An improved method for channel sounding and an apparatus using the method

Field of the invention

The invention relates to a method for radio channel measurement and parameter estimation. It addresses an improvement of the dynamic range of channel sounding employing a parameter estimation technique, such as EM (Expectation Maximisation) based approaches. The SAGE (Space Altering Generalized Expectation maximization) and the RIMAX algorithms are examples of EM-based parameter estimation techniques. The invention also relates to an arrangement which utilizes the method for estimation of radio channel characteristics of wireless signal paths.

Background of the invention

Parameter estimation techniques of known art cover a variety of different methods and algorithms.

Model-based estimation techniques, often called super-resolution techniques, are used in applications in which resolution better than the classical resolution is desired. For example, the classical delay resolution for a bandwidth of 100 MHz is 1/100 MHz, i.e. 10 ns. However, the super-resolution techniques can offer a resolution of 1 ns or better in favourable conditions.

The model-based techniques employ deconvolution techniques by which e.g. the bandwidth limitation can be mitigated. For example, if an impulse which has extremely wide bandwidth is transmitted through a transmitter (TX) having a bandwidth of B, the output signal (filtered impulse) has a bandwidth of B. In time domain the filtering effect can be expressed with convolution:

x(τ) = δ(τ)*f(τ) (1 )

where ( ^* ) denotes the convolution and δ(τ) is the impulse (Dirac's delta-function) and f(τ) is the impulse response of the transmitter. In frequency domain (1 ) becomes a simple multiplication and reads:

X(f) = δ(f) F(f) (2)

where δ(f) and F(f) are the Fourier transforms of δ(τ) and f(τ). Since δ(f) is constant for all frequencies the signal X(f) consists of those frequencies which are passed by the transmitter frequency response F(f).

As a result of the convolution the extremely short impulse δ(τ) is widened to a smooth signal waveform x(τ) the duration of which is determined by the bandwidth of the transmitter. Accordingly, the classical delay resolution is defined by this bandwidth. However, if the receiver (RX) has an exact model of f(τ) it can deconvolve the effect of the transmitter impulse response from the received signal (the deconvolution in time domain corresponds to division in frequency domain:

Y(f) = X(f) / F(f) = δ(f) F(f) / F(f) = δ(f) (3)

The equation indicates that an extremely narrow signal is obtained if the receiver has the exact model of the transmitted signal. Thus, the classical resolution can be dramatically exceeded and an extremely accurate delay resolution is possible. Instead of the Dirac's delta function also other wideband signal waveforms can be applied, e.g. the chirp signal which is often applied in radar techniques.

High dynamic range is an essential feature of radio channel sounding devices. It is limited mainly by the maximum transmit power, transmitter and receiver antenna gains and the receiver sensitivity. Channel sounding devices are suffering from low transmit power due to the limitations related to the antenna switch. They also suffer small antenna gains when patch antennas with wide patterns are applied.

The SAGE algorithm is a post processing tool used in channel sounding devices. It requires rather good signal-to-noise ratio (SNR) to operate reliably. Usually the SNR should be better than 8-10 dB when SAGE is applied. If only limited transmit power is available, which is usually the case with a prior art RF switch with a transmit power of about 26 dBm, this means that the SNR requirement limits the maximum measurement range significantly. This is especially true if wideband multi-antenna data is measured at high carrier frequencies. Thus a basic problem of the prior art technique is the limited dynamic range of measured impulse responses.

In MIMO application (multiple-input multiple-output communication) the SAGE algorithm operates in the antenna domain extracting for example the following characteristics of the wideband radio channel: number of significant propagation paths,

Doppler frequency, propagation delays, angle of arrival at the receiver (AoA), i.e. both elevation and azimuth, angle of departure at the transmitter (AoD), i.e. in both elevation and azimuth, polarization matrix of the propagation paths and rotation direction of polarization.

