Method for automatically obtaining, in closed form, the coefficients of an equalizing network in a system of data transmission of Orthogonal Frequency Division Multiplexing (OFDM) type

The present invention refers to a method for the calculation of the interference coefficients in a data transmission system of the OFDM (Orthogonal Frequency Division Multiplexing) type with only the knowledge of the pulse response of the transmission channel.

It is known that a OFDM system transmits a block of data in a finite time T using N subcarriers equally spaced in frequency, on every of them being transmitted a symbol having a duration not greater than T, using a QAM (Quadrature Amplitude Modulation) system and a shaping pulse g_T(t), where g_T (t)=1 for 0≤t≤T and g_T (t)=0 elsewhere.

The signal, passing through the transmission channel is distorted and each symbol is affected by an interference, generally due both to the symbols of the same block, and to the symbols of the preceding blocks.

At the reception side, in order to reduce to a minimum the distortion, it is necessary to use equalisers of the DFE (Decision Feedback Equaliser) type wherein, as known, the interference coefficients are obtained using adaptive, very complicated and expensive methods.

Aim of the present invention is to indicate a method for determining the interference coefficients, which only needs the knowledge of the pulse response of the channel and allows to obtain equalisers showing very simple and cheap structures.

In order to reach the said aim the present invention has as its subject a method for determining the interference coefficients showing the features which are indicated in the first claim.

Further characteristics and advantages of the present invention will be clear from the following description and from the attached drawings, which are given only as an indicative and not limiting example.

In the drawings: <ul id="ul0001" list-style="dash" compact="compact"><li>figure 1 shows the simplified block diagram of an OFDM system without channel distortion;</li><li>figure 2 shows the time behaviours of the pulses of the data in transmission and in reception;</li><li>figure 3 shows the block diagram of a known DFE equaliser;</li><li>figure 4 shows the block diagram of the transmitter and of the receiver of data according to the invention;</li><li>figure 5 shows in detail the receiver according to the invention.</li></ul>

In order to facilitate the reading, in the following the symbols and the terms used are listed. <ul id="ul0002" list-style="none" compact="compact"><li>Δf distance between two subcarriers;</li><li>f_P R.F. carrier;</li><li><maths id="math0001" num=""><math display="inline"><mrow><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>o</mtext></mrow></msub><mtext>=1/Δf</mtext></mrow></math><img file="EP0793369A2_D0001.tif" /></maths> time interval between two subcarriers;</li><li>T duration of a data block;</li><li><maths id="math0002" num=""><math display="inline"><mrow><mtext>B=N Δf</mtext></mrow></math><img file="EP0793369A2_D0002.tif" /></maths> total band of the modulated signal;</li><li><maths id="math0003" num=""><math display="inline"><mrow><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>G</mtext></mrow></msub><msub><mrow><mtext>=T-T</mtext></mrow><mrow><mtext>o</mtext></mrow></msub></mrow></math><img file="EP0793369A2_D0003.tif" /></maths>guard interval between consecutive blocks;</li><li>g_T (t) data shaping pulse in transmission of duration T;</li><li>g_R (t) data shaping pulse in reception of duration T_o.</li></ul> The symbol * used as an index means 'conjugated complex', used as an operator between two functions means 'convolution'.

Indicating with <maths id="math0004" num=""><math display="inline"><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>=a</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>+jb</mtext></mrow><mrow><mtext>n</mtext></mrow></msub></mrow></math><img file="EP0793369A2_D0004.tif" /></maths> the constellation point associated to the symbol having position n, to each block corresponds the following pulse signal modulated on a carrier frequency f_P:<maths id="math0005" num=""><img file="EP0793369A2_D0005.tif" /></maths> with <maths id="math0006" num=""><math display="inline"><mrow><msub><mrow><mtext>f'</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>=f</mtext></mrow><mrow><mtext>p</mtext></mrow></msub><msub><mrow><mtext>+f</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>   f</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext>=(n-N/2)Δf</mtext></mrow></math><img file="EP0793369A2_D0006.tif" /></maths>

The complex envelope of 1) is given by<maths id="math0007" num=""><img file="EP0793369A2_D0007.tif" /></maths> wherein:<maths id="math0008" num=""><math display="block"><mrow><msub><mrow><mtext>ϕ</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>(t) = g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msup><mrow><mtext>(t) e</mtext></mrow><mrow><mtext>j2π(n-N/2)Δft</mtext></mrow></msup></mrow></math><img file="EP0793369A2_D0008.tif" /></maths>

In order that the symbols can be recovered at reception side, it is necessary that the family j_n(t), where n=0,...,N-1, is orthogonal, or, more generally, it is advisable that a family exists y_k(t), where k=0,...,N-1, so that<maths id="math0009" num=""><img file="EP0793369A2_D0009.tif" /></maths> (here and in the following two functions indicated with a small letter and the corresponding capital are intended as connected by a Fourier transform)

In the said assumption in fact it results that the value of a symbol received in a position m is given by the interrelation between <u>x</u>(t) and y_m(t); in fact from the 3) we obtain:<maths id="math0010" num=""><img file="EP0793369A2_D0010.tif" /></maths> and if:<maths id="math0011" num="3'')"><math display="block"><mrow><msub><mrow><mtext>ψ</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>(t) = g</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msup><mrow><mtext>(t) e</mtext></mrow><mrow><mtext>j2π(n-N/2)Δft</mtext></mrow></msup></mrow></math><img file="EP0793369A2_D0011.tif" /></maths> the preceding orthogonality condition becomes:<maths id="math0012" num=""><img file="EP0793369A2_D0012.tif" /></maths>

In order to evaluate the band occupied by the complex envelope of the signal of a block, it is necessary to calculate the Fourier transform of the 2):<maths id="math0013" num=""><img file="EP0793369A2_D0013.tif" /></maths>

The band occupation is therefore N times that of the transmission data pulse G_T(f), and as the latter in praxis is equal to S times Δf, where S equal to a few units, it is possible to maintain the band of <u>x</u>(t) within a value which is equal to NΔf if the first and the last S symbols of the block are null, i.e. if the relevant carriers are suppressed. With this assumption, which has been considered valid in the following, we may assume that the total band of <u>x</u>(t) is given by <maths id="math0014" num=""><math display="inline"><mrow><mtext>B=NΔf</mtext></mrow></math><img file="EP0793369A2_D0014.tif" /></maths>, and therefore <u>x</u>(t) is described by its samples taken at distances 1/NΔf:<maths id="math0015" num=""><math display="block"><mrow><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>m</mtext></mrow></msub><mtext>= </mtext><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><mtext>(t-m/NΔ f)</mtext></mrow></math><img file="EP0793369A2_D0015.tif" /></maths> and, with a few calculations, we obtain:<maths id="math0016" num=""><math display="block"><mrow><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>m</mtext></mrow></msub><mtext>= </mtext><apply>&lt;sum /&gt;<lowlimit><mtext>n=S</mtext></lowlimit><uplimit><mtext>N-S</mtext></uplimit><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msubsup><mrow><mtext> W</mtext></mrow><mrow><mtext>N</mtext></mrow><mrow><mtext>mn</mtext></mrow></msubsup></mrow></apply></mrow></math><img file="EP0793369A2_D0016.tif" /></maths>

As can be seen the said samples are substantially given by the IDTF of the sequence (c_o....c_N-1) periodic with period N.

In particular, each sample of <u>x</u>(t) is represented by a term of the sequence multiplied by the corresponding sample of the transmission data pulse (taken with alternation of the sign, which represents a frequency translation of NΔf/2, necessary for centring the spectrum around the origin) and is different from zero in the index interval from S to N-S.

Taking in account the Parseval theorem, the 3'') and the orthogonality condition, through a few calculations we obtain:<maths id="math0017" num=""><img file="EP0793369A2_D0017.tif" /></maths>

From the 6) we see that, in case of a channel without transmission distortion, a generic transmitted symbol can be obtained from the DFT of the sequence of duration N, which in turn is obtained substantially adding samples of the received signal, alternately with changed sign, and multiplied by the corresponding samples of g_R(t).

In figure 1 it is represented the synthesis of the transmitter and the receiver according to the 5) and 6). When the <u>x</u>(t) signal passes through a channel having the characteristic of a low-pass filter LP, the output is given by <maths id="math0018" num=""><math display="inline"><mrow><munder accentunder="true"><mrow><mtext>y</mtext></mrow><mo>̲</mo></munder><mtext>(t)= </mtext><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><mtext>(t)* </mtext><munder accentunder="true"><mrow><mtext>h</mtext></mrow><mo>̲</mo></munder><mtext>(t)</mtext></mrow></math><img file="EP0793369A2_D0018.tif" /></maths>, wherein <u>h</u>(t) is the pulse response of the filter, which we assume different from zero in the interval (0-T_h).

It is possible to demonstrate that the symbol c_k, on the basis of what said before, becomes at the reception side:<maths id="math0019" num="7)"><math display="block"><mrow><msub><mrow><mtext>^C</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext> = </mtext><apply>&lt;sum /&gt;<lowlimit><mtext>n=S</mtext></lowlimit><uplimit><mtext>N-S</mtext></uplimit><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext> Ω (k-n;n) con S≤k≤N-S</mtext></mrow></apply></mrow></math><img file="EP0793369A2_D0019.tif" /></maths>

If<maths id="math0020" num=""><math display="block"><mrow><mtext>g'(t;n) = </mtext><apply>&lt;int /&gt;<lowlimit><mtext>τ=o</mtext></lowlimit><uplimit><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>h</mtext></mrow></msub></uplimit><mrow><munder accentunder="true"><mrow><mtext>h</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>(τ) g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msup><mrow><mtext>(t-τ) e</mtext></mrow><mrow><mtext>-j2π(n-N/2)Δfτ</mtext></mrow></msup></mrow></apply><mtext>   dτ =</mtext><mspace linebreak="newline" /><msub><mrow><mtext> = g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><mtext>(t)*(</mtext><munder accentunder="true"><mrow><mtext>h</mtext></mrow><mo>̲</mo></munder><msup><mrow><mtext>(t) e</mtext></mrow><mrow><mtext>-j2π(n-N/2)Δft)</mtext></mrow></msup></mrow></math><img file="EP0793369A2_D0020.tif" /></maths> and<maths id="math0021" num=""><math display="block"><mrow><mtext>Ω (m,n) = </mtext><apply>&lt;int /&gt;<lowlimit><mtext>o</mtext></lowlimit><uplimit><mtext>T</mtext></uplimit><mrow><msub><mrow><mtext>g'</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msub><mrow><mtext>(t,n) g</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msup><mrow><mtext>(t) e</mtext></mrow><mrow><mtext>-j2πnmΔft</mtext></mrow></msup><mtext> dt</mtext></mrow></apply><mtext>   m = k-n</mtext></mrow></math><img file="EP0793369A2_D0021.tif" /></maths> we obtain<maths id="math0022" num="7')"><math display="block"><mrow><msub><mrow><mtext>Ω(0,n)≅H(f</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext>)</mtext></mrow></math><img file="EP0793369A2_D0022.tif" /></maths>

The 7) proves that in general a transformation LP on <u>x</u>(t) gives at the reception side a cosymbolic interference, i.e. the k-th symbol depends on the symbols of the same block and on their distance from the considered symbol.

Let's consider the case of a traditional OFDM wherein (see fig. 2) g_T(t) and g_R(t) show the classic raised cosine behaviour and putting, for the sake of orthogonality,<maths id="math0023" num=""><img file="EP0793369A2_D0023.tif" /></maths> and, if T_G≥ T_h, from the previous formulas we obtain<maths id="math0024" num="8)"><math display="block"><mrow><msub><mrow><mtext>^C</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext> = </mtext><apply>&lt;sum /&gt;<lowlimit><mtext>n=S</mtext></lowlimit><uplimit><mtext>N-S</mtext></uplimit><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><munder accentunder="true"><mrow><mtext>H</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>(f</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext>)δ(k-n)</mtext></mrow></apply><msub><mrow><mtext>   cioè ñ</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext> = c</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext></mtext><munder accentunder="true"><mrow><mtext>H</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>(f</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>)</mtext></mrow></math><img file="EP0793369A2_D0024.tif" /></maths> where ∂(m)= Kronecker pulse.

From the 8) we see that, if the guard interval T_G is greater than the duration T_h of the pulse response of the channel, the symbol received in the k position is equal to that transmitted, multiplied by the value of the transfer function in correspondence of the carrier f_k, without any cosymbolic interference.

This result is however no more valid if the duration of the pulse response is longer than the guard interval; assuming e.g. that the pulse response is represented by an echo, delayed by D, where D&gt;T_G, we obtain<maths id="math0025" num=""><math display="block"><mrow><mtext>h(t) = δ(t) + </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><mtext> δ(t-D)</mtext></mrow></math><img file="EP0793369A2_D0025.tif" /></maths> where <u>a</u>= echo attenuation: and putting<maths id="math0026" num=""><math display="block"><mrow><mtable><mtr><mtd><mrow><mtable><mtr><mtd><mrow><munder accentunder="true"><mrow><mtext>α</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext> = </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><msup><mrow><mtext> e</mtext></mrow><mrow><mtext>-j2π(n-N/2)ΔfD</mtext></mrow></msup></mrow></mtd></mtr><mtr><mtd><mrow><mtext>ξ(m)= </mtext><apply>&lt;int /&gt;<lowlimit><mtext>o</mtext></lowlimit><uplimit><msub><mrow><mtext>T=T</mtext></mrow><mrow><mtext>G</mtext></mrow></msub><mtext>+To</mtext></uplimit><mrow><msub><mrow><mtext>g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msub><mrow><mtext>(t-D)g</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msup><mrow><mtext>(t)e</mtext></mrow><mrow><mtext>-j2πmΔft</mtext></mrow></msup><msub><mrow><mtext>dt=[G</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msup><mrow><mtext>(f)e</mtext></mrow><mrow><mtext>-j2πfD</mtext></mrow></msup><msub><mrow><mtext>]*G</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(f)|</mtext></mrow><mrow><mtext>f=mΔf</mtext></mrow></msub></mrow></apply></mrow></mtd></mtr></mtable></mrow></mtd></mtr></mtable></mrow></math><img file="EP0793369A2_D0026.tif" /></maths> we obtain<maths id="math0027" num="9)"><math display="block"><mrow><msub><mrow><mtext>^C</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>= </mtext><apply>&lt;sum /&gt;<lowlimit><mtext>n=S</mtext></lowlimit><uplimit><mtext>N-S</mtext></uplimit><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext> [δ(m+αnξ(m)]</mtext></mrow></apply><mtext>   m=k-n</mtext></mrow></math><img file="EP0793369A2_D0027.tif" /></maths>

From the 9) we can see that a symbol at a distance m from that interfered shows an interference coefficient which depends on the distance and on the position of the interfering symbol.

The 7) can be written<maths id="math0028" num=""><img file="EP0793369A2_D0028.tif" /></maths> where<maths id="math0029" num=""><math display="block"><mrow><mtext>η(m,k) = </mtext><mfrac><mrow><mtext>Ω(m;k-m)</mtext></mrow><mrow><mtext>Ω(0,k)</mtext></mrow></mfrac><mtext> = ICI coefficients (of cosymbolic interference)</mtext></mrow></math><img file="EP0793369A2_D0029.tif" /></maths>

In the real case, assuming that the duration of the pulse response is not greater than the duration T of the block, receiving a block we shall only take in account the previous block, as it is clear from fig. 2.

In fact, if T_h &gt; T, the symbols of a block may at the most trespass into the next block mixing with the symbols of the latter.

If we indicate with [c'n...c'N-1] the sequence of the previous block and with<maths id="math0030" num=""><img file="EP0793369A2_D0030.tif" /></maths> the relevant complex envelope, we obtain the general expression of the k-th symbol at the reception side<maths id="math0031" num=""><img file="EP0793369A2_D0031.tif" /></maths> where<maths id="math0032" num=""><math display="block"><mrow><mtable><mtr><mtd><mrow><mtable><mtr><mtd><mrow><msup><mrow><mtext>Ω'(m,n) = e</mtext></mrow><mrow><mtext>j2πfnT</mtext></mrow></msup><mtext></mtext><apply>&lt;int /&gt;<lowlimit><mtext>o</mtext></lowlimit><uplimit><mtext>T</mtext></uplimit><mrow><msub><mrow><mtext>g''</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msub><mrow><mtext>(t,n) g</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msup><mrow><mtext>(t) e</mtext></mrow><mrow><mtext>-j2πmΔft</mtext></mrow></msup><mtext> dt</mtext></mrow></apply></mrow></mtd></mtr><mtr><mtd><mrow><msub><mrow><mtext>g"</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msub><mrow><mtext>(t,n) =∫h(τ) g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msup><mrow><mtext>(t+T-τ) e</mtext></mrow><mrow><mtext>-j2πfnτ</mtext></mrow></msup><mtext> dτ</mtext></mrow></mtd></mtr></mtable></mrow></mtd></mtr></mtable></mrow></math><img file="EP0793369A2_D0032.tif" /></maths>

In a similar way to what we see before, the 10) can be written<maths id="math0033" num=""><math display="block"><mrow><mtable><mtr><mtd><mrow><mtable><mtr><mtd><mrow><mtable><mlabeledtr><mtext>10')</mtext><mtd><mrow><msub><mrow><mtext>^C</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext> = Ω(0,k)[c</mtext></mrow><mrow><mtext>k+</mtext></mrow></msub><apply>&lt;sum /&gt;<lowlimit><mtext>m≠o</mtext></lowlimit>&lt;uplimit /&gt;<mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>k-m η(m;k) +</mtext></mrow></msub></mrow></apply><mtext></mtext><apply>&lt;sum /&gt;<lowlimit><mtext>m</mtext></lowlimit>&lt;uplimit /&gt;<mrow><msub><mrow><mtext> c</mtext></mrow><mrow><mtext>k-m η'(m;k)</mtext></mrow></msub></mrow></apply><mtext>]</mtext></mrow></mtd></mlabeledtr></mtable></mrow></mtd></mtr><mtr><mtd><mrow><mtext>η'(m,k) = </mtext><mfrac><mrow><mtext>Ω'(m;k-m)</mtext></mrow><mrow><mtext>Ω(0,k)</mtext></mrow></mfrac><mtext> = ISI coefficients (of intersymbolic interference),</mtext></mrow></mtd></mtr></mtable></mrow></mtd></mtr></mtable></mrow></math><img file="EP0793369A2_D0033.tif" /></maths> depending on the symbols of the previous block.

It is possible to demonstrate that<maths id="math0034" num=""><img file="EP0793369A2_D0034.tif" /></maths>

In order to eliminate the ICI and ISI interferences it is necessary to use at the output of the DFT an equaliser, e.g. of the DFE (Decision Feedback Equaliser) type, which means a quite complicated structure, which is shown in fig. 3, wherein we have assumed <maths id="math0035" num=""><math display="inline"><mrow><mtext>h'(0,k)=0</mtext></mrow></math><img file="EP0793369A2_D0035.tif" /></maths>, due to the 7').

The said equaliser shows a "feedforward" part (left side) which defines the h(m,k) coefficients in order to minimise the ICI and a "feedback" part (right side) wherein the h'(m,k) coefficients are calculated in order to minimise the ISI.

The D block represents the decisional element which assigns to each estimated symbol c_k the nearest position c_k in the constellation.

<maths id="math0036" num=""><math display="inline"><mrow><msub><mrow><mtext>e</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>=c</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>-c</mtext></mrow><mrow><mtext>k</mtext></mrow></msub></mrow></math><img file="EP0793369A2_D0036.tif" /></maths> represents the symbol error. Usually after 15-20 blocks a known block is transmitted, whose symbols are memorised in the memory M of the receiver; e_k is periodically "refreshed" with real and not with estimated values.

The h coefficients are calculated using the LMS (Least Mean Square) algorithm:<maths id="math0037" num=""><math display="block"><mrow><mtable><mtr><mtd><mrow><mtable><mtr><mtd><mrow><msub><mrow><mtext>η</mtext></mrow><mrow><mtext>m,k</mtext></mrow></msub><msub><mrow><mtext>(t+T) = η</mtext></mrow><mrow><mtext>m,k</mtext></mrow></msub><msub><mrow><mtext>(t) - µ^C</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>*(t)e</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>(t)</mtext></mrow></mtd></mtr><mtr><mtd><mrow><msub><mrow><mtext>η'</mtext></mrow><mrow><mtext>m,k</mtext></mrow></msub><msub><mrow><mtext>(t+T) = η'</mtext></mrow><mrow><mtext>m,k</mtext></mrow></msub><msub><mrow><mtext>(t) + µc'</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>* e</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>(t)</mtext></mrow></mtd></mtr></mtable></mrow></mtd></mtr></mtable></mrow></math><img file="EP0793369A2_D0037.tif" /></maths> where <ul id="ul0003" list-style="none" compact="compact"><li>µ = adaptation parameter</li><li>t + T indicates the next block to the current block.</li></ul>

In the fig. 3 there is shown the equalisation of the K-th symbol assuming that the interference is due only to the two adjacent symbols.

Of course the pattern must be repeated for each of the N symbols so that it is evident the complexity of the known equalisation systems, which must use a high number of adders, multipliers and memories (delays).

Coming back to the already seen case of a channel whose pulse response shows only one echo (but the results, correctly evaluated, have a general validity) having an attenuation <u>a</u> and a delay D.

The pulse response is:<maths id="math0038" num=""><math display="block"><mrow><munder accentunder="true"><mrow><mtext>h</mtext></mrow><mo>̲</mo></munder><mtext>(t) = δ(t) + </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><mtext> δ(t-D)</mtext></mrow></math><img file="EP0793369A2_D0038.tif" /></maths> and therefore<maths id="math0039" num=""><img file="EP0793369A2_D0039.tif" /></maths> where<maths id="math0040" num=""><math display="block"><mrow><msub><mrow><mtext>α</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext> = </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><msup><mrow><mtext> e</mtext></mrow><mrow><mtext>-j2πf</mtext></mrow></msup><msup><mrow><mtext>k</mtext></mrow><mrow><mtext>D</mtext></mrow></msup><mtext>   ξ(m) = </mtext><apply>&lt;int /&gt;<lowlimit><mtext>o</mtext></lowlimit><uplimit><mtext>T</mtext></uplimit><mrow><msub><mrow><mtext>g</mtext></mrow><mrow><mtext>T</mtext></mrow></msub><msup><mrow><mtext>(t-D)gR(t)e</mtext></mrow><mrow><mtext>-j2nmΔft</mtext></mrow></msup><mtext> dt</mtext></mrow></apply></mrow></math><img file="EP0793369A2_D0040.tif" /></maths>

The calculation of the interference coefficients needs therefore only the knowledge of the z(m) coefficients.

If D≤T_G the product g_T(t-D) g_R(t) is a raised cosine function whose amplitude is 1/T_O, whose duration is T_O and which is centred at <maths id="math0041" num=""><math display="inline"><mrow><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>C</mtext></mrow></msub><msub><mrow><mtext> = T</mtext></mrow><mrow><mtext>G</mtext></mrow></msub><msub><mrow><mtext> +T</mtext></mrow><mrow><mtext>O</mtext></mrow></msub><mtext>/2</mtext></mrow></math><img file="EP0793369A2_D0041.tif" /></maths>; indicating with G_TΔ(f) the Fourier transform of such a function having a duration T_Δ and centred at the origin, we obtain:<maths id="math0042" num=""><img file="EP0793369A2_D0042.tif" /></maths>

In the case of D&gt;T_G, the said product shows a Fourier transform which cannot be expressed in a simple way in closed form; however, if D≥T_G + 2g T_O, then the product is still a raised cosine function having amplitude 1/T_O, but a duration <maths id="math0043" num=""><math display="inline"><mrow><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>Δ</mtext></mrow></msub><msub><mrow><mtext> = T + 2g T</mtext></mrow><mrow><mtext>O</mtext></mrow></msub><mtext> - D</mtext></mrow></math><img file="EP0793369A2_D0043.tif" /></maths> and centred at <maths id="math0044" num=""><math display="inline"><mrow><msub><mrow><mtext>T</mtext></mrow><mrow><mtext>C</mtext></mrow></msub><msub><mrow><mtext>=(T -2g T</mtext></mrow><mrow><mtext>O</mtext></mrow></msub><mtext>+ D) /2</mtext></mrow></math><img file="EP0793369A2_D0044.tif" /></maths>; in this case the coefficients z(m)<maths id="math0045" num=""><math display="block"><mrow><msub><mrow><mtext>ξ(m) = G</mtext></mrow><mrow><mtext>TΔ</mtext></mrow></msub><mtext>(mΔf) e-</mtext><msup><mrow><mtext>​</mtext></mrow><mrow><msub><mrow><mtext>j2πmΔfT</mtext></mrow><mrow><mtext>C</mtext></mrow></msub></mrow></msup></mrow></math><img file="EP0793369A2_D0045.tif" /></maths> and, more explicitly<maths id="math0046" num=""><img file="EP0793369A2_D0046.tif" /></maths>

The expression 13) is important because it allows, through the 11) and the 12), to obtain in analytic way (i.e. without an adaptive procedure) the interference coefficients starting only from the knowledge of the pulse response of the channel.

The latter can be estimated quite easily providing to build up once a data block to be transmitted as a Kronecker pulse and considering the received signal as representing the pulse response of the channel.

Knowing the interference coefficients and writing the 10') as follows<maths id="math0047" num=""><img file="EP0793369A2_D0047.tif" /></maths> the equaliser structure may be of the type shown in fig. 4, where in a similar way to what has been done in figure 3, only the coefficients relating to the symbols adjacent to the K-th are considered significant, and, for the sake of causality, in the FIR part the received symbols have been considered as an approximation of those equalised. Comparing the diagrams of figures 3) and 4) it appears clear how much the invention is advantageous in terms of simplification and, as a consequence, of cost.

For a better understanding of the invention, some further indications are provided.

The invention can be applied if the following assumptions are met: <ul id="ul0004" list-style="dash" compact="compact"><li>the transmission channel must be linear, i.e. <maths id="math0048" num=""><math display="inline"><mrow><mtext>y(t) = x(t)*h(t)</mtext></mrow></math><img file="EP0793369A2_D0048.tif" /></maths>, where h(t) is the impulse response of the system; it is known that the Fourier transform of h(t) is the transfer function of the channel; the impulse function in the analogue field is the known Dirac function <maths id="math0049" num=""><math display="inline"><mrow><mtext>d(t) = 1</mtext></mrow></math><img file="EP0793369A2_D0049.tif" /></maths> for t = 0 and <maths id="math0050" num=""><math display="inline"><mrow><mtext>d(t) = 0</mtext></mrow></math><img file="EP0793369A2_D0050.tif" /></maths> for t ≠ 0; the Kronecker impulse is the equivalent in the digital field <maths id="math0051" num=""><math display="inline"><mrow><mtext>d(n) = 1</mtext></mrow></math><img file="EP0793369A2_D0051.tif" /></maths> for n = 0 and <maths id="math0052" num=""><math display="inline"><mrow><mtext>d(n) = 0</mtext></mrow></math><img file="EP0793369A2_D0052.tif" /></maths> for n ≠ 0.</li><li>the transmission channel must be of the FIR (Finite Impulse Response) type, i.e. the response to the Kronecker impulse must be limited in time (limited number of echoes).</li><li>the distortion of the channel is represented by an echo delayed by D and attenuated by <u>a</u>. Therefore if at the input (transmission side) we have e. g. d(t), at the output (reception side) we have <maths id="math0053" num=""><math display="inline"><mrow><mtext>h(t) = d(t) + </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><mtext> d(t-D)</mtext></mrow></math><img file="EP0793369A2_D0053.tif" /></maths>, or, in digital notation, <maths id="math0054" num=""><math display="inline"><mrow><mtext>h(m) = d(m) + </mtext><munder accentunder="true"><mrow><mtext>a</mtext></mrow><mo>̲</mo></munder><mtext> d(m-k).</mtext></mrow></math><img file="EP0793369A2_D0054.tif" /></maths></li></ul>

Conclusions are valid also in the general case (multiple echoes).

It is however necessary that D&lt;T in order to limit the considerations to two contiguous blocks.

The invention starts from equations 11), 12) and 13), wherefore it is clear that the coefficients h and 'h , which up to now are obtained through an adaptive procedure, as in fig. 3, become known, if <u>a</u> and D are known, i.e. attenuation and delay at the reception side.

Practically we shall proceed as indicated in the block diagram of figure 4: Within a determined time interval T (which is periodically repeated, in order to take in account possible variations of the channel response), <u>x</u>_n, all null but one, are sent, i.e. the channel is tested using the Kronecker function.

This can be obtained, for instance, using a block of <maths id="math0055" num=""><math display="inline"><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext> = 1/N</mtext></mrow></math><img file="EP0793369A2_D0055.tif" /></maths> so that, from the 5), we obtain <maths id="math0056" num=""><math display="inline"><mrow><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>O</mtext></mrow></msub><mtext>=1</mtext></mrow></math><img file="EP0793369A2_D0056.tif" /></maths> (not considering the constant) and all other <u>x</u>=0.

If we want to have <maths id="math0057" num=""><math display="inline"><mrow><munder accentunder="true"><mrow><mtext>x</mtext></mrow><mo>̲</mo></munder><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>=1</mtext></mrow></math><img file="EP0793369A2_D0057.tif" /></maths> it must be <maths id="math0058" num=""><math display="inline"><mrow><msub><mrow><mtext>c</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msup><mrow><mtext> = 1/N e</mtext></mrow><mrow><mtext>-j2πnk/N</mtext></mrow></msup></mrow></math><img file="EP0793369A2_D0058.tif" /></maths>.

At the reception side there are present only <u>x</u>_k and <u>x</u>_k+D , i.e. the impulse and its echo, delayed by D and attenuated by <u>a</u>, where D and <u>a</u> are easily measurable.

By means of a microprocessor, usually available in the system for executing other functions, formulas 11) 12) and 13) are processed once in order to get the coefficients h (ICI) and h' (ISI). It is no more necessary any special circuitry, as it is clear from fig. 4.

The equaliser's structure can be of the type shown in detail in fig. 5, where, in a similar way to what has been done in figure 3, only the coefficients relating to the symbols adjacent to the k-th are considered significant, and, for the sake of causality, in the FIR part the received symbols have been considered as an approximation of those equalised.

We may summarise that, if the distortion of the channel is represented by one or more attenuated echoes, which is the most frequent case, equations 11), 12) and 13), allow to determine the ICI and ISI coefficients having the knowledge of the attenuation and the delay. The quickest way to know them is to transmit a single impulse (Kronecker) and measure the received signal.

From the above description the characteristics of the method according to the present invention will be clear, as clear are also the advantages thereof.

In particular the method according to the invention allows, in respect of the known methods, a simplification of the necessary apparatuses and therefore a cost reduction, as it is evident comparing figure 3 and figure 5.

It is clear that several variations may be applied to the method according to the present invention, without exiting from the novelty principles of the inventive idea, and it is also clear that in the implementing of the invention components and shapes of the details shown may be different and that the same may be substituted with technically equivalent elements.