samples per chip wherein for each these increased number of window displacements (Kos),

-) performing an absolute value squaring and non-coherently accumulation forming a first value ( β_m (K₀₈ ) ),

d) selecting (a reduced) number of window displacements (k) (for each possible signature code) and for each of these

e) applying signature filters to, absolute squaring and non-coherently adding interpolated values forming a sum function value (β_D2(S,K_0S)),

f) resolving the window displacement (k) corresponding to the maximum value of the sum function (β_D2(S,K_os)).

3. Method according to claim 1 or 2, wherein under the step of interpolation ((c);(c)) for every second sample value, interpolated values are formed according to

when Jc is even

when k is odd

4. Method according to any previous claim, wherein the code stream sequence consist of a non-repetitive scrambling code (S_r.p_re,_n(k)) and a repetitive signature code (C_sig,s(k)).

5. Method according to any previous claim, wherein the offset position is used to detect the presence of the given sought code stream sequence (c(k)).

6. Method according to claim 2, wherein

-) the incoming signal is de-multiplexed into multiple streams as (eq. 1)

-) the streams are slit into segments as (eq. 2)

-) each segment is despread as (eq. 3)

-) all samples in each despread segment are added, each segment is indi- vidually as (eq. 4).

7. Method according to claim 2 or 6, wherein

-) the interpolation is used such that the number of samples for each chip is increased, each segment individually as (eq. 5)

-) now again for each window placement Kos:

-) all segments are absolute squared individually

-) all segments are added as (eq. 6)

8. Method according to claim 2 or 7, wherein

-) a number of window displacements are selected and for each of these selected window placements:

-) signature filters are used on the interpolated values as (eq. 10)

-) all values absolute squared individually and non-coherently added as (eq.

11)

9. Method according to claim 6, wherein the number of segments is 1.

10. Method according to claim 6, wherein the number of segments are 2.

11. Method according to claim 6, wherein the number of segments are 4.

12. Method according to claim 2 wherein the selection of a reduced number of window placements are performed by selecting the (N_D2 ) delays with the largest value of the first value (β_m(K_0S) ).

13. Method according to claim 2, wherein the selection of a reduced number of window placements are performed by selecting the delays for which the first value ( β_DΪ (K_os) ) exceeds a threshold.