In the prior art, channel sounding is performed antenna element by antenna element, which is depicted in Fig. 1. The channel sounder 19 comprises an antenna switching unit (not shown in Fig. 1 ) which switches one antenna of the antenna array 13 at a time to the primary channel sounder device in the receiving end. Therefore, in the depicted example of Fig. 1 , each channel between a single transmitter antenna and a single receiver antenna is measured at a time. It causes the antenna gain at transmitter and receiver to be very small (roughly 0-1 dBi) and the dynamic range of the measurement is therefore very limited.

Fig. 1 depicts a principle of basic antenna-domain receiver or transmitter arrangement 10 known in the prior art in evaluating the direction-of-arrival (DoA) in azimuth plane. The antenna-domain arrangement scans all the possible azimuth angles of a patch antenna array 13 e.g. with steps of two degrees and takes into account the complex gain άk, reference 11a, of each antenna of the antenna array 13 in each direction φk, reference 12. Each of the patch antennas of the antenna array 13 is at a time connected in antenna domain 14 to a measurement device 19 in which for example SAGE algorithm is utilized. The measurement device 19 includes calibration files of antenna patterns 11 of individual antennas of the antenna array 13 which belongs either to a transmitter or receiver used. The measurement device 19 utilizes said calibration files and received signals when it estimates transmission channels between the transmitter and receiver under test using SAGE algorithm.

To improve the situation of basic antenna-domain SAGE algorithm following alternatives have been utilized. First alternative is to increase the transmit power. Second alternative is to increase the code length (large processing gain achieved) and third alternative is to average in the time domain.

Increasing the transmit power is very costly due to the fact that high-power antenna switches are very expensive because 10 W wideband switch costs about 50,000 €.

Increasing code length makes the measurement slow and reduces maximum measurable Doppler frequency significantly. Moreover, it leads to large amounts of data.

Averaging in time domain gets complicated if the transmitter or receiver is moving.

The super-resolution techniques perform the better the higher the SNR is. Thus it is important to develop measurement and data analysis methodology in a way that provides as high SNR as possible.

Summary of the invention

An object of the invention is to provide in MIMO environment a channel sounding arrangement and channel sounding method by which dynamic range of the channel sounding can be increased.

The object of the present invention is fulfilled by providing a channel sounding method comprising

- transmitting a sounding signal and

- receiving the sounding signal with multiple antennas, in which method

- in transmission

- several phase shifts are provided to the sounding signal to be transmitted if several transmitter antennas are utilized, and

- in reception

- different phase shifts are provided to the received sounding signals of different antennas of the receiver,

- the phase shifted sounding signals are combined coherently for creating a distinct receiver beam and angle of arrival, and

- MIMO channel parameters are estimated using the coherently combined sounding signals in a measurement apparatus, which utilizes a parameter estimation technique.

The object of the present invention is fulfilled also by providing a channel sounding arrangement comprising

- a transmitter with at least one antenna,

- a receiver with multiple antennas, where

- the transmitter comprises

- means for providing several phase shifts to a sounding signal to be transmitted if several transmitter antennas are utilized, and

- the receiver comprises

- means for providing different phase shifts to the received sounding signals of different antennas of the receiver,

- means for combining the phase shifted sounding signals coherently for creating a distinct receiver beam and angle of arrival, and

- means for estimating MIMO channel parameters using the coherently combined sounding signals in a measurement apparatus which utilizes a parameter estimation technique.

The object of the present invention is fulfilled also by providing a MIMO channel sounder transmitter antenna array which comprises means for providing for each antenna a different phase shift to the sounding signal to be transmitted.

The object of the present invention is fulfilled also by providing a MIMO channel sounder receiver antenna array which comprises

- means for providing different phase shifts to the received sounding signals of different antennas, and

- means for combining the phase shifted sounding signals coherently for creating a distinct receiver beam and angle of arrival.

Some advantageous embodiments of the invention are disclosed in the dependent claims.

An advantage of the invention is that the link budget of the measurement can be greatly increased by antenna array gain, which leads to larger measurement range or alternatively to better dynamic range at a specific sounding distance. Thus it improves the quality of radio channel sounding especially in outdoor environments

Another advantage of the invention is that it is simple, robust and low-cost.

A further advantage of the invention is that it does not require any changes in the estimation algorithm if SAGE algorithm is utilized.

A further advantage of the invention is that the antenna gain of an ULA array (Uniform Linear patch Antenna) using the invention can be 7-8 dBi while it is about 0-1

dBi for the prior art patch antenna array. Thus, for example, due to the increased beam gain of a 6-element ULA the received signal power can be up to 14-16 dB larger for a single propagation path compared to a case in which patch antennas are employed at the transmitter and the receiver.

A further advantage of the invention is that it allows simple dipole antenna arrays with metal backplanes (reflectors) to be used. If ULAs of dipole antennas are used for channel sounding the gain in dynamic range can be up to 20-30 dB.

Yet another advantage of the invention is that it allows also the use of two- dimensional planar arrays. Then the beams can be formed and optimised in both azimuth and elevation domains.

The idea of the invention is basically as follows: The invention combines the effective fixed beam approach with the SAGE parameter estimation method. The method is called hereafter beamformer-SAGE method. The method enables efficient utilisation of the specific properties of the SAGE using coherently combined antenna-domain signals in desired directions. In the beamformer-SAGE method the receiver has knowledge of the transmitter and receiver beam patterns. In practise this means that the complex beam patterns have been measured when calibrating the transmitter and receiver antenna arrays with the corresponding phasing matrices. Therefore the radio channel is measured by switching the transmitter and receiver beams instead of switching between single antennas like during prior art antenna-domain SAGE process. Thus each impulse response is measured in a co-ordinated fashion employing all the antenna elements of the antenna array at both the transmitter and receiver end. All antenna signals are summed up coherently at the transmitter and receiver which bring high array gains. This reduces any need to use higher transmitter power if better dynamic sounding range is needed.

In general, the invention includes an idea of combining beamforming with some known parameter estimation algorithm. Parameter estimation algorithms can be based on super-resolution algorithms. Examples of usable super-resolution algorithms are for example SAGE and RIMAX algorithms as well as ESPRIT, MUSIC and CLEAN algorithms and their derivatives. SAGE and RIMAX are also examples of expectation maximisation algorithms, which are also usable in the current invention.

Further scope of applicability of the present invention will become apparent from the detailed description given hereafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.

Brief description of the drawings

The present invention will become more fully understood from the detailed description given herein below and from the accompanying drawings which are given by way of illustration only, and thus are not limitative of the present invention and wherein

Fig. 1 shows a prior art channel sounding antenna arrangement,

Fig. 2 shows as an example a channel sounding antenna array arrangement according to the invention,

Fig. 3a shows an exemplary receiver antenna array arrangement according to the invention,

Fig. 3b shows as an example a Butler matrix suitable for an antenna array of four antennas, and

Fig. 4 shows as an exemplary a flow chart including main stages of the channel sounding using the beamformer-SAGE method.

Detailed description

Fig. 1 was discussed in conjunction with the description of the prior art.

In the beamformer-SAGE method channel sounding is performed by using high- gain narrow beams rather than low-gain, wide-beam antennas as is the case with known patch antennas. The narrow beams of beamformer-SAGE method are advantageously applied both to the transmitter and the receiver. The basic SAGE algorithm is flexible and therefore it can employ different kinds of calibrated antenna patterns in transmission channel parameter estimation. Therefore, SAGE can also be applied to narrow beams as long as the calibrated beam patterns have been measured and are known to the transmitter and the receiver antenna array.

In practise the beamformer-SAGE method requires specific antenna arrays and beam forming blocks. One simple antenna constellation includes a Uniform Linear Array (ULA) with inter-element spacing of one half of a wavelength and a phase shifting Butler-matrix. A M x M Butler-matrix provides such relative phase shifts between different ULA antennas that create M spatially orthogonal beams. M equals 2 ^π and where n = 1 , 2, 3,...

For example, a 4 x 4 Butler-matrix has the following phase shifts:

0 - 135 - 270 - 405

0 - 45 - 90 - 135

W = ( 4)

0 45 90 135

0 135 270 405

The relative phase shifts are the same for one antenna, a reference antenna (column 1 ), and are linearly increasing for the other antennas. In equation (4) the first row corresponds to the relative phase shifts for beam 1 , the second row corresponds to beam 2, and so on. Beams are directed symmetrically on both sides of the ULA broadside.

The Butler-matrix can be integrated into the antenna unit as an analog phase shift network, i.e. an analog delay line. Thus it is very simple and does not require online calibration. Only once-in-a-life-time calibration is needed in the production phase. ULA antenna arrays offer significant gains in the link budget, which determines the dynamic range of the measured impulse responses. If omnidirectional sounding is needed then e.g. three ULAs can be applied in a 3-sector fashion.

The beamformer-SAGE method according to the invention is by no means limited to the Butler-matrix and the corresponding orthogonal beams. Any phasing matrix will apply as long as the corresponding beam patterns are appropriately modelled for the channel parameter estimation phase.

The beamformer-SAGE method according to the invention gives significant gain in the dynamic range of channel sounding. It improves the quality of radio channel sounding especially in outdoor environments. For example, if a 6-element ULA is applied the array gain of 8 dB is achieved both in transmitter and receiver. Thus, by assuming constant transmitter power the link budget (dynamic range of the

measurement) is improved by 16 dB. The gain can be even larger if two- dimensional planar arrays are employed or if reflector type dipoles are used in the ULA. Alternatively, for a specific distance range, the performance and reliability of the SAGE algorithm is greatly improved.

However, the inventive idea is also applicable in a situation where only the receiving end utilizes beamformer-SAGE method. In that case the measurement apparatus comprises calibration files of prior art antenna domain transmitter and calibration files of the beam domain receiver.

Fig. 2 illustrates an example of an antenna array 23 arrangement 20 utilized in the beam domain according to the invention both in the transmitter and receiver.

In beamformer-SAGE approach, when utilized in a receiver, all received antenna signals are first fed in antenna domain 24 to a beamformer 25. The beamformer 25 causes a different phase shift to the signals from different antennas of the antenna array 23. With specific phase shifts at different antennas, the transmitted sounding signal combines coherently in the radio channel into the specific transmit direction. The coherent combination of the phase-shifted signals causes a formation of narrow receiver beam 21.

Using the beamformer 25 it is possible to get a narrow receiver antenna beam with a composite gain β _k , reference 21a, to each direction φ _k , reference 22. The composite gain is:

β _k = Mά _k , (6)

where M is the number of antennas in the antenna array 23 and όκ is the complex gain 21a of an individual antenna belonging to the antenna array 23. From the beamformer 25 the antenna signals are fed in a beam domain 26, one beam at a time, to a measurement device 19 where SAGE algorithm is applied.

It is noted that the beamforming gain can be obtained also in the transmitter. In the transmission end sounding signals to be transmitted are inputted from the transmitter device to the beamformer 25 one beam domain channel at a time. The inputted signals are then phase shifted in the beamformer 25 before conveying them to the antennas. In the radio channel excited radio waves from different antennas sum up coherently, which causes a narrow transmitter beam 21 to be formed. The

formed transmitter beam 21 has a complex antenna gain 21a and a distinct azimuth and elevation angle 22. The antenna gain and said angles of the beam 21 depend on the number of the antennas and the phase shifts used between the antennas.

In one advantageous embodiment the beamformer 25 used either in the transmitter or receiver is an analog Butler matrix.

However, in principle any kind of phasing matrixes can be applied instead of the Butler matrix. For example, the number of beams can be different from the number of antennas in the ULA. In that embodiment is utilized a beamforming matrix which forms more transmitter and/or receiver beams than there are antennas in the transmitter and receiver arrays.

The beamforming phase shifts can be readily calculated for any direction of arrival or direction of departure using the well-known array response vector for a linear uniform array:

a(θ) = [1 , exp(-j 2πd sin(θ)/λ), exp(-j 4πd sin(θ)/λ) .. exp(-j2π(M-1 )d sin(θ)/λ)] (7)

in which a(θ) is the array steering vector (antenna specific phase shifts) to the azimuth direction of θ, d is the inter-element distance between adjacent antennas in the array, M is the number of antennas in the antenna array, and λ is the wavelength of the carrier frequency. Thus it is possible to choose any azimuth direction θ and calculate the corresponding antenna-specific phase shifts to steer the beam to the direction of θ. If N beam directions are desired the N array steering vectors form an NxM beamforming matrix. Moreover, it is possible to shape the beam patterns by using the well-known antenna aperture tapering (windowing) techniques in selecting the amplitude weights of different antenna elements. For example, a Gaussian, Chebyshev, Bartlett, Hamming or Blackman window function can be applied for the amplitude tapering of the beamforming weight vectors. The window functions are employed to smooth the beam patterns and to decrease the level of the side-lobes.

Fig. 3a depicts a block diagram of an exemplary implementation of the invention. An antenna array 33 is connected in antenna domain 34 to a beamformer 35. The antenna array 33 can comprise for example patch antennas, dipole antennas or reflector antennas.

In one advantageous embodiment the antenna array 33 is two-dimensional. In this embodiment the beam of the antenna array can be formed both in azimuth and elevation domain.

The beamformer 35 comprises advantageously phase shifting circuits known in the art. After the phase shift the signals from different antennas are combined coherently in the radio channel to form beam domain signals according to the invention, which are utilized in the channel parameter estimation.

The beamformer 35 can be integrated as part of the antenna array 33. From the beamformer 35 phase-shifted and combined antenna signals are fed in the beam domain to an antenna switching unit 37. The antenna switching unit 37 connects one of the beamformer 35 outputs to a measurement apparatus 39 at a time, for example PROPSOUND™CS, which makes the channel sounding. It applies advantageously SAGE algorithm for fulfilling the channel sounding. The measurement apparatus 39 can estimate using SAGE algorithm for example the following channel characteristics: number of significant propagation paths, Doppler frequency, propagation delays, angle of arrival at the receiver, i.e. both elevation and azimuth, angle of departure at the transmitter, i.e. in both elevation and azimuth, polarization matrix of the propagation paths and rotation direction of polarization.

Fig. 3b depicts one possible solution for the beamformer 35 of a transmitter or receiver. This exemplary embodiment comprises a 4 x 4 Butler matrix 35a, which is essentially an analog phase shift network. In the depicted example of Fig. 3b four discrete antennas (not shown in Fig. 3b) are connected in the antenna domain 34 to the Butler matrix 35a. The connection is fulfilled by outputs (transmitter) or inputs (receiver) of the Butler matrix, references 34a, 34b, 34c and 34d. The Butler matrix comprises four phase shifting circuits, references 351-354. Each of the phase shifting circuits 351-354 comprises four phase shift elements having different phase shifts. Each of the phase shift circuits comprises in the beam domain 36 an input (transmitter) or output (receiver), references 36a, 36b, 36c and 36d. The inputs or outputs 36a, 36b, 36c and 36d are connected to the antenna switching unit (not shown in Fig. 3b) in the beam domain 36. In the transmitter the antenna switching unit connects one of the inputs of the Butler matrix 35a to the transmitter at a time. In receiver the antenna switching unit connects one of the outputs of the Butler matrix 35a to the channel sounder 39.

The exemplary phase shifts are depicted in Fig. 3b with four columns in each phase shifting circuit 351-354. The phase shifts of a certain column in each phase shifting circuit correspond to rows of the Butler matrix according to the equation (4). For example the phase shifts in the first columns are the zero phase shift of the reference antenna. Phase shifts for the second antenna can be found in the second column. Phase shifts for the third antenna can be found in the third column and the phase shifts for the fourth antenna can be found in the fourth column of the Butler matrix 35a.

The Butler matrix produces spatially orthogonal beams 21 when the phase-shifted signals are coherently combined. The number of the orthogonal beams 21 is determined by the order of the Butler matrix. However, any kind of phase shift network could advantageously be applied to optimize the number of beam, beam widths, gains and beam directions. The number of beams could also be smaller than the number of antennas in the array.

The utilization of the inventive approach to a typical channel sounder measurement can include for example the following steps:

- designing a phasing matrix for the transmitter antenna array

- designing a phasing matrix for the receiver antenna array

- connecting the designed phasing networks to the corresponding antenna arrays

- making calibration measurements for the transmitter and receiver beams in order to get a model for complex gains of the narrow beams

- providing the calibrated transmitter and receiver beam pattern models to the measurement apparatus, and

- making a radio channel measurement using the narrow beams of the transmitter and receiver and applying an estimation algorithm, for example SAGE algorithm, to the beam domain.

Fig. 4 depicts as an exemplary flow chart actual channel sounding steps when implemented in a MIMO environment. The preceding steps before actual measurement are fulfilled before starting a channel sounding process according to the invention.

The channel sounding is started in step 40. Both the transmitter and receiver utilized in the channel sounding are on standby mode. In step 41 the transmitter begins transmission using a narrow beam 21 transmission in a defined direction. The

narrow beam 21 is accomplished by using several antennas to which a measurement signal is inputted through one input of a beamformer.

The beamformer comprises some known phasing matrix. In one advantageous embodiment the phase shift is accomplished using a Butler matrix of size N x N, where N equals 2 ^π and where n = 1 , 2, 3,... The matrix size N equals the number of available antennas of the antenna array 33. The transmitter antenna gain, beam width, azimuth or elevation angle of the beam in beam domain is determined by the number of transmitter antennas and utilized phase shifts between the antennas.

In step 42 the multipath signals arrive to antennas of the receiver system. The number of antennas utilized can be for example M. The received antenna signals are transferred in antenna domain 34 into a beamformer 35 according to the invention. The beamformer 35 can advantageously be accomplished using a Butler matrix of size M x M.

Then, in step 43 the antenna signals are phase-shifted in the beamformer and after that coherently combined to beam domain signals 36. The coherent combination of the antenna signals in the beamformer causes narrowing of the receiver beam 21 in the beam domain. At the same time the antenna gain increases in the direction of the beam 21. Also an azimuth or elevation angle of the beam domain is determined by the phase shift network. The number of outputted beam domain signals can advantageously be equal to the number of antennas, i.e. M.

In step 44 beam domain signals of one beamformer 35 output 36a-36d are switched through the antenna switching unit 37 to the channel sounder 39. The channel sounder advantageously utilizes some parameter estimation technique in the channel estimation. Advantageously it applies SAGE algorithm to the beam domain signals.

After estimating a received channel, it is checked in step 45, if the channel sounding has been fulfilled for all transmission and receiving beam combinations. If all transmitter beams have been used in the measurement and all those transmissions have been received using all receiver beams, the channel sounding ends in step 46. However, if some of the receiver beams have not been used in the estimation, or if some of the transmitter beams are not yet transmitted, the process returns to step 41 and the process repeats the foregoing process.

In the description above is mainly described the usage of SAGE algorithm for the parameter estimation. It must be understood that SAGE is only one example of super-resolution-based algorithms, and any other applicable algorithm, such as RIMAX can also be used. These two are also examples of expectation maximization algorithms. In addition, the parameter estimation can be accomplished with any applicable super-resolution algorithm. Examples of such super-resolution algorithms are ESPRIT, MUSIC and CLEAN and their counterparts.

The invention being thus described, it will be obvious that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the invention, and all such modifications as would be obvious to one skilled in the art are intended to be included within the scope of the following claims